1 2021-11-20 Jan Hubicka <hubicka@ucw.cz>
4 * ipa-modref.c (ignore_nondeterminism_p): Allow looping pure/cont.
5 (merge_call_side_effects): Improve debug output.
7 2021-11-20 Jan Hubicka <hubicka@ucw.cz>
10 * ipa-pure-const.c (propagate_pure_const): Fix merging of loping flag.
12 2021-11-20 Jeff Law <jeffreyalaw@gmail.com>
14 PR tree-optimization/103226
15 * config/bfin/bfin.md (doloop pattern, splitter and expander): Clobber
18 2021-11-20 Andrew Pinski <apinski@marvell.com>
20 PR tree-optimization/103220
21 * match.pd ((type) X bitop CST): Don't check if CST
22 fits into the type if only the sign changes.
24 2021-11-20 Alexandre Oliva <oliva@adacore.com>
26 PR tree-optimization/102988
27 * gimple-harden-conditionals.cc (detach_value): Copy SSA_NAME
30 2021-11-19 Iain Sandoe <iain@sandoe.co.uk>
32 * doc/invoke.texi: Remove whitespace after an @option.
34 2021-11-19 Paul A. Clarke <pc@us.ibm.com>
36 * config/rs6000/emmintrin.h (_mm_sad_epu8): Use vec_absd when
37 _ARCH_PWR9, optimize vec_sum2s when LE.
39 2021-11-19 Iain Sandoe <iain@sandoe.co.uk>
42 * config/darwin-driver.c (darwin_driver_init): Handle exported
43 symbols and symbol lists (suppress automatic export of the TLS
45 * config/darwin.c (darwin_rename_builtins): Remove workaround.
46 * config/darwin.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
47 (REAL_LIBGCC_SPEC): Handle revised library uses.
48 * config/darwin.opt (nodefaultexport): New.
49 * config/i386/darwin.h (PR80556_WORKAROUND): Remove.
50 * config/i386/darwin32-biarch.h (PR80556_WORKAROUND): Likewise.
51 * config/i386/darwin64-biarch.h (PR80556_WORKAROUND): Likewise.
53 2021-11-19 Martin Jambor <mjambor@suse.cz>
55 * opts.c (default_options_table): Switch off
56 flag_semantic_interposition at Ofast.
57 * doc/invoke.texi (Optimize Options): Document that Ofast switches off
58 -fsemantic-interposition.
60 2021-11-19 Jan Hubicka <hubicka@ucw.cz>
62 * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Do not guard modref
63 by !gimple_call_chain.
65 2021-11-19 Martin Sebor <msebor@redhat.com>
69 * doc/invoke.texi (-Waddress): Update.
71 2021-11-19 Andrew MacLeod <amacleod@redhat.com>
73 PR tree-optimization/103254
74 * gimple-range-gori.cc (range_def_chain::get_def_chain): Limit the
75 depth for all statements with multple ssa names.
77 2021-11-19 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
79 * config/s390/s390.md (define_peephole2): Variable insn points
80 to the first matched insn. Use peep2_next_insn(1) to refer to
81 the second matched insn.
83 2021-11-19 Tamar Christina <tamar.christina@arm.com>
85 PR tree-optimization/103311
87 * tree-vect-slp-patterns.c (vect_validate_multiplication): Fix CONJ
89 (complex_mul_pattern::matches): Move check downwards.
91 2021-11-19 Martin Liska <mliska@suse.cz>
94 2021-11-19 Martin Liska <mliska@suse.cz>
96 * cfgexpand.c (pass_expand::execute): Use option directly.
97 * function.c (allocate_struct_function): Likewise.
98 * gimple-low.c (lower_function_body): Likewise.
99 (lower_stmt): Likewise.
100 * gimple-ssa-backprop.c (backprop::prepare_change): Likewise.
101 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call): Likewise.
102 * ipa-split.c (split_function): Likewise.
103 * lto-streamer-in.c (input_function): Likewise.
104 * sese.c (sese_insert_phis_for_liveouts): Likewise.
105 * ssa-iterators.h (num_imm_uses): Likewise.
106 * tree-cfg.c (make_blocks): Likewise.
107 (gimple_merge_blocks): Likewise.
108 * tree-inline.c (tree_function_versioning): Likewise.
109 * tree-loop-distribution.c (generate_loops_for_partition): Likewise.
110 * tree-sra.c (analyze_access_subtree): Likewise.
111 * tree-ssa-dce.c (remove_dead_stmt): Likewise.
112 * tree-ssa-loop-ivopts.c (remove_unused_ivs): Likewise.
113 * tree-ssa-phiopt.c (spaceship_replacement): Likewise.
114 * tree-ssa-reassoc.c (reassoc_remove_stmt): Likewise.
115 * tree-ssa-tail-merge.c (tail_merge_optimize): Likewise.
116 * tree-ssa-threadedge.c (propagate_threaded_block_debug_into): Likewise.
117 * tree-ssa.c (gimple_replace_ssa_lhs): Likewise.
118 (target_for_debug_bind): Likewise.
119 (insert_debug_temp_for_var_def): Likewise.
120 (insert_debug_temps_for_defs): Likewise.
121 (reset_debug_uses): Likewise.
122 * tree-ssanames.c (release_ssa_name_fn): Likewise.
123 * tree-vect-loop-manip.c (adjust_vec_debug_stmts): Likewise.
124 (adjust_debug_stmts): Likewise.
125 (adjust_phi_and_debug_stmts): Likewise.
126 (vect_do_peeling): Likewise.
127 * tree-vect-loop.c (vect_transform_loop_stmt): Likewise.
128 (vect_transform_loop): Likewise.
129 * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): Remove
130 (MAY_HAVE_DEBUG_BIND_STMTS): Remove.
131 (MAY_HAVE_DEBUG_STMTS): Use options directly.
133 2021-11-19 Giuliano Belinassi <gbelinassi@suse.de>
135 * gcc.c (process_command): Skip dumpdir override if file is a
137 * doc/invoke.texi: Update -dumpdir documentation.
139 2021-11-19 Andrew Pinski <apinski@marvell.com>
141 PR tree-optimization/103314
142 * match.pd ((type) X op CST): Restrict the equal
143 TYPE_PRECISION case to GIMPLE only.
145 2021-11-19 Martin Liska <mliska@suse.cz>
148 * ipa-modref-tree.h (struct modref_parm_map): Add default
150 * ipa-modref.c (ipa_merge_modref_summary_after_inlining): Use it.
152 2021-11-19 Richard Biener <rguenther@suse.de>
155 * tree-eh.c (operation_could_trap_helper_p): Properly handle
156 fixed-point RDIV_EXPR.
158 2021-11-19 Richard Biener <rguenther@suse.de>
160 PR tree-optimization/102436
161 * tree-ssa-loop-im.c (execute_sm_if_changed): Add mode
162 to just create the if structure and return the then block.
163 (execute_sm): Add flag to indicate the var will re-use
165 (hoist_memory_references): Support a single conditional
166 block with all stores as special case.
168 2021-11-19 Andrew Pinski <apinski@marvell.com>
170 PR tree-optimization/103317
171 * tree-ssa-phiopt.c (minmax_replacement): For the non empty
172 middle bb case, check to make sure it has a single predecessor.
174 2021-11-19 Andrew Pinski <apinski@marvell.com>
176 PR tree-optimization/103257
178 ((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0):
179 Disable until !canonicalize_math_p ().
181 2021-11-19 Marek Polacek <polacek@redhat.com>
185 * doc/invoke.texi: Update documentation for -Wuninitialized.
186 * tree.c (stabilize_reference): Set location.
188 2021-11-19 liuhongt <hongtao.liu@intel.com>
191 * config/i386/x86-tune-costs.h (skylake_cost): Reduce cost of
192 storing 256/512-bit SSE register to be equal to cost of
193 unaligned store to avoid odd alignment peeling.
194 (icelake_cost): Ditto.
196 2021-11-18 David Edelsohn <dje.gcc@gmail.com>
198 * config/rs6000/predicates.md (current_file_function_operand):
199 Add flag_semantic_interposition to call of decl_replaceable_p.
201 2021-11-18 Martin Liska <mliska@suse.cz>
203 * ipa-modref.c (analyze_function): Do not execute the code
204 only if dump_file != NULL.
206 2021-11-18 Martin Liska <mliska@suse.cz>
208 * ipa-modref.c (analyze_function): Use fnode instead of repeated
209 cgraph_node::get (current_function_decl).
211 2021-11-18 Jan Hubicka <hubicka@ucw.cz>
213 * cgraph.c (cgraph_node::get_availability): Update call of
215 (cgraph_node::verify_node): Verify that semantic_interposition flag
217 * cgraph.h: (symtab_node): Add semantic_interposition flag.
218 * cgraphclones.c (set_new_clone_decl_and_node_flags): Clear
219 semantic_interposition flag.
220 * cgraphunit.c (cgraph_node::finalize_function): Set
221 semantic_interposition flag.
222 (cgraph_node::add_new_function): Likewise.
223 (varpool_node::finalize_decl): Likewise.
224 (cgraph_node::create_wrapper): Likewise.
225 * common.opt (fsemantic-interposition): Turn to optimization node.
226 * lto-cgraph.c (lto_output_node): Stream semantic_interposition.
227 (lto_output_varpool_node): Likewise.
228 (input_overwrite_node): Likewise.
229 (input_varpool_node): Likewise.
230 * symtab.c (symtab_node::dump_base): Dump new flag.
231 * varasm.c (decl_replaceable_p): Add semantic_interposition_p
233 * varasm.h (decl_replaceable_p): Update declaration.
234 * varpool.c (varpool_node::ctor_useable_for_folding_p):
235 Use semantic_interposition flag.
236 (varpool_node::get_availability): Likewise.
237 (varpool_node::create_alias): Copy semantic_interposition flag.
239 2021-11-18 Jan Hubicka <hubicka@ucw.cz>
242 * ipa-modref.c (modref_eaf_analysis::merge_call_lhs_flags): Unused
243 parameter may still be returned.
244 (modref_eaf_analysis::analyze_ssa_name): Call merge_call_lhs_flags
245 even for unused function args.
247 2021-11-18 Tamar Christina <tamar.christina@arm.com>
249 PR tree-optimization/103311
250 * tree-vect-slp-patterns.c (complex_mul_pattern::matches,
251 complex_fms_pattern::matches): Check for multiplications.
253 2021-11-18 H.J. Lu <hjl.tools@gmail.com>
256 * config/i386/i386.c (ix86_output_jmp_thunk_or_indirect): Emit
257 CS prefix for -mindirect-branch-cs-prefix.
258 (ix86_output_indirect_branch_via_reg): Likewise.
259 * config/i386/i386.opt: Add -mindirect-branch-cs-prefix.
260 * doc/invoke.texi: Document -mindirect-branch-cs-prefix.
262 2021-11-18 David Edelsohn <dje.gcc@gmail.com>
264 * config/rs6000/driver-rs6000.c (detect_processor_aix): Add
267 2021-11-18 Tamar Christina <tamar.christina@arm.com>
269 PR tree-optimization/103253
270 * tree-ssa-math-opts.c (convert_mult_to_fma): Check for LHS.
272 2021-11-18 Matthias Kretz <m.kretz@gsi.de>
274 * doc/extend.texi: Document __builtin_assoc_barrier.
276 2021-11-18 Martin Liska <mliska@suse.cz>
278 * cfgexpand.c (pass_expand::execute): Use option directly.
279 * function.c (allocate_struct_function): Likewise.
280 * gimple-low.c (lower_function_body): Likewise.
281 (lower_stmt): Likewise.
282 * gimple-ssa-backprop.c (backprop::prepare_change): Likewise.
283 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call): Likewise.
284 * ipa-split.c (split_function): Likewise.
285 * lto-streamer-in.c (input_function): Likewise.
286 * sese.c (sese_insert_phis_for_liveouts): Likewise.
287 * ssa-iterators.h (num_imm_uses): Likewise.
288 * tree-cfg.c (make_blocks): Likewise.
289 (gimple_merge_blocks): Likewise.
290 * tree-inline.c (tree_function_versioning): Likewise.
291 * tree-loop-distribution.c (generate_loops_for_partition): Likewise.
292 * tree-sra.c (analyze_access_subtree): Likewise.
293 * tree-ssa-dce.c (remove_dead_stmt): Likewise.
294 * tree-ssa-loop-ivopts.c (remove_unused_ivs): Likewise.
295 * tree-ssa-phiopt.c (spaceship_replacement): Likewise.
296 * tree-ssa-reassoc.c (reassoc_remove_stmt): Likewise.
297 * tree-ssa-tail-merge.c (tail_merge_optimize): Likewise.
298 * tree-ssa-threadedge.c (propagate_threaded_block_debug_into): Likewise.
299 * tree-ssa.c (gimple_replace_ssa_lhs): Likewise.
300 (target_for_debug_bind): Likewise.
301 (insert_debug_temp_for_var_def): Likewise.
302 (insert_debug_temps_for_defs): Likewise.
303 (reset_debug_uses): Likewise.
304 * tree-ssanames.c (release_ssa_name_fn): Likewise.
305 * tree-vect-loop-manip.c (adjust_vec_debug_stmts): Likewise.
306 (adjust_debug_stmts): Likewise.
307 (adjust_phi_and_debug_stmts): Likewise.
308 (vect_do_peeling): Likewise.
309 * tree-vect-loop.c (vect_transform_loop_stmt): Likewise.
310 (vect_transform_loop): Likewise.
311 * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): Remove
312 (MAY_HAVE_DEBUG_BIND_STMTS): Remove.
313 (MAY_HAVE_DEBUG_STMTS): Use options directly.
315 2021-11-18 Richard Biener <rguenther@suse.de>
317 PR tree-optimization/103277
318 * tree-ssa-dse.c (need_ab_cleanup): New.
319 (dse_optimize_redundant_stores): Adjust.
320 (delete_dead_or_redundant_assignment): Get extra
321 need_ab_cleanup argument and set when abnormal cleanup is
323 (dse_optimize_call): Adjust.
324 (dse_optimize_stmt): Likewise.
325 (pass_dse::execute): Allocate and deallocate need_ab_cleanup.
326 Perform abnormal cleanup.
327 * tree-ssa-dse.h (delete_dead_or_redundant_assignment): Adjust.
329 2021-11-18 Hongyu Wang <hongyu.wang@intel.com>
331 * config/i386/i386-expand.c (ix86_expand_atomic_fetch_op_loop):
332 Adjust generated cfg to avoid infinite loop.
334 2021-11-18 konglin1 <lingling.kong@intel.com>
336 * config/i386/avx512fp16intrin.h (_mm512_mul_pch): Add alias for _mm512_fmul_pch.
337 (_mm512_mask_mul_pch): Likewise.
338 (_mm512_maskz_mul_pch): Likewise.
339 (_mm512_mul_round_pch): Likewise.
340 (_mm512_mask_mul_round_pch): Likewise.
341 (_mm512_maskz_mul_round_pch): Likewise.
342 (_mm512_cmul_pch): Likewise.
343 (_mm512_mask_cmul_pch): Likewise.
344 (_mm512_maskz_cmul_pch): Likewise.
345 (_mm512_cmul_round_pch): Likewise.
346 (_mm512_mask_cmul_round_pch): Likewise.
347 (_mm512_maskz_cmul_round_pch): Likewise.
348 (_mm_mul_sch): Likewise.
349 (_mm_mask_mul_sch): Likewise.
350 (_mm_maskz_mul_sch): Likewise.
351 (_mm_mul_round_sch): Likewise.
352 (_mm_mask_mul_round_sch): Likewise.
353 (_mm_maskz_mul_round_sch): Likewise.
354 (_mm_cmul_sch): Likewise.
355 (_mm_mask_cmul_sch): Likewise.
356 (_mm_maskz_cmul_sch): Likewise.
357 (_mm_cmul_round_sch): Likewise.
358 (_mm_mask_cmul_round_sch): Likewise.
359 (_mm_maskz_cmul_round_sch): Likewise.
360 * config/i386/avx512fp16vlintrin.h (_mm_mul_pch): Likewise.
361 (_mm_mask_mul_pch): Likewise.
362 (_mm_maskz_mul_pch): Likewise.
363 (_mm256_mul_pch): Likewise.
364 (_mm256_mask_mul_pch): Likewise.
365 (_mm256_maskz_mul_pch): Likewise.
366 (_mm_cmul_pch): Likewise.
367 (_mm_mask_cmul_pch): Likewise.
368 (_mm_maskz_cmul_pch): Likewise.
369 (_mm256_cmul_pch): Likewise.
370 (_mm256_mask_cmul_pch): Likewise.
371 (_mm256_maskz_cmul_pch): Likewise.
373 2021-11-17 Andrew Pinski <apinski@marvell.com>
375 PR tree-optimization/103228
376 PR tree-optimization/55177
377 * match.pd ((type) X bitop CST): Also do this
378 transformation for nop conversions.
380 2021-11-17 Martin Sebor <msebor@redhat.com>
382 PR tree-optimization/102759
383 * gimple-array-bounds.cc (build_printable_array_type): Move...
384 * gimple-ssa-warn-access.cc (build_printable_array_type): Avoid
385 pathological function redeclarations that remove a previously
387 Improve formatting of function arguments in informational notes.
388 * pointer-query.cc (build_printable_array_type): ...to here.
389 * pointer-query.h (build_printable_array_type): Declared.
391 2021-11-17 H.J. Lu <hjl.tools@gmail.com>
394 * config/i386/i386-opts.h (harden_sls): New enum.
395 * config/i386/i386.c (output_indirect_thunk): Mitigate against
396 SLS for function return.
397 (ix86_output_function_return): Likewise.
398 (ix86_output_jmp_thunk_or_indirect): Mitigate against indirect
400 (ix86_output_indirect_jmp): Likewise.
401 (ix86_output_call_insn): Likewise.
402 * config/i386/i386.opt: Add -mharden-sls=.
403 * doc/invoke.texi: Document -mharden-sls=.
405 2021-11-17 H.J. Lu <hjl.tools@gmail.com>
408 * config/i386/i386.c (ix86_code_end): Remove "%!" before ret.
409 (ix86_output_function_return): Likewise.
410 * config/i386/i386.md (simple_return_pop_internal): Likewise.
412 2021-11-17 Jan Hubicka <hubicka@ucw.cz>
415 * ipa-modref.c (read_modref_records): Fix streaminig in of every_access
418 2021-11-17 Uroš Bizjak <ubizjak@gmail.com>
420 * config/i386/i386.c (indirect_thunks_used): Redefine as HARD_REG_SET.
421 (ix86_code_end): Use TEST_HARD_REG_BIT on indirect_thunks_used.
422 (ix86_output_indirect_branch_via_reg): Use SET_HARD_REG_BIT
423 on indirect_thunks_used.
424 (ix86_output_indirect_function_return): Ditto.
426 2021-11-17 Jan Hubicka <hubicka@ucw.cz>
428 * ipa-modref-tree.c: Include cgraph.h and tree-streamer.h.
429 (modref_access_node::stream_out): New member function.
430 (modref_access_node::stream_in): New member function.
431 * ipa-modref-tree.h (modref_access_node::stream_out,
432 modref_access_node::stream_in): Declare.
433 * ipa-modref.c (modref_summary_lto::useful_p): Free useless kills.
434 (modref_summary_lto::dump): Dump kills.
435 (analyze_store): Record kills for LTO
436 (analyze_stmt): Likewise.
437 (modref_summaries_lto::duplicate): Duplicate kills.
438 (write_modref_records): Use new stream_out member function.
439 (read_modref_records): Likewise.
440 (modref_write): Stream out kills.
441 (read_section): Stream in kills
442 (remap_kills): New function.
443 (update_signature): Use it.
445 2021-11-17 Uroš Bizjak <ubizjak@gmail.com>
447 * config/i386/i386.h (LEGACY_SSE_REGNO_P): New predicate.
448 (SSE_REGNO_P): Use LEGACY_SSE_REGNO_P predicate.
449 * config/i386/i386.c (zero_all_vector_registers):
450 Use LEGACY_SSE_REGNO_P predicate.
451 (ix86_register_priority): Use REX_INT_REGNO_P, REX_SSE_REGNO_P
452 and EXT_REG_SSE_REGNO_P predicates.
453 (ix86_hard_regno_call_part_clobbered): Use REX_SSE_REGNO_P
454 and LEGACY_SSE_REGNO_P predicates.
456 2021-11-17 Jason Merrill <jason@redhat.com>
458 * doc/invoke.texi (C++ Dialect Options): Document
459 -fimplicit-constexpr.
461 2021-11-17 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
463 * config/aarch64/aarch64-modes.def (VECTOR_MODE): New V8DI mode.
464 * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Handle
466 * config/aarch64/iterators.md (define_mode_attr nunits): Add entry
469 2021-11-17 Martin Uecker <uecker@gcc.gnu.org>
473 * gimplify.c (gimplify_var_or_parm_decl): Update comment.
474 (gimplify_compound_lval): Gimplify base expression first.
475 (gimplify_target_expr): Add comment.
477 2021-11-17 Jakub Jelinek <jakub@redhat.com>
479 PR tree-optimization/103192
480 * tree-ssa-loop-im.c (move_computations_worker): Use
481 reset_flow_sensitive_info instead of manually clearing
482 SSA_NAME_RANGE_INFO and do it for all SSA_NAMEs, not just ones
485 2021-11-17 Jakub Jelinek <jakub@redhat.com>
487 PR tree-optimization/103255
488 * gimple-range-fold.cc (fold_using_range::range_of_address): Return
489 range_nonzero rather than unadjusted base's range. Formatting fixes.
491 2021-11-17 Richard Sandiford <richard.sandiford@arm.com>
493 * doc/md.texi (cond_fmin@var{mode}, cond_fmax@var{mode}): Document.
494 * optabs.def (cond_fmin_optab, cond_fmax_optab): New optabs.
495 * internal-fn.def (COND_FMIN, COND_FMAX): New functions.
496 * internal-fn.c (first_commutative_argument): Handle them.
497 (FOR_EACH_COND_FN_PAIR): Likewise.
498 * match.pd (UNCOND_BINARY, COND_BINARY): Likewise.
499 * config/aarch64/aarch64-sve.md (cond_<fmaxmin><mode>): New
502 2021-11-17 Kewen Lin <linkw@linux.ibm.com>
504 * config/i386/i386.md (*add<dwi>3_doubleword, *addv<dwi>4_doubleword,
505 *addv<dwi>4_doubleword_1, *sub<dwi>3_doubleword,
506 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1,
507 *add<dwi>3_doubleword_cc_overflow_1, *divmodsi4_const,
508 *neg<dwi>2_doubleword, *tls_dynamic_gnu2_combine_64_<mode>): Fix split
511 2021-11-17 Andrew Pinski <apinski@marvell.com>
513 PR tree-optimization/103288
514 * tree-ssa-phiopt.c (value_replacement): Return early if middle
515 block has more than one pred.
517 2021-11-17 Kewen Lin <linkw@linux.ibm.com>
519 * config/visium/visium.md (*add<mode>3_insn, *addsi3_insn, *addi3_insn,
520 *sub<mode>3_insn, *subsi3_insn, *subdi3_insn, *neg<mode>2_insn,
521 *negdi2_insn, *and<mode>3_insn, *ior<mode>3_insn, *xor<mode>3_insn,
522 *one_cmpl<mode>2_insn, *ashl<mode>3_insn, *ashr<mode>3_insn,
523 *lshr<mode>3_insn, *trunchiqi2_insn, *truncsihi2_insn,
524 *truncdisi2_insn, *extendqihi2_insn, *extendqisi2_insn,
525 *extendhisi2_insn, *extendsidi2_insn, *zero_extendqihi2_insn,
526 *zero_extendqisi2_insn, *zero_extendsidi2_insn): Fix split condition.
528 2021-11-17 Marek Polacek <polacek@redhat.com>
530 PR preprocessor/103026
531 * doc/invoke.texi: Document -Wbidi-chars.
533 2021-11-17 Jan Hubicka <hubicka@ucw.cz>
536 * ipa-modref.c (ipa_merge_modref_summary_after_inlining): Fix clearing
539 2021-11-16 Jan Hubicka <jh@suse.cz>
541 * ipa-modref.c (get_modref_function_summary): Declare.
542 * ipa-modref.h (get_modref_function_summary): New function.
543 * tree-ssa-dse.c (clear_live_bytes_for_ref): Break out from ...
544 (clear_bytes_written_by): ... here; also clear memory killed by
547 2021-11-16 Iain Sandoe <iain@sandoe.co.uk>
549 * ggc-common.c (gt_pch_save): If we cannot find a suitable
550 memory segment for save, then error-out, do not try to
552 (gt_pch_restore): Save the existing line table, and when
553 the replacement is being read, use that when constructing
556 2021-11-16 Peter Bergner <bergner@linux.ibm.com>
559 * config/rs6000/mma.md (*vsx_assemble_pair): Add early-clobber for
561 (*mma_assemble_acc): Likewise.
563 2021-11-16 Martin Sebor <msebor@redhat.com>
565 PR tree-optimization/102960
566 * gimple-fold.c (get_range_strlen): Take bitmap as an argument rather
567 than a pointer to it.
568 (get_range_strlen_tree): Same. Remove bitmap allocation. Use
570 (get_maxval_strlen): Use an auto_bitmap.
571 * tree-ssa-strlen.c (get_range_strlen_dynamic): Factor out PHI
573 (get_range_strlen_phi): ...into this function.
574 Avoid assuming maximum string length is constant
575 (printf_strlen_execute): Dump pointer query cache contents when
576 details are requisted.
578 2021-11-16 Jason Merrill <jason@redhat.com>
580 * langhooks.h (struct lang_hooks): Adjust comment.
581 * print-tree.c (print_node): Also call print_xnode hook for
584 2021-11-16 Andrew Pinski <apinski@marvell.com>
586 PR tree-optimization/103218
587 * match.pd: New pattern for "((type)(a<0)) << SIGNBITOFA".
589 2021-11-16 Claudiu Zissulescu <claziss@synopsys.com>
591 * config/arc/arc.md (maddhisi4): Use a single move to accumulator.
592 (umaddhisi4): Likewise.
593 (machi): Update pattern.
596 2021-11-16 Richard Biener <rguenther@suse.de>
598 PR tree-optimization/102880
599 * tree-ssa-dce.c (sort_phi_args): New function.
600 (make_forwarders_with_degenerate_phis): Likewise.
601 (perform_tree_ssa_dce): Call
602 make_forwarders_with_degenerate_phis.
604 2021-11-16 Richard Biener <rguenther@suse.de>
606 PR tree-optimization/102880
607 * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Push
608 single_pred (bb1) condition to places that really need it.
609 (match_simplify_replacement): Likewise.
610 (value_replacement): Likewise.
611 (replace_phi_edge_with_variable): Deal with extra edges
614 2021-11-16 Martin Jambor <mjambor@suse.cz>
616 * cfgexpand.c (expand_gimple_basic_block): Use build_debug_expr_decl,
617 add a fixme note about the mode assignment perhaps being unnecessary.
618 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call):
620 (ipa_param_body_adjustments::mark_dead_statements): Likewise.
621 (ipa_param_body_adjustments::reset_debug_stmts): Likewise.
622 * tree-inline.c (remap_ssa_name): Likewise.
623 (tree_function_versioning): Likewise.
624 * tree-into-ssa.c (rewrite_debug_stmt_uses): Likewise.
625 * tree-ssa-loop-ivopts.c (remove_unused_ivs): Likewise.
626 * tree-ssa.c (insert_debug_temp_for_var_def): Likewise.
628 2021-11-16 Jakub Jelinek <jakub@redhat.com>
630 PR tree-optimization/103208
631 * omp-expand.c (expand_omp_build_cond): New function.
632 (expand_omp_for_init_counts, expand_omp_for_init_vars,
633 expand_omp_for_static_nochunk, expand_omp_for_static_chunk): Use it.
635 2021-11-16 Jakub Jelinek <jakub@redhat.com>
637 PR tree-optimization/102009
638 * gimple-ssa-warn-access.cc (pass_waccess::check_alloc_size_call):
639 Punt if any of alloc_size arguments is out of bounds vs. number of
642 2021-11-16 Roger Sayle <roger@nextmovesoftware.com>
644 * config/i386/i386.md (*bmi2_rorx<mode3>_1): Make conditional
645 on !optimize_function_for_size_p.
646 (*<any_rotate><mode>3_1): Add preferred_for_size attribute.
647 (define_splits): Conditionalize on !optimize_function_for_size_p.
648 (*bmi2_rorxsi3_1_zext): Likewise.
649 (*<any_rotate>si2_1_zext): Add preferred_for_size attribute.
650 (define_splits): Conditionalize on !optimize_function_for_size_p.
652 2021-11-16 Jan Hubicka <jh@suse.cz>
655 * ipa-modref.c (merge_call_side_effects): Fix uninitialized
658 2021-11-16 Andrew Pinski <apinski@marvell.com>
660 PR tree-optimization/103245
661 * match.pd: Combine the abs pattern matching using multiplication.
662 Adding optional nop_convert too.
664 2021-11-16 H.J. Lu <hjl.tools@gmail.com>
667 * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Add a missing
670 2021-11-15 Siddhesh Poyarekar <siddhesh@gotplt.org>
672 * gimple-fold.c (gimple_fold_builtin_strncat): Use ranges to
673 determine if it is safe to transform to strcat.
674 (gimple_fold_builtin_snprintf): Likewise.
676 2021-11-15 Siddhesh Poyarekar <siddhesh@gotplt.org>
678 * gimple-fold.c (known_lower): New function.
679 (gimple_fold_builtin_strncat_chk,
680 gimple_fold_builtin_memory_chk, gimple_fold_builtin_stxcpy_chk,
681 gimple_fold_builtin_stxncpy_chk,
682 gimple_fold_builtin_snprintf_chk,
683 gimple_fold_builtin_sprintf_chk): Use it.
685 2021-11-15 Siddhesh Poyarekar <siddhesh@gotplt.org>
687 * gimple-fold.c (dump_transformation): New function.
688 (gimple_fold_builtin_stxcpy_chk,
689 gimple_fold_builtin_stxncpy_chk): Use it. Simplify to
690 BUILT_IN_STRNCPY if return value is not used.
692 2021-11-15 H.J. Lu <hjl.tools@gmail.com>
695 * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Check optab
696 before transforming equivalent, but slighly different cases to
697 their canonical forms.
699 2021-11-15 Iain Sandoe <iain@sandoe.co.uk>
702 * config/darwin.h (TARGET_DTORS_FROM_CXA_ATEXIT): New.
703 * doc/tm.texi: Regenerated.
704 * doc/tm.texi.in: Add TARGET_DTORS_FROM_CXA_ATEXIT hook.
705 * ipa.c (cgraph_build_static_cdtor_1): Return the built
707 (build_cxa_atexit_decl): New.
708 (build_dso_handle_decl): New.
709 (build_cxa_dtor_registrations): New.
710 (compare_cdtor_tu_order): New.
711 (build_cxa_atexit_fns): New.
712 (ipa_cdtor_merge): If dtors_from_cxa_atexit is set,
713 process the DTORs/CTORs accordingly.
714 (pass_ipa_cdtor_merge::gate): Also run if
715 dtors_from_cxa_atexit is set.
716 * target.def (dtors_from_cxa_atexit): New hook.
718 2021-11-15 Iain Sandoe <iain@sandoe.co.uk>
720 * config.in: Regenerate.
721 * configure: Regenerate.
722 * configure.ac: Test ld64 for -platform-version support.
724 2021-11-15 Thomas Schwinge <thomas@codesourcery.com>
726 * diagnostic-spec.h (typedef xint_hash_t)
727 (typedef xint_hash_map_t): Replace with...
728 (typedef nowarn_map_t): ... this.
729 (nowarn_map): Adjust.
730 * diagnostic-spec.c (nowarn_map, suppress_warning_at): Likewise.
732 2021-11-15 Thomas Schwinge <thomas@codesourcery.com>
734 * profile.c (branch_prob): Use 'location_hash' for
737 2021-11-15 Aldy Hernandez <aldyh@redhat.com>
739 PR tree-optimization/103207
740 * value-range.cc (irange::set): Drop overflow.
742 2021-11-15 Jakub Jelinek <jakub@redhat.com>
744 * gimplify.c (optimize_target_teams): Only add OMP_CLAUSE_THREAD_LIMIT
745 to OMP_TARGET_CLAUSES if it isn't there already.
747 2021-11-15 Aldy Hernandez <aldyh@redhat.com>
749 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
751 (path_range_query::ssa_defined_in_bb): New.
752 (path_range_query::ssa_range_in_phi): Avoid fold_range call that
753 could trigger additional lookups.
754 Do not use the cache for ARGs defined in this block.
755 (path_range_query::compute_ranges_in_block): Use ssa_defined_in_bb.
756 (path_range_query::maybe_register_phi_relation): Same.
757 (path_range_query::range_of_stmt): Adjust comment.
758 * gimple-range-path.h (ssa_defined_in_bb): New.
760 2021-11-15 Aldy Hernandez <aldyh@redhat.com>
762 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
763 Default to global range if nothing found.
765 2021-11-15 Richard Biener <rguenther@suse.de>
767 PR tree-optimization/103237
768 * tree-vect-loop.c (vect_is_simple_reduction): Fail for
769 double reductions with multiple inner loop LC PHI nodes.
771 2021-11-15 Hongyu Wang <hongyu.wang@intel.com>
774 * config/i386/i386-expand.c (ix86_expand_atomic_fetch_op_loop):
776 * config/i386/i386-options.c (ix86_target_string): Add
777 -mrelax-cmpxchg-loop flag.
778 (ix86_valid_target_attribute_inner_p): Likewise.
779 * config/i386/i386-protos.h (ix86_expand_atomic_fetch_op_loop):
780 New expand function prototype.
781 * config/i386/i386.opt: Add -mrelax-cmpxchg-loop.
782 * config/i386/sync.md (atomic_fetch_<logic><mode>): New expander
784 (atomic_<logic>_fetch<mode>): Likewise.
785 (atomic_fetch_nand<mode>): Likewise.
786 (atomic_nand_fetch<mode>): Likewise.
787 (atomic_fetch_<logic><mode>): New expander for DI,TI modes.
788 (atomic_<logic>_fetch<mode>): Likewise.
789 (atomic_fetch_nand<mode>): Likewise.
790 (atomic_nand_fetch<mode>): Likewise.
791 * doc/invoke.texi: Document -mrelax-cmpxchg-loop.
793 2021-11-15 Richard Biener <rguenther@suse.de>
795 PR tree-optimization/103219
796 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Use single_exit
797 to determine the exit for the VN region.
799 2021-11-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
801 * tree-ssa-loop.c (pass_vectorize): Move to tree-vectorizer.c.
802 (pass_data_vectorize): Likewise.
803 (make_pass_vectorize): Likewise.
804 * tree-vectorizer.c (vectorize_loops): Merge with
805 pass_vectorize::execute and replace cfun occurences with fun param.
806 (adjust_simduid_builtins): Add fun param, replace cfun occurences with
807 fun, and adjust callers approrpiately.
808 (note_simd_array_uses): Likewise.
809 (vect_loop_dist_alias_call): Likewise.
810 (set_uid_loop_bbs): Likewise.
811 (vect_transform_loops): Likewise.
812 (try_vectorize_loop_1): Likewise.
813 (try_vectorize_loop): Likewise.
815 2021-11-15 Jakub Jelinek <jakub@redhat.com>
818 * config/i386/sync.md (atomic_bit_test_and_set<mode>,
819 atomic_bit_test_and_complement<mode>,
820 atomic_bit_test_and_reset<mode>): Use OPTAB_WIDEN instead of
823 2021-11-15 Maciej W. Rozycki <macro@embecosm.com>
825 * config/vax/vax.c (vax_lra_p): New prototype and function.
826 (TARGET_LRA_P): Wire it.
827 * config/vax/vax.opt (mlra): New option.
828 * doc/invoke.texi (Option Summary, VAX Options): Document the
831 2021-11-14 Jan Hubicka <jh@suse.cz>
833 * ipa-modref.h (struct modref_summary): Add nondeterministic
834 and calls_interposable flags.
835 * ipa-modref.c (modref_summary::modref_summary): Initialize new flags.
836 (modref_summary::useful_p): Check new flags.
837 (struct modref_summary_lto): Add nondeterministic and
838 calls_interposable flags.
839 (modref_summary_lto::modref_summary_lto): Initialize new flags.
840 (modref_summary_lto::useful_p): Check new flags.
841 (modref_summary::dump): Dump new flags.
842 (modref_summary_lto::dump): Dump new flags.
843 (ignore_nondeterminism_p): New function.
844 (merge_call_side_effects): Merge new flags.
845 (process_fnspec): Likewise.
846 (analyze_load): Volatile access is nondeterministic.
847 (analyze_store): Liekwise.
848 (analyze_stmt): Volatile ASM is nondeterministic.
849 (analyze_function): Clear new flags.
850 (modref_summaries::duplicate): Duplicate new flags.
851 (modref_summaries_lto::duplicate): Duplicate new flags.
852 (modref_write): Stream new flags.
853 (read_section): Stream new flags.
854 (propagate_unknown_call): Update new flags.
855 (modref_propagate_in_scc): Propagate new flags.
856 * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Check
858 * tree-ssa-structalias.c (determine_global_memory_access):
861 2021-11-14 Maciej W. Rozycki <macro@embecosm.com>
863 * config/vax/vax.h (SET_RATIO): New macro.
864 * config/vax/vax.md (UNSPEC_SETMEM_FILL): New constant.
865 (setmemhi): New expander.
866 (setmemhi1): New insn and splitter.
867 (*setmemhi1): New insn.
869 2021-11-14 Jan Hubicka <hubicka@ucw.cz>
871 * ipa-modref-tree.c (modref_access_node::update_for_kills): New
873 (modref_access_node::merge_for_kills): Likewise.
874 (modref_access_node::insert_kill): Likewise.
875 * ipa-modref-tree.h (modref_access_node::update_for_kills,
876 modref_access_node::merge_for_kills, modref_access_node::insert_kill):
878 (modref_access_node::useful_for_kill): New member function.
879 * ipa-modref.c (modref_summary::useful_p): Release useless kills.
880 (lto_modref_summary): Add kills.
881 (modref_summary::dump): Dump kills.
882 (record_access): Add mdoref_access_node parameter.
883 (record_access_lto): Likewise.
884 (merge_call_side_effects): Merge kills.
885 (analyze_call): Add ALWAYS_EXECUTED param and pass it around.
886 (struct summary_ptrs): Add always_executed filed.
887 (analyze_load): Update.
888 (analyze_store): Update; record kills.
889 (analyze_stmt): Add always_executed; record kills in clobbers.
890 (analyze_function): Track always_executed.
891 (modref_summaries::duplicate): Duplicate kills.
892 (update_signature): Release kills.
893 * ipa-modref.h (struct modref_summary): Add kills.
894 * tree-ssa-alias.c (alias_stats): Add kill stats.
895 (dump_alias_stats): Dump kill stats.
896 (store_kills_ref_p): Break out from ...
897 (stmt_kills_ref_p): Use it; handle modref info based kills.
899 2021-11-14 Aldy Hernandez <aldyh@redhat.com>
901 PR tree-optimization/103229
902 * gimple-range-cache.cc (ssa_global_cache::clear): Do not pass
903 null value to memset.
905 2021-11-14 Jan Hubicka <hubicka@ucw.cz>
907 * ipa-modref-tree.c (modref_access_node::get_call_arg): New member
909 (modref_access_node::get_ao_ref): Likewise.
910 * ipa-modref-tree.h (modref_access_node::get_call_arg): Declare.
911 (modref_access_node::get_ao_ref): Declare.
912 * tree-ssa-alias.c (modref_may_conflict): Use new accessors.
913 * tree-ssa-dse.c (dse_optimize_call): Use new accessors.
915 2021-11-13 Jan Hubicka <jh@suse.cz>
918 * dbgcnt.def (ipa_attr): New counters.
919 * ipa-pure-const.c: Include dbgcnt.c
920 (ipa_make_function_const): Use debug counter.
921 (ipa_make_function_pure): Likewise.
922 (propagate_pure_const): Fix bug in my previous change.
924 2021-11-13 Jan Hubicka <hubicka@ucw.cz>
926 * ipa-modref-tree.c (modref_access_node::range_info_useful_p):
927 Offline from ipa-modref-tree.h.
928 (modref_access_node::dump): Move from ipa-modref.c; make member
930 * ipa-modref-tree.h (modref_access_node::range_info_useful_p.
931 modref_access_node::dump): Declare.
932 * ipa-modref.c (dump_access): Remove.
933 (dump_records): Update.
934 (dump_lto_records): Update.
935 (record_access): Update.
936 (record_access_lto): Update.
938 2021-11-13 Jan Hubicka <hubicka@ucw.cz>
940 * ipa-modref.c (modref_summary::modref_summary): Clear new flags.
941 (modref_summary::dump): Dump try_dse.
942 (modref_summary::finalize): Add FUN attribute; compute try-dse.
943 (analyze_function): Update.
944 (read_section): Update.
945 (update_signature): Update.
946 (pass_ipa_modref::execute): Update.
947 * ipa-modref.h (struct modref_summary):
948 * tree-ssa-alias.c (ao_ref_init_from_ptr_and_range): Export.
949 * tree-ssa-alias.h (ao_ref_init_from_ptr_and_range): Declare.
950 * tree-ssa-dse.c (dse_optimize_call): New function.
951 (dse_optimize_stmt): Use it.
953 2021-11-13 Jan Hubicka <hubicka@ucw.cz>
955 * ipa-modref-tree.c: Move #if CHECKING_P to proper place.
957 2021-11-13 Jan Hubicka <jh@suse.cz>
960 (struct modref_access_node): Move longer member functions to
962 (modref_ref_node::try_merge_with): Turn into modreef_acces_node member
964 * ipa-modref-tree.c (modref_access_node::contains): Move here
965 from ipa-modref-tree.h.
966 (modref_access_node::update): Likewise.
967 (modref_access_node::merge): Likewise.
968 (modref_access_node::closer_pair_p): Likewise.
969 (modref_access_node::forced_merge): Likewise.
970 (modref_access_node::update2): Likewise.
971 (modref_access_node::combined_offsets): Likewise.
972 (modref_access_node::try_merge_with): Likewise.
973 (modref_access_node::insert): Likewise.
975 2021-11-13 Jan Hubicka <jh@suse.cz>
977 * ipa-modref.c (modref_summary::global_memory_read_p): Remove.
978 (modref_summary::global_memory_written_p): Remove.
979 (modref_summary::dump): Dump new flags.
980 (modref_summary::finalize): New member function.
981 (analyze_function): Call it.
982 (read_section): Call it.
983 (update_signature): Call it.
984 (pass_ipa_modref::execute): Call it.
985 * ipa-modref.h (struct modref_summary): Remove
986 global_memory_read_p and global_memory_written_p.
987 Add global_memory_read, global_memory_written.
988 * tree-ssa-structalias.c (determine_global_memory_access):
991 2021-11-13 Jan Hubicka <jh@suse.cz>
993 * ipa-fnsummary.c (compute_fn_summary): Use type_attribut_allowed_p
994 * ipa-param-manipulation.c
995 (ipa_param_adjustments::type_attribute_allowed_p):
997 (drop_type_attribute_if_params_changed_p): New function.
998 (build_adjusted_function_type): Use it.
999 * ipa-param-manipulation.h: Add type_attribute_allowed_p.
1001 2021-11-13 David Malcolm <dmalcolm@redhat.com>
1003 * doc/invoke.texi (Static Analyzer Options): Add
1004 -Wno-analyzer-tainted-allocation-size,
1005 -Wno-analyzer-tainted-divisor, -Wno-analyzer-tainted-offset, and
1006 -Wno-analyzer-tainted-size to list. Add
1007 -Wanalyzer-tainted-allocation-size, -Wanalyzer-tainted-divisor,
1008 -Wanalyzer-tainted-offset, and -Wanalyzer-tainted-size to list
1009 of options effectively enabled by -fanalyzer.
1010 (-Wanalyzer-tainted-allocation-size): New.
1011 (-Wanalyzer-tainted-array-index): Tweak wording; add link to CWE.
1012 (-Wanalyzer-tainted-divisor): New.
1013 (-Wanalyzer-tainted-offset): New.
1014 (-Wanalyzer-tainted-size): New.
1016 2021-11-13 Jan Hubicka <jh@suse.cz>
1018 * attr-fnspec.h (attr_fnspec::arg_eaf_flags): Break out from ...
1019 * gimple.c (gimple_call_arg_flags): ... here.
1020 * ipa-modref.c (analyze_parms): Record flags known from fnspec.
1021 (modref_merge_call_site_flags): Use arg_eaf_flags.
1023 2021-11-13 Aldy Hernandez <aldyh@redhat.com>
1025 PR tree-optimization/103222
1026 * gimple-range-path.cc (path_range_query::compute_ranges_in_phis):
1028 (path_range_query::compute_ranges_in_block): Call
1029 compute_ranges_in_phis.
1030 * gimple-range-path.h (path_range_query::compute_ranges_in_phis):
1033 2021-11-13 Jan Hubicka <jh@suse.cz>
1035 * ipa-fnsummary.c (compute_fn_summary): Do not give up on signature
1036 changes on "fn spec" attribute; give up on varadic types.
1037 * ipa-param-manipulation.c: Include attribs.h.
1038 (build_adjusted_function_type): New parameter ARG_MODIFIED; if it is
1039 true remove "fn spec" attribute.
1040 (ipa_param_adjustments::build_new_function_type): Update.
1041 (ipa_param_body_adjustments::modify_formal_parameters): update.
1042 * ipa-sra.c: Include attribs.h.
1043 (ipa_sra_preliminary_function_checks): Do not check for TYPE_ATTRIBUTES.
1045 2021-11-13 Aldy Hernandez <aldyh@redhat.com>
1047 * gimple-range-path.cc (path_range_query::path_range_query): Merge
1049 (path_range_query::import_p): Move from header file.
1050 (path_range_query::~path_range_query): Adjust for combined ctors.
1051 * gimple-range-path.h: Merge ctors.
1052 (path_range_query::import_p): Move to .cc file.
1054 2021-11-13 Jan Hubicka <jh@suse.cz>
1056 PR tree-optimization/103182
1057 * ipa-modref.c (callee_to_caller_flags): Fix merging of flags.
1058 (modref_eaf_analysis::analyze_ssa_name): Fix merging of flags.
1060 2021-11-12 Stafford Horne <shorne@gmail.com>
1062 * config/or1k/or1k-protos.h (or1k_profile_hook): New function.
1063 * config/or1k/or1k.h (PROFILE_HOOK): Change macro to reference
1064 new function or1k_profile_hook.
1065 * config/or1k/or1k.c (struct machine_function): Add new field
1066 set_mcount_arg_insn.
1067 (or1k_profile_hook): New function.
1068 (or1k_init_pic_reg): Update to inject pic rtx after _mcount arg
1070 (or1k_frame_pointer_required): Frame pointer no longer needed
1073 2021-11-12 Jan Hubicka <jh@suse.cz>
1075 PR tree-optimization/103209
1076 * tree-ssa-structalias.c (find_func_aliases_for_call): Fix
1077 use of handle_rhs_call
1079 2021-11-12 Aldy Hernandez <aldyh@redhat.com>
1081 PR tree-optimization/103202
1082 * gimple-range-path.cc
1083 (path_range_query::compute_ranges_in_block): Solve PHI imports first.
1085 2021-11-12 Jan Hubicka <jh@suse.cz>
1087 * ipa-pure-const.c (propagate_pure_const): Remove redundant check;
1088 fix call of ipa_make_function_const and ipa_make_function_pure.
1090 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
1092 * config/aarch64/aarch64.c (aarch64_vector_op::n_advsimd_ops): Delete.
1093 (aarch64_vector_op::m_seen_loads): Likewise.
1094 (aarch64_vector_costs::aarch64_vector_costs): Don't push to
1096 (aarch64_vector_op::count_ops): Remove vectype and factor parameters.
1097 Remove code that tries to predict different vec_flags from the
1099 (aarch64_vector_costs::add_stmt_cost): Update accordingly.
1100 Remove m_advsimd_ops handling.
1102 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
1104 * config/aarch64/aarch64.c (aarch64_vector_costs::m_saw_sve_only_op)
1105 (aarch64_sve_only_stmt_p): Delete.
1106 (aarch64_vector_costs::prefer_unrolled_loop): New function,
1107 extracted from adjust_body_cost.
1108 (aarch64_vector_costs::better_main_loop_than_p): New function,
1109 using heuristics extracted from adjust_body_cost and
1110 adjust_body_cost_sve.
1111 (aarch64_vector_costs::adjust_body_cost_sve): Remove
1112 advsimd_cycles_per_iter and could_use_advsimd parameters.
1113 Update after changes above.
1114 (aarch64_vector_costs::adjust_body_cost): Update after changes above.
1116 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
1118 * config/aarch64/aarch64.c (aarch64_vec_op_count::m_vf_factor):
1119 New member variable.
1120 (aarch64_vec_op_count::aarch64_vec_op_count): Add a parameter for it.
1121 (aarch64_vec_op_count::vf_factor): New function.
1122 (aarch64_vector_costs::aarch64_vector_costs): When costing for
1123 neoverse-512tvb, pass a vf_factor of 2 for the Neoverse V1 version
1125 (aarch64_vector_costs::adjust_body_cost): Read the vf factor
1126 instead of hard-coding 2.
1128 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
1130 * config/aarch64/aarch64.c
1131 (aarch64_vec_op_count::rename_cycles_per_iter): New function.
1132 (aarch64_vec_op_count::min_nonpred_cycles_per_iter): Likewise.
1133 (aarch64_vec_op_count::min_pred_cycles_per_iter): Likewise.
1134 (aarch64_vec_op_count::min_cycles_per_iter): Likewise.
1135 (aarch64_vec_op_count::dump): Move earlier in file. Dump the
1136 above properties too.
1137 (aarch64_estimate_min_cycles_per_iter): Delete.
1138 (adjust_body_cost): Use aarch64_vec_op_count::min_cycles_per_iter
1139 instead of aarch64_estimate_min_cycles_per_iter. Rely on the dump
1140 routine to print CPI estimates.
1141 (adjust_body_cost_sve): Likewise. Use the other functions above
1142 instead of doing the work inline.
1144 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
1146 * config/aarch64/aarch64.c (aarch64_vec_op_count): Allow default
1148 (aarch64_vec_op_count::base_issue_info): Remove handling of null
1150 (aarch64_vec_op_count::simd_issue_info): Likewise.
1151 (aarch64_vec_op_count::sve_issue_info): Likewise.
1152 (aarch64_vector_costs::m_ops): Turn into a vector.
1153 (aarch64_vector_costs::m_advsimd_ops): Likewise.
1154 (aarch64_vector_costs::aarch64_vector_costs): Add entries to
1155 the vectors based on aarch64_tune_params.
1156 (aarch64_vector_costs::analyze_loop_vinfo): Update the pred_ops
1157 of all entries in m_ops.
1158 (aarch64_vector_costs::add_stmt_cost): Call count_ops for all
1160 (aarch64_estimate_min_cycles_per_iter): Remove issue_info
1161 parameter and get the information from the ops instead.
1162 (aarch64_vector_costs::adjust_body_cost_sve): Take a
1163 aarch64_vec_issue_info instead of a aarch64_vec_op_count.
1164 (aarch64_vector_costs::adjust_body_cost): Update call accordingly.
1165 Exit earlier if m_ops is empty for either cost structure.
1167 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
1169 * config/aarch64/aarch64.c (aarch64_vector_costs::m_scalar_ops)
1170 (aarch64_vector_costs::m_sve_ops): Replace with...
1171 (aarch64_vector_costs::m_ops): ...this.
1172 (aarch64_vector_costs::analyze_loop_vinfo): Update accordingly.
1173 (aarch64_vector_costs::adjust_body_cost_sve): Likewise.
1174 (aarch64_vector_costs::aarch64_vector_costs): Likewise.
1175 Initialize m_vec_flags here rather than in add_stmt_cost.
1176 (aarch64_vector_costs::count_ops): Test for scalar reductions too.
1177 Allow vectype to be null.
1178 (aarch64_vector_costs::add_stmt_cost): Call count_ops for scalar
1179 code too. Don't require vectype to be nonnull.
1180 (aarch64_vector_costs::adjust_body_cost): Take the loop_vec_info
1181 and scalar costs as parameters. Use the scalar costs to determine
1182 the cycles per iteration of the scalar loop, then multiply it
1183 by the estimated VF.
1184 (aarch64_vector_costs::finish_cost): Update call accordingly.
1186 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
1188 * config/aarch64/aarch64.c (aarch64_dr_type): New function.
1189 (aarch64_vector_costs::count_ops): Use it rather than the
1190 vectype to determine floatness.
1192 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
1194 * config/aarch64/aarch64.c (aarch64_sve_in_loop_reduction_latency):
1195 Remove vectype parameter and get floatness from the type of the
1197 (arch64_in_loop_reduction_latency): Likewise.
1198 (aarch64_detect_vector_stmt_subtype): Update caller.
1199 (aarch64_vector_costs::count_ops): Likewise.
1201 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
1203 * config/aarch64/aarch64.c (aarch64_sve_op_count): Fold into...
1204 (aarch64_vec_op_count): ...this. Add a constructor.
1205 (aarch64_vec_op_count::vec_flags): New function.
1206 (aarch64_vec_op_count::base_issue_info): Likewise.
1207 (aarch64_vec_op_count::simd_issue_info): Likewise.
1208 (aarch64_vec_op_count::sve_issue_info): Likewise.
1209 (aarch64_vec_op_count::m_issue_info): New member variable.
1210 (aarch64_vec_op_count::m_vec_flags): Likewise.
1211 (aarch64_vector_costs): Add a constructor.
1212 (aarch64_vector_costs::m_sve_ops): Change type to aarch64_vec_op_count.
1213 (aarch64_vector_costs::aarch64_vector_costs): New function.
1214 Initialize m_scalar_ops, m_advsimd_ops and m_sve_ops.
1215 (aarch64_vector_costs::count_ops): Remove vec_flags and
1216 issue_info parameters, using the new aarch64_vec_op_count
1218 (aarch64_vector_costs::add_stmt_cost): Update call accordingly.
1219 (aarch64_sve_op_count::dump): Fold into...
1220 (aarch64_vec_op_count::dump): ..here.
1222 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
1224 * config/aarch64/aarch64.c: Include tree-dfa.h.
1225 (aarch64_check_consecutive_mems): New function that takes MEM_EXPR
1226 and MEM_OFFSET into account.
1227 (aarch64_swap_ldrstr_operands): Use it.
1228 (aarch64_operands_ok_for_ldpstp): Likewise. Check that the
1229 address of the second memory doesn't depend on the result of
1232 2021-11-12 Jan Hubicka <hubicka@ucw.cz>
1234 * ipa-pure-const.c (ipa_make_function_pure): Fix exit condition.
1236 2021-11-12 Jan Hubicka <jh@suse.cz>
1238 PR tree-optimization/103175
1239 * ipa-modref.c (modref_lattice::merge): Add sanity check.
1240 (callee_to_caller_flags): Make flags adjustment sane.
1241 (modref_eaf_analysis::analyze_ssa_name): Likewise.
1243 2021-11-12 Jan Hubicka <jh@suse.cz>
1246 * ipa-modref.c (analyze_function, modref_propagate_in_scc): Do
1247 not mark pure/const function if there are side-effects.
1249 2021-11-12 Chung-Lin Tang <cltang@codesourcery.com>
1251 * tree.h (OMP_CLAUSE_MAP_RUNTIME_IMPLICIT_P): New access macro for
1252 'implicit' bit, using 'base.deprecated_flag' field of tree_node.
1253 * tree-pretty-print.c (dump_omp_clause): Add support for printing
1254 implicit attribute in tree dumping.
1255 * gimplify.c (gimplify_adjust_omp_clauses_1):
1256 Set OMP_CLAUSE_MAP_RUNTIME_IMPLICIT_P to 1 if map clause is implicitly
1258 (gimplify_adjust_omp_clauses): Adjust place of adding implicitly created
1259 clauses, from simple append, to starting of list, after non-map clauses.
1260 * omp-low.c (lower_omp_target): Add GOMP_MAP_IMPLICIT bits into kind
1261 values passed to libgomp for implicit maps.
1263 2021-11-12 Jakub Jelinek <jakub@redhat.com>
1265 * omp-builtins.def (BUILT_IN_GOMP_TEAMS): Remove.
1266 (BUILT_IN_GOMP_TEAMS4): New.
1267 * builtin-types.def (BT_FN_VOID_UINT_UINT): Remove.
1268 (BT_FN_BOOL_UINT_UINT_UINT_BOOL): New.
1269 * omp-low.c (lower_omp_teams): Use GOMP_teams4 instead of
1270 GOMP_teams, pass to it also num_teams lower-bound expression
1271 or a dup of upper-bound if it is missing and a flag whether
1272 it is the first call or not.
1274 2021-11-12 Martin Liska <mliska@suse.cz>
1276 PR tree-optimization/102497
1277 * gimple-predicate-analysis.cc (add_pred): Remove unused
1280 2021-11-12 Richard Biener <rguenther@suse.de>
1282 PR tree-optimization/103204
1283 * tree-ssa-sccvn.c (valueize_refs_1): Re-valueize the
1284 top operand after folding in an address.
1286 2021-11-11 Aldy Hernandez <aldyh@redhat.com>
1288 * gimple-range-path.cc (path_range_query::path_range_query): New
1289 ctor without a ranger.
1290 (path_range_query::~path_range_query): Free ranger if necessary.
1291 (path_range_query::range_on_path_entry): Adjust m_ranger for pointer.
1292 (path_range_query::ssa_range_in_phi): Same.
1293 (path_range_query::compute_ranges_in_block): Same.
1294 (path_range_query::compute_imports): Same.
1295 (path_range_query::compute_ranges): Same.
1296 (path_range_query::range_of_stmt): Same.
1297 (path_range_query::compute_outgoing_relations): Same.
1298 * gimple-range-path.h (class path_range_query): New ctor.
1299 * tree-ssa-loop-ch.c (ch_base::copy_headers): Remove gimple_ranger
1300 as path_range_query allocates one.
1301 * tree-ssa-threadbackward.c (class back_threader): Remove m_ranger.
1302 (back_threader::~back_threader): Same.
1304 2021-11-11 Aldy Hernandez <aldyh@redhat.com>
1306 * tree-ssa-threadbackward.c
1307 (back_threader_profitability::profitable_path_p): Remove loop
1308 crossing restriction.
1310 2021-11-11 Jan Hubicka <hubicka@ucw.cz>
1312 * ipa-modref.c (analyze_function): Do pure/const discovery, return
1314 (pass_modref::execute): If pure/const is discovered fixup cfg.
1315 (ignore_edge): Do not ignore pure/const edges.
1316 (modref_propagate_in_scc): Do pure/const discovery, return true if
1317 cdtor was promoted pure/const.
1318 (pass_ipa_modref::execute): If needed remove unreachable functions.
1319 * ipa-pure-const.c (warn_function_noreturn): Fix whitespace.
1320 (warn_function_cold): Likewise.
1321 (skip_function_for_local_pure_const): Move earlier.
1322 (ipa_make_function_const): Break out from ...
1323 (ipa_make_function_pure): Break out from ...
1324 (propagate_pure_const): ... here.
1325 (pass_local_pure_const::execute): Use it.
1326 * ipa-utils.h (ipa_make_function_const): Declare.
1327 (ipa_make_function_pure): Declare.
1328 * passes.def: Move early modref after pure-const.
1330 2021-11-11 David Malcolm <dmalcolm@redhat.com>
1333 * diagnostic-show-locus.c (def_policy): Use def_tabstop.
1335 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
1337 * config/aarch64/aarch64-builtins.c (TYPES_COMBINE): Delete.
1338 (TYPES_COMBINEP): Delete.
1339 * config/aarch64/aarch64-simd-builtins.def: Declare type-
1340 qualified builtins for vcombine_* intrinsics.
1341 * config/aarch64/arm_neon.h (vcombine_s8): Remove unnecessary
1343 (vcombine_s16): Likewise.
1344 (vcombine_s32): Likewise.
1345 (vcombine_f32): Likewise.
1346 (vcombine_u8): Use type-qualified builtin and remove casts.
1347 (vcombine_u16): Likewise.
1348 (vcombine_u32): Likewise.
1349 (vcombine_u64): Likewise.
1350 (vcombine_p8): Likewise.
1351 (vcombine_p16): Likewise.
1352 (vcombine_p64): Likewise.
1353 (vcombine_bf16): Remove unnecessary cast.
1354 * config/aarch64/iterators.md (VD_I): New mode iterator.
1355 (VDC_P): New mode iterator.
1357 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
1359 * config/aarch64/aarch64-builtins.c (TYPES_LOAD1_U): Define.
1360 (TYPES_LOAD1_P): Define.
1361 (TYPES_STORE1_U): Define.
1362 (TYPES_STORE1P): Rename to...
1363 (TYPES_STORE1_P): This.
1364 (get_mem_type_for_load_store): Add unsigned and poly types.
1365 (aarch64_general_gimple_fold_builtin): Add unsigned and poly
1366 type-qualified builtin declarations.
1367 * config/aarch64/aarch64-simd-builtins.def: Declare type-
1368 qualified builtins for LD1/ST1.
1369 * config/aarch64/arm_neon.h (vld1_p8): Use type-qualified
1370 builtin and remove cast.
1371 (vld1_p16): Likewise.
1372 (vld1_u8): Likewise.
1373 (vld1_u16): Likewise.
1374 (vld1_u32): Likewise.
1375 (vld1q_p8): Likewise.
1376 (vld1q_p16): Likewise.
1377 (vld1q_p64): Likewise.
1378 (vld1q_u8): Likewise.
1379 (vld1q_u16): Likewise.
1380 (vld1q_u32): Likewise.
1381 (vld1q_u64): Likewise.
1382 (vst1_p8): Likewise.
1383 (vst1_p16): Likewise.
1384 (vst1_u8): Likewise.
1385 (vst1_u16): Likewise.
1386 (vst1_u32): Likewise.
1387 (vst1q_p8): Likewise.
1388 (vst1q_p16): Likewise.
1389 (vst1q_p64): Likewise.
1390 (vst1q_u8): Likewise.
1391 (vst1q_u16): Likewise.
1392 (vst1q_u32): Likewise.
1393 (vst1q_u64): Likewise.
1394 * config/aarch64/iterators.md (VALLP_NO_DI): New iterator.
1396 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
1398 * config/aarch64/aarch64-simd-builtins.def: Declare unsigned
1399 builtins for vector reduction.
1400 * config/aarch64/arm_neon.h (vaddv_u8): Use type-qualified
1401 builtin and remove casts.
1402 (vaddv_u16): Likewise.
1403 (vaddv_u32): Likewise.
1404 (vaddvq_u8): Likewise.
1405 (vaddvq_u16): Likewise.
1406 (vaddvq_u32): Likewise.
1407 (vaddvq_u64): Likewise.
1409 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
1411 * config/aarch64/aarch64-simd-builtins.def:
1412 * config/aarch64/arm_neon.h (vpaddq_u8): Use type-qualified
1413 builtin and remove casts.
1414 (vpaddq_u16): Likewise.
1415 (vpaddq_u32): Likewise.
1416 (vpaddq_u64): Likewise.
1417 (vpadd_u8): Likewise.
1418 (vpadd_u16): Likewise.
1419 (vpadd_u32): Likewise.
1420 (vpaddd_u64): Likewise.
1422 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
1424 * config/aarch64/aarch64-simd-builtins.def: Declare unsigned
1425 builtins for [r]subhn[2].
1426 * config/aarch64/arm_neon.h (vsubhn_s16): Remove unnecessary
1428 (vsubhn_s32): Likewise.
1429 (vsubhn_s64): Likewise.
1430 (vsubhn_u16): Use type-qualified builtin and remove casts.
1431 (vsubhn_u32): Likewise.
1432 (vsubhn_u64): Likewise.
1433 (vrsubhn_s16): Remove unnecessary cast.
1434 (vrsubhn_s32): Likewise.
1435 (vrsubhn_s64): Likewise.
1436 (vrsubhn_u16): Use type-qualified builtin and remove casts.
1437 (vrsubhn_u32): Likewise.
1438 (vrsubhn_u64): Likewise.
1439 (vrsubhn_high_s16): Remove unnecessary cast.
1440 (vrsubhn_high_s32): Likewise.
1441 (vrsubhn_high_s64): Likewise.
1442 (vrsubhn_high_u16): Use type-qualified builtin and remove
1444 (vrsubhn_high_u32): Likewise.
1445 (vrsubhn_high_u64): Likewise.
1446 (vsubhn_high_s16): Remove unnecessary cast.
1447 (vsubhn_high_s32): Likewise.
1448 (vsubhn_high_s64): Likewise.
1449 (vsubhn_high_u16): Use type-qualified builtin and remove
1451 (vsubhn_high_u32): Likewise.
1452 (vsubhn_high_u64): Likewise.
1454 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
1456 * config/aarch64/aarch64-simd-builtins.def: Declare unsigned
1457 builtins for [r]addhn[2].
1458 * config/aarch64/arm_neon.h (vaddhn_s16): Remove unnecessary
1460 (vaddhn_s32): Likewise.
1461 (vaddhn_s64): Likewise.
1462 (vaddhn_u16): Use type-qualified builtin and remove casts.
1463 (vaddhn_u32): Likewise.
1464 (vaddhn_u64): Likewise.
1465 (vraddhn_s16): Remove unnecessary cast.
1466 (vraddhn_s32): Likewise.
1467 (vraddhn_s64): Likewise.
1468 (vraddhn_u16): Use type-qualified builtin and remove casts.
1469 (vraddhn_u32): Likewise.
1470 (vraddhn_u64): Likewise.
1471 (vaddhn_high_s16): Remove unnecessary cast.
1472 (vaddhn_high_s32): Likewise.
1473 (vaddhn_high_s64): Likewise.
1474 (vaddhn_high_u16): Use type-qualified builtin and remove
1476 (vaddhn_high_u32): Likewise.
1477 (vaddhn_high_u64): Likewise.
1478 (vraddhn_high_s16): Remove unnecessary cast.
1479 (vraddhn_high_s32): Likewise.
1480 (vraddhn_high_s64): Likewise.
1481 (vraddhn_high_u16): Use type-qualified builtin and remove
1483 (vraddhn_high_u32): Likewise.
1484 (vraddhn_high_u64): Likewise.
1486 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
1488 * config/aarch64/aarch64-simd-builtins.def: Use BINOPU type
1489 qualifiers in generator macros for uhsub builtins.
1490 * config/aarch64/arm_neon.h (vhsub_s8): Remove unnecessary
1492 (vhsub_s16): Likewise.
1493 (vhsub_s32): Likewise.
1494 (vhsub_u8): Use type-qualified builtin and remove casts.
1495 (vhsub_u16): Likewise.
1496 (vhsub_u32): Likewise.
1497 (vhsubq_s8): Remove unnecessary cast.
1498 (vhsubq_s16): Likewise.
1499 (vhsubq_s32): Likewise.
1500 (vhsubq_u8): Use type-qualified builtin and remove casts.
1501 (vhsubq_u16): Likewise.
1502 (vhsubq_u32): Likewise.
1504 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
1506 * config/aarch64/aarch64-simd-builtins.def: Use BINOPU type
1507 qualifiers in generator macros for u[r]hadd builtins.
1508 * config/aarch64/arm_neon.h (vhadd_s8): Remove unnecessary
1510 (vhadd_s16): Likewise.
1511 (vhadd_s32): Likewise.
1512 (vhadd_u8): Use type-qualified builtin and remove casts.
1513 (vhadd_u16): Likewise.
1514 (vhadd_u32): Likewise.
1515 (vhaddq_s8): Remove unnecessary cast.
1516 (vhaddq_s16): Likewise.
1517 (vhaddq_s32): Likewise.
1518 (vhaddq_u8): Use type-qualified builtin and remove casts.
1519 (vhaddq_u16): Likewise.
1520 (vhaddq_u32): Likewise.
1521 (vrhadd_s8): Remove unnecessary cast.
1522 (vrhadd_s16): Likewise.
1523 (vrhadd_s32): Likewise.
1524 (vrhadd_u8): Use type-qualified builtin and remove casts.
1525 (vrhadd_u16): Likewise.
1526 (vrhadd_u32): Likewise.
1527 (vrhaddq_s8): Remove unnecessary cast.
1528 (vrhaddq_s16): Likewise.
1529 (vrhaddq_s32): Likewise.
1530 (vrhaddq_u8): Use type-wualified builtin and remove casts.
1531 (vrhaddq_u16): Likewise.
1532 (vrhaddq_u32): Likewise.
1534 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
1536 * config/aarch64/aarch64-simd-builtins.def: Use BINOPU type
1537 qualifiers in generator macros for usub[lw][2] builtins.
1538 * config/aarch64/arm_neon.h (vsubl_s8): Remove unnecessary
1540 (vsubl_s16): Likewise.
1541 (vsubl_s32): Likewise.
1542 (vsubl_u8): Use type-qualified builtin and remove casts.
1543 (vsubl_u16): Likewise.
1544 (vsubl_u32): Likewise.
1545 (vsubl_high_s8): Remove unnecessary cast.
1546 (vsubl_high_s16): Likewise.
1547 (vsubl_high_s32): Likewise.
1548 (vsubl_high_u8): Use type-qualified builtin and remove casts.
1549 (vsubl_high_u16): Likewise.
1550 (vsubl_high_u32): Likewise.
1551 (vsubw_s8): Remove unnecessary casts.
1552 (vsubw_s16): Likewise.
1553 (vsubw_s32): Likewise.
1554 (vsubw_u8): Use type-qualified builtin and remove casts.
1555 (vsubw_u16): Likewise.
1556 (vsubw_u32): Likewise.
1557 (vsubw_high_s8): Remove unnecessary cast.
1558 (vsubw_high_s16): Likewise.
1559 (vsubw_high_s32): Likewise.
1560 (vsubw_high_u8): Use type-qualified builtin and remove casts.
1561 (vsubw_high_u16): Likewise.
1562 (vsubw_high_u32): Likewise.
1564 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
1566 * config/aarch64/aarch64-simd-builtins.def: Use BINOPU type
1567 qualifiers in generator macros for uadd[lw][2] builtins.
1568 * config/aarch64/arm_neon.h (vaddl_s8): Remove unnecessary
1570 (vaddl_s16): Likewise.
1571 (vaddl_s32): Likewise.
1572 (vaddl_u8): Use type-qualified builtin and remove casts.
1573 (vaddl_u16): Likewise.
1574 (vaddl_u32): Likewise.
1575 (vaddl_high_s8): Remove unnecessary cast.
1576 (vaddl_high_s16): Likewise.
1577 (vaddl_high_s32): Likewise.
1578 (vaddl_high_u8): Use type-qualified builtin and remove casts.
1579 (vaddl_high_u16): Likewise.
1580 (vaddl_high_u32): Likewise.
1581 (vaddw_s8): Remove unnecessary cast.
1582 (vaddw_s16): Likewise.
1583 (vaddw_s32): Likewise.
1584 (vaddw_u8): Use type-qualified builtin and remove casts.
1585 (vaddw_u16): Likewise.
1586 (vaddw_u32): Likewise.
1587 (vaddw_high_s8): Remove unnecessary cast.
1588 (vaddw_high_s16): Likewise.
1589 (vaddw_high_s32): Likewise.
1590 (vaddw_high_u8): Use type-qualified builtin and remove casts.
1591 (vaddw_high_u16): Likewise.
1592 (vaddw_high_u32): Likewise.
1594 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
1596 * config/aarch64/aarch64-simd-builtins.def: Declare type-
1597 qualified builtins for [R]SHRN[2].
1598 * config/aarch64/arm_neon.h (vshrn_n_u16): Use type-qualified
1599 builtin and remove casts.
1600 (vshrn_n_u32): Likewise.
1601 (vshrn_n_u64): Likewise.
1602 (vrshrn_high_n_u16): Likewise.
1603 (vrshrn_high_n_u32): Likewise.
1604 (vrshrn_high_n_u64): Likewise.
1605 (vrshrn_n_u16): Likewise.
1606 (vrshrn_n_u32): Likewise.
1607 (vrshrn_n_u64): Likewise.
1608 (vshrn_high_n_u16): Likewise.
1609 (vshrn_high_n_u32): Likewise.
1610 (vshrn_high_n_u64): Likewise.
1612 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
1614 * config/aarch64/aarch64-simd-builtins.def: Declare unsigned
1615 type-qualified builtins for XTN[2].
1616 * config/aarch64/arm_neon.h (vmovn_high_u16): Use type-
1617 qualified builtin and remove casts.
1618 (vmovn_high_u32): Likewise.
1619 (vmovn_high_u64): Likewise.
1620 (vmovn_u16): Likewise.
1621 (vmovn_u32): Likewise.
1622 (vmovn_u64): Likewise.
1624 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
1626 * config/aarch64/aarch64-simd-builtins.def: Use poly type
1627 qualifier in builtin generator macros.
1628 * config/aarch64/arm_neon.h (vmul_p8): Use type-qualified
1629 builtin and remove casts.
1630 (vmulq_p8): Likewise.
1631 (vmull_high_p8): Likewise.
1632 (vmull_p8): Likewise.
1634 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
1636 * config/aarch64/aarch64-simd-builtins.def: Declare type-
1637 qualified builtin generators for unsigned MLA/MLS intrinsics.
1638 * config/aarch64/arm_neon.h (vmla_n_u16): Use type-qualified
1640 (vmla_n_u32): Likewise.
1641 (vmla_u8): Likewise.
1642 (vmla_u16): Likewise.
1643 (vmla_u32): Likewise.
1644 (vmlaq_n_u16): Likewise.
1645 (vmlaq_n_u32): Likewise.
1646 (vmlaq_u8): Likewise.
1647 (vmlaq_u16): Likewise.
1648 (vmlaq_u32): Likewise.
1649 (vmls_n_u16): Likewise.
1650 (vmls_n_u32): Likewise.
1651 (vmls_u8): Likewise.
1652 (vmls_u16): Likewise.
1653 (vmls_u32): Likewise.
1654 (vmlsq_n_u16): Likewise.
1655 (vmlsq_n_u32): Likewise.
1656 (vmlsq_u8): Likewise.
1657 (vmlsq_u16): Likewise.
1658 (vmlsq_u32): Likewise.
1660 2021-11-11 Jan Hubicka <hubicka@ucw.cz>
1662 * ipa-modref.c (modref_summary::useful_p): Check also for side-effects
1663 with looping const/pure.
1664 (modref_summary_lto::useful_p): Likewise.
1665 (merge_call_side_effects): Merge side effects before early exit
1667 (process_fnspec): Also handle pure functions.
1668 (analyze_call): Do not early exit on looping pure const.
1669 (propagate_unknown_call): Also handle nontrivial SCC as side-effect.
1670 (modref_propagate_in_scc): Update.
1672 2021-11-11 Richard Biener <rguenther@suse.de>
1674 PR tree-optimization/103190
1675 * tree-ssa-reassoc.c (insert_stmt_after): Only assert on asm goto.
1677 2021-11-11 Aldy Hernandez <aldyh@redhat.com>
1679 * gimple-range-path.cc (path_range_query::add_copies_to_imports):
1681 (path_range_query::compute_imports): ...this. Adapt it so it can
1682 be passed the imports bitmap instead of working on m_imports.
1683 (path_range_query::compute_ranges): Call compute_imports in all
1684 cases unless an imports bitmap is passed.
1685 * gimple-range-path.h (path_range_query::compute_imports): New.
1686 (path_range_query::add_copies_to_imports): Remove.
1687 * tree-ssa-threadbackward.c (back_threader::resolve_def): Remove.
1688 (back_threader::find_paths_to_names): Inline resolve_def.
1689 (back_threader::find_paths): Call compute_imports.
1690 (back_threader::resolve_phi): Adjust comment.
1692 2021-11-11 Richard Biener <rguenther@suse.de>
1694 PR tree-optimization/103188
1695 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p):
1696 Remove query parameter, split out check for size
1698 (ch_base::m_ranger, cb_base::m_query): Remove.
1699 (ch_base::copy_headers): Split processing loop into
1700 analysis around which we allocate and use ranger and
1701 transform where we do not.
1702 (pass_ch::execute): Do not allocate/free ranger here.
1703 (pass_ch_vect::execute): Likewise.
1705 2021-11-11 Jan Hubicka <hubicka@ucw.cz>
1707 * ipa-pure-const.c (propagate_pure_const): Self recursion is
1710 2021-11-11 Jan Hubicka <hubicka@ucw.cz>
1712 * cgraph.c (set_noreturn_flag_1): New function.
1713 (cgraph_node::set_noreturn_flag): New member function
1714 * cgraph.h (cgraph_node::set_noreturn_flags): Declare.
1715 * ipa-pure-const.c (pass_local_pure_const::execute): Use it.
1717 2021-11-11 Aldy Hernandez <aldyh@redhat.com>
1718 Richard Biener <rguenther@suse.de>
1720 * tree-ssa-loop-ch.c (entry_loop_condition_is_static): Resolve
1721 statically to the edge remaining in the loop.
1723 2021-11-11 Richard Biener <rguenther@suse.de>
1725 PR middle-end/103181
1726 * tree-eh.c (operation_could_trap_helper_p): Properly
1727 check vector constants for a zero element for integer
1728 division. Separate floating point and integer division code.
1730 2021-11-11 Jakub Jelinek <jakub@redhat.com>
1733 * dwarf2out.c (field_byte_offset): Do the PCC_BITFIELD_TYPE_MATTERS
1734 handling only for DECL_BIT_FIELD_TYPE decls.
1736 2021-11-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1739 * config/aarch64/aarch64.c (aarch64_process_target_attr): Check if
1740 token is arch extension without leading '+' and emit appropriate
1741 diagnostic for the same.
1743 2021-11-11 Jakub Jelinek <jakub@redhat.com>
1745 * tree.h (OMP_CLAUSE_NUM_TEAMS_EXPR): Rename to ...
1746 (OMP_CLAUSE_NUM_TEAMS_UPPER_EXPR): ... this.
1747 (OMP_CLAUSE_NUM_TEAMS_LOWER_EXPR): Define.
1748 * tree.c (omp_clause_num_ops): Increase num ops for
1749 OMP_CLAUSE_NUM_TEAMS to 2.
1750 * tree-pretty-print.c (dump_omp_clause): Print optional lower bound
1751 for OMP_CLAUSE_NUM_TEAMS.
1752 * gimplify.c (gimplify_scan_omp_clauses): Gimplify
1753 OMP_CLAUSE_NUM_TEAMS_LOWER_EXPR if non-NULL.
1754 (optimize_target_teams): Use OMP_CLAUSE_NUM_TEAMS_UPPER_EXPR instead
1755 of OMP_CLAUSE_NUM_TEAMS_EXPR. Handle OMP_CLAUSE_NUM_TEAMS_LOWER_EXPR.
1756 * omp-low.c (lower_omp_teams): Use OMP_CLAUSE_NUM_TEAMS_UPPER_EXPR
1757 instead of OMP_CLAUSE_NUM_TEAMS_EXPR.
1758 * omp-expand.c (expand_teams_call, get_target_arguments): Likewise.
1760 2021-11-11 Richard Biener <rguenther@suse.de>
1762 * cfganal.c (find_pdom): Remove.
1763 (control_dependences::find_control_dependence): Remove
1764 special-casing of entry block, call get_immediate_dominator
1766 * gimple-predicate-analysis.cc (find_pdom): Remove.
1767 (find_dom): Likewise.
1768 (find_control_equiv_block): Call get_immediate_dominator
1770 (compute_control_dep_chain): Likewise.
1771 (predicate::init_from_phi_def): Likewise.
1773 2021-11-11 Richard Biener <rguenther@suse.de>
1775 * cfganal.h (control_dependences::control_dependence_map):
1777 (control_dependences::m_bitmaps): New.
1778 * cfganal.c (control_dependences::set_control_dependence_map_bit):
1780 (control_dependences::clear_control_dependence_bitmap):
1782 (control_dependences::find_control_dependence): Do not
1783 find_edge for the abnormal edge test.
1784 (control_dependences::control_dependences): Instead do not
1785 add abnormal edges to the edge list. Adjust.
1786 (control_dependences::~control_dependences): Likewise.
1787 (control_dependences::get_edges_dependent_on): Likewise.
1788 * function-tests.c: Include bitmap.h.
1790 2021-11-11 Kewen Lin <linkw@linux.ibm.com>
1792 * doc/invoke.texi: Change references to "future cpu" to "power10",
1793 "-mcpu=future" to "-mcpu=power10". Adjust words for float128.
1795 2021-11-11 Cui,Lili <lili.cui@intel.com>
1797 * config/i386/i386-options.c (m_CORE_AVX2): Remove Alderlake
1799 (processor_cost_table): Use alderlake_cost for Alderlake.
1800 * config/i386/i386.c (ix86_sched_init_global): Handle Alderlake.
1801 * config/i386/x86-tune-costs.h (struct processor_costs): Add alderlake
1803 * config/i386/x86-tune-sched.c (ix86_issue_rate): Change Alderlake
1805 (ix86_adjust_cost): Handle Alderlake.
1806 * config/i386/x86-tune.def (X86_TUNE_SCHEDULE): Enable for Alderlake.
1807 (X86_TUNE_PARTIAL_REG_DEPENDENCY): Likewise.
1808 (X86_TUNE_SSE_PARTIAL_REG_DEPENDENCY): Likewise.
1809 (X86_TUNE_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY): Likewise.
1810 (X86_TUNE_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
1811 (X86_TUNE_MEMORY_MISMATCH_STALL): Likewise.
1812 (X86_TUNE_USE_LEAVE): Likewise.
1813 (X86_TUNE_PUSH_MEMORY): Likewise.
1814 (X86_TUNE_USE_INCDEC): Likewise.
1815 (X86_TUNE_INTEGER_DFMODE_MOVES): Likewise.
1816 (X86_TUNE_MISALIGNED_MOVE_STRING_PRO_EPILOGUES): Likewise.
1817 (X86_TUNE_USE_SAHF): Likewise.
1818 (X86_TUNE_USE_BT): Likewise.
1819 (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI): Likewise.
1820 (X86_TUNE_ONE_IF_CONV_INSN): Likewise.
1821 (X86_TUNE_AVOID_MFENCE): Likewise.
1822 (X86_TUNE_USE_SIMODE_FIOP): Likewise.
1823 (X86_TUNE_EXT_80387_CONSTANTS): Likewise.
1824 (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL): Likewise.
1825 (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Likewise.
1826 (X86_TUNE_SSE_TYPELESS_STORES): Likewise.
1827 (X86_TUNE_SSE_LOAD0_BY_PXOR): Likewise.
1828 (X86_TUNE_AVOID_4BYTE_PREFIXES): Likewise.
1829 (X86_TUNE_USE_GATHER): Disable for Alderlake.
1830 (X86_TUNE_AVX256_MOVE_BY_PIECES): Likewise.
1831 (X86_TUNE_AVX256_STORE_BY_PIECES): Likewise.
1833 2021-11-11 liuhongt <hongtao.liu@intel.com>
1836 * config/i386/sse.md (V_128_256): Extend to V8HF/V16HF.
1837 (avxsizesuffix): Ditto.
1839 2021-11-11 Kito Cheng <kito.cheng@sifive.com>
1841 * common/config/riscv/riscv-common.c (riscv_subset_list::to_string): Fix
1842 wrong marco checking.
1844 2021-11-10 Aldy Hernandez <aldyh@redhat.com>
1846 PR tree-optimization/102906
1847 * tree-ssa-loop-ch.c (entry_loop_condition_is_static): New.
1848 (should_duplicate_loop_header_p): Call entry_loop_condition_is_static.
1849 (class ch_base): Add m_ranger and m_query.
1850 (ch_base::copy_headers): Pass m_query to
1851 entry_loop_condition_is_static.
1852 (pass_ch::execute): Allocate and deallocate m_ranger and
1854 (pass_ch_vect::execute): Same.
1856 2021-11-10 Andrew Pinski <apinski@marvell.com>
1859 * config/aarch64/aarch64-simd.md (aarch64_simd_dup<mode>):
1860 Use vwcore iterator for the r constraint output string.
1862 2021-11-10 qing zhao <qing.zhao@oracle.com>
1864 * internal-fn.c (expand_DEFERRED_INIT): Apply pattern initialization
1865 only when have_insn_for return true for the mode. Fix a memory leak.
1867 2021-11-10 Christophe Lyon <christophe.lyon@foss.st.com>
1869 * config/arm/arm.c (cortexa9_extra_costs, cortexa8_extra_costs,
1870 cortexa5_extra_costs, cortexa7_extra_costs,
1871 cortexa12_extra_costs, cortexa15_extra_costs, v7m_extra_costs):
1872 Initialize movi, dup and extract costing fields.
1874 2021-11-10 Aldy Hernandez <aldyh@redhat.com>
1876 * gimple-range-path.cc (path_range_query::path_range_query): Do
1878 (path_range_query::dump): Change m_path uses to non-pointer.
1879 (path_range_query::defined_outside_path): Same.
1880 (path_range_query::set_path): Same.
1881 (path_range_query::add_copies_to_imports): Same.
1882 (path_range_query::range_of_stmt): Same.
1883 (path_range_query::compute_outgoing_relations): Same.
1884 (path_range_query::compute_ranges): Imports are now optional.
1885 Implement overload that takes an edge.
1886 * gimple-range-path.h (class path_range_query): Make imports
1887 optional for compute_ranges. Add compute_ranges(edge) overload.
1888 Make m_path an auto_vec instead of a pointer and adjust
1891 2021-11-10 Tamar Christina <tamar.christina@arm.com>
1893 * tree-vectorizer.h (struct scalar_cond_masked_key): Add inverted_p.
1894 (default_hash_traits<scalar_conf_masked_key>): Likewise.
1895 * tree-vect-stmts.c (vectorizable_condition): Check if inverse of mask
1897 * tree-vectorizer.c (scalar_cond_masked_key::get_cond_ops_from_tree):
1898 Register mask inverses.
1900 2021-11-10 Tamar Christina <tamar.christina@arm.com>
1902 * tree-vectorizer.c (vectorize_loops): Do local CSE through RPVN upon
1903 successful vectorization.
1905 2021-11-10 Andrew MacLeod <amacleod@redhat.com>
1907 * gimple-range-cache.cc (sbr_vector::grow): New.
1908 (sbr_vector::set_bb_range): Call grow.
1909 (sbr_vector::get_bb_range): Same.
1910 (sbr_vector::bb_range_p): Remove assert.
1912 2021-11-10 Tamar Christina <tamar.christina@arm.com>
1914 * config/aarch64/aarch64-simd.md (*aarch64_topbits_shuffle<mode>_le
1915 ,*aarch64_topbits_shuffle<mode>_be): Remove.
1917 2021-11-10 Jan Hubicka <jh@suse.cz>
1919 * ipa-modref.c: Include tree-eh.h
1920 (modref_summary::modref_summary): Initialize side_effects.
1921 (struct modref_summary_lto): New bool field side_effects.
1922 (modref_summary_lto::modref_summary_lto): Initialize side_effects.
1923 (modref_summary::dump): Dump side_effects.
1924 (modref_summary_lto::dump): Dump side_effects.
1925 (merge_call_side_effects): Merge side effects.
1926 (process_fnspec): Calls to non-const/pure or looping
1927 function is a side effect.
1928 (analyze_call): Self-recursion is a side-effect; handle
1930 (analyze_load): Watch for volatile and throwing memory.
1931 (analyze_store): Likewise.
1932 (analyze_stmt): Watch for volatitle asm.
1933 (analyze_function): Handle side_effects.
1934 (modref_summaries::duplicate): Duplicate side_effects.
1935 (modref_summaries_lto::duplicate): Likewise.
1936 (modref_write): Stream side_effects.
1937 (read_section): Likewise.
1938 (update_signature): Update.
1939 (propagate_unknown_call): Handle side_effects.
1940 (modref_propagate_in_scc): Likewise.
1941 * ipa-modref.h (struct modref_summary): Add side_effects.
1942 * ipa-pure-const.c (special_builtin_state): Rename to ...
1943 (builtin_safe_for_const_function_p): ... this one.
1944 (check_call): Update.
1945 (finite_function_p): Break out from ...
1946 (propagate_pure_const): ... here
1947 * ipa-utils.h (finite_function): Declare.
1949 2021-11-10 Lucas A. M. Magalhães <lamm@linux.ibm.com>
1951 * config.gcc (powerpc*-*-*): Remove -rpath from
1952 --with-advance-toolchain.
1954 2021-11-10 Marek Polacek <polacek@redhat.com>
1957 * attribs.c (struct scoped_attributes): Add a bool member.
1958 (lookup_scoped_attribute_spec): Forward declare.
1959 (register_scoped_attributes): New bool parameter, defaulted to
1961 (handle_ignored_attributes_option): New function.
1962 (free_attr_data): New function.
1963 (init_attributes): Call handle_ignored_attributes_option.
1964 (attr_namespace_ignored_p): New function.
1965 (decl_attributes): Check attr_namespace_ignored_p before
1967 * attribs.h (free_attr_data): Declare.
1968 (register_scoped_attributes): Adjust declaration.
1969 (handle_ignored_attributes_option): Declare.
1970 (canonicalize_attr_name): New function template.
1971 (canonicalize_attr_name): Use it.
1972 * common.opt (Wattributes=): New option with a variable.
1973 * doc/extend.texi: Document #pragma GCC diagnostic ignored_attributes.
1974 * doc/invoke.texi: Document -Wno-attributes=.
1975 * opts.c (common_handle_option) <case OPT_Wattributes_>: Handle.
1976 * plugin.h (register_scoped_attributes): Adjust declaration.
1977 * toplev.c (compile_file): Call free_attr_data.
1979 2021-11-10 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
1981 * config/arm/arm-cpus.in (cortex-a710): New CPU.
1982 * config/arm/arm-tables.opt: Regenerate.
1983 * config/arm/arm-tune.md: Regenerate.
1984 * doc/invoke.texi: Update docs.
1986 2021-11-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
1988 * config/aarch64/aarch64-builtins.c
1989 (aarch64_general_gimple_fold_builtin): Mark argument as unused.
1991 2021-11-10 Martin Liska <mliska@suse.cz>
1993 * lto-wrapper.c (merge_and_complain): Make the first argument
1996 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
1998 * config/aarch64/iterators.md (optab): Use fmax_nan instead of
1999 smax_nan and fmin_nan instead of smin_nan.
2000 (maxmin_uns): Rename to...
2001 (fmaxmin): ...this and make the same changes. Remove entries
2002 unrelated to fmax* and fmin*.
2003 * config/aarch64/aarch64.md (<maxmin_uns><mode>3): Rename to...
2004 (<fmaxmin><mode>3): ...this.
2005 * config/aarch64/aarch64-simd.md (aarch64_<maxmin_uns>p<mode>):
2007 (aarch64_<optab>p<mode>): ...this.
2008 (<maxmin_uns><mode>3): Rename to...
2009 (<fmaxmin><mode>3): ...this.
2010 (reduc_<maxmin_uns>_scal_<mode>): Rename to...
2011 (reduc_<optab>_scal_<mode>): ...this and update gen* call.
2012 (aarch64_reduc_<maxmin_uns>_internal<mode>): Rename to...
2013 (aarch64_reduc_<optab>_internal<mode>): ...this.
2014 (aarch64_reduc_<maxmin_uns>_internalv2si): Rename to...
2015 (aarch64_reduc_<optab>_internalv2si): ...this.
2016 * config/aarch64/aarch64-sve.md (<maxmin_uns><mode>3): Rename to...
2017 (<fmaxmin><mode>3): ...this.
2018 * config/aarch64/aarch64-simd-builtins.def (smax_nan, smin_nan)
2020 (fmax_nan, fmin_nan): ...this.
2021 * config/aarch64/arm_neon.h (vmax_f32, vmax_f64, vmaxq_f32, vmaxq_f64)
2022 (vmin_f32, vmin_f64, vminq_f32, vminq_f64, vmax_f16, vmaxq_f16)
2023 (vmin_f16, vminq_f16): Update accordingly.
2025 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
2027 * tree-vectorizer.h (vector_costs::finish_cost): Take the
2028 corresponding scalar costs as a parameter.
2029 (finish_cost): Likewise.
2030 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
2031 (vect_estimate_min_profitable_iters): Update accordingly.
2032 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Likewise.
2033 * tree-vectorizer.c (vector_costs::finish_cost): Likewise.
2034 * config/aarch64/aarch64.c (aarch64_vector_costs::finish_cost):
2036 * config/rs6000/rs6000.c (rs6000_cost_data::finish_cost): Likewise.
2038 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
2040 * tree-vectorizer.h (_loop_vec_info::scalar_costs): New member
2042 (_loop_vec_info::single_scalar_iteration_cost): Delete.
2043 (LOOP_VINFO_SINGLE_SCALAR_ITERATION_COST): Delete.
2044 (vector_costs::total_cost): New function.
2045 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Update
2046 after above changes.
2047 (_loop_vec_info::~_loop_vec_info): Delete scalar_costs.
2048 (vect_compute_single_scalar_iteration_cost): Store the costs
2049 in loop_vinfo->scalar_costs.
2050 (vect_estimate_min_profitable_iters): Get the scalar cost from
2051 loop_vinfo->scalar_costs.
2053 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
2055 * tree-vectorizer.h (vector_costs::better_main_loop_than_p)
2056 (vector_costs::better_epilogue_loop_than_p)
2057 (vector_costs::compare_inside_loop_cost)
2058 (vector_costs::compare_outside_loop_cost): Likewise.
2059 * tree-vectorizer.c (vector_costs::better_main_loop_than_p)
2060 (vector_costs::better_epilogue_loop_than_p)
2061 (vector_costs::compare_inside_loop_cost)
2062 (vector_costs::compare_outside_loop_cost): New functions,
2063 containing code moved from...
2064 * tree-vect-loop.c (vect_better_loop_vinfo_p): ...here.
2066 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
2068 * tree-vectorizer.h (_loop_vec_info): Remove vec_outside_cost
2069 and vec_inside_cost.
2070 (vector_costs::outside_cost): New function.
2071 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Update
2073 (vect_estimate_min_profitable_iters): Likewise.
2074 (vect_better_loop_vinfo_p): Get the inside and outside costs
2075 from the loop_vec_infos' vector_costs.
2077 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
2079 * tree-vectorizer.h (vec_info::target_cost_data): Replace with...
2080 (_loop_vec_info::vector_costs): ...this.
2081 (LOOP_VINFO_TARGET_COST_DATA): Delete.
2082 * tree-vectorizer.c (vec_info::vec_info): Remove target_cost_data
2084 (vec_info::~vec_info): Remove corresponding delete.
2085 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
2086 vector_costs to null.
2087 (_loop_vec_info::~_loop_vec_info): Delete vector_costs.
2088 (vect_analyze_loop_operations): Update after above changes.
2089 (vect_analyze_loop_2): Likewise.
2090 (vect_estimate_min_profitable_iters): Likewise.
2091 * tree-vect-slp.c (vect_slp_analyze_operations): Likewise.
2093 2021-11-10 Jan Hubicka <hubicka@ucw.cz>
2095 * tree-core.h (EAF_DIRECT): Remove.
2096 (EAF_NOCLOBBER): Remove.
2097 (EAF_UNUSED): Remove.
2098 (EAF_NOESCAPE): Remove.
2099 (EAF_NO_DIRECT_CLOBBER): New.
2100 (EAF_NO_INDIRECT_CLOBBER): New.
2101 (EAF_NODIRECTESCAPE): Remove.
2102 (EAF_NO_DIRECT_ESCAPE): New.
2103 (EAF_NO_INDIRECT_ESCAPE): New.
2104 (EAF_NOT_RETURNED): Remove.
2105 (EAF_NOT_RETURNED_INDIRECTLY): New.
2106 (EAF_NOREAD): Remove.
2107 (EAF_NO_DIRECT_READ): New.
2108 (EAF_NO_INDIRECT_READ): New.
2109 * gimple.c (gimple_call_arg_flags): Update for new flags.
2110 (gimple_call_retslot_flags): Update for new flags.
2111 * ipa-modref.c (dump_eaf_flags): Likewise.
2112 (remove_useless_eaf_flags): Likewise.
2113 (deref_flags): Likewise.
2114 (modref_lattice::init): Likewise.
2115 (modref_lattice::merge): Likewise.
2116 (modref_lattice::merge_direct_load): Likewise.
2117 (modref_lattice::merge_direct_store): Likewise.
2118 (modref_eaf_analysis::merge_call_lhs_flags): Likewise.
2119 (callee_to_caller_flags): Likewise.
2120 (modref_eaf_analysis::analyze_ssa_name): Likewise.
2121 (modref_eaf_analysis::propagate): Likewise.
2122 (modref_merge_call_site_flags): Likewise.
2123 * ipa-modref.h (interposable_eaf_flags): Likewise.
2124 * tree-ssa-alias.c: (ref_maybe_used_by_call_p_1) Likewise.
2125 * tree-ssa-structalias.c (handle_call_arg): Likewise.
2126 (handle_rhs_call): Likewise.
2127 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Likewise.
2129 2021-11-10 Tamar Christina <tamar.christina@arm.com>
2131 * tree-vect-slp-patterns.c (complex_mul_pattern::matches): Remove l1node.
2133 2021-11-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
2135 * config/aarch64/aarch64-builtins.c
2136 (aarch64_general_gimple_fold_builtin): Change pointer alignment and
2139 2021-11-10 Jan Hubicka <jh@suse.cz>
2141 * ipa-modref-tree.h (modref_tree::remap_params): Fix off-by-one error.
2143 2021-11-10 H.J. Lu <hongjiu.lu@intel.com>
2144 Hongtao Liu <hongtao.liu@intel.com>
2146 PR middle-end/102566
2147 * match.pd (nop_atomic_bit_test_and_p): New match.
2148 * tree-ssa-ccp.c (convert_atomic_bit_not): New function.
2149 (gimple_nop_atomic_bit_test_and_p): New prototype.
2150 (optimize_atomic_bit_test_and): Transform equivalent, but slighly
2151 different cases to their canonical forms.
2153 2021-11-10 liuhongt <hongtao.liu@intel.com>
2155 PR tree-optimization/103126
2156 * tree-vect-loop.c (neutral_op_for_reduction): Remove static.
2157 * tree-vectorizer.h (neutral_op_for_reduction): Declare.
2158 * tree-if-conv.c : Include tree-vectorizer.h.
2159 (is_cond_scalar_reduction): Handle
2160 BIT_XOR_EXPR/BIT_IOR_EXPR/BIT_AND_EXPR.
2161 (convert_scalar_cond_reduction): Ditto.
2163 2021-11-10 konglin1 <lingling.kong@intel.com>
2165 * config/i386/sse.md (cmul<conj_op><mode>3): add new define_expand.
2166 (cmla<conj_op><mode>4): Likewise
2168 2021-11-10 Aldy Hernandez <aldyh@redhat.com>
2170 * tree-ssa-threadedge.c: Do not include
2171 gimple-ssa-evrp-analyze.h.
2172 * value-pointer-equiv.cc: Same.
2174 2021-11-10 Aldy Hernandez <aldyh@redhat.com>
2176 * tree-ssa-threadbackward.c
2177 (back_threader::maybe_register_path_dump): Abstract path dumping...
2178 (dump_path): ...here.
2179 (back_threader::resolve_phi): Call dump_path.
2182 2021-11-10 konglin1 <lingling.kong@intel.com>
2184 * config/i386/sse.md (fma_<complexpairopname>_<mode>_pair):
2185 Add new define_insn.
2186 (fma_<mode>_fmaddc_bcst): Add new define_insn_and_split.
2187 (fma_<mode>_fcmaddc_bcst): Likewise
2189 2021-11-10 liuhongt <hongtao.liu@intel.com>
2192 * match.pd: Simplify (trunc)fmax/fmin((extend)a, (extend)b) to
2195 2021-11-10 Andrew Pinski <apinski@marvell.com>
2198 * config/aarch64/aarch64.c (aarch64_evpc_ins): Don't use target
2199 as an input, use original one.
2201 2021-11-10 Sandra Loosemore <sandra@codesourcery.com>
2203 * config/nios2/nios2.c (nios2_can_inline_p): New.
2204 (TARGET_CAN_INLINE_P): Define.
2206 2021-11-09 Bill Schmidt <wschmidt@linux.ibm.com>
2208 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_builtin):
2209 Disable gimple fold for RS6000_BIF_{XVMINDP,XVMINSP,VMINFP} and
2210 RS6000_BIF_{XVMAXDP,XVMAXSP,VMAXFP} when fast-math is not set.
2211 (lxvrse_expand_builtin): Modify the expansion for sign extension.
2212 All extensions are done within VSX registers.
2214 2021-11-09 Andrew Pinski <apinski@marvell.com>
2216 PR tree-optimization/10352
2217 * match.pd: Remove check of TYPE_PRECISION for
2220 2021-11-09 Andrew MacLeod <amacleod@redhat.com>
2222 * function.c (allocate_struct_function): Don't set x_range_query.
2223 * function.h (get_range_query): Move to value-query.h.
2224 * gimple-range.cc (enable_ranger): Check that query is currently NULL.
2225 (disable_ranger): Clear function current query field.
2226 * value-query.cc (get_global_range_query): Relocate to:
2227 * value-query.h (get_global_range_query): Here and inline.
2228 (get_range_query): Relocate here from function.h.
2230 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
2232 * tree-ssa-threadbackward.c
2233 (back_threader::maybe_register_path_dump): New.
2234 (back_threader::maybe_register_path): Call maybe_register_path_dump.
2236 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
2238 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
2239 Return NULL when unprofitable.
2241 2021-11-09 Martin Jambor <mjambor@suse.cz>
2243 * tree.h (build_debug_expr_decl): Declare.
2244 * tree.c (build_debug_expr_decl): New function.
2245 * cfgexpand.c (avoid_deep_ter_for_debug): Use build_debug_expr_decl
2246 instead of building a DEBUG_EXPR_DECL.
2247 * ipa-param-manipulation.c
2248 (ipa_param_body_adjustments::prepare_debug_expressions): Likewise.
2249 * omp-simd-clone.c (ipa_simd_modify_stmt_ops): Likewise.
2250 * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
2251 * tree-ssa-phiopt.c (spaceship_replacement): Likewise.
2252 * tree-ssa-reassoc.c (make_new_ssa_for_def): Likewise.
2254 2021-11-09 Bill Schmidt <wschmidt@linux.ibm.com>
2256 * config/rs6000/rs6000-call.c (rs6000_type_string): New function.
2257 (def_builtin): Change debug formatting for easier parsing and
2258 include more information.
2259 (rs6000_init_builtins): Add dump of autogenerated builtins.
2260 (altivec_init_builtins): Dump __builtin_altivec_mask_for_load for
2263 2021-11-09 Marek Polacek <polacek@redhat.com>
2265 * ipa-modref.c (analyze_parms): Fix a typo.
2267 2021-11-09 Kito Cheng <kito.cheng@sifive.com>
2270 * common/config/riscv/riscv-common.c (multi_letter_subset_rank): Remove
2271 assertion for Z*-ext.
2273 2021-11-09 Jan Hubicka <hubicka@ucw.cz>
2275 * ipa-modref.c (analyze_parms): Add past_flags, past_retslot_flags
2276 and past_static_chain; merge past summary with current one.
2277 (analyze_function): Update.
2279 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
2281 * doc/invoke.texi (Invoking GCC): Document --param=threader-debug.
2283 2021-11-09 Martin Liska <mliska@suse.cz>
2285 * print-rtl.c (rtx_writer::rtx_writer): Make the compilation
2286 conditional based on
2287 * print-rtl.h (class rtx_writer): Likewise.
2289 2021-11-09 Thomas Schwinge <thomas@codesourcery.com>
2291 * input.h (location_hash): Use 'BUILTINS_LOCATION' as spare value
2292 for 'Deleted'. Turn into a '#define'.
2294 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
2296 * gimple-range-path.cc (path_range_query::dump): Clean up.
2297 (path_range_query::compute_ranges): Same.
2298 * value-relation.cc (path_oracle::dump): Same.
2300 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
2302 * dumpfile.c (dump_options): Remove TDF_THREADING entry.
2303 * dumpfile.h (enum dump_flag): Remove TDF_THREADING and adjust
2305 * flag-types.h (enum threader_debug): New.
2306 * gimple-range-path.cc (DEBUG_SOLVER): Use param_threader_debug.
2307 * params.opt: Add entry for --param=threader-debug=.
2309 2021-11-09 Thomas Schwinge <thomas@codesourcery.com>
2311 PR middle-end/101204
2313 * diagnostic-spec.h (typedef xint_hash_t): Turn into...
2314 (struct xint_hash_t): ... this.
2315 * doc/gty.texi: Update.
2317 2021-11-09 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
2319 * config/arm/arm-cpus.in (armv9): New define.
2320 (ARMv9a): New group.
2321 (armv9-a): New arch definition.
2322 * config/arm/arm-tables.opt: Regenerate.
2323 * config/arm/arm.h (BASE_ARCH_9A): New arch enum value.
2324 * config/arm/t-aprofile: Added armv9-a and armv9+simd.
2325 * config/arm/t-arm-elf: Added arm9-a, v9_fps and all_v9_archs
2326 to MULTILIB_MATCHES.
2327 * config/arm/t-multilib: Added v9_a_nosimd_variants and
2328 v9_a_simd_variants to MULTILIB_MATCHES.
2329 * doc/invoke.texi: Update docs.
2331 2021-11-09 Martin Jambor <mjambor@suse.cz>
2334 * ipa-param-manipulation.c (replace_with_mapped_expr): Early
2335 return with error_mark_mode when part of expression is mapped to
2337 (ipa_param_body_adjustments::remap_with_debug_expressions): Set
2338 mapped value to NULL if walk_tree returns error_mark_mode.
2340 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
2342 * tree-ssa-strlen.c (compare_nonzero_chars): Pass statement
2344 (get_addr_stridx): Same.
2346 (get_range_strlen_dynamic): Same.
2347 (handle_builtin_strlen): Same.
2348 (handle_builtin_strchr): Same.
2349 (handle_builtin_strcpy): Same.
2350 (maybe_diag_stxncpy_trunc): Same.
2351 (handle_builtin_stxncpy_strncat): Same.
2352 (handle_builtin_memcpy): Same.
2353 (handle_builtin_strcat): Same.
2354 (handle_alloc_call): Same.
2355 (handle_builtin_memset): Same.
2356 (handle_builtin_string_cmp): Same.
2357 (handle_pointer_plus): Same.
2358 (count_nonzero_bytes_addr): Same.
2359 (count_nonzero_bytes): Same.
2360 (handle_store): Same.
2361 (fold_strstr_to_strncmp): Same.
2362 (handle_integral_assign): Same.
2363 (check_and_optimize_stmt): Same.
2364 (class strlen_dom_walker): Replace evrp with ranger.
2365 (strlen_dom_walker::before_dom_children): Remove evrp.
2366 (strlen_dom_walker::after_dom_children): Remove evrp.
2367 * gimple-ssa-warn-access.cc (maybe_check_access_sizes):
2368 Restrict sprintf output.
2370 2021-11-09 Martin Liska <mliska@suse.cz>
2372 * genconditions.c (write_one_condition): Add const qualifier
2375 2021-11-09 Jeff Law <jeffreyalaw@gmail.com>
2377 * match.pd: New pattern to simplify (1 << n) & M ==/!= 0 for M
2380 2021-11-08 Bill Schmidt <wschmidt@linux.ibm.com>
2382 * config/rs6000/rs6000-builtin-new.def (VEC_INIT_V16QI): Use
2383 escape-newline support.
2384 (VEC_INIT_V4SI): Likewise.
2385 (VEC_INIT_V8HI): Likewise.
2386 (PACK_V1TI): Likewise.
2388 (VFIRSTMISMATCHOREOSINDEX_V16QI): Likewise.
2389 (VFIRSTMISMATCHOREOSINDEX_V8HI): Likewise.
2390 (VFIRSTMISMATCHOREOSINDEX_V4SI): Likewise.
2394 (FMAF128_ODD): Likewise.
2395 (VSCEQPUO): Likewise.
2397 (VSIEQPF): Likewise.
2398 (VSTDCQP): Likewise.
2399 (PACK_TD): Likewise.
2400 (TABORTDC): Likewise.
2401 (TABORTDCI): Likewise.
2402 (SE_LXVRBX): Likewise.
2403 (SE_LXVRHX): Likewise.
2404 (SE_LXVRWX): Likewise.
2405 (SE_LXVRDX): Likewise.
2406 (VREPLACE_UN_UV2DI): Likewise.
2407 (VREPLACE_UN_UV4SI): Likewise.
2408 (VREPLACE_UN_V2DI): Likewise.
2409 (VREPLACE_ELT_UV2DI): Likewise.
2410 (VREPLACE_ELT_V2DI): Likewise.
2411 (ZE_LXVRBX): Likewise.
2412 (ZE_LXVRHX): Likewise.
2413 (ZE_LXVRWX): Likewise.
2414 (ZE_LXVRDX): Likewise.
2416 (CNTLZDM): Likewise.
2417 (CNTTZDM): Likewise.
2420 (PMXVBF16GER2): Likewise.
2421 (PMXVBF16GER2_INTERNAL): Likewise.
2422 (PMXVBF16GER2NN): Likewise.
2423 (PMXVBF16GER2NN_INTERNAL): Likewise.
2424 (PMXVBF16GER2NP): Likewise.
2425 (PMXVBF16GER2NP_INTERNAL): Likewise.
2426 (PMXVBF16GER2PN): Likewise.
2427 (PMXVBF16GER2PN_INTERNAL): Likewise.
2428 (PMXVBF16GER2PP): Likewise.
2429 (PMXVBF16GER2PP_INTERNAL): Likewise.
2430 (PMXVF16GER2): Likewise.
2431 (PMXVF16GER2_INTERNAL): Likewise.
2432 (PMXVF16GER2NN): Likewise.
2433 (PMXVF16GER2NN_INTERNAL): Likewise.
2434 (PMXVF16GER2NP): Likewise.
2435 (PMXVF16GER2NP_INTERNAL): Likewise.
2436 (PMXVF16GER2PN): Likewise.
2437 (PMXVF16GER2PN_INTERNAL): Likewise.
2438 (PMXVF16GER2PP): Likewise.
2439 (PMXVF16GER2PP_INTERNAL): Likewise.
2440 (PMXVF32GER_INTERNAL): Likewise.
2441 (PMXVF32GERNN): Likewise.
2442 (PMXVF32GERNN_INTERNAL): Likewise.
2443 (PMXVF32GERNP): Likewise.
2444 (PMXVF32GERNP_INTERNAL): Likewise.
2445 (PMXVF32GERPN): Likewise.
2446 (PMXVF32GERPN_INTERNAL): Likewise.
2447 (PMXVF32GERPP): Likewise.
2448 (PMXVF32GERPP_INTERNAL): Likewise.
2449 (PMXVF64GER): Likewise.
2450 (PMXVF64GER_INTERNAL): Likewise.
2451 (PMXVF64GERNN): Likewise.
2452 (PMXVF64GERNN_INTERNAL): Likewise.
2453 (PMXVF64GERNP): Likewise.
2454 (PMXVF64GERNP_INTERNAL): Likewise.
2455 (PMXVF64GERPN): Likewise.
2456 (PMXVF64GERPN_INTERNAL): Likewise.
2457 (PMXVF64GERPP): Likewise.
2458 (PMXVF64GERPP_INTERNAL): Likewise.
2459 (PMXVI16GER2): Likewise.
2460 (PMXVI16GER2_INTERNAL): Likewise.
2461 (PMXVI16GER2PP): Likewise.
2462 (PMXVI16GER2PP_INTERNAL): Likewise.
2463 (PMXVI16GER2S): Likewise.
2464 (PMXVI16GER2S_INTERNAL): Likewise.
2465 (PMXVI16GER2SPP): Likewise.
2466 (PMXVI16GER2SPP_INTERNAL): Likewise.
2467 (PMXVI4GER8): Likewise.
2468 (PMXVI4GER8_INTERNAL): Likewise.
2469 (PMXVI4GER8PP): Likewise.
2470 (PMXVI4GER8PP_INTERNAL): Likewise.
2471 (PMXVI8GER4): Likewise.
2472 (PMXVI8GER4_INTERNAL): Likewise.
2473 (PMXVI8GER4PP): Likewise.
2474 (PMXVI8GER4PP_INTERNAL): Likewise.
2475 (PMXVI8GER4SPP): Likewise.
2476 (PMXVI8GER4SPP_INTERNAL): Likewise.
2477 * config/rs6000/rs6000-gen-builtins.c (MAXLINES): New macro.
2478 (linebuf): Increase size.
2479 (lines): New variable.
2480 (lastline): Likewise.
2481 (real_line_pos): New function.
2482 (diag): Change signature.
2483 (bif_diag): Change signature; support escape-newline handling.
2484 (ovld_diag): Likewise.
2485 (fatal): Move earlier.
2486 (consume_whitespace): Adjust diag call.
2487 (advance_line): Add escape-newline handling; call fatal.
2488 (safe_inc_pos): Adjust diag call.
2489 (match_identifier): Likewise.
2490 (match_integer): Likewise.
2491 (match_to_right_bracket): Call fatal instead of diag; adjust diag
2493 (match_basetype): Adjust diag calls.
2494 (match_bracketed_pair): Likewise.
2495 (match_const_restriction): Likewise.
2496 (match_type): Likewise.
2497 (parse_args): Likewise.
2498 (parse_bif_attrs): Likewise.
2499 (complete_vector_type): Likewise.
2500 (complete_base_type): Likewise.
2501 (parse_prototype): Likewise.
2502 (parse_bif_entry): Likewise.
2503 (parse_bif_stanza): Likewise.
2504 (parse_ovld_entry): Likewise.
2505 (parse_ovld_stanza): Likewise.
2506 (main): Allocate buffers for lines[].
2508 2021-11-08 Eric Botcazou <ebotcazou@adacore.com>
2510 * config/i386/i386.c (legitimize_pic_address): Adjust comment and
2511 use the REG argument on the CM_LARGE_PIC code path as well.
2512 * config/i386/predicates.md (gotoff_operand): Do not treat VxWorks
2513 specially with the large code models.
2515 2021-11-08 Jan Hubicka <hubicka@ucw.cz>
2517 * gimple.c (gimple_call_static_chain_flags): Revert the workaround
2518 allowing interposition since issues with binds_to_local_def were
2521 2021-11-08 Andrew MacLeod <amacleod@redhat.com>
2523 PR tree-optimization/103122
2524 * gimple-range.cc (gimple_ranger::range_of_expr): Request the cache
2525 entry with "calulate new values" set to false.
2527 2021-11-08 Jan Hubicka <hubicka@ucw.cz>
2529 * builtins.c (is_simple_builtin): Add builitin_dwarf_cfa
2530 and builtin_return_address.
2531 (builtin_fnspec): Annotate builtin_return,
2532 bulitin_eh_pointer, builtin_eh_filter, builtin_unwind_resume,
2533 builtin_cxa_end_cleanup, builtin_eh_copy_values,
2534 builtin_frame_address, builtin_apply_args,
2535 builtin_asan_before_dynamic_init, builtin_asan_after_dynamic_init,
2536 builtin_prefetch, builtin_dwarf_cfa, builtin_return_addrss
2538 * ipa-pure-const.c (special_builtin_state): Add builtin_dwarf_cfa
2539 and builtin_return_address.
2541 2021-11-08 Jan Hubicka <hubicka@ucw.cz>
2543 PR tree-optimization/103177
2544 * passes.def: Move uncprop after pure/const and modref.
2546 2021-11-08 Martin Jambor <mjambor@suse.cz>
2550 * tree-inline.c (remap_gimple_stmt): Unshare the expression without
2551 location before invoking remap_with_debug_expressions on it.
2552 * ipa-param-manipulation.c
2553 (ipa_param_body_adjustments::prepare_debug_expressions): Likewise.
2555 2021-11-08 David Edelsohn <dje.gcc@gmail.com>
2557 * config/rs6000/vsx.md (vsx_splat_v4si_di): Revert "wa"
2560 2021-11-08 Richard Biener <rguenther@suse.de>
2562 * cfgloop.c (verify_loop_structure): Use a temporary BB flag
2563 instead of an sbitmap to cache irreducible state.
2565 2021-11-08 Aldy Hernandez <aldyh@redhat.com>
2567 PR tree-optimization/103120
2568 * gimple-range-path.cc (path_range_query::range_defined_in_block):
2569 Bail if there's a cache entry.
2571 2021-11-08 Bill Schmidt <wschmidt@linux.ibm.com>
2573 * config/rs6000/rs6000.c (rs6000_builtin_reciprocal): Use
2574 rs6000_builtin_decls_x when appropriate.
2575 (add_condition_to_bb): Likewise.
2576 (rs6000_atomic_assign_expand_fenv): Likewise.
2578 2021-11-08 Bill Schmidt <wschmidt@linux.ibm.com>
2580 * config/rs6000/rs6000-call.c (rs6000_new_builtin_decl): New function.
2581 (rs6000_builtin_decl): Call it.
2583 2021-11-08 Martin Liska <mliska@suse.cz>
2585 * cgraph.c (cgraph_node::dump): Dump it from decl.
2587 2021-11-08 Martin Liska <mliska@suse.cz>
2589 PR gcov-profile/100520
2590 * coverage.c (coverage_compute_profile_id): Strip .gk when
2591 compare debug is used.
2592 * system.h (endswith): New function.
2594 2021-11-08 Martin Liska <mliska@suse.cz>
2596 * cgraph.c (cgraph_node::dump): Dump static_chain_decl.
2598 2021-11-08 Thomas Schwinge <thomas@codesourcery.com>
2600 * config/rs6000/rbtree.c: Fix 'Copyright (C) 2020-21' into '2020-2021'
2601 * config/rs6000/rbtree.h: Likewise.
2602 * config/rs6000/rs6000-builtin-new.def: Likewise.
2603 * config/rs6000/rs6000-gen-builtins.c: Likewise.
2604 * config/rs6000/rs6000-overload.def: Likewise.
2606 2021-11-08 Richard Sandiford <richard.sandiford@arm.com>
2608 * config/aarch64/aarch64.c (aarch64_ldn_stn_vectors): New function.
2609 (aarch64_address_cost): Use it instead of testing for CImode and
2612 2021-11-08 Richard Sandiford <richard.sandiford@arm.com>
2614 * genmodes.c (emit_insn_modes_h): Define NUM_MODE_* macros.
2615 * expmed.h (NUM_MODE_INT): Delete in favor of genmodes definitions.
2616 (NUM_MODE_PARTIAL_INT, NUM_MODE_VECTOR_INT): Likewise.
2617 * real.h (real_format_for_mode): Use NUM_MODE_FLOAT and
2618 NUM_MODE_DECIMAL_FLOAT.
2619 (REAL_MODE_FORMAT): Likewise.
2621 2021-11-08 Andre Vieira <andre.simoesdiasvieira@arm.com>
2623 * tree-vect-loop.c (vect_better_loop_vinfo_p): Change how epilogue loop
2626 2021-11-08 Richard Biener <rguenther@suse.de>
2628 * tree-vectorizer.h (vect_create_loop_vinfo): Add main_loop_info
2630 * tree-vect-loop.c (vect_create_loop_vinfo): Likewise. Set
2631 LOOP_VINFO_ORIG_LOOP_INFO and conditionalize set of
2632 LOOP_VINFO_NITERS_ASSUMPTIONS.
2633 (vect_analyze_loop_1): Adjust.
2634 (vect_analyze_loop): Move loop constraint setting and
2635 SCEV/niter reset here from vect_create_loop_vinfo to perform
2637 (vect_analyze_loop_form): Move dumping of symbolic niters
2638 here from vect_create_loop_vinfo.
2640 2021-11-08 Jan Hubicka <hubicka@ucw.cz>
2642 * ipa-modref-tree.h (enum modref_special_parms): New enum.
2643 (struct modref_access_node): update for special parms.
2644 (struct modref_ref_node): Likewise.
2645 (struct modref_parm_map): Likewise.
2646 (struct modref_tree): Likewise.
2647 * ipa-modref.c (dump_access): Likewise.
2648 (get_access): Detect static chain.
2649 (parm_map_for_arg): Take tree as arg instead of
2651 (merge_call_side_effects): Compute map for static chain.
2652 (process_fnspec): Update.
2653 (struct escape_point): Remove retslot_arg and static_chain_arg.
2654 (analyze_parms): Update.
2655 (compute_parm_map): Update.
2656 (propagate_unknown_call): Update.
2657 (modref_propagate_in_scc): Update.
2658 (modref_merge_call_site_flags): Update.
2659 (ipa_merge_modref_summary_after_inlining): Update.
2660 * tree-ssa-alias.c (modref_may_conflict): Handle static chain.
2661 * ipa-modref-tree.c (test_merge): Update.
2663 2021-11-08 Haochen Gui <guihaoc@gcc.gnu.org>
2665 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_builtin): Disable
2666 gimple fold for VSX_BUILTIN_XVMINDP, ALTIVEC_BUILTIN_VMINFP,
2667 VSX_BUILTIN_XVMAXDP, ALTIVEC_BUILTIN_VMAXFP when fast-math is not
2670 2021-11-08 liuhongt <hongtao.liu@intel.com>
2672 PR tree-optimization/103077
2673 * doc/invoke.texi (Options That Control Optimization):
2674 Update documentation for -ftree-loop-vectorize and
2675 -ftree-slp-vectorize which are enabled by default at -02.
2677 2021-11-08 liuhongt <hongtao.liu@intel.com>
2680 * match.pd (Simplifcation (trunc)copysign((extend)a, (extend)b)
2681 to .COPYSIGN (a, b)): Add !HONOR_SNANS.
2683 2021-11-08 liuhongt <hongtao.liu@intel.com>
2686 * match.pd: Simplify
2687 (trunc)fma ((extend)a, (extend)b, (extend)c) to IFN_FMA (a, b,
2688 c) under flag_unsafe_math_optimizations.
2690 2021-11-07 Jan Hubicka <hubicka@ucw.cz>
2693 * params.opt (modref-max-depth): Add range.
2694 (modref-max-adjustments): Fix range.
2696 2021-11-07 Aldy Hernandez <aldyh@redhat.com>
2698 * tree-pass.h (make_pass_vrp_threader): Remove.
2699 * tree-ssa-threadbackward.c
2700 (back_threader_profitability::profitable_path_p): Remove
2701 ASSERT_EXPR references.
2702 * tree-ssa-threadedge.c (jt_state::register_equivs_stmt): Same.
2703 * tree-vrp.c (vrp_folder::simplify_casted_conds): Same.
2704 (execute_vrp): Same.
2705 (class hybrid_threader): Remove.
2706 (hybrid_threader::hybrid_threader): Remove.
2707 (hybrid_threader::~hybrid_threader): Remove.
2708 (hybrid_threader::before_dom_children): Remove.
2709 (hybrid_threader::after_dom_children): Remove.
2710 (execute_vrp_threader): Remove.
2711 (class pass_vrp_threader): Remove.
2712 (make_pass_vrp_threader): Remove.
2714 2021-11-07 Jan Hubicka <hubicka@ucw.cz>
2716 * gimple.c (gimple_call_arg_flags): Use interposable_eaf_flags.
2717 (gimple_call_retslot_flags): Likewise.
2718 (gimple_call_static_chain_flags): Likewise.
2719 * ipa-modref.c (remove_useless_eaf_flags): Do not remove everything for
2721 (modref_summary::useful_p): Likewise.
2722 (modref_summary_lto::useful_p): Likewise.
2723 (analyze_parms): Do not give up on NOVOPS.
2724 (analyze_function): When dumping report chnages in EAF flags
2725 between IPA and local pass.
2726 (modref_merge_call_site_flags): Compute implicit eaf flags
2727 based on callee ecf_flags and fnspec; if the function does not
2728 bind to current defs use interposable_eaf_flags.
2729 (modref_propagate_flags_in_scc): Update.
2730 * ipa-modref.h (interposable_eaf_flags): New function.
2732 2021-11-07 Bill Schmidt <wschmidt@linux.ibm.com>
2734 * config/rs6000/rs6000-call.c (rs6000_expand_new_builtin): New
2736 (rs6000_invalid_new_builtin): New function.
2737 (rs6000_expand_builtin): Call rs6000_expand_new_builtin.
2738 (rs6000_expand_ldst_mask): New function.
2739 (new_cpu_expand_builtin): Likewise.
2740 (elemrev_icode): Likewise.
2741 (ldv_expand_builtin): Likewise.
2742 (lxvrse_expand_builtin): Likewise.
2743 (lxvrze_expand_builtin): Likewise.
2744 (stv_expand_builtin): Likewise.
2745 (new_mma_expand_builtin): Likewise.
2746 (new_htm_spr_num): Likewise.
2747 (new_htm_expand_builtin): Likewise.
2748 (rs6000_expand_new_builtin): Likewise.
2749 (rs6000_init_builtins): Initialize altivec_builtin_mask_for_load.
2751 2021-11-07 Jan Hubicka <hubicka@ucw.cz>
2753 * ipa-modref.c (modref_lattice): Add do_dataflow,
2754 changed and propagate_to fields.
2755 (modref_lattice::release): Free propagate_to
2756 (modref_lattice::merge): Do not give up early on unknown
2758 (modref_lattice::merge_deref): Likewise.
2759 (modref_eaf_analysis): Update toplevel comment.
2760 (modref_eaf_analysis::analyze_ssa_name): Record postponned ssa names;
2761 do optimistic dataflow initialization.
2762 (modref_eaf_analysis::merge_with_ssa_name): Build dataflow graph.
2763 (modref_eaf_analysis::propagate): New member function.
2764 (analyze_parms): Update to new API of modref_eaf_analysis.
2766 2021-11-06 Jan Hubicka <hubicka@ucw.cz>
2768 * cgraph.h (cgraph_node::can_be_discarded_p): Do not
2769 return true on functions from other partition.
2771 2021-11-06 David Edelsohn <dje.gcc@gmail.com>
2773 * config/rs6000/vsx.md (vsx_splat_v4si): Change constraints to "wa".
2774 (vsx_splat_v4si_di): Change constraint to "wa".
2776 2021-11-06 Aldy Hernandez <aldyh@redhat.com>
2778 PR tree-optimization/103061
2779 * value-relation.cc (path_oracle::path_oracle): Initialize
2781 (path_oracle::killing_def): Set m_killed_defs.
2782 (path_oracle::query_relation): Do not look at the root oracle for
2784 * value-relation.h (class path_oracle): Add m_killed_defs.
2786 2021-11-06 Aldy Hernandez <aldyh@redhat.com>
2788 * tree-ssa-threadbackward.c (back_threader::find_paths_to_names):
2789 Remove gotos and other cleanups.
2791 2021-11-05 Jan Hubicka <hubicka@ucw.cz>
2794 * ipa-modref-tree.h (modref_tree::insert): Do nothing for
2795 paradoxical and zero sized accesses.
2797 2021-11-05 Jan Hubicka <hubicka@ucw.cz>
2800 * ipa-modref-tree.h (struct modref_access_node): Avoid left shift
2803 2021-11-05 Iain Sandoe <iain@sandoe.co.uk>
2805 * config/darwin.h (ASM_GENERATE_INTERNAL_LABEL): Add LTRAMP
2806 to the list of symbol prefixes that must be made linker-
2809 2021-11-05 Iain Sandoe <iain@sandoe.co.uk>
2811 * config.host: Add support for aarch64-*-darwin.
2812 * config/aarch64/host-aarch64-darwin.c: New file.
2813 * config/aarch64/x-darwin: New file.
2815 2021-11-05 Andrew MacLeod <amacleod@redhat.com>
2817 PR tree-optimization/103093
2818 * gimple-range-gori.cc (range_def_chain::get_imports): Remove assert.
2820 2021-11-05 Andrew MacLeod <amacleod@redhat.com>
2822 PR tree-optimization/102943
2823 * gimple-range-cache.cc (class update_list): New.
2824 (update_list::add): Replace add_to_update.
2825 (update_list::pop): New.
2826 (ranger_cache::ranger_cache): Adjust.
2827 (ranger_cache::~ranger_cache): Adjust.
2828 (ranger_cache::add_to_update): Delete.
2829 (ranger_cache::propagate_cache): Adjust to new class.
2830 (ranger_cache::propagate_updated_value): Ditto.
2831 (ranger_cache::fill_block_cache): Ditto.
2832 * gimple-range-cache.h (class ranger_cache): Adjust to update class.
2834 2021-11-05 Richard Biener <rguenther@suse.de>
2836 * tree-vect-loop.c (vect_analyze_loop): Remove obsolete
2837 comment and expand on another one. Combine nested if.
2839 2021-11-05 John David Anglin <danglin@gcc.gnu.org>
2842 * config/pa/pa-modes.def: Add OImode integer type.
2843 * config/pa/pa.c (pa_scalar_mode_supported_p): Allow TImode
2845 * config/pa/pa.h (MIN_UNITS_PER_WORD) Define to MIN_UNITS_PER_WORD
2846 to UNITS_PER_WORD if IN_LIBGCC2.
2847 * config/pa/pa.md (addti3, addvti3, subti3, subvti3, negti2,
2848 negvti2, ashlti3, shrpd_internal): New patterns.
2849 Change some multi instruction types to multi.
2851 2021-11-05 Jakub Jelinek <jakub@redhat.com>
2854 * config/i386/i386.h
2855 (stringop_algs::stringop_strategy::stringop_strategy): Make the ctor
2858 2021-11-05 Wilco Dijkstra <wdijkstr@arm.com>
2861 * config/aarch64/aarch64.c (aarch64_mov_operand_p): Strip the salt
2863 * config/aarch64/constraints.md: Support const in Usw.
2865 2021-11-05 John David Anglin <danglin@gcc.gnu.org>
2867 * config/pa/pa.h (PREFERRED_DEBUGGING_TYPE): Define to DWARF2_DEBUG.
2868 * config/pa/pa64-hpux.h (PREFERRED_DEBUGGING_TYPE): Remove define.
2870 2021-11-05 Richard Biener <rguenther@suse.de>
2872 * tree-vectorizer.h (vec_info_shared::n_stmts): Add.
2873 (LOOP_VINFO_N_STMTS): Likewise.
2874 (vec_info_for_bb): Remove unused function.
2875 * tree-vectorizer.c (vec_info_shared::vec_info_shared):
2876 Initialize n_stmts member.
2877 * tree-vect-loop.c: Remove INCLUDE_FUNCTIONAL.
2878 (vect_create_loop_vinfo): Do not set loop->aux.
2879 (vect_analyze_loop_2): Do not get n_stmts as argument,
2880 instead use LOOP_VINFO_N_STMTS. Set LOOP_VINFO_VECTORIZABLE_P
2882 (vect_analyze_loop_1): Remove callback, get the mode iterator
2883 and autodetected_vector_mode as argument, advancing the
2884 iterator and initializing autodetected_vector_mode here.
2885 (vect_analyze_loop): Split analysis loop into two, first
2886 processing main loops only and then epilogues.
2888 2021-11-05 Martin Jambor <mjambor@suse.cz>
2890 * ipa-prop.c (compute_complex_assign_jump_func): Remove
2891 unnecessary check for RECORD_TYPE.
2893 2021-11-05 Gerald Pfeifer <gerald@pfeifer.com>
2895 * doc/install.texi2html: Do not generate old.html any longer.
2897 2021-11-05 Martin Liska <mliska@suse.cz>
2900 * opts.c (finish_options): Reset flag_gtoggle when it is used.
2902 2021-11-05 Jakub Jelinek <jakub@redhat.com>
2905 * dwarf2out.c (add_const_value_attribute): Add MODE argument, use it
2906 in CONST_WIDE_INT handling. Adjust recursive calls.
2907 (add_location_or_const_value_attribute): Pass DECL_MODE (decl) to
2908 new add_const_value_attribute argument.
2909 (tree_add_const_value_attribute): Pass TYPE_MODE (type) to new
2910 add_const_value_attribute argument.
2912 2021-11-05 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
2914 * config/vx-common.h: Test value of TARGET_VXWORKS7 rather
2917 2021-11-05 Richard Biener <rguenther@suse.de>
2919 * tree-vectorizer.h (struct vect_loop_form_info): New.
2920 (vect_analyze_loop_form): Adjust.
2921 (vect_create_loop_vinfo): New.
2922 * tree-parloops.c (gather_scalar_reductions): Adjust for
2923 vect_analyze_loop_form API change.
2924 * tree-vect-loop.c: Include <functional>.
2925 (vect_analyze_loop_form_1): Rename to vect_analyze_loop_form,
2926 take struct vect_loop_form_info as output parameter and adjust.
2927 (vect_analyze_loop_form): Rename to vect_create_loop_vinfo and
2928 split out call to the original vect_analyze_loop_form_1.
2929 (vect_reanalyze_as_main_loop): Rename to...
2930 (vect_analyze_loop_1): ... this, factor out the call to
2931 vect_analyze_loop_form and generalize to be able to use it twice ...
2932 (vect_analyze_loop): ... here. Perform vect_analyze_loop_form
2935 2021-11-05 Xionghu Luo <luoxhu@linux.ibm.com>
2938 * config/rs6000/fusion.md: Regenerate.
2939 * config/rs6000/genfusion.pl: Fix incorrect clobber constraint.
2941 2021-11-04 Andreas Krebbel <krebbel@linux.ibm.com>
2943 * config/s390/s390.h (STACK_CHECK_MOVING_SP): New macro
2946 2021-11-04 Tamar Christina <tamar.christina@arm.com>
2948 * config/aarch64/aarch64-builtins.c
2949 (aarch64_general_gimple_fold_builtin): Add ashl, sshl, ushl, ashr,
2950 ashr_simd, lshr, lshr_simd.
2951 * config/aarch64/aarch64-simd-builtins.def (lshr): Use USHIFTIMM.
2952 * config/aarch64/arm_neon.h (vshr_n_u8, vshr_n_u16, vshr_n_u32,
2953 vshrq_n_u8, vshrq_n_u16, vshrq_n_u32, vshrq_n_u64): Fix type hack.
2955 2021-11-04 Tamar Christina <tamar.christina@arm.com>
2957 * match.pd: New negate+shift pattern.
2959 2021-11-04 Andrew MacLeod <amacleod@redhat.com>
2961 PR tree-optimization/103079
2962 * gimple-range-gori.cc (gimple_range_calc_op1): Treat undefined as
2964 (gimple_range_calc_op2): Ditto.
2966 2021-11-04 Martin Jambor <mjambor@suse.cz>
2969 * ipa-param-manipulation.h (class ipa_param_body_adjustments): New
2970 members remap_with_debug_expressions, m_dead_ssa_debug_equiv,
2971 m_dead_stmt_debug_equiv and prepare_debug_expressions. Added
2972 parameter to mark_dead_statements.
2973 * ipa-param-manipulation.c: Include tree-phinodes.h and cfgexpand.h.
2974 (ipa_param_body_adjustments::mark_dead_statements): New parameter
2975 debugstack, push into it all SSA names used in debug statements,
2976 produce m_dead_ssa_debug_equiv mapping for the removed param.
2977 (replace_with_mapped_expr): New function.
2978 (ipa_param_body_adjustments::remap_with_debug_expressions): Likewise.
2979 (ipa_param_body_adjustments::prepare_debug_expressions): Likewise.
2980 (ipa_param_body_adjustments::common_initialization): Gather and
2981 procecc SSA which will be removed but are in debug statements. Simplify.
2982 (ipa_param_body_adjustments::ipa_param_body_adjustments): Initialize
2984 * tree-inline.c (remap_gimple_stmt): Create a debug bind when possible
2985 when avoiding a copy of an unnecessary statement. Remap removed SSA
2986 names in existing debug statements.
2987 (tree_function_versioning): Do not create DEBUG_EXPR_DECL for removed
2988 parameters if we have already done so.
2990 2021-11-04 Jan Hubicka <hubicka@ucw.cz>
2993 * gimple.c (gimple_call_static_chain_flags): Handle case when
2994 nested function does not bind locally.
2996 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
2998 * config/aarch64/aarch64.c (aarch64_function_value): Generate
2999 a register rtx for Neon vector-tuple modes.
3000 (aarch64_layout_arg): Likewise.
3002 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
3004 * lower-subreg.c (simple_move): Prevent decomposition if
3005 modes are not tieable.
3007 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
3008 Richard Sandiford <richard.sandiford@arm.com>
3010 * config/aarch64/aarch64-builtins.c (v2x8qi_UP): Define.
3011 (v2x4hi_UP): Likewise.
3012 (v2x4hf_UP): Likewise.
3013 (v2x4bf_UP): Likewise.
3014 (v2x2si_UP): Likewise.
3015 (v2x2sf_UP): Likewise.
3016 (v2x1di_UP): Likewise.
3017 (v2x1df_UP): Likewise.
3018 (v2x16qi_UP): Likewise.
3019 (v2x8hi_UP): Likewise.
3020 (v2x8hf_UP): Likewise.
3021 (v2x8bf_UP): Likewise.
3022 (v2x4si_UP): Likewise.
3023 (v2x4sf_UP): Likewise.
3024 (v2x2di_UP): Likewise.
3025 (v2x2df_UP): Likewise.
3026 (v3x8qi_UP): Likewise.
3027 (v3x4hi_UP): Likewise.
3028 (v3x4hf_UP): Likewise.
3029 (v3x4bf_UP): Likewise.
3030 (v3x2si_UP): Likewise.
3031 (v3x2sf_UP): Likewise.
3032 (v3x1di_UP): Likewise.
3033 (v3x1df_UP): Likewise.
3034 (v3x16qi_UP): Likewise.
3035 (v3x8hi_UP): Likewise.
3036 (v3x8hf_UP): Likewise.
3037 (v3x8bf_UP): Likewise.
3038 (v3x4si_UP): Likewise.
3039 (v3x4sf_UP): Likewise.
3040 (v3x2di_UP): Likewise.
3041 (v3x2df_UP): Likewise.
3042 (v4x8qi_UP): Likewise.
3043 (v4x4hi_UP): Likewise.
3044 (v4x4hf_UP): Likewise.
3045 (v4x4bf_UP): Likewise.
3046 (v4x2si_UP): Likewise.
3047 (v4x2sf_UP): Likewise.
3048 (v4x1di_UP): Likewise.
3049 (v4x1df_UP): Likewise.
3050 (v4x16qi_UP): Likewise.
3051 (v4x8hi_UP): Likewise.
3052 (v4x8hf_UP): Likewise.
3053 (v4x8bf_UP): Likewise.
3054 (v4x4si_UP): Likewise.
3055 (v4x4sf_UP): Likewise.
3056 (v4x2di_UP): Likewise.
3057 (v4x2df_UP): Likewise.
3058 (TYPES_GETREGP): Delete.
3059 (TYPES_SETREGP): Likewise.
3060 (TYPES_LOADSTRUCT_U): Define.
3061 (TYPES_LOADSTRUCT_P): Likewise.
3062 (TYPES_LOADSTRUCT_LANE_U): Likewise.
3063 (TYPES_LOADSTRUCT_LANE_P): Likewise.
3064 (TYPES_STORE1P): Move for consistency.
3065 (TYPES_STORESTRUCT_U): Define.
3066 (TYPES_STORESTRUCT_P): Likewise.
3067 (TYPES_STORESTRUCT_LANE_U): Likewise.
3068 (TYPES_STORESTRUCT_LANE_P): Likewise.
3069 (aarch64_simd_tuple_types): Define.
3070 (aarch64_lookup_simd_builtin_type): Handle tuple type lookup.
3071 (aarch64_init_simd_builtin_functions): Update frontend lookup
3072 for builtin functions after handling arm_neon.h pragma.
3073 (register_tuple_type): Manually set modes of single-integer
3074 tuple types. Record tuple types.
3075 * config/aarch64/aarch64-modes.def
3076 (ADV_SIMD_D_REG_STRUCT_MODES): Define D-register tuple modes.
3077 (ADV_SIMD_Q_REG_STRUCT_MODES): Define Q-register tuple modes.
3078 (SVE_MODES): Give single-vector modes priority over vector-
3080 (VECTOR_MODES_WITH_PREFIX): Set partial-vector mode order to
3081 be after all single-vector modes.
3082 * config/aarch64/aarch64-simd-builtins.def: Update builtin
3083 generator macros to reflect modifications to the backend
3085 * config/aarch64/aarch64-simd.md (aarch64_simd_ld2<mode>):
3086 Use vector-tuple mode iterator and rename to...
3087 (aarch64_simd_ld2<vstruct_elt>): This.
3088 (aarch64_simd_ld2r<mode>): Use vector-tuple mode iterator and
3090 (aarch64_simd_ld2r<vstruct_elt>): This.
3091 (aarch64_vec_load_lanesoi_lane<mode>): Use vector-tuple mode
3092 iterator and rename to...
3093 (aarch64_vec_load_lanes<mode>_lane<vstruct_elt>): This.
3094 (vec_load_lanesoi<mode>): Use vector-tuple mode iterator and
3096 (vec_load_lanes<mode><vstruct_elt>): This.
3097 (aarch64_simd_st2<mode>): Use vector-tuple mode iterator and
3099 (aarch64_simd_st2<vstruct_elt>): This.
3100 (aarch64_vec_store_lanesoi_lane<mode>): Use vector-tuple mode
3101 iterator and rename to...
3102 (aarch64_vec_store_lanes<mode>_lane<vstruct_elt>): This.
3103 (vec_store_lanesoi<mode>): Use vector-tuple mode iterator and
3105 (vec_store_lanes<mode><vstruct_elt>): This.
3106 (aarch64_simd_ld3<mode>): Use vector-tuple mode iterator and
3108 (aarch64_simd_ld3<vstruct_elt>): This.
3109 (aarch64_simd_ld3r<mode>): Use vector-tuple mode iterator and
3111 (aarch64_simd_ld3r<vstruct_elt>): This.
3112 (aarch64_vec_load_lanesci_lane<mode>): Use vector-tuple mode
3113 iterator and rename to...
3114 (vec_load_lanesci<mode>): This.
3115 (aarch64_simd_st3<mode>): Use vector-tuple mode iterator and
3117 (aarch64_simd_st3<vstruct_elt>): This.
3118 (aarch64_vec_store_lanesci_lane<mode>): Use vector-tuple mode
3119 iterator and rename to...
3120 (vec_store_lanesci<mode>): This.
3121 (aarch64_simd_ld4<mode>): Use vector-tuple mode iterator and
3123 (aarch64_simd_ld4<vstruct_elt>): This.
3124 (aarch64_simd_ld4r<mode>): Use vector-tuple mode iterator and
3126 (aarch64_simd_ld4r<vstruct_elt>): This.
3127 (aarch64_vec_load_lanesxi_lane<mode>): Use vector-tuple mode
3128 iterator and rename to...
3129 (vec_load_lanesxi<mode>): This.
3130 (aarch64_simd_st4<mode>): Use vector-tuple mode iterator and
3132 (aarch64_simd_st4<vstruct_elt>): This.
3133 (aarch64_vec_store_lanesxi_lane<mode>): Use vector-tuple mode
3134 iterator and rename to...
3135 (vec_store_lanesxi<mode>): This.
3136 (mov<mode>): Define for Neon vector-tuple modes.
3137 (aarch64_ld1x3<VALLDIF:mode>): Use vector-tuple mode iterator
3139 (aarch64_ld1x3<vstruct_elt>): This.
3140 (aarch64_ld1_x3_<mode>): Use vector-tuple mode iterator and
3142 (aarch64_ld1_x3_<vstruct_elt>): This.
3143 (aarch64_ld1x4<VALLDIF:mode>): Use vector-tuple mode iterator
3145 (aarch64_ld1x4<vstruct_elt>): This.
3146 (aarch64_ld1_x4_<mode>): Use vector-tuple mode iterator and
3148 (aarch64_ld1_x4_<vstruct_elt>): This.
3149 (aarch64_st1x2<VALLDIF:mode>): Use vector-tuple mode iterator
3151 (aarch64_st1x2<vstruct_elt>): This.
3152 (aarch64_st1_x2_<mode>): Use vector-tuple mode iterator and
3154 (aarch64_st1_x2_<vstruct_elt>): This.
3155 (aarch64_st1x3<VALLDIF:mode>): Use vector-tuple mode iterator
3157 (aarch64_st1x3<vstruct_elt>): This.
3158 (aarch64_st1_x3_<mode>): Use vector-tuple mode iterator and
3160 (aarch64_st1_x3_<vstruct_elt>): This.
3161 (aarch64_st1x4<VALLDIF:mode>): Use vector-tuple mode iterator
3163 (aarch64_st1x4<vstruct_elt>): This.
3164 (aarch64_st1_x4_<mode>): Use vector-tuple mode iterator and
3166 (aarch64_st1_x4_<vstruct_elt>): This.
3167 (*aarch64_mov<mode>): Define for vector-tuple modes.
3168 (*aarch64_be_mov<mode>): Likewise.
3169 (aarch64_ld<VSTRUCT:nregs>r<VALLDIF:mode>): Use vector-tuple
3170 mode iterator and rename to...
3171 (aarch64_ld<nregs>r<vstruct_elt>): This.
3172 (aarch64_ld2<mode>_dreg): Use vector-tuple mode iterator and
3174 (aarch64_ld2<vstruct_elt>_dreg): This.
3175 (aarch64_ld3<mode>_dreg): Use vector-tuple mode iterator and
3177 (aarch64_ld3<vstruct_elt>_dreg): This.
3178 (aarch64_ld4<mode>_dreg): Use vector-tuple mode iterator and
3180 (aarch64_ld4<vstruct_elt>_dreg): This.
3181 (aarch64_ld<VSTRUCT:nregs><VDC:mode>): Use vector-tuple mode
3182 iterator and rename to...
3183 (aarch64_ld<nregs><vstruct_elt>): Use vector-tuple mode
3184 iterator and rename to...
3185 (aarch64_ld<VSTRUCT:nregs><VQ:mode>): Use vector-tuple mode
3186 (aarch64_ld1x2<VQ:mode>): Delete.
3187 (aarch64_ld1x2<VDC:mode>): Use vector-tuple mode iterator and
3189 (aarch64_ld1x2<vstruct_elt>): This.
3190 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use vector-
3191 tuple mode iterator and rename to...
3192 (aarch64_ld<nregs>_lane<vstruct_elt>): This.
3193 (aarch64_get_dreg<VSTRUCT:mode><VDC:mode>): Delete.
3194 (aarch64_get_qreg<VSTRUCT:mode><VQ:mode>): Likewise.
3195 (aarch64_st2<mode>_dreg): Use vector-tuple mode iterator and
3197 (aarch64_st2<vstruct_elt>_dreg): This.
3198 (aarch64_st3<mode>_dreg): Use vector-tuple mode iterator and
3200 (aarch64_st3<vstruct_elt>_dreg): This.
3201 (aarch64_st4<mode>_dreg): Use vector-tuple mode iterator and
3203 (aarch64_st4<vstruct_elt>_dreg): This.
3204 (aarch64_st<VSTRUCT:nregs><VDC:mode>): Use vector-tuple mode
3205 iterator and rename to...
3206 (aarch64_st<nregs><vstruct_elt>): This.
3207 (aarch64_st<VSTRUCT:nregs><VQ:mode>): Use vector-tuple mode
3208 iterator and rename to aarch64_st<nregs><vstruct_elt>.
3209 (aarch64_st<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use vector-
3210 tuple mode iterator and rename to...
3211 (aarch64_st<nregs>_lane<vstruct_elt>): This.
3212 (aarch64_set_qreg<VSTRUCT:mode><VQ:mode>): Delete.
3213 (aarch64_simd_ld1<mode>_x2): Use vector-tuple mode iterator
3215 (aarch64_simd_ld1<vstruct_elt>_x2): This.
3216 * config/aarch64/aarch64.c (aarch64_advsimd_struct_mode_p):
3217 Refactor to include new vector-tuple modes.
3218 (aarch64_classify_vector_mode): Add cases for new vector-
3220 (aarch64_advsimd_partial_struct_mode_p): Define.
3221 (aarch64_advsimd_full_struct_mode_p): Likewise.
3222 (aarch64_advsimd_vector_array_mode): Likewise.
3223 (aarch64_sve_data_mode): Change location in file.
3224 (aarch64_array_mode): Handle case of Neon vector-tuple modes.
3225 (aarch64_hard_regno_nregs): Handle case of partial Neon
3227 (aarch64_classify_address): Refactor to include handling of
3228 Neon vector-tuple modes.
3229 (aarch64_print_operand): Print "d" for "%R" for a partial
3230 Neon vector structure.
3231 (aarch64_expand_vec_perm_1): Use new vector-tuple mode.
3232 (aarch64_modes_tieable_p): Prevent tieing Neon partial struct
3233 modes with scalar machines modes larger than 8 bytes.
3234 (aarch64_can_change_mode_class): Don't allow changes between
3235 partial and full Neon vector-structure modes.
3236 * config/aarch64/arm_neon.h (vst2_lane_f16): Use updated
3237 builtin and remove boiler-plate code for opaque mode.
3238 (vst2_lane_f32): Likewise.
3239 (vst2_lane_f64): Likewise.
3240 (vst2_lane_p8): Likewise.
3241 (vst2_lane_p16): Likewise.
3242 (vst2_lane_p64): Likewise.
3243 (vst2_lane_s8): Likewise.
3244 (vst2_lane_s16): Likewise.
3245 (vst2_lane_s32): Likewise.
3246 (vst2_lane_s64): Likewise.
3247 (vst2_lane_u8): Likewise.
3248 (vst2_lane_u16): Likewise.
3249 (vst2_lane_u32): Likewise.
3250 (vst2_lane_u64): Likewise.
3251 (vst2q_lane_f16): Likewise.
3252 (vst2q_lane_f32): Likewise.
3253 (vst2q_lane_f64): Likewise.
3254 (vst2q_lane_p8): Likewise.
3255 (vst2q_lane_p16): Likewise.
3256 (vst2q_lane_p64): Likewise.
3257 (vst2q_lane_s8): Likewise.
3258 (vst2q_lane_s16): Likewise.
3259 (vst2q_lane_s32): Likewise.
3260 (vst2q_lane_s64): Likewise.
3261 (vst2q_lane_u8): Likewise.
3262 (vst2q_lane_u16): Likewise.
3263 (vst2q_lane_u32): Likewise.
3264 (vst2q_lane_u64): Likewise.
3265 (vst3_lane_f16): Likewise.
3266 (vst3_lane_f32): Likewise.
3267 (vst3_lane_f64): Likewise.
3268 (vst3_lane_p8): Likewise.
3269 (vst3_lane_p16): Likewise.
3270 (vst3_lane_p64): Likewise.
3271 (vst3_lane_s8): Likewise.
3272 (vst3_lane_s16): Likewise.
3273 (vst3_lane_s32): Likewise.
3274 (vst3_lane_s64): Likewise.
3275 (vst3_lane_u8): Likewise.
3276 (vst3_lane_u16): Likewise.
3277 (vst3_lane_u32): Likewise.
3278 (vst3_lane_u64): Likewise.
3279 (vst3q_lane_f16): Likewise.
3280 (vst3q_lane_f32): Likewise.
3281 (vst3q_lane_f64): Likewise.
3282 (vst3q_lane_p8): Likewise.
3283 (vst3q_lane_p16): Likewise.
3284 (vst3q_lane_p64): Likewise.
3285 (vst3q_lane_s8): Likewise.
3286 (vst3q_lane_s16): Likewise.
3287 (vst3q_lane_s32): Likewise.
3288 (vst3q_lane_s64): Likewise.
3289 (vst3q_lane_u8): Likewise.
3290 (vst3q_lane_u16): Likewise.
3291 (vst3q_lane_u32): Likewise.
3292 (vst3q_lane_u64): Likewise.
3293 (vst4_lane_f16): Likewise.
3294 (vst4_lane_f32): Likewise.
3295 (vst4_lane_f64): Likewise.
3296 (vst4_lane_p8): Likewise.
3297 (vst4_lane_p16): Likewise.
3298 (vst4_lane_p64): Likewise.
3299 (vst4_lane_s8): Likewise.
3300 (vst4_lane_s16): Likewise.
3301 (vst4_lane_s32): Likewise.
3302 (vst4_lane_s64): Likewise.
3303 (vst4_lane_u8): Likewise.
3304 (vst4_lane_u16): Likewise.
3305 (vst4_lane_u32): Likewise.
3306 (vst4_lane_u64): Likewise.
3307 (vst4q_lane_f16): Likewise.
3308 (vst4q_lane_f32): Likewise.
3309 (vst4q_lane_f64): Likewise.
3310 (vst4q_lane_p8): Likewise.
3311 (vst4q_lane_p16): Likewise.
3312 (vst4q_lane_p64): Likewise.
3313 (vst4q_lane_s8): Likewise.
3314 (vst4q_lane_s16): Likewise.
3315 (vst4q_lane_s32): Likewise.
3316 (vst4q_lane_s64): Likewise.
3317 (vst4q_lane_u8): Likewise.
3318 (vst4q_lane_u16): Likewise.
3319 (vst4q_lane_u32): Likewise.
3320 (vst4q_lane_u64): Likewise.
3321 (vtbl3_s8): Likewise.
3322 (vtbl3_u8): Likewise.
3323 (vtbl3_p8): Likewise.
3324 (vtbl4_s8): Likewise.
3325 (vtbl4_u8): Likewise.
3326 (vtbl4_p8): Likewise.
3327 (vld1_u8_x3): Likewise.
3328 (vld1_s8_x3): Likewise.
3329 (vld1_u16_x3): Likewise.
3330 (vld1_s16_x3): Likewise.
3331 (vld1_u32_x3): Likewise.
3332 (vld1_s32_x3): Likewise.
3333 (vld1_u64_x3): Likewise.
3334 (vld1_s64_x3): Likewise.
3335 (vld1_f16_x3): Likewise.
3336 (vld1_f32_x3): Likewise.
3337 (vld1_f64_x3): Likewise.
3338 (vld1_p8_x3): Likewise.
3339 (vld1_p16_x3): Likewise.
3340 (vld1_p64_x3): Likewise.
3341 (vld1q_u8_x3): Likewise.
3342 (vld1q_s8_x3): Likewise.
3343 (vld1q_u16_x3): Likewise.
3344 (vld1q_s16_x3): Likewise.
3345 (vld1q_u32_x3): Likewise.
3346 (vld1q_s32_x3): Likewise.
3347 (vld1q_u64_x3): Likewise.
3348 (vld1q_s64_x3): Likewise.
3349 (vld1q_f16_x3): Likewise.
3350 (vld1q_f32_x3): Likewise.
3351 (vld1q_f64_x3): Likewise.
3352 (vld1q_p8_x3): Likewise.
3353 (vld1q_p16_x3): Likewise.
3354 (vld1q_p64_x3): Likewise.
3355 (vld1_u8_x2): Likewise.
3356 (vld1_s8_x2): Likewise.
3357 (vld1_u16_x2): Likewise.
3358 (vld1_s16_x2): Likewise.
3359 (vld1_u32_x2): Likewise.
3360 (vld1_s32_x2): Likewise.
3361 (vld1_u64_x2): Likewise.
3362 (vld1_s64_x2): Likewise.
3363 (vld1_f16_x2): Likewise.
3364 (vld1_f32_x2): Likewise.
3365 (vld1_f64_x2): Likewise.
3366 (vld1_p8_x2): Likewise.
3367 (vld1_p16_x2): Likewise.
3368 (vld1_p64_x2): Likewise.
3369 (vld1q_u8_x2): Likewise.
3370 (vld1q_s8_x2): Likewise.
3371 (vld1q_u16_x2): Likewise.
3372 (vld1q_s16_x2): Likewise.
3373 (vld1q_u32_x2): Likewise.
3374 (vld1q_s32_x2): Likewise.
3375 (vld1q_u64_x2): Likewise.
3376 (vld1q_s64_x2): Likewise.
3377 (vld1q_f16_x2): Likewise.
3378 (vld1q_f32_x2): Likewise.
3379 (vld1q_f64_x2): Likewise.
3380 (vld1q_p8_x2): Likewise.
3381 (vld1q_p16_x2): Likewise.
3382 (vld1q_p64_x2): Likewise.
3383 (vld1_s8_x4): Likewise.
3384 (vld1q_s8_x4): Likewise.
3385 (vld1_s16_x4): Likewise.
3386 (vld1q_s16_x4): Likewise.
3387 (vld1_s32_x4): Likewise.
3388 (vld1q_s32_x4): Likewise.
3389 (vld1_u8_x4): Likewise.
3390 (vld1q_u8_x4): Likewise.
3391 (vld1_u16_x4): Likewise.
3392 (vld1q_u16_x4): Likewise.
3393 (vld1_u32_x4): Likewise.
3394 (vld1q_u32_x4): Likewise.
3395 (vld1_f16_x4): Likewise.
3396 (vld1q_f16_x4): Likewise.
3397 (vld1_f32_x4): Likewise.
3398 (vld1q_f32_x4): Likewise.
3399 (vld1_p8_x4): Likewise.
3400 (vld1q_p8_x4): Likewise.
3401 (vld1_p16_x4): Likewise.
3402 (vld1q_p16_x4): Likewise.
3403 (vld1_s64_x4): Likewise.
3404 (vld1_u64_x4): Likewise.
3405 (vld1_p64_x4): Likewise.
3406 (vld1q_s64_x4): Likewise.
3407 (vld1q_u64_x4): Likewise.
3408 (vld1q_p64_x4): Likewise.
3409 (vld1_f64_x4): Likewise.
3410 (vld1q_f64_x4): Likewise.
3411 (vld2_s64): Likewise.
3412 (vld2_u64): Likewise.
3413 (vld2_f64): Likewise.
3414 (vld2_s8): Likewise.
3415 (vld2_p8): Likewise.
3416 (vld2_p64): Likewise.
3417 (vld2_s16): Likewise.
3418 (vld2_p16): Likewise.
3419 (vld2_s32): Likewise.
3420 (vld2_u8): Likewise.
3421 (vld2_u16): Likewise.
3422 (vld2_u32): Likewise.
3423 (vld2_f16): Likewise.
3424 (vld2_f32): Likewise.
3425 (vld2q_s8): Likewise.
3426 (vld2q_p8): Likewise.
3427 (vld2q_s16): Likewise.
3428 (vld2q_p16): Likewise.
3429 (vld2q_p64): Likewise.
3430 (vld2q_s32): Likewise.
3431 (vld2q_s64): Likewise.
3432 (vld2q_u8): Likewise.
3433 (vld2q_u16): Likewise.
3434 (vld2q_u32): Likewise.
3435 (vld2q_u64): Likewise.
3436 (vld2q_f16): Likewise.
3437 (vld2q_f32): Likewise.
3438 (vld2q_f64): Likewise.
3439 (vld3_s64): Likewise.
3440 (vld3_u64): Likewise.
3441 (vld3_f64): Likewise.
3442 (vld3_s8): Likewise.
3443 (vld3_p8): Likewise.
3444 (vld3_s16): Likewise.
3445 (vld3_p16): Likewise.
3446 (vld3_s32): Likewise.
3447 (vld3_u8): Likewise.
3448 (vld3_u16): Likewise.
3449 (vld3_u32): Likewise.
3450 (vld3_f16): Likewise.
3451 (vld3_f32): Likewise.
3452 (vld3_p64): Likewise.
3453 (vld3q_s8): Likewise.
3454 (vld3q_p8): Likewise.
3455 (vld3q_s16): Likewise.
3456 (vld3q_p16): Likewise.
3457 (vld3q_s32): Likewise.
3458 (vld3q_s64): Likewise.
3459 (vld3q_u8): Likewise.
3460 (vld3q_u16): Likewise.
3461 (vld3q_u32): Likewise.
3462 (vld3q_u64): Likewise.
3463 (vld3q_f16): Likewise.
3464 (vld3q_f32): Likewise.
3465 (vld3q_f64): Likewise.
3466 (vld3q_p64): Likewise.
3467 (vld4_s64): Likewise.
3468 (vld4_u64): Likewise.
3469 (vld4_f64): Likewise.
3470 (vld4_s8): Likewise.
3471 (vld4_p8): Likewise.
3472 (vld4_s16): Likewise.
3473 (vld4_p16): Likewise.
3474 (vld4_s32): Likewise.
3475 (vld4_u8): Likewise.
3476 (vld4_u16): Likewise.
3477 (vld4_u32): Likewise.
3478 (vld4_f16): Likewise.
3479 (vld4_f32): Likewise.
3480 (vld4_p64): Likewise.
3481 (vld4q_s8): Likewise.
3482 (vld4q_p8): Likewise.
3483 (vld4q_s16): Likewise.
3484 (vld4q_p16): Likewise.
3485 (vld4q_s32): Likewise.
3486 (vld4q_s64): Likewise.
3487 (vld4q_u8): Likewise.
3488 (vld4q_u16): Likewise.
3489 (vld4q_u32): Likewise.
3490 (vld4q_u64): Likewise.
3491 (vld4q_f16): Likewise.
3492 (vld4q_f32): Likewise.
3493 (vld4q_f64): Likewise.
3494 (vld4q_p64): Likewise.
3495 (vld2_dup_s8): Likewise.
3496 (vld2_dup_s16): Likewise.
3497 (vld2_dup_s32): Likewise.
3498 (vld2_dup_f16): Likewise.
3499 (vld2_dup_f32): Likewise.
3500 (vld2_dup_f64): Likewise.
3501 (vld2_dup_u8): Likewise.
3502 (vld2_dup_u16): Likewise.
3503 (vld2_dup_u32): Likewise.
3504 (vld2_dup_p8): Likewise.
3505 (vld2_dup_p16): Likewise.
3506 (vld2_dup_p64): Likewise.
3507 (vld2_dup_s64): Likewise.
3508 (vld2_dup_u64): Likewise.
3509 (vld2q_dup_s8): Likewise.
3510 (vld2q_dup_p8): Likewise.
3511 (vld2q_dup_s16): Likewise.
3512 (vld2q_dup_p16): Likewise.
3513 (vld2q_dup_s32): Likewise.
3514 (vld2q_dup_s64): Likewise.
3515 (vld2q_dup_u8): Likewise.
3516 (vld2q_dup_u16): Likewise.
3517 (vld2q_dup_u32): Likewise.
3518 (vld2q_dup_u64): Likewise.
3519 (vld2q_dup_f16): Likewise.
3520 (vld2q_dup_f32): Likewise.
3521 (vld2q_dup_f64): Likewise.
3522 (vld2q_dup_p64): Likewise.
3523 (vld3_dup_s64): Likewise.
3524 (vld3_dup_u64): Likewise.
3525 (vld3_dup_f64): Likewise.
3526 (vld3_dup_s8): Likewise.
3527 (vld3_dup_p8): Likewise.
3528 (vld3_dup_s16): Likewise.
3529 (vld3_dup_p16): Likewise.
3530 (vld3_dup_s32): Likewise.
3531 (vld3_dup_u8): Likewise.
3532 (vld3_dup_u16): Likewise.
3533 (vld3_dup_u32): Likewise.
3534 (vld3_dup_f16): Likewise.
3535 (vld3_dup_f32): Likewise.
3536 (vld3_dup_p64): Likewise.
3537 (vld3q_dup_s8): Likewise.
3538 (vld3q_dup_p8): Likewise.
3539 (vld3q_dup_s16): Likewise.
3540 (vld3q_dup_p16): Likewise.
3541 (vld3q_dup_s32): Likewise.
3542 (vld3q_dup_s64): Likewise.
3543 (vld3q_dup_u8): Likewise.
3544 (vld3q_dup_u16): Likewise.
3545 (vld3q_dup_u32): Likewise.
3546 (vld3q_dup_u64): Likewise.
3547 (vld3q_dup_f16): Likewise.
3548 (vld3q_dup_f32): Likewise.
3549 (vld3q_dup_f64): Likewise.
3550 (vld3q_dup_p64): Likewise.
3551 (vld4_dup_s64): Likewise.
3552 (vld4_dup_u64): Likewise.
3553 (vld4_dup_f64): Likewise.
3554 (vld4_dup_s8): Likewise.
3555 (vld4_dup_p8): Likewise.
3556 (vld4_dup_s16): Likewise.
3557 (vld4_dup_p16): Likewise.
3558 (vld4_dup_s32): Likewise.
3559 (vld4_dup_u8): Likewise.
3560 (vld4_dup_u16): Likewise.
3561 (vld4_dup_u32): Likewise.
3562 (vld4_dup_f16): Likewise.
3563 (vld4_dup_f32): Likewise.
3564 (vld4_dup_p64): Likewise.
3565 (vld4q_dup_s8): Likewise.
3566 (vld4q_dup_p8): Likewise.
3567 (vld4q_dup_s16): Likewise.
3568 (vld4q_dup_p16): Likewise.
3569 (vld4q_dup_s32): Likewise.
3570 (vld4q_dup_s64): Likewise.
3571 (vld4q_dup_u8): Likewise.
3572 (vld4q_dup_u16): Likewise.
3573 (vld4q_dup_u32): Likewise.
3574 (vld4q_dup_u64): Likewise.
3575 (vld4q_dup_f16): Likewise.
3576 (vld4q_dup_f32): Likewise.
3577 (vld4q_dup_f64): Likewise.
3578 (vld4q_dup_p64): Likewise.
3579 (vld2_lane_u8): Likewise.
3580 (vld2_lane_u16): Likewise.
3581 (vld2_lane_u32): Likewise.
3582 (vld2_lane_u64): Likewise.
3583 (vld2_lane_s8): Likewise.
3584 (vld2_lane_s16): Likewise.
3585 (vld2_lane_s32): Likewise.
3586 (vld2_lane_s64): Likewise.
3587 (vld2_lane_f16): Likewise.
3588 (vld2_lane_f32): Likewise.
3589 (vld2_lane_f64): Likewise.
3590 (vld2_lane_p8): Likewise.
3591 (vld2_lane_p16): Likewise.
3592 (vld2_lane_p64): Likewise.
3593 (vld2q_lane_u8): Likewise.
3594 (vld2q_lane_u16): Likewise.
3595 (vld2q_lane_u32): Likewise.
3596 (vld2q_lane_u64): Likewise.
3597 (vld2q_lane_s8): Likewise.
3598 (vld2q_lane_s16): Likewise.
3599 (vld2q_lane_s32): Likewise.
3600 (vld2q_lane_s64): Likewise.
3601 (vld2q_lane_f16): Likewise.
3602 (vld2q_lane_f32): Likewise.
3603 (vld2q_lane_f64): Likewise.
3604 (vld2q_lane_p8): Likewise.
3605 (vld2q_lane_p16): Likewise.
3606 (vld2q_lane_p64): Likewise.
3607 (vld3_lane_u8): Likewise.
3608 (vld3_lane_u16): Likewise.
3609 (vld3_lane_u32): Likewise.
3610 (vld3_lane_u64): Likewise.
3611 (vld3_lane_s8): Likewise.
3612 (vld3_lane_s16): Likewise.
3613 (vld3_lane_s32): Likewise.
3614 (vld3_lane_s64): Likewise.
3615 (vld3_lane_f16): Likewise.
3616 (vld3_lane_f32): Likewise.
3617 (vld3_lane_f64): Likewise.
3618 (vld3_lane_p8): Likewise.
3619 (vld3_lane_p16): Likewise.
3620 (vld3_lane_p64): Likewise.
3621 (vld3q_lane_u8): Likewise.
3622 (vld3q_lane_u16): Likewise.
3623 (vld3q_lane_u32): Likewise.
3624 (vld3q_lane_u64): Likewise.
3625 (vld3q_lane_s8): Likewise.
3626 (vld3q_lane_s16): Likewise.
3627 (vld3q_lane_s32): Likewise.
3628 (vld3q_lane_s64): Likewise.
3629 (vld3q_lane_f16): Likewise.
3630 (vld3q_lane_f32): Likewise.
3631 (vld3q_lane_f64): Likewise.
3632 (vld3q_lane_p8): Likewise.
3633 (vld3q_lane_p16): Likewise.
3634 (vld3q_lane_p64): Likewise.
3635 (vld4_lane_u8): Likewise.
3636 (vld4_lane_u16): Likewise.
3637 (vld4_lane_u32): Likewise.
3638 (vld4_lane_u64): Likewise.
3639 (vld4_lane_s8): Likewise.
3640 (vld4_lane_s16): Likewise.
3641 (vld4_lane_s32): Likewise.
3642 (vld4_lane_s64): Likewise.
3643 (vld4_lane_f16): Likewise.
3644 (vld4_lane_f32): Likewise.
3645 (vld4_lane_f64): Likewise.
3646 (vld4_lane_p8): Likewise.
3647 (vld4_lane_p16): Likewise.
3648 (vld4_lane_p64): Likewise.
3649 (vld4q_lane_u8): Likewise.
3650 (vld4q_lane_u16): Likewise.
3651 (vld4q_lane_u32): Likewise.
3652 (vld4q_lane_u64): Likewise.
3653 (vld4q_lane_s8): Likewise.
3654 (vld4q_lane_s16): Likewise.
3655 (vld4q_lane_s32): Likewise.
3656 (vld4q_lane_s64): Likewise.
3657 (vld4q_lane_f16): Likewise.
3658 (vld4q_lane_f32): Likewise.
3659 (vld4q_lane_f64): Likewise.
3660 (vld4q_lane_p8): Likewise.
3661 (vld4q_lane_p16): Likewise.
3662 (vld4q_lane_p64): Likewise.
3663 (vqtbl2_s8): Likewise.
3664 (vqtbl2_u8): Likewise.
3665 (vqtbl2_p8): Likewise.
3666 (vqtbl2q_s8): Likewise.
3667 (vqtbl2q_u8): Likewise.
3668 (vqtbl2q_p8): Likewise.
3669 (vqtbl3_s8): Likewise.
3670 (vqtbl3_u8): Likewise.
3671 (vqtbl3_p8): Likewise.
3672 (vqtbl3q_s8): Likewise.
3673 (vqtbl3q_u8): Likewise.
3674 (vqtbl3q_p8): Likewise.
3675 (vqtbl4_s8): Likewise.
3676 (vqtbl4_u8): Likewise.
3677 (vqtbl4_p8): Likewise.
3678 (vqtbl4q_s8): Likewise.
3679 (vqtbl4q_u8): Likewise.
3680 (vqtbl4q_p8): Likewise.
3681 (vqtbx2_s8): Likewise.
3682 (vqtbx2_u8): Likewise.
3683 (vqtbx2_p8): Likewise.
3684 (vqtbx2q_s8): Likewise.
3685 (vqtbx2q_u8): Likewise.
3686 (vqtbx2q_p8): Likewise.
3687 (vqtbx3_s8): Likewise.
3688 (vqtbx3_u8): Likewise.
3689 (vqtbx3_p8): Likewise.
3690 (vqtbx3q_s8): Likewise.
3691 (vqtbx3q_u8): Likewise.
3692 (vqtbx3q_p8): Likewise.
3693 (vqtbx4_s8): Likewise.
3694 (vqtbx4_u8): Likewise.
3695 (vqtbx4_p8): Likewise.
3696 (vqtbx4q_s8): Likewise.
3697 (vqtbx4q_u8): Likewise.
3698 (vqtbx4q_p8): Likewise.
3699 (vst1_s64_x2): Likewise.
3700 (vst1_u64_x2): Likewise.
3701 (vst1_f64_x2): Likewise.
3702 (vst1_s8_x2): Likewise.
3703 (vst1_p8_x2): Likewise.
3704 (vst1_s16_x2): Likewise.
3705 (vst1_p16_x2): Likewise.
3706 (vst1_s32_x2): Likewise.
3707 (vst1_u8_x2): Likewise.
3708 (vst1_u16_x2): Likewise.
3709 (vst1_u32_x2): Likewise.
3710 (vst1_f16_x2): Likewise.
3711 (vst1_f32_x2): Likewise.
3712 (vst1_p64_x2): Likewise.
3713 (vst1q_s8_x2): Likewise.
3714 (vst1q_p8_x2): Likewise.
3715 (vst1q_s16_x2): Likewise.
3716 (vst1q_p16_x2): Likewise.
3717 (vst1q_s32_x2): Likewise.
3718 (vst1q_s64_x2): Likewise.
3719 (vst1q_u8_x2): Likewise.
3720 (vst1q_u16_x2): Likewise.
3721 (vst1q_u32_x2): Likewise.
3722 (vst1q_u64_x2): Likewise.
3723 (vst1q_f16_x2): Likewise.
3724 (vst1q_f32_x2): Likewise.
3725 (vst1q_f64_x2): Likewise.
3726 (vst1q_p64_x2): Likewise.
3727 (vst1_s64_x3): Likewise.
3728 (vst1_u64_x3): Likewise.
3729 (vst1_f64_x3): Likewise.
3730 (vst1_s8_x3): Likewise.
3731 (vst1_p8_x3): Likewise.
3732 (vst1_s16_x3): Likewise.
3733 (vst1_p16_x3): Likewise.
3734 (vst1_s32_x3): Likewise.
3735 (vst1_u8_x3): Likewise.
3736 (vst1_u16_x3): Likewise.
3737 (vst1_u32_x3): Likewise.
3738 (vst1_f16_x3): Likewise.
3739 (vst1_f32_x3): Likewise.
3740 (vst1_p64_x3): Likewise.
3741 (vst1q_s8_x3): Likewise.
3742 (vst1q_p8_x3): Likewise.
3743 (vst1q_s16_x3): Likewise.
3744 (vst1q_p16_x3): Likewise.
3745 (vst1q_s32_x3): Likewise.
3746 (vst1q_s64_x3): Likewise.
3747 (vst1q_u8_x3): Likewise.
3748 (vst1q_u16_x3): Likewise.
3749 (vst1q_u32_x3): Likewise.
3750 (vst1q_u64_x3): Likewise.
3751 (vst1q_f16_x3): Likewise.
3752 (vst1q_f32_x3): Likewise.
3753 (vst1q_f64_x3): Likewise.
3754 (vst1q_p64_x3): Likewise.
3755 (vst1_s8_x4): Likewise.
3756 (vst1q_s8_x4): Likewise.
3757 (vst1_s16_x4): Likewise.
3758 (vst1q_s16_x4): Likewise.
3759 (vst1_s32_x4): Likewise.
3760 (vst1q_s32_x4): Likewise.
3761 (vst1_u8_x4): Likewise.
3762 (vst1q_u8_x4): Likewise.
3763 (vst1_u16_x4): Likewise.
3764 (vst1q_u16_x4): Likewise.
3765 (vst1_u32_x4): Likewise.
3766 (vst1q_u32_x4): Likewise.
3767 (vst1_f16_x4): Likewise.
3768 (vst1q_f16_x4): Likewise.
3769 (vst1_f32_x4): Likewise.
3770 (vst1q_f32_x4): Likewise.
3771 (vst1_p8_x4): Likewise.
3772 (vst1q_p8_x4): Likewise.
3773 (vst1_p16_x4): Likewise.
3774 (vst1q_p16_x4): Likewise.
3775 (vst1_s64_x4): Likewise.
3776 (vst1_u64_x4): Likewise.
3777 (vst1_p64_x4): Likewise.
3778 (vst1q_s64_x4): Likewise.
3779 (vst1q_u64_x4): Likewise.
3780 (vst1q_p64_x4): Likewise.
3781 (vst1_f64_x4): Likewise.
3782 (vst1q_f64_x4): Likewise.
3783 (vst2_s64): Likewise.
3784 (vst2_u64): Likewise.
3785 (vst2_f64): Likewise.
3786 (vst2_s8): Likewise.
3787 (vst2_p8): Likewise.
3788 (vst2_s16): Likewise.
3789 (vst2_p16): Likewise.
3790 (vst2_s32): Likewise.
3791 (vst2_u8): Likewise.
3792 (vst2_u16): Likewise.
3793 (vst2_u32): Likewise.
3794 (vst2_f16): Likewise.
3795 (vst2_f32): Likewise.
3796 (vst2_p64): Likewise.
3797 (vst2q_s8): Likewise.
3798 (vst2q_p8): Likewise.
3799 (vst2q_s16): Likewise.
3800 (vst2q_p16): Likewise.
3801 (vst2q_s32): Likewise.
3802 (vst2q_s64): Likewise.
3803 (vst2q_u8): Likewise.
3804 (vst2q_u16): Likewise.
3805 (vst2q_u32): Likewise.
3806 (vst2q_u64): Likewise.
3807 (vst2q_f16): Likewise.
3808 (vst2q_f32): Likewise.
3809 (vst2q_f64): Likewise.
3810 (vst2q_p64): Likewise.
3811 (vst3_s64): Likewise.
3812 (vst3_u64): Likewise.
3813 (vst3_f64): Likewise.
3814 (vst3_s8): Likewise.
3815 (vst3_p8): Likewise.
3816 (vst3_s16): Likewise.
3817 (vst3_p16): Likewise.
3818 (vst3_s32): Likewise.
3819 (vst3_u8): Likewise.
3820 (vst3_u16): Likewise.
3821 (vst3_u32): Likewise.
3822 (vst3_f16): Likewise.
3823 (vst3_f32): Likewise.
3824 (vst3_p64): Likewise.
3825 (vst3q_s8): Likewise.
3826 (vst3q_p8): Likewise.
3827 (vst3q_s16): Likewise.
3828 (vst3q_p16): Likewise.
3829 (vst3q_s32): Likewise.
3830 (vst3q_s64): Likewise.
3831 (vst3q_u8): Likewise.
3832 (vst3q_u16): Likewise.
3833 (vst3q_u32): Likewise.
3834 (vst3q_u64): Likewise.
3835 (vst3q_f16): Likewise.
3836 (vst3q_f32): Likewise.
3837 (vst3q_f64): Likewise.
3838 (vst3q_p64): Likewise.
3839 (vst4_s64): Likewise.
3840 (vst4_u64): Likewise.
3841 (vst4_f64): Likewise.
3842 (vst4_s8): Likewise.
3843 (vst4_p8): Likewise.
3844 (vst4_s16): Likewise.
3845 (vst4_p16): Likewise.
3846 (vst4_s32): Likewise.
3847 (vst4_u8): Likewise.
3848 (vst4_u16): Likewise.
3849 (vst4_u32): Likewise.
3850 (vst4_f16): Likewise.
3851 (vst4_f32): Likewise.
3852 (vst4_p64): Likewise.
3853 (vst4q_s8): Likewise.
3854 (vst4q_p8): Likewise.
3855 (vst4q_s16): Likewise.
3856 (vst4q_p16): Likewise.
3857 (vst4q_s32): Likewise.
3858 (vst4q_s64): Likewise.
3859 (vst4q_u8): Likewise.
3860 (vst4q_u16): Likewise.
3861 (vst4q_u32): Likewise.
3862 (vst4q_u64): Likewise.
3863 (vst4q_f16): Likewise.
3864 (vst4q_f32): Likewise.
3865 (vst4q_f64): Likewise.
3866 (vst4q_p64): Likewise.
3867 (vtbx4_s8): Likewise.
3868 (vtbx4_u8): Likewise.
3869 (vtbx4_p8): Likewise.
3870 (vld1_bf16_x2): Likewise.
3871 (vld1q_bf16_x2): Likewise.
3872 (vld1_bf16_x3): Likewise.
3873 (vld1q_bf16_x3): Likewise.
3874 (vld1_bf16_x4): Likewise.
3875 (vld1q_bf16_x4): Likewise.
3876 (vld2_bf16): Likewise.
3877 (vld2q_bf16): Likewise.
3878 (vld2_dup_bf16): Likewise.
3879 (vld2q_dup_bf16): Likewise.
3880 (vld3_bf16): Likewise.
3881 (vld3q_bf16): Likewise.
3882 (vld3_dup_bf16): Likewise.
3883 (vld3q_dup_bf16): Likewise.
3884 (vld4_bf16): Likewise.
3885 (vld4q_bf16): Likewise.
3886 (vld4_dup_bf16): Likewise.
3887 (vld4q_dup_bf16): Likewise.
3888 (vst1_bf16_x2): Likewise.
3889 (vst1q_bf16_x2): Likewise.
3890 (vst1_bf16_x3): Likewise.
3891 (vst1q_bf16_x3): Likewise.
3892 (vst1_bf16_x4): Likewise.
3893 (vst1q_bf16_x4): Likewise.
3894 (vst2_bf16): Likewise.
3895 (vst2q_bf16): Likewise.
3896 (vst3_bf16): Likewise.
3897 (vst3q_bf16): Likewise.
3898 (vst4_bf16): Likewise.
3899 (vst4q_bf16): Likewise.
3900 (vld2_lane_bf16): Likewise.
3901 (vld2q_lane_bf16): Likewise.
3902 (vld3_lane_bf16): Likewise.
3903 (vld3q_lane_bf16): Likewise.
3904 (vld4_lane_bf16): Likewise.
3905 (vld4q_lane_bf16): Likewise.
3906 (vst2_lane_bf16): Likewise.
3907 (vst2q_lane_bf16): Likewise.
3908 (vst3_lane_bf16): Likewise.
3909 (vst3q_lane_bf16): Likewise.
3910 (vst4_lane_bf16): Likewise.
3911 (vst4q_lane_bf16): Likewise.
3912 * config/aarch64/geniterators.sh: Modify iterator regex to
3913 match new vector-tuple modes.
3914 * config/aarch64/iterators.md (insn_count): Extend mode
3915 attribute with vector-tuple type information.
3917 (Vendreg): Likewise.
3920 (VSTRUCT_2D): New mode iterator.
3921 (VSTRUCT_2DNX): Likewise.
3922 (VSTRUCT_2DX): Likewise.
3923 (VSTRUCT_2Q): Likewise.
3924 (VSTRUCT_2QD): Likewise.
3925 (VSTRUCT_3D): Likewise.
3926 (VSTRUCT_3DNX): Likewise.
3927 (VSTRUCT_3DX): Likewise.
3928 (VSTRUCT_3Q): Likewise.
3929 (VSTRUCT_3QD): Likewise.
3930 (VSTRUCT_4D): Likewise.
3931 (VSTRUCT_4DNX): Likewise.
3932 (VSTRUCT_4DX): Likewise.
3933 (VSTRUCT_4Q): Likewise.
3934 (VSTRUCT_4QD): Likewise.
3935 (VSTRUCT_D): Likewise.
3936 (VSTRUCT_Q): Likewise.
3937 (VSTRUCT_QD): Likewise.
3938 (VSTRUCT_ELT): New mode attribute.
3939 (vstruct_elt): Likewise.
3940 * genmodes.c (VECTOR_MODE): Add default prefix and order
3942 (VECTOR_MODE_WITH_PREFIX): Define.
3943 (make_vector_mode): Add mode prefix and order parameters.
3945 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
3947 * expmed.c (extract_bit_field_1): Ensure modes are tieable.
3949 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
3951 * expr.c (emit_group_load_1): Remove historic workaround.
3953 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
3955 * config/aarch64/aarch64-builtins.c (aarch64_init_simd_builtins):
3956 Factor out main loop to...
3957 (aarch64_init_simd_builtin_functions): This new function.
3958 (register_tuple_type): Define.
3959 (aarch64_scalar_builtin_type_p): Define.
3960 (handle_arm_neon_h): Define.
3961 * config/aarch64/aarch64-c.c (aarch64_pragma_aarch64): Handle
3962 pragma for arm_neon.h.
3963 * config/aarch64/aarch64-protos.h (aarch64_advsimd_struct_mode_p):
3965 (handle_arm_neon_h): Likewise.
3966 * config/aarch64/aarch64.c (aarch64_advsimd_struct_mode_p):
3967 Remove static modifier.
3968 * config/aarch64/arm_neon.h (target): Remove Neon vector
3969 structure type definitions.
3971 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
3973 PR tree-optimization/102943
3974 * gimple-range-path.cc (path_range_query::range_on_path_entry):
3975 Prefer range_of_expr unless there are no statements in the BB.
3977 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
3979 PR tree-optimization/102943
3980 * tree-ssa-threadbackward.c (back_threader::find_paths_to_names):
3981 Avoid duplicate calculation of paths.
3983 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
3985 PR tree-optimization/102943
3986 * gimple-range-path.cc (path_range_query::compute_phi_relations):
3987 Only compute relations for SSA names in the import list.
3988 (path_range_query::compute_outgoing_relations): Same.
3989 * gimple-range-path.h (path_range_query::import_p): New.
3991 2021-11-04 Richard Biener <rguenther@suse.de>
3993 PR rtl-optimization/103075
3994 * simplify-rtx.c (exact_int_to_float_conversion_p): Return
3995 false for a VOIDmode operand.
3997 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
3999 * config/aarch64/aarch64.c (aarch64_vector_costs): Make member
4000 variables private and add "m_" to their names. Remove is_loop.
4001 (aarch64_record_potential_advsimd_unrolling): Replace with...
4002 (aarch64_vector_costs::record_potential_advsimd_unrolling): ...this.
4003 (aarch64_analyze_loop_vinfo): Replace with...
4004 (aarch64_vector_costs::analyze_loop_vinfo): ...this.
4005 Move initialization of (m_)vec_flags to add_stmt_cost.
4006 (aarch64_analyze_bb_vinfo): Delete.
4007 (aarch64_count_ops): Replace with...
4008 (aarch64_vector_costs::count_ops): ...this.
4009 (aarch64_vector_costs::add_stmt_cost): Set m_vec_flags,
4010 using m_costing_for_scalar to test whether we're costing
4011 scalar or vector code.
4012 (aarch64_adjust_body_cost_sve): Replace with...
4013 (aarch64_vector_costs::adjust_body_cost_sve): ...this.
4014 (aarch64_adjust_body_cost): Replace with...
4015 (aarch64_vector_costs::adjust_body_cost): ...this.
4016 (aarch64_vector_costs::finish_cost): Use m_vinfo instead of is_loop.
4018 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
4020 * target.def (targetm.vectorize.init_cost): Replace with...
4021 (targetm.vectorize.create_costs): ...this.
4022 (targetm.vectorize.add_stmt_cost): Delete.
4023 (targetm.vectorize.finish_cost): Likewise.
4024 (targetm.vectorize.destroy_cost_data): Likewise.
4025 * doc/tm.texi.in (TARGET_VECTORIZE_INIT_COST): Replace with...
4026 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
4027 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
4028 (TARGET_VECTORIZE_FINISH_COST): Likewise.
4029 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
4030 * doc/tm.texi: Regenerate.
4031 * tree-vectorizer.h (vec_info::vec_info): Remove target_cost_data
4033 (vec_info::target_cost_data): Change from a void * to a vector_costs *.
4034 (vector_costs): New class.
4035 (init_cost): Take a vec_info and return a vector_costs.
4036 (dump_stmt_cost): Remove data parameter.
4037 (add_stmt_cost): Replace vinfo and data parameters with a vector_costs.
4038 (add_stmt_costs): Likewise.
4039 (finish_cost): Replace data parameter with a vector_costs.
4040 (destroy_cost_data): Delete.
4041 * tree-vectorizer.c (dump_stmt_cost): Remove data argument and
4043 (vec_info::vec_info): Remove the target_cost_data parameter and
4044 initialize the member variable to null instead.
4045 (vec_info::~vec_info): Delete target_cost_data instead of calling
4047 (vector_costs::add_stmt_cost): New function.
4048 (vector_costs::finish_cost): Likewise.
4049 (vector_costs::record_stmt_cost): Likewise.
4050 (vector_costs::adjust_cost_for_freq): Likewise.
4051 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Update
4052 call to vec_info::vec_info.
4053 (vect_compute_single_scalar_iteration_cost): Update after above
4054 changes to costing interface.
4055 (vect_analyze_loop_operations): Likewise.
4056 (vect_estimate_min_profitable_iters): Likewise.
4057 (vect_analyze_loop_2): Initialize LOOP_VINFO_TARGET_COST_DATA
4058 at the start_over point, where it needs to be recreated after
4059 trying without slp. Update retry code accordingly.
4060 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Update call
4061 to vec_info::vec_info.
4062 (vect_slp_analyze_operation): Update after above changes to costing
4064 (vect_bb_vectorization_profitable_p): Likewise.
4065 * targhooks.h (default_init_cost): Replace with...
4066 (default_vectorize_create_costs): ...this.
4067 (default_add_stmt_cost): Delete.
4068 (default_finish_cost, default_destroy_cost_data): Likewise.
4069 * targhooks.c (default_init_cost): Replace with...
4070 (default_vectorize_create_costs): ...this.
4071 (default_add_stmt_cost): Delete, moving logic to vector_costs instead.
4072 (default_finish_cost, default_destroy_cost_data): Delete.
4073 * config/aarch64/aarch64.c (aarch64_vector_costs): Inherit from
4074 vector_costs. Add a constructor.
4075 (aarch64_init_cost): Replace with...
4076 (aarch64_vectorize_create_costs): ...this.
4077 (aarch64_add_stmt_cost): Replace with...
4078 (aarch64_vector_costs::add_stmt_cost): ...this. Use record_stmt_cost
4079 to adjust the cost for inner loops.
4080 (aarch64_finish_cost): Replace with...
4081 (aarch64_vector_costs::finish_cost): ...this.
4082 (aarch64_destroy_cost_data): Delete.
4083 (TARGET_VECTORIZE_INIT_COST): Replace with...
4084 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
4085 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
4086 (TARGET_VECTORIZE_FINISH_COST): Likewise.
4087 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
4088 * config/i386/i386.c (ix86_vector_costs): New structure.
4089 (ix86_init_cost): Replace with...
4090 (ix86_vectorize_create_costs): ...this.
4091 (ix86_add_stmt_cost): Replace with...
4092 (ix86_vector_costs::add_stmt_cost): ...this. Use adjust_cost_for_freq
4093 to adjust the cost for inner loops.
4094 (ix86_finish_cost, ix86_destroy_cost_data): Delete.
4095 (TARGET_VECTORIZE_INIT_COST): Replace with...
4096 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
4097 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
4098 (TARGET_VECTORIZE_FINISH_COST): Likewise.
4099 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
4100 * config/rs6000/rs6000.c (TARGET_VECTORIZE_INIT_COST): Replace with...
4101 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
4102 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
4103 (TARGET_VECTORIZE_FINISH_COST): Likewise.
4104 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
4105 (rs6000_cost_data): Inherit from vector_costs.
4106 Add a constructor. Drop loop_info, cost and costing_for_scalar
4107 in favor of the corresponding vector_costs member variables.
4108 Add "m_" to the names of the remaining member variables and
4110 (rs6000_density_test): Replace with...
4111 (rs6000_cost_data::density_test): ...this.
4112 (rs6000_init_cost): Replace with...
4113 (rs6000_vectorize_create_costs): ...this.
4114 (rs6000_update_target_cost_per_stmt): Replace with...
4115 (rs6000_cost_data::update_target_cost_per_stmt): ...this.
4116 (rs6000_add_stmt_cost): Replace with...
4117 (rs6000_cost_data::add_stmt_cost): ...this. Use adjust_cost_for_freq
4118 to adjust the cost for inner loops.
4119 (rs6000_adjust_vect_cost_per_loop): Replace with...
4120 (rs6000_cost_data::adjust_vect_cost_per_loop): ...this.
4121 (rs6000_finish_cost): Replace with...
4122 (rs6000_cost_data::finish_cost): ...this. Group loop code
4123 into a single if statement and pass the loop_vinfo down to
4125 (rs6000_destroy_cost_data): Delete.
4127 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
4129 PR tree-optimization/103062
4130 PR tree-optimization/103062
4131 * value-pointer-equiv.cc (ssa_equiv_stack::ssa_equiv_stack):
4132 Increase size of allocation by 1.
4133 (ssa_equiv_stack::push_replacement): Grow as needed.
4134 (ssa_equiv_stack::get_replacement): Same.
4135 (pointer_equiv_analyzer::pointer_equiv_analyzer): Same.
4136 (pointer_equiv_analyzer::~pointer_equiv_analyzer): Remove delete.
4137 (pointer_equiv_analyzer::set_global_equiv): Grow as needed.
4138 (pointer_equiv_analyzer::get_equiv): Same.
4139 (pointer_equiv_analyzer::get_equiv_expr): Remove const.
4140 * value-pointer-equiv.h (class pointer_equiv_analyzer): Remove
4141 const markers. Use auto_vec instead of tree *.
4143 2021-11-04 Richard Biener <rguenther@suse.de>
4145 * tree-ssa-sccvn.c (vn_nary_op_insert_into): Remove always
4146 true parameter and inline valueization.
4147 (vn_nary_op_lookup_1): Inline valueization from ...
4148 (vn_nary_op_compute_hash): ... here and remove it here.
4149 * tree-ssa-pre.c (phi_translate_1): Do not valueize
4150 before vn_nary_lookup_pieces.
4151 (get_representative_for): Mark created SSA representatives
4154 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
4156 * simplify-rtx.c (simplify_context::simplify_gen_vec_select): Assert
4157 that the operand has a vector mode. Use subreg_lowpart_offset
4158 to test whether an index corresponds to the low part.
4160 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
4162 * read-rtl.c: Remove dead !GENERATOR_FILE block.
4163 * read-rtl-function.c (function_reader::consolidate_singletons):
4164 Generate canonical CONST_VECTORs.
4166 2021-11-04 liuhongt <hongtao.liu@intel.com>
4169 * config/i386/predicates.md (reg_or_notreg_operand): Rename to ..
4170 (regmem_or_bitnot_regmem_operand): .. and extend to handle
4172 * config/i386/sse.md (*<avx512>_vpternlog<mode>_1): Force_reg
4173 the operands which are required to be register_operand.
4174 (*<avx512>_vpternlog<mode>_2): Ditto.
4175 (*<avx512>_vpternlog<mode>_3): Ditto.
4176 (*<avx512>_vternlog<mode>_all): Disallow embeded broadcast for
4177 vector HFmodes since it's not a real AVX512FP16 instruction.
4179 2021-11-04 liuhongt <hongtao.liu@intel.com>
4182 * match.pd: simplify (trunc)copysign((extend)a, (extend)b) to
4183 .COPYSIGN (a,b) when a and b are same type as the truncation
4184 type and has less precision than extend type.
4186 2021-11-04 Richard Biener <rguenther@suse.de>
4188 * doc/generic.texi: Update TARGET_MEM_REF and MEM_REF
4191 2021-11-04 Hongyu Wang <hongyu.wang@intel.com>
4193 * config/i386/sse.md (VI2_AVX512VNNIBW): New mode iterator.
4194 (VI1_AVX512VNNI): Likewise.
4195 (SDOT_VPDP_SUF): New mode_attr.
4198 (sdot_prod<mode>): Use VI2_AVX512F iterator, expand to
4199 vpdpwssd when VNNI targets available.
4200 (usdot_prod<mode>): New expander for vector QImode.
4202 2021-11-04 Hongyu Wang <hongyu.wang@intel.com>
4204 * config/i386/amxtileintrin.h (_tile_loadd_internal): Add
4205 parentheses to base and stride.
4206 (_tile_stream_loadd_internal): Likewise.
4207 (_tile_stored_internal): Likewise.
4209 2021-11-03 Maciej W. Rozycki <macro@embecosm.com>
4211 * config/riscv/riscv.c (riscv_class_max_nregs): Swap the
4212 arguments to `reg_class_subset_p'.
4214 2021-11-03 Joseph Myers <joseph@codesourcery.com>
4217 * fold-const.c (fold_init): New function.
4218 * fold-const.h (fold_init): New prototype.
4220 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
4222 * params.opt (param_vrp2_mode): Make ranger the default for VRP2.
4224 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
4226 * gimple-range-fold.cc (fold_using_range::range_of_phi): Don't import
4227 a range from edge if arg == phidef.
4229 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
4231 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Test
4232 for constant before any other processing.
4234 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
4236 * flag-types.h (RANGER_DEBUG_ALL): Fix values.
4238 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
4240 * gimple-range.cc (gimple_ranger::gimple_ranger): Initialize current_bb.
4241 (gimple_ranger::range_of_expr): Pick up range_on_entry when there is
4242 no explcit context and current_bb is set.
4243 (gimple_ranger::fold_stmt): New.
4244 * gimple-range.h (current_bb, fold_stmt): New.
4245 * tree-vrp.c (rvrp_folder::fold_stmt): Call ranger's fold_stmt.
4247 2021-11-03 Richard Biener <rguenther@suse.de>
4249 PR tree-optimization/102970
4250 * tree-ssa-pre.c (phi_translate_1): Drop clique and base
4251 when translating a MEM_REF over a backedge.
4253 2021-11-03 Philipp Tomsich <philipp.tomsich@vrull.eu>
4255 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New Ampere-1 core.
4256 * config/aarch64/aarch64-tune.md: Regenerate.
4257 * config/aarch64/aarch64-cost-tables.h: Add extra costs for Ampere-1.
4258 * config/aarch64/aarch64.c: Add tuning structures for Ampere-1.
4259 * doc/invoke.texi: Add documentation for Ampere-1 core.
4261 2021-11-03 Wilco Dijkstra <wdijkstr@arm.com>
4263 * config/aarch64/aarch64.md (movsi): Add alternative for GOT accesses.
4265 (ldr_got_small_<mode>): Remove pattern.
4266 (ldr_got_small_sidi): Likewise.
4267 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately): Keep
4268 GOT accesses as moves.
4269 (aarch64_print_operand): Correctly print got_lo12 in L specifier.
4270 (aarch64_mov_operand_p): Make GOT accesses valid move operands.
4271 * config/aarch64/constraints.md: Add new constraint Usw for GOT access.
4273 2021-11-03 Martin Liska <mliska@suse.cz>
4275 * gcov.c (read_line): Remove dead variable.
4277 2021-11-03 Martin Liska <mliska@suse.cz>
4280 * ipa-fnsummary.c (edge_predicate_pool): Rename predicate class to ipa_predicate.
4281 (ipa_fn_summary::account_size_time): Likewise.
4282 (edge_set_predicate): Likewise.
4283 (set_hint_predicate): Likewise.
4284 (add_freqcounting_predicate): Likewise.
4285 (evaluate_conditions_for_known_args): Likewise.
4286 (evaluate_properties_for_edge): Likewise.
4287 (remap_freqcounting_preds_after_dup): Likewise.
4288 (ipa_fn_summary_t::duplicate): Likewise.
4289 (set_cond_stmt_execution_predicate): Likewise.
4290 (set_switch_stmt_execution_predicate): Likewise.
4291 (compute_bb_predicates): Likewise.
4292 (will_be_nonconstant_expr_predicate): Likewise.
4293 (will_be_nonconstant_predicate): Likewise.
4294 (phi_result_unknown_predicate): Likewise.
4295 (predicate_for_phi_result): Likewise.
4296 (analyze_function_body): Likewise.
4297 (compute_fn_summary): Likewise.
4298 (summarize_calls_size_and_time): Likewise.
4299 (estimate_calls_size_and_time): Likewise.
4300 (ipa_call_context::estimate_size_and_time): Likewise.
4301 (remap_edge_summaries): Likewise.
4302 (remap_freqcounting_predicate): Likewise.
4303 (ipa_merge_fn_summary_after_inlining): Likewise.
4304 (ipa_update_overall_fn_summary): Likewise.
4305 (read_ipa_call_summary): Likewise.
4306 (inline_read_section): Likewise.
4307 * ipa-fnsummary.h (struct ipa_freqcounting_predicate): Likewise.
4308 * ipa-predicate.c (predicate::add_clause): Likewise.
4309 (ipa_predicate::add_clause): Likewise.
4310 (predicate::or_with): Likewise.
4311 (ipa_predicate::or_with): Likewise.
4312 (predicate::evaluate): Likewise.
4313 (ipa_predicate::evaluate): Likewise.
4314 (predicate::probability): Likewise.
4315 (ipa_predicate::probability): Likewise.
4316 (dump_condition): Likewise.
4317 (dump_clause): Likewise.
4318 (predicate::dump): Likewise.
4319 (ipa_predicate::dump): Likewise.
4320 (predicate::debug): Likewise.
4321 (ipa_predicate::debug): Likewise.
4322 (predicate::remap_after_duplication): Likewise.
4323 (ipa_predicate::remap_after_duplication): Likewise.
4324 (predicate::remap_after_inlining): Likewise.
4325 (ipa_predicate::remap_after_inlining): Likewise.
4326 (predicate::stream_in): Likewise.
4327 (ipa_predicate::stream_in): Likewise.
4328 (predicate::stream_out): Likewise.
4329 (ipa_predicate::stream_out): Likewise.
4330 (add_condition): Likewise.
4331 * ipa-predicate.h (class predicate): Likewise.
4332 (class ipa_predicate): Likewise.
4333 (add_condition): Likewise.
4335 2021-11-03 Richard Biener <rguenther@suse.de>
4337 * bitmap.h (bitmap_bit_p): Change the return type to bool.
4338 * bitmap.c (bitmap_bit_p): Likewise.
4339 * sbitmap.h (bitmap_bit_p): Likewise.
4340 (bitmap_set_bit): Return whether the bit changed.
4341 (bitmap_clear_bit): Likewise.
4342 * tree-ssa.c (verify_vssa): Make use of the changed state
4343 from bitmap_set_bit.
4345 2021-11-03 Richard Biener <rguenther@suse.de>
4347 PR middle-end/103033
4348 * internal-fn.c (expand_DEFERRED_INIT): Elide the
4349 native_interpret_expr path in favor of folding the
4350 VIEW_CONVERT_EXPR generated when punning the RHS.
4352 2021-11-03 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
4354 * config/s390/s390.c (s390_loop_unroll_adjust): In case of early
4357 2021-11-03 Jan Hubicka <hubicka@ucw.cz>
4360 * ipa-modref.c (callee_to_caller_flags): New function.
4361 (modref_eaf_analysis::analyze_ssa_name): Use it.
4362 (ipa_merge_modref_summary_after_inlining): Fix whitespace.
4364 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
4366 * ipa-modref.c (modref_eaf_analysis::analyze_ssa_name): Revert
4369 2021-11-02 Roger Sayle <roger@nextmovesoftware.com>
4370 Uroš Bizjak <ubizjak@gmail.com>
4372 * config/i386/i386.md (<any_rotate>ti3): Provide expansion for
4373 rotations by non-constant amounts.
4375 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
4377 * ipa-modref.c: Fix anonymous namespace placement.
4378 (class modref_eaf_analysis): New class.
4379 (analyze_ssa_name_flags): Turn to ...
4380 (modref_eaf_analysis::analyze_ssa_name): ... this one.
4381 (merge_call_lhs_flags): Turn to ...
4382 (modref_eaf_analysis::merge_call_lhs_flags): .. this one
4383 (modref_eaf_analysis::merge_with_ssa_name): New member function.
4384 (record_escape_points): Turn to ...
4385 (modref_eaf_analysis::record_escape_points): ... this one.
4386 (analyze_parms): Updat
4387 (ipa_merge_modref_summary_after_inlining): Move to the end of file.
4389 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
4391 * gimple.c (gimple_call_static_chain_flags): New function.
4392 * gimple.h (gimple_call_static_chain_flags): Declare
4393 * ipa-modref.c (modref_summary::modref_summary): Initialize
4395 (modref_summary_lto::modref_summary_lto): Likewise.
4396 (modref_summary::useful_p): Test static_chain_flags.
4397 (modref_summary_lto::useful_p): Likewise.
4398 (struct modref_summary_lto): Add static_chain_flags.
4399 (modref_summary::dump): Dump static_chain_flags.
4400 (modref_summary_lto::dump): Likewise.
4401 (struct escape_point): Add static_cahin_arg.
4402 (analyze_ssa_name_flags): Use gimple_call_static_chain_flags.
4403 (analyze_parms): Handle static chains.
4404 (modref_summaries::duplicate): Duplicate static_chain_flags.
4405 (modref_summaries_lto::duplicate): Likewise.
4406 (modref_write): Stream static_chain_flags.
4407 (read_section): Likewise.
4408 (modref_merge_call_site_flags): Handle static_chain_flags.
4409 * ipa-modref.h (struct modref_summary): Add static_chain_flags.
4410 * tree-ssa-structalias.c (handle_rhs_call): Use
4411 gimple_static_chain_flags.
4413 2021-11-02 Richard Biener <rguenther@suse.de>
4415 PR tree-optimization/103029
4416 * tree-vect-loop-manip.c (vect_loop_versioning): Ensure
4417 the PHI nodes in the loop maintain their original operand
4420 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
4422 * tree-core.h (EAF_NOT_RETURNED_DIRECTLY): New flag.
4423 (EAF_NOREAD): Renumber.
4424 * ipa-modref.c (dump_eaf_flags): Dump EAF_NOT_RETURNED_DIRECTLY.
4425 (remove_useless_eaf_flags): Handle EAF_NOT_RETURNED_DIRECTLY
4426 (deref_flags): Likewise.
4427 (modref_lattice::init): Likewise.
4428 (modref_lattice::merge): Likewise.
4429 (merge_call_lhs_flags): Likewise.
4430 (analyze_ssa_name_flags): Likewise.
4431 (modref_merge_call_site_flags): Likewise.
4432 * tree-ssa-structalias.c (handle_call_arg): Likewise.
4434 2021-11-02 Maciej W. Rozycki <macro@embecosm.com>
4436 * config/riscv/riscv.c (riscv_rtx_costs): Correct a CONST_INT_P
4437 check and remove an unused local variable with shNadd/shNadd.uw
4440 2021-11-02 Tamar Christina <tamar.christina@arm.com>
4442 PR tree-optimization/103007
4443 * tree-vect-slp-patterns.c (complex_fms_pattern::matches): Add elem
4446 2021-11-02 Richard Biener <rguenther@suse.de>
4448 PR middle-end/103038
4449 * fold-const.c (native_interpret_expr): Handle OFFSET_TYPE.
4450 (can_native_interpret_type_p): Likewise.
4451 * internal-fn.c (expand_DEFERRED_INIT): View-convert the
4452 RHS if the LHS is an SSA name.
4454 2021-11-02 Richard Sandiford <richard.sandiford@arm.com>
4456 * langhooks.h (lang_hooks_for_types::simulate_record_decl): New hook.
4457 * langhooks-def.h (lhd_simulate_record_decl): Declare.
4458 (LANG_HOOKS_SIMULATE_RECORD_DECL): Define.
4459 (LANG_HOOKS_FOR_TYPES_INITIALIZER): Include it.
4460 * langhooks.c (lhd_simulate_record_decl): New function.
4462 2021-11-02 Jakub Jelinek <jakub@redhat.com>
4465 * config/i386/i386.c (ix86_vector_mode_supported_p): Reject vector
4466 modes with TImode inner mode if 32-bit.
4468 2021-11-02 liuhongt <hongtao.liu@intel.com>
4470 * doc/sourcebuild.texi (vect_slp_v4qi_store_unalign,
4471 vect_slp_v2hi_store_unalign, vect_slp_v4hi_store_unalign,
4472 vect_slp_v4si_store_unalign): Document efficient target.
4473 (vect_slp_v4qi_store_unalign_1, vect_slp_v8qi_store_unalign_1,
4474 vect_slp_v16qi_store_unalign_1): Ditto.
4475 (vect_slp_v2hi_store_align,vect_slp_v2qi_store_align,
4476 vect_slp_v2si_store_align, vect_slp_v4qi_store_align): Ditto.
4477 (struct_4char_block_move, struct_8char_block_move,
4478 struct_16char_block_move): Ditto.
4480 2021-11-02 Roger Sayle <roger@nextmovesoftware.com>
4481 Jakub Jelinek <jakub@redhat.com>
4484 * config/i386/i386-expand.c (ix86_expand_v1ti_to_ti,
4485 ix86_expand_ti_to_v1ti): New helper functions.
4486 (ix86_expand_v1ti_shift): Check if the amount operand is an
4487 integer constant, and expand as a TImode shift if it isn't.
4488 (ix86_expand_v1ti_rotate): Check if the amount operand is an
4489 integer constant, and expand as a TImode rotate if it isn't.
4490 (ix86_expand_v1ti_ashiftrt): New function to expand arithmetic
4491 right shifts of V1TImode quantities.
4492 * config/i386/i386-protos.h (ix86_expand_v1ti_ashift): Prototype.
4493 * config/i386/sse.md (ashlv1ti3, lshrv1ti3): Change constraints
4494 to QImode general_operand, and let the helper functions lower
4495 shifts by non-constant operands, as TImode shifts. Make
4496 conditional on TARGET_64BIT.
4497 (ashrv1ti3): New expander calling ix86_expand_v1ti_ashiftrt.
4498 (rotlv1ti3, rotrv1ti3): Change shift operand to QImode.
4499 Make conditional on TARGET_64BIT.
4501 2021-11-02 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
4503 * config/s390/s390.md ("*cc_to_int", "tabort", "*tabort_1",
4504 "*tabort_1_plus"): Remove operands non-null check.
4506 2021-11-01 Martin Liska <mliska@suse.cz>
4508 * opt-functions.awk: Add new sanity checking.
4509 * optc-gen.awk: Add new argument to integer_range_info.
4510 * params.opt: Update 2 params which have negative IntegerRange.
4512 2021-11-01 qing zhao <qing.zhao@oracle.com>
4514 * gimplify.c (gimplify_decl_expr): Do not add call to
4515 __builtin_clear_padding when a variable is a gimple register
4516 or it might not have padding.
4517 (gimplify_init_constructor): Likewise.
4519 2021-11-01 Tamar Christina <tamar.christina@arm.com>
4521 * config/arm/aarch-common-protos.h (struct vector_cost_table): Add
4522 movi, dup and extract costing fields.
4523 * config/aarch64/aarch64-cost-tables.h (qdf24xx_extra_costs,
4524 thunderx_extra_costs, thunderx2t99_extra_costs,
4525 thunderx3t110_extra_costs, tsv110_extra_costs, a64fx_extra_costs): Use
4527 * config/arm/aarch-cost-tables.h (generic_extra_costs,
4528 cortexa53_extra_costs, cortexa57_extra_costs, cortexa76_extra_costs,
4529 exynosm1_extra_costs, xgene1_extra_costs): Likewise
4530 * config/aarch64/aarch64-simd.md (aarch64_simd_dup<mode>): Add r->w dup.
4531 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add extra costs.
4533 2021-11-01 Tamar Christina <tamar.christina@arm.com>
4535 * cse.c (add_to_set): New.
4536 (find_sets_in_insn): Register constants in sets.
4537 (canonicalize_insn): Use auto_vec instead.
4538 (cse_insn): Try materializing using vec_dup.
4539 * rtl.h (simplify_context::simplify_gen_vec_select,
4540 simplify_gen_vec_select): New.
4541 * simplify-rtx.c (simplify_context::simplify_gen_vec_select): New.
4543 2021-11-01 David Malcolm <dmalcolm@redhat.com>
4545 * common.opt (fdiagnostics-escape-format=): New.
4546 (diagnostics_escape_format): New enum.
4547 (DIAGNOSTICS_ESCAPE_FORMAT_UNICODE): New enum value.
4548 (DIAGNOSTICS_ESCAPE_FORMAT_BYTES): Likewise.
4549 * diagnostic-format-json.cc (json_end_diagnostic): Add
4550 "escape-source" attribute.
4551 * diagnostic-show-locus.c
4552 (exploc_with_display_col::exploc_with_display_col): Replace
4553 "tabstop" param with a cpp_char_column_policy and add an "aspect"
4554 param. Use these to compute m_display_col accordingly.
4555 (struct char_display_policy): New struct.
4556 (layout::m_policy): New field.
4557 (layout::m_escape_on_output): New field.
4558 (def_policy): New function.
4559 (make_range): Update for changes to exploc_with_display_col ctor.
4560 (default_print_decoded_ch): New.
4561 (width_per_escaped_byte): New.
4562 (escape_as_bytes_width): New.
4563 (escape_as_bytes_print): New.
4564 (escape_as_unicode_width): New.
4565 (escape_as_unicode_print): New.
4567 (layout::layout): Initialize new fields. Update m_exploc ctor
4568 call for above change to ctor.
4569 (layout::maybe_add_location_range): Update for changes to
4570 exploc_with_display_col ctor.
4571 (layout::calculate_x_offset_display): Update for change to
4573 (layout::print_source_line): Pass policy
4574 to cpp_display_width_computation. Capture cpp_decoded_char when
4575 calling process_next_codepoint. Move printing of source code to
4576 m_policy.m_print_cb.
4577 (line_label::line_label): Pass in policy rather than context.
4578 (layout::print_any_labels): Update for change to line_label ctor.
4579 (get_affected_range): Pass in policy rather than context, updating
4580 calls to location_compute_display_column accordingly.
4581 (get_printed_columns): Likewise, also for cpp_display_width.
4582 (correction::correction): Pass in policy rather than tabstop.
4583 (correction::compute_display_cols): Pass m_policy rather than
4584 m_tabstop to cpp_display_width.
4585 (correction::m_tabstop): Replace with...
4586 (correction::m_policy): ...this.
4587 (line_corrections::line_corrections): Pass in policy rather than
4589 (line_corrections::m_context): Replace with...
4590 (line_corrections::m_policy): ...this.
4591 (line_corrections::add_hint): Update to use m_policy rather than
4593 (line_corrections::add_hint): Likewise.
4594 (layout::print_trailing_fixits): Likewise.
4595 (selftest::test_display_widths): New.
4596 (selftest::test_layout_x_offset_display_utf8): Update to use
4597 policy rather than tabstop.
4598 (selftest::test_one_liner_labels_utf8): Add test of escaping
4600 (selftest::test_diagnostic_show_locus_one_liner_utf8): Update to
4601 use policy rather than tabstop.
4602 (selftest::test_overlapped_fixit_printing): Likewise.
4603 (selftest::test_overlapped_fixit_printing_utf8): Likewise.
4604 (selftest::test_overlapped_fixit_printing_2): Likewise.
4605 (selftest::test_tab_expansion): Likewise.
4606 (selftest::test_escaping_bytes_1): New.
4607 (selftest::test_escaping_bytes_2): New.
4608 (selftest::diagnostic_show_locus_c_tests): Call the new tests.
4609 * diagnostic.c (diagnostic_initialize): Initialize
4610 context->escape_format.
4611 (convert_column_unit): Update to use default character width policy.
4612 (selftest::test_diagnostic_get_location_text): Likewise.
4613 * diagnostic.h (enum diagnostics_escape_format): New enum.
4614 (diagnostic_context::escape_format): New field.
4615 * doc/invoke.texi (-fdiagnostics-escape-format=): New option.
4616 (-fdiagnostics-format=): Add "escape-source" attribute to examples
4617 of JSON output, and document it.
4618 * input.c (location_compute_display_column): Pass in "policy"
4619 rather than "tabstop", passing to
4620 cpp_byte_column_to_display_column.
4621 (selftest::test_cpp_utf8): Update to use cpp_char_column_policy.
4622 * input.h (class cpp_char_column_policy): New forward decl.
4623 (location_compute_display_column): Pass in "policy" rather than
4625 * opts.c (common_handle_option): Handle
4626 OPT_fdiagnostics_escape_format_.
4627 * selftest.c (temp_source_file::temp_source_file): New ctor
4628 overload taking a size_t.
4629 * selftest.h (temp_source_file::temp_source_file): Likewise.
4631 2021-11-01 Aldy Hernandez <aldyh@redhat.com>
4633 * dbgcnt.def: Add debug counter for back_thread[12] and
4634 back_threadfull[12].
4635 * passes.def: Pass "first" argument to each back threading pass.
4636 * tree-ssa-threadbackward.c (back_threader::back_threader): Add
4638 (back_threader::debug_counter): New.
4639 (back_threader::maybe_register_path): Call debug_counter.
4641 2021-11-01 Aldy Hernandez <aldyh@redhat.com>
4643 * tree-ssa-threadbackward.c (BT_NONE): New.
4646 (back_threader::back_threader): Add flags.
4647 Move loop initialization here.
4648 (back_threader::~back_threader): New.
4649 (back_threader::find_taken_edge_switch): Change solver and ranger
4651 (back_threader::find_taken_edge_cond): Same.
4652 (back_threader::find_paths_to_names): Same.
4653 (back_threader::find_paths): Same.
4654 (back_threader::dump): Same.
4655 (try_thread_blocks): Merge into thread_blocks.
4656 (back_threader::thread_blocks): New.
4657 (do_early_thread_jumps): Merge into thread_blocks.
4658 (do_thread_jumps): Merge into thread_blocks.
4659 (back_threader::thread_through_all_blocks): Remove.
4661 2021-11-01 Andrew MacLeod <amacleod@redhat.com>
4663 PR tree-optimization/103003
4664 * value-relation.cc (dom_oracle::register_relation): If the 2
4665 ssa names are the same, don't register any relation.
4667 2021-11-01 Dan Li <ashimida@linux.alibaba.com>
4669 * config/aarch64/aarch64.c (aarch64_expand_epilogue): Remove
4670 redundant check for calls_eh_return.
4671 * config/aarch64/aarch64.md (*do_return): Likewise.
4673 2021-11-01 Xionghu Luo <luoxhu@linux.ibm.com>
4675 * cfghooks.c (cfg_hook_duplicate_loop_to_header_edge): Rename
4676 duplicate_loop_to_header_edge to
4677 duplicate_loop_body_to_header_edge.
4678 (cfg_hook_duplicate_loop_body_to_header_edge): Likewise.
4679 * cfghooks.h (struct cfg_hooks): Likewise.
4680 (cfg_hook_duplicate_loop_body_to_header_edge): Likewise.
4681 * cfgloopmanip.c (duplicate_loop_body_to_header_edge): Likewise.
4682 (clone_loop_to_header_edge): Likewise.
4683 * cfgloopmanip.h (duplicate_loop_body_to_header_edge): Likewise.
4684 * cfgrtl.c (struct cfg_hooks): Likewise.
4685 * doc/loop.texi: Likewise.
4686 * loop-unroll.c (unroll_loop_constant_iterations): Likewise.
4687 (unroll_loop_runtime_iterations): Likewise.
4688 (unroll_loop_stupid): Likewise.
4689 (apply_opt_in_copies): Likewise.
4690 * tree-cfg.c (struct cfg_hooks): Likewise.
4691 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Likewise.
4692 (try_peel_loop): Likewise.
4693 * tree-ssa-loop-manip.c (copy_phi_node_args): Likewise.
4694 (gimple_duplicate_loop_body_to_header_edge): Likewise.
4695 (tree_transform_and_unroll_loop): Likewise.
4696 * tree-ssa-loop-manip.h (gimple_duplicate_loop_body_to_header_edge):
4699 2021-11-01 Xionghu Luo <luoxhu@linux.ibm.com>
4701 * cfgloopmanip.c (loop_version): Refactor loopify to
4702 loop_version. Move condition generation after loopify.
4704 * cfgloopmanip.h (loopify): Delete.
4706 2021-10-31 Jan Hubicka <hubicka@ucw.cz>
4708 * ipa-fnsummary.c: Include tree-dfa.h.
4709 (points_to_local_or_readonly_memory_p): Return true on return
4711 * ipa-modref.c (analyze_ssa_name_flags): Fix handling of copy
4714 2021-10-30 Tobias Burnus <tobias@codesourcery.com>
4716 PR middle-end/102972
4717 * omp-low.c (omp_runtime_api_call): Use DECL_ASSEMBLER_NAME to get
4718 internal Fortran name; new permit_num_teams arg to permit
4719 omp_get_num_teams and omp_get_team_num.
4720 (scan_omp_1_stmt): Update call to it, add missing call for
4721 reverse offload, and check for strictly nested API calls in teams.
4723 2021-10-30 Jakub Jelinek <jakub@redhat.com>
4725 * gimplify.c (gimplify_omp_for): Diagnose threadprivate iterators.
4727 2021-10-29 David Malcolm <dmalcolm@redhat.com>
4729 * selftest.c (assert_streq): Add newlines when emitting non-equal
4732 2021-10-29 David Malcolm <dmalcolm@redhat.com>
4734 * Makefile.in: Fix syntax for reference to LIBDEPS in
4737 2021-10-29 Bernhard Reutner-Fischer <aldot@gcc.gnu.org>
4739 * doc/install.texi: Bump required minimum DejaGnu version.
4741 2021-10-29 Aldy Hernandez <aldyh@redhat.com>
4742 Andrew MacLeod <amacleod@redhat.com>
4744 * value-relation.cc (path_oracle::killing_def): Add a
4745 self-equivalence so we don't look to the root oracle.
4747 2021-10-29 Aldy Hernandez <aldyh@redhat.com>
4749 * passes.def: Replace the pass_thread_jumps before VRP* with
4750 pass_thread_jumps_full. Remove all pass_vrp_threader instances.
4751 * tree-ssa-threadbackward.c (pass_data_thread_jumps_full):
4752 Remove hyphen from "thread-full" name.
4754 2021-10-29 Andrew MacLeod <amacleod@redhat.com>
4756 PR tree-optimization/102983
4757 * gimple-range-cache.h (propagate_updated_value): Make public.
4758 * gimple-range.cc (gimple_ranger::range_of_stmt): Propagate exports
4759 when processing gcond stmts.
4761 2021-10-29 Jan Hubicka <hubicka@ucw.cz>
4763 * gimple.c (gimple_call_retslot_flags): New function.
4764 * gimple.h (gimple_call_retslot_flags): Declare.
4765 * ipa-modref.c: Include tree-cfg.h.
4766 (struct escape_entry): Turn parm_index to signed.
4767 (modref_summary_lto::modref_summary_lto): Add retslot_flags.
4768 (modref_summary::modref_summary): Initialize retslot_flags.
4769 (struct modref_summary_lto): Likewise.
4770 (modref_summary::useful_p): Check retslot_flags.
4771 (modref_summary_lto::useful_p): Likewise.
4772 (modref_summary::dump): Dump retslot_flags.
4773 (modref_summary_lto::dump): Likewise.
4774 (struct escape_point): Add hidden_args enum.
4775 (analyze_ssa_name_flags): Ignore return slot return;
4776 use gimple_call_retslot_flags.
4777 (record_escape_points): Break out from ...
4778 (analyze_parms): ... here; handle retslot_flags.
4779 (modref_summaries::duplicate): Duplicate retslot_flags.
4780 (modref_summaries_lto::duplicate): Likewise.
4781 (modref_write_escape_summary): Stream parm_index as signed.
4782 (modref_read_escape_summary): Likewise.
4783 (modref_write): Stream retslot_flags.
4784 (read_section): Likewise.
4785 (struct escape_map): Fix typo in comment.
4786 (update_escape_summary_1): Fix whitespace.
4787 (ipa_merge_modref_summary_after_inlining): Drop retslot_flags.
4788 (modref_merge_call_site_flags): Merge retslot_flags.
4789 * ipa-modref.h (struct modref_summary): Add retslot_flags.
4790 * tree-ssa-structalias.c (handle_rhs_call): Handle retslot_flags.
4792 2021-10-29 Tamar Christina <tamar.christina@arm.com>
4794 PR tree-optimization/102977
4795 * tree-vect-slp-patterns.c (vect_match_call_p): Remove.
4796 (vect_detect_pair_op): Add crosslane check.
4797 (vect_match_call_complex_mla): Remove.
4798 (class complex_mul_pattern): Update comment.
4799 (complex_mul_pattern::matches): Update detection.
4800 (class complex_fma_pattern): Remove.
4801 (complex_fma_pattern::matches): Remove.
4802 (complex_fma_pattern::recognize): Remove.
4803 (complex_fma_pattern::build): Remove.
4804 (class complex_fms_pattern): Update comment.
4805 (complex_fms_pattern::matches): Remove.
4806 (complex_operations_pattern::recognize): Remove complex_fma_pattern
4808 2021-10-29 Jakub Jelinek <jakub@redhat.com>
4810 * gimple-fold.c (gimple_fold_builtin_memset): Copy over location from
4813 2021-10-29 Xionghu Luo <luoxhu@linux.ibm.com>
4816 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Add
4817 patterns match and emit for VSX xxpermdi.
4819 2021-10-29 liuhongt <hongtao.liu@intel.com>
4822 * config/i386/i386-builtin-types.def (V8HF_FTYPE_V8HF): New
4824 (V16HF_FTYPE_V16HF): Ditto.
4825 (V32HF_FTYPE_V32HF): Ditto.
4826 (V8HF_FTYPE_V8HF_ROUND): Ditto.
4827 (V16HF_FTYPE_V16HF_ROUND): Ditto.
4828 (V32HF_FTYPE_V32HF_ROUND): Ditto.
4829 * config/i386/i386-builtin.def ( IX86_BUILTIN_FLOORPH,
4830 IX86_BUILTIN_CEILPH, IX86_BUILTIN_TRUNCPH,
4831 IX86_BUILTIN_FLOORPH256, IX86_BUILTIN_CEILPH256,
4832 IX86_BUILTIN_TRUNCPH256, IX86_BUILTIN_FLOORPH512,
4833 IX86_BUILTIN_CEILPH512, IX86_BUILTIN_TRUNCPH512): New builtin.
4834 * config/i386/i386-builtins.c
4835 (ix86_builtin_vectorized_function): Enable vectorization for
4836 HFmode FLOOR/CEIL/TRUNC operation.
4837 * config/i386/i386-expand.c (ix86_expand_args_builtin): Handle
4839 * config/i386/sse.md (rint<mode>2, nearbyint<mode>2): Extend
4842 2021-10-28 Aldy Hernandez <aldyh@redhat.com>
4843 Andrew MacLeod <amacleod@redhat.com>
4845 * value-relation.cc (path_oracle::killing_def): Walk the
4846 equivalency list and remove SSA from any equivalencies.
4848 2021-10-28 Stafford Horne <shorne@gmail.com>
4850 * config/or1k/or1k.h (PROFILE_HOOK): Add return address argument
4853 2021-10-28 Jakub Jelinek <jakub@redhat.com>
4855 PR tree-optimization/102951
4856 * fold-const.h (address_compare): Declare.
4857 * fold-const.c (address_compare): New function.
4858 * match.pd (cmp (convert1?@2 addr@0) (convert2? addr@1)): Use
4859 address_compare helper.
4860 (minmax cmp (convert1?@2 addr@0) (convert2?@3 addr@1)): New
4863 2021-10-28 Andrew MacLeod <amacleod@redhat.com>
4865 * vr-values.c (simplify_using_ranges::fold_cond): Change fold message.
4867 2021-10-28 Andrew MacLeod <amacleod@redhat.com>
4869 PR tree-optimization/102940
4870 * tree-vrp.c (execute_ranger_vrp): Reset scev.
4872 2021-10-28 Richard Purdie <richard.purdie@linuxfoundation.org>
4874 * config/nios2/linux.h (MUSL_DYNAMIC_LINKER): Add musl linker
4876 2021-10-28 Richard Purdie <richard.purdie@linuxfoundation.org>
4878 * configure: Regenerate.
4879 * configure.ac: Use CPPFLAGS_FOR_BUILD for GMPINC
4881 2021-10-28 Eric Botcazou <ebotcazou@adacore.com>
4883 * doc/invoke.texi (%X): Remove obsolete reference to -Wl.
4885 2021-10-28 Richard Biener <rguenther@suse.de>
4888 * fold-const.c (fold_convert_const): Avoid int to float
4889 constant folding with -frounding-math and inexact result.
4890 * simplify-rtx.c (simplify_const_unary_operation): Likewise
4891 for both float and unsigned_float.
4893 2021-10-28 Aldy Hernandez <aldyh@redhat.com>
4895 * tree-ssa-threadbackward.c
4896 (back_threader::find_taken_edge_switch): Use find_case_label_range
4897 instead of find_taken_edge.
4899 2021-10-28 Aldy Hernandez <aldyh@redhat.com>
4901 * tree-ssa-threadbackward.c (class back_threader_registry):
4902 Inherit from back_jt_path_registry.
4903 (back_threader_registry::thread_through_all_blocks): Remove.
4904 (back_threader_registry::register_path): Remove
4905 m_lowlevel_registry prefix.
4907 2021-10-28 Richard Biener <rguenther@suse.de>
4910 * fold-const.c (fold_convert_const_real_from_real): Honor
4911 -frounding-math if the conversion is not exact.
4912 * simplify-rtx.c (simplify_const_unary_operation): Do not
4913 simplify FLOAT_TRUNCATE with sign dependent rounding.
4915 2021-10-28 Richard Biener <rguenther@suse.de>
4917 PR tree-optimization/102949
4918 * tree-vect-stmts.c (ensure_base_align): Look at the
4919 dr_info of a group leader and assert we are looking at
4920 one with analyzed alignment.
4922 2021-10-28 Kewen Lin <linkw@linux.ibm.com>
4925 * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Consider
4926 V1T1 mode for unaligned load and store.
4928 2021-10-28 Kito Cheng <kito.cheng@sifive.com>
4930 * config/riscv/riscv.md (zero_extendsidi2_internal): Allow ZBB
4933 2021-10-28 Kito Cheng <kito.cheng@sifive.com>
4935 * config/riscv/arch-canonicalize (CANONICAL_ORDER): Add `i` to
4938 2021-10-28 Alexandre Oliva <oliva@adacore.com>
4940 * common.opt (fharden-compares): New.
4941 (fharden-conditional-branches): New.
4942 * doc/invoke.texi: Document new options.
4943 * gimple-harden-conditionals.cc: New.
4944 * Makefile.in (OBJS): Build it.
4945 * passes.def: Add new passes.
4946 * tree-pass.h (make_pass_harden_compares): Declare.
4947 (make_pass_harden_conditional_branches): Declare.
4949 2021-10-28 Xionghu Luo <luoxhu@linux.ibm.com>
4952 * config/rs6000/altivec.md: Add vsx register constraints.
4953 * config/rs6000/vsx.md (vsx_xxsel<mode>): Delete.
4954 (vsx_xxsel<mode>2): Likewise.
4955 (vsx_xxsel<mode>3): Likewise.
4956 (vsx_xxsel<mode>4): Likewise.
4958 2021-10-28 Xionghu Luo <luoxhu@linux.ibm.com>
4961 * config/rs6000/altivec.md (*altivec_vsel<mode>): Change to ...
4962 (altivec_vsel<mode>): ... this and update define.
4963 (*altivec_vsel<mode>_uns): Delete.
4964 (altivec_vsel<mode>2): New define_insn.
4965 (altivec_vsel<mode>3): Likewise.
4966 (altivec_vsel<mode>4): Likewise.
4967 * config/rs6000/rs6000-call.c (altivec_expand_vec_sel_builtin): New.
4968 (altivec_expand_builtin): Call altivec_expand_vec_sel_builtin to expand
4970 * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Use bit-wise
4971 selection instead of per element.
4972 * config/rs6000/vector.md:
4973 * config/rs6000/vsx.md (*vsx_xxsel<mode>): Change to ...
4974 (vsx_xxsel<mode>): ... this and update define.
4975 (*vsx_xxsel<mode>_uns): Delete.
4976 (vsx_xxsel<mode>2): New define_insn.
4977 (vsx_xxsel<mode>3): Likewise.
4978 (vsx_xxsel<mode>4): Likewise.
4980 2021-10-28 Hongyu Wang <hongyu.wang@intel.com>
4982 * config/i386/i386.c (use_rsqrt_p): Add mode parameter, enable
4983 HFmode rsqrt without TARGET_SSE_MATH.
4984 (ix86_optab_supported_p): Refactor rint, adjust floor, ceil,
4985 btrunc condition to be restricted by -ftrapping-math, adjust
4986 use_rsqrt_p function call.
4987 * config/i386/i386.md (rcphf2): New define_insn.
4988 (rsqrthf2): Likewise.
4989 * config/i386/sse.md (div<mode>3): Change VF2H to VF2.
4990 (div<mode>3): New expander for HF mode.
4991 (rsqrt<mode>2): Likewise.
4992 (*avx512fp16_vmrcpv8hf2): New define_insn for rpad pass.
4993 (*avx512fp16_vmrsqrtv8hf2): Likewise.
4995 2021-10-27 Saagar Jha <saagar@saagarjha.com>
4997 * config.gcc: Adjust for Darwin21.
4998 * config/darwin-c.c (macosx_version_as_macro): Likewise.
4999 * config/darwin-driver.c (validate_macosx_version_min):
5001 (darwin_find_version_from_kernel): Likewise.
5003 2021-10-27 Aldy Hernandez <aldyh@redhat.com>
5005 * gimple-range-path.cc
5006 (path_range_query::range_defined_in_block): Call killing_def.
5008 2021-10-27 Aldy Hernandez <aldyh@redhat.com>
5010 * gimple-range-fold.cc (fold_using_range::range_of_range_op): Dump
5011 operands as well as relation.
5012 * gimple-range-path.cc
5013 (path_range_query::compute_ranges_in_block): Compute PHI relations
5014 first. Compute outgoing relations at the end.
5015 (path_range_query::compute_ranges): Remove call to compute_relations.
5016 (path_range_query::compute_relations): Remove.
5017 (path_range_query::maybe_register_phi_relation): New.
5018 (path_range_query::compute_phi_relations): Abstract out
5019 registering one PHI relation to...
5020 (path_range_query::compute_outgoing_relations): ...here.
5021 * gimple-range-path.h (class path_range_query): Remove
5023 Add maybe_register_phi_relation.
5025 2021-10-27 Aldy Hernandez <aldyh@redhat.com>
5026 Andrew MacLeod <amacleod@redhat.com>
5028 * value-relation.cc (path_oracle::killing_def): Kill second
5031 2021-10-27 Martin Jambor <mjambor@suse.cz>
5033 * ipa-cp.c (good_cloning_opportunity_p): Decide whether to use
5034 profile feedback depending on their local availability.
5036 2021-10-27 Martin Jambor <mjambor@suse.cz>
5038 * params.opt (param_ipa_cp_profile_count_base): New parameter.
5039 * doc/invoke.texi (Optimize Options): Add entry for
5040 ipa-cp-profile-count-base.
5041 * ipa-cp.c (max_count): Replace with base_count, replace all
5042 occurrences too, unless otherwise stated.
5043 (ipcp_cloning_candidate_p): identify mostly-directly called
5044 functions based on their counts, not max_count.
5045 (compare_edge_profile_counts): New function.
5046 (ipcp_propagate_stage): Instead of setting max_count, find the
5047 appropriate edge count in a sorted vector of counts of eligible
5048 edges and make it the base_count.
5050 2021-10-27 Martin Jambor <mjambor@suse.cz>
5052 * ipa-cp.c (struct caller_statistics): New fields rec_count_sum,
5053 n_nonrec_calls and itself, document all fields.
5054 (init_caller_stats): Initialize the above new fields.
5055 (gather_caller_stats): Gather self-recursive counts and calls number.
5056 (get_info_about_necessary_edges): Gather counts of self-recursive and
5057 other edges bringing in the requested value separately.
5058 (dump_profile_updates): Rework to dump info about a single node only.
5059 (lenient_count_portion_handling): New function.
5060 (struct gather_other_count_struct): New type.
5061 (gather_count_of_non_rec_edges): New function.
5062 (struct desc_incoming_count_struct): New type.
5063 (analyze_clone_icoming_counts): New function.
5064 (adjust_clone_incoming_counts): Likewise.
5065 (update_counts_for_self_gen_clones): Likewise.
5066 (update_profiling_info): Rewritten.
5067 (update_specialized_profile): Adjust call to dump_profile_updates.
5068 (create_specialized_node): Do not update profiling info.
5069 (decide_about_value): New parameter self_gen_clones, either push new
5070 clones into it or updat their profile counts. For self-recursively
5071 generated values, use a portion of the node count instead of count
5072 from self-recursive edges to estimate goodness.
5073 (decide_whether_version_node): Gather clones for self-generated values
5074 in a new vector, update their profiles at once at the end.
5076 2021-10-27 Richard Biener <rguenther@suse.de>
5078 * tree-vectorizer.c (vect_transform_loops): New function,
5080 (try_vectorize_loop_1): ... here. Simplify as epilogues
5081 are now fully handled in the split part.
5083 2021-10-27 Jakub Jelinek <jakub@redhat.com>
5085 * omp-expand.c (expand_omp_for_init_counts): Handle non-rectangular
5086 iterators with pointer types.
5087 (expand_omp_for_init_vars, extract_omp_for_update_vars): Likewise.
5089 2021-10-26 Martin Sebor <msebor@redhat.com>
5091 PR tree-optimization/102238
5092 PR tree-optimization/102919
5093 * gimple-ssa-sprintf.c (get_string_length): Add an argument.
5094 (array_elt_at_offset): Move to pointer-query.
5095 (set_aggregate_size_and_offset): New function.
5096 (field_at_offset): Move to pointer-query.
5097 (get_origin_and_offset): Rename...
5098 (get_origin_and_offset_r): this. Add an argument. Make aggregate
5099 handling more robust.
5100 (get_origin_and_offset): New.
5101 (alias_offset): Add an argument.
5102 (format_string): Use subobject size determined by get_origin_and_offset.
5103 * pointer-query.cc (field_at_offset): Move from gimple-ssa-sprintf.c.
5104 Improve/correct handling of aggregates.
5105 (array_elt_at_offset): Same.
5106 * pointer-query.h (field_at_offset): Declare.
5107 (array_elt_at_offset): Declare.
5109 2021-10-26 Martin Sebor <msebor@redhat.com>
5111 * builtins.c (check_strncat_sizes): Pass access_data ctor additional
5113 (expand_builtin_memcmp): Move code to gimple-ssa-warn-access.cc.
5114 (expand_builtin_fork_or_exec): Same.
5115 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Pass
5116 compute_objsize additional arguments.
5117 (inbounds_memaccess_p): Same.
5118 (array_bounds_checker::check_array_bounds): Add an assert. Stash
5119 statement in a member.
5120 (check_array_bounds_dom_walker::before_dom_children): Same.
5121 * gimple-array-bounds.h (array_bounds_checker::m_stmt): New member.
5122 * gimple-ssa-sprintf.c (get_destination_size): Add an argument.
5123 (handle_printf_call): Pass a new argument.
5124 * gimple-ssa-warn-access.cc (get_size_range): Add an argument.
5125 (check_access): Add an argument and pass it along to callees.
5126 (check_read_access): Make a member function.
5127 (pass_waccess::check_strcat): Pass access_data ctor additional
5129 (pass_waccess::check_strncat): Same.
5130 (pass_waccess::check_stxcpy): Same.
5131 (pass_waccess::check_stxncpy): Same.
5132 (pass_waccess::check_strncmp): Same.
5133 (pass_waccess::check_read_access): Same.
5134 (pass_waccess::check_builtin): Same.
5135 (pass_waccess::maybe_check_access_sizes): Same.
5136 (pass_waccess::maybe_check_dealloc_call): Same.
5137 * gimple-ssa-warn-access.h (check_read_access): Declare a new
5139 * pointer-query.cc (compute_objsize_r): Add an argument.
5140 (gimple_call_return_array): Same.
5141 (gimple_call_alloc_size): Same.
5142 (access_ref::access_ref): Same.
5143 (access_ref::get_ref): Same.
5144 (pointer_query::get_ref): Same.
5145 (handle_min_max_size): Pass an arguments to callees.
5146 (handle_array_ref): Add an argument.
5147 (handle_mem_ref): Same.
5148 (compute_objsize): Same.
5149 * pointer-query.h (struct access_ref): Adjust signatures.
5150 (struct access_data): Same.
5151 (gimple_call_alloc_size): Add an argument.
5152 (gimple_parm_array_size): Same.
5153 (compute_objsize): Same.
5154 * tree-ssa-strlen.c (strlen_pass::adjust_last_stmt): Pass an additional
5155 argument to compute_objsize.
5156 (strlen_pass::maybe_warn_overflow): Same.
5157 (maybe_diag_stxncpy_trunc): Same.
5159 2021-10-26 Martin Sebor <msebor@redhat.com>
5161 PR middle-end/102453
5162 * gimple-ssa-warn-access.cc (pass_waccess::check_atomic_builtin): New.
5163 (pass_waccess::check_atomic_builtin): Call it.
5165 2021-10-26 Vladimir N. Makarov <vmakarov@redhat.com>
5167 PR rtl-optimization/102842
5168 * lra-constraints.c (match_reload): Ignore out in checking values
5170 (curr_insn_transform): Collect outputs before doing reloads of operands.
5172 2021-10-26 Paul A. Clarke <pc@us.ibm.com>
5175 * config/rs6000/x86intrin.h: Move some included headers to new
5176 headers. Include new immintrin.h instead of those headers.
5177 * config/rs6000/immintrin.h: New.
5178 * config/rs6000/x86gprintrin.h: New.
5179 * config.gcc (powerpc*-*-*): Add new headers to extra_headers.
5181 2021-10-26 Richard Biener <rguenther@suse.de>
5183 * tree-vectorizer.h (vect_create_addr_base_for_vector_ref):
5184 Remove byte_offset parameter.
5185 (vect_create_data_ref_ptr): Likewise.
5186 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref):
5188 (vect_create_data_ref_ptr): Likewise.
5189 * tree-vect-stmts.c (vectorizable_store): Adjust.
5190 (vectorizable_load): Likewise.
5192 2021-10-26 Richard Biener <rguenther@suse.de>
5194 PR tree-optimization/96109
5195 * tree-vectorizer.h (dr_misalignment): Add optional offset
5197 * tree-vect-data-refs.c (dr_misalignment): Likewise. Remove
5198 offset applied for negative stride accesses.
5199 (vect_enhance_data_refs_alignment): Compute negative stride
5200 access offset and pass it to dr_misalignment.
5201 * tree-vect-stmts.c (get_negative_load_store_type): Pass
5202 negative offset to dr_misalignment.
5203 (get_group_load_store_type): Likewise.
5204 (get_load_store_type): Likewise.
5205 (vectorizable_store): Remove asserts about alignment.
5206 (vectorizable_load): Likewise.
5208 2021-10-26 Kewen Lin <linkw@linux.ibm.com>
5210 PR tree-optimization/102897
5211 * tree-ssa-forwprop.c (simplify_permutation): Remove a wrong assertion.
5213 2021-10-26 Richard Biener <rguenther@suse.de>
5215 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref):
5216 Take offset in bytes.
5217 (vect_create_data_ref_ptr): Likewise.
5218 * tree-vect-loop-manip.c (get_misalign_in_elems): Multiply
5219 offset by element size.
5220 (vect_create_cond_for_align_checks): Likewise.
5221 * tree-vect-stmts.c (get_negative_load_store_type): Likewise.
5222 (vectorizable_load): Remove duplicate leftover from merge
5225 2021-10-26 Roger Sayle <roger@nextmovesoftware.com>
5227 * config/i386/i386-expand.c (ix86_expand_v1ti_shift): New helper
5228 function to expand V1TI mode logical shifts by integer constants.
5229 (ix86_expand_v1ti_rotate): New helper function to expand V1TI
5230 mode rotations by integer constants.
5231 * config/i386/i386-protos.h (ix86_expand_v1ti_shift,
5232 ix86_expand_v1ti_rotate): Prototype new functions here.
5233 * config/i386/sse.md (ashlv1ti3, lshrv1ti3, rotlv1ti3, rotrv1ti3):
5234 New TARGET_SSE2 expanders to implement V1TI shifts and rotations.
5236 2021-10-26 Aldy Hernandez <aldyh@redhat.com>
5238 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
5239 Avoid threading circular paths.
5241 2021-10-26 Aldy Hernandez <aldyh@redhat.com>
5243 * tree-ssa-threadbackward.c (back_threader::resolve_phi):
5244 Attempt to resolve all incoming paths to a PHI.
5245 (back_threader::resolve_def): Always return true for PHIs.
5247 2021-10-26 Aldy Hernandez <aldyh@redhat.com>
5249 * tree-ssa-threadbackward.c (back_threader::find_paths_to_names):
5250 Always try to resolve path without looking back.
5251 * tree-ssa-threadupdate.c (dump_jump_thread): Indidicate whether
5252 edge is a back edge.
5254 2021-10-26 Kewen Lin <linkw@linux.ibm.com>
5256 PR tree-optimization/102789
5257 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
5258 update inits of simd_lane_access.
5260 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
5262 * tree-vrp.c (vrp_simplify_cond_using_ranges): Add return type and
5263 move to vr-values.c.
5264 (simplify_casted_conds): Move to vrp_folder class.
5265 (execute_vrp): Call via vrp_folder now.
5266 * vr-values.c (simplify_cond_using_ranges_1): Call simplify_casted_cond.
5267 (simplify_using_ranges::simplify_casted_cond): Relocate from tree-vrp.c.
5268 * vr-values.h (simplify_casted_cond): Add prototype.
5270 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
5272 * tree-vrp.c (rvrp_folder::fold_stmt): If simplification fails, try
5275 2021-10-25 Paul A. Clarke <pc@us.ibm.com>
5277 * config/rs6000/smmintrin.h (_mm_testz_si128): Add "extern" to
5279 (_mm_testc_si128): Likewise.
5280 (_mm_testnzc_si128): Likewise.
5281 (_mm_blend_ps): Likewise.
5282 (_mm_blendv_ps): Likewise.
5283 (_mm_blend_pd): Likewise.
5284 (_mm_blendv_pd): Likewise.
5285 (_mm_ceil_pd): Likewise.
5286 (_mm_ceil_sd): Likewise.
5287 (_mm_ceil_ps): Likewise.
5288 (_mm_ceil_ss): Likewise.
5289 (_mm_floor_pd): Likewise.
5290 (_mm_floor_sd): Likewise.
5291 (_mm_floor_ps): Likewise.
5292 (_mm_floor_ss): Likewise.
5293 (_mm_minpos_epu16): Likewise.
5294 (_mm_mul_epi32): Likewise.
5295 (_mm_cvtepi8_epi16): Likewise.
5296 (_mm_packus_epi32): Likewise.
5297 (_mm_cmpgt_epi64): Likewise.
5299 2021-10-25 Roger Sayle <roger@nextmovesoftware.com>
5301 * simplify-rtx.c (simplify_binary_operation_1) [SS_ASHIFT]: Simplify
5302 shifts of the mode's smin_value and smax_value when the bit count
5303 operand doesn't have side-effects.
5304 [US_ASHIFT]: Likewise, simplify shifts of the mode's umax_value
5305 when the bit count operand doesn't have side-effects.
5306 (simplify_const_binary_operation) [SS_ASHIFT, US_ASHIFT]: Perform
5307 compile-time evaluation of saturating left shifts with constant
5310 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
5312 * gimple-range.cc (gimple_ranger::export_global_ranges): Remove check
5315 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
5317 * flag-types.h (enum ranger_debug): Adjust values.
5318 * params.opt (ranger_debug): Ditto.
5320 2021-10-25 Martin Jambor <mjambor@suse.cz>
5322 PR tree-optimization/102886
5323 * tree-sra.c (totally_scalarize_subtree): Fix the out of
5326 2021-10-25 Andrew Pinski <apinski@marvell.com>
5328 * tree-ssa-dce.c (simple_dce_from_worklist):
5329 Check stmt_unremovable_because_of_non_call_eh_p also
5330 before removing the statement.
5332 2021-10-25 Richard Biener <rguenther@suse.de>
5334 PR tree-optimization/102905
5335 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
5336 Use vect_supportable_dr_alignment again to determine whether
5337 an access is supported when not aligned.
5339 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
5341 * config/riscv/riscv.c (riscv_rtx_costs): Handle cost model
5344 2021-10-25 Jim Wilson <jimw@sifive.com>
5345 Kito Cheng <kito.cheng@sifive.com>
5346 Jia-Wei Chen <jiawei@iscas.ac.cn>
5347 Shi-Hua Liao <shihua@iscas.ac.cn>
5349 * config/riscv/bitmanip.md (shiftm1): New.
5350 (*bset<mode>): Ditto.
5351 (*bset<mode>_mask): Ditto.
5352 (*bset<mode>_1): Ditto.
5353 (*bset<mode>_1_mask): Ditto.
5354 (*bseti<mode>): Ditto.
5355 (*bclr<mode>): Ditto.
5356 (*bclri<mode>): Ditto.
5357 (*binv<mode>): Ditto.
5358 (*binvi<mode>): Ditto.
5359 (*bext<mode>): Ditto.
5361 * config/riscv/predicates.md (splittable_const_int_operand):
5363 (single_bit_mask_operand): New.
5364 (not_single_bit_mask_operand): Ditto.
5365 (const31_operand): Ditto.
5366 (const63_operand): Ditto.
5367 * config/riscv/riscv.c (riscv_build_integer_1): Handle bseti.
5368 (riscv_output_move): Ditto.
5369 (riscv_print_operand): Handle new operand type: T and S.
5370 * config/riscv/riscv.h (SINGLE_BIT_MASK_OPERAND): New.
5372 2021-10-25 Jim Wilson <jimw@sifive.com>
5374 * config/riscv/riscv.c (riscv_build_integer_1): Build integer
5377 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
5379 * config/riscv/riscv.c (riscv_extend_cost): Handle cost model
5381 (riscv_rtx_costs): Ditto.
5383 2021-10-25 Jim Wilson <jimw@sifive.com>
5384 Kito Cheng <kito.cheng@sifive.com>
5385 Jia-Wei Chen <jiawei@iscas.ac.cn>
5387 * config/riscv/bitmanip.md (bitmanip_bitwise): New.
5388 (bitmanip_minmax): New.
5389 (clz_ctz_pcnt): New.
5390 (bitmanip_optab): New.
5391 (bitmanip_insn): New.
5392 (*<optab>_not<mode>): New.
5393 (*xor_not<mode>): New.
5394 (<bitmanip_optab>si2): New.
5395 (*<bitmanip_optab>disi2): New.
5396 (<bitmanip_optab>di2): New.
5397 (*zero_extendhi<GPR:mode>2_bitmanip): New.
5398 (*extend<SHORT:mode><SUPERQI:mode>2_zbb): New.
5399 (*zero_extendhi<GPR:mode>2_zbb): New.
5402 (rotrsi3_sext): New.
5405 (rotlsi3_sext): New.
5406 (bswap<mode>2): New.
5407 (<bitmanip_optab><mode>3): New.
5408 * config/riscv/riscv.md (type): Add rotate.
5409 (zero_extendhi<GPR:mode>2): Change to define_expand pattern.
5410 (*zero_extendhi<GPR:mode>2): New.
5411 (extend<SHORT:mode><SUPERQI:mode>2): Change to define_expand pattern.
5412 (*extend<SHORT:mode><SUPERQI:mode>2): New.
5414 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
5416 * config/riscv/riscv.c (riscv_extend_cost): Handle cost model
5418 (riscv_rtx_costs): Ditto.
5420 2021-10-25 Jim Wilson <jimw@sifive.com>
5421 Kito Cheng <kito.cheng@sifive.com>
5422 Jia-Wei Chen <jiawei@iscas.ac.cn>
5424 * config/riscv/bitmanip.md (*zero_extendsidi2_bitmanip): New.
5429 (riscv_rtx_costs): Ditto.
5430 * config/riscv/riscv.md: Include bitmanip.md
5431 (type): Add bitmanip bype.
5432 (zero_extendsidi2): Change to define_expand pattern.
5433 (*zero_extendsidi2_internal): New.
5434 (zero_extendsidi2_shifted): Disable for ZBA.
5436 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
5438 * common/config/riscv/riscv-common.c (riscv_ext_version_table):
5439 Add zba, zbb, zbc and zbs.
5440 (riscv_ext_flag_table): Ditto.
5441 * config/riscv/riscv-opts.h (MASK_ZBA): New.
5445 (TARGET_ZBA): Ditto.
5446 (TARGET_ZBB): Ditto.
5447 (TARGET_ZBC): Ditto.
5448 (TARGET_ZBS): Ditto.
5449 * config/riscv/riscv.opt (riscv_zb_subext): New.
5451 2021-10-25 liuhongt <hongtao.liu@intel.com>
5454 * match.pd: Simplify (_Float16) sqrtf((float) a) to .SQRT(a)
5455 when direct_internal_fn_supported_p, similar for sqrt/sqrtl.
5457 2021-10-25 Richard Biener <rguenther@suse.de>
5459 PR tree-optimization/102920
5460 * tree-ssa-sccvn.h (expressions_equal_p): Add argument
5461 controlling VN_TOP matching behavior.
5462 * tree-ssa-sccvn.c (expressions_equal_p): Likewise.
5463 (vn_phi_eq): Do not optimistically match VN_TOP.
5465 2021-10-25 konglin1 <lingling.kong@intel.com>
5467 * config/i386/sse.md (fma_<mode>_fadd_fmul): Add new
5468 define_insn_and_split.
5469 (fma_<mode>_fadd_fcmul):Likewise
5470 (fma_<complexopname>_<mode>_fma_zero):Likewise
5472 2021-10-24 John David Anglin <danglin@gcc.gnu.org>
5474 * config/pa/pa-d.c (pa_d_handle_target_float_abi): Don't check
5475 TARGET_DISABLE_FPREGS.
5476 * config/pa/pa.c (fix_range): Use MASK_SOFT_FLOAT instead of
5477 MASK_DISABLE_FPREGS.
5478 (hppa_rtx_costs): Don't check TARGET_DISABLE_FPREGS. Adjust
5479 cost of hardware integer multiplication.
5480 (pa_conditional_register_usage): Don't check TARGET_DISABLE_FPREGS.
5481 * config/pa/pa.h (INT14_OK_STRICT): Likewise.
5482 * config/pa/pa.md: Don't check TARGET_DISABLE_FPREGS. Check
5483 TARGET_SOFT_FLOAT in patterns that use xmpyu instruction.
5484 * config/pa/pa.opt (mdisable-fpregs): Change target mask to
5485 SOFT_FLOAT. Revise comment.
5486 (msoft-float): New option.
5488 2021-10-24 John David Anglin <danglin@gcc.gnu.org>
5490 * config/pa/pa.md: Don't use 'G' constraint in integer move patterns.
5492 2021-10-24 Gerald Pfeifer <gerald@pfeifer.com>
5494 * doc/install.texi (Specific): Remove obsolete details
5495 around GNU/Linux on Itanium.
5496 (Specific): Remove reference to Windows for Itanium.
5498 2021-10-23 Bernhard Reutner-Fischer <aldot@gcc.gnu.org>
5500 * config/i386/x86-tune-sched-bd.c (dispatch_group): Commentary
5503 2021-10-23 Jan Hubicka <hubicka@ucw.cz>
5505 * tree-ssa-structalias.c (compute_points_to_sets): Cleanup.
5507 2021-10-23 Roger Sayle <roger@nextmovesoftware.com>
5509 * config/i386/sse.md (<any_logic>v1ti3): New define_insn to
5510 implement V1TImode AND, IOR and XOR on TARGET_SSE2 (and above).
5511 (one_cmplv1ti2): New define expand.
5513 2021-10-22 Eric Gallager <egallager@gcc.gnu.org>
5516 * Makefile.in: Handle dvidir and install-dvi target.
5517 * configure: Regenerate.
5518 * configure.ac: Add install-dvi to target_list.
5520 2021-10-22 Gerald Pfeifer <gerald@pfeifer.com>
5522 * doc/install.texi (Binaries): Convert mingw-w64.org to https.
5525 2021-10-22 Richard Biener <rguenther@suse.de>
5527 PR tree-optimization/102893
5528 * tree-ssa-dce.c (find_obviously_necessary_stmts): Fix the
5529 test for an exit edge.
5531 2021-10-22 Aldy Hernandez <aldyh@redhat.com>
5532 Andrew MacLeod <amacleod@redhat.com>
5534 * gimple-range-path.cc (path_range_query::compute_phi_relations):
5535 Kill any global relations we may know before registering a new
5537 * value-relation.cc (path_oracle::killing_def): New.
5538 * value-relation.h (path_oracle::killing_def): New.
5540 2021-10-22 Richard Biener <rguenther@suse.de>
5543 * tree-ssa-sccvn.c (vn_phi_insert): For undefined SSA args
5545 (vn_phi_lookup): Likewise.
5547 2021-10-21 H.J. Lu <hjl.tools@gmail.com>
5550 * doc/invoke.texi: Document -fcf-protection requires i686 or
5553 2021-10-21 Eric Botcazou <ebotcazou@adacore.com>
5555 PR middle-end/102764
5556 * cfgexpand.c (expand_gimple_basic_block): Robustify latest change.
5558 2021-10-21 Jonathan Wright <jonathan.wright@arm.com>
5560 * config/aarch64/arm_neon.h (__STRUCTN): Delete function
5561 macro and all invocations.
5563 2021-10-21 Andrew MacLeod <amacleod@redhat.com>
5565 * doc/invoke.texi (ranger-debug): Document.
5566 * flag-types.h (enum ranger_debug): New.
5567 (enum evrp_mode): Remove debug values.
5568 * gimple-range-cache.cc (DEBUG_RANGE_CACHE): Use new debug flag.
5569 * gimple-range-gori.cc (gori_compute::gori_compute): Ditto.
5570 * gimple-range.cc (gimple_ranger::gimple_ranger): Ditto.
5571 * gimple-ssa-evrp.c (hybrid_folder::choose_value): Ditto.
5572 (execute_early_vrp): Use evrp-mode directly.
5573 * params.opt (enum evrp_mode): Remove debug values.
5574 (ranger-debug): New.
5575 (ranger-logical-depth): Relocate to be in alphabetical order.
5577 2021-10-21 Andrew MacLeod <amacleod@redhat.com>
5579 * doc/invoke.texi: (vrp1-mode, vrp2-mode): Document.
5580 * flag-types.h: (enum vrp_mode): New.
5581 * params.opt: (vrp1-mode, vrp2-mode): New.
5582 * tree-vrp.c (vrp_pass_num): New.
5583 (pass_vrp::pass_vrp): Set pass number.
5584 (pass_vrp::execute): Choose which VRP mode to execute.
5586 2021-10-21 Andrew MacLeod <amacleod@redhat.com>
5588 * gimple-ssa-evrp.c (class rvrp_folder): Move to tree-vrp.c.
5589 (execute_early_vrp): For ranger only mode, invoke ranger_vrp.
5590 * tree-vrp.c (class rvrp_folder): Relocate here.
5591 (execute_ranger_vrp): New.
5592 * tree-vrp.h (execute_ranger_vrp): Export.
5594 2021-10-21 Martin Liska <mliska@suse.cz>
5598 * opts.c (finish_options): Process flag_var_tracking* options
5599 here as they can be adjusted by optimize attribute.
5600 Process also flag_syntax_only and flag_gtoggle.
5601 * toplev.c (process_options): Remove it here.
5602 * common.opt: Make debug_nonbind_markers_p as PerFunction
5603 attribute as it depends on optimization level.
5605 2021-10-21 Martin Jambor <mjambor@suse.cz>
5607 PR tree-optimization/102505
5608 * tree-sra.c (totally_scalarize_subtree): Check that the
5609 encountered field fits within the acces we would like to put it
5612 2021-10-21 Aldy Hernandez <aldyh@redhat.com>
5614 * tree-ssa-threadbackward.c
5615 (back_threader::maybe_register_path): Remove circular paths check.
5617 2021-10-21 Richard Biener <rguenther@suse.de>
5619 * toplev.c (process_options): Move the initial debug_hooks
5621 (toplev::main): ... before the call of the post_options
5624 2021-10-21 Richard Biener <rguenther@suse.de>
5626 PR tree-optimization/102847
5627 * tree-vect-stmts.c (vect_model_load_cost): Add the scalar
5628 load cost in the prologue for VMAT_INVARIANT.
5630 2021-10-21 Richard Biener <rguenther@suse.de>
5632 PR tree-optimization/102847
5633 * tree-vect-stmts.c (vect_model_load_cost): Explicitely
5634 handle VMAT_INVARIANT as a splat in the prologue.
5636 2021-10-21 Hongyu Wang <hongyu.wang@intel.com>
5639 * config/i386/i386.c (ix86_get_ssemov): Adjust HFmode vector
5640 move to use the same logic as HImode.
5642 2021-10-21 Richard Biener <rguenther@suse.de>
5644 * tree-vect-slp.c (vect_build_slp_tree_1): Remove
5645 superfluous gimple_call_nothrow_p check.
5647 2021-10-21 Andrew Pinski <apinski@marvell.com>
5649 * tree-cfg.c (maybe_remove_writeonly_store): Add dce_ssa_names argument.
5650 Mark the ssa-name of the rhs as one to be removed.
5651 (execute_fixup_cfg): Update call to maybe_remove_writeonly_store.
5652 Call simple_dce_from_worklist at the end to a simple dce.
5654 2021-10-21 Andrew Pinski <apinski@marvell.com>
5656 * tree-cfg.c (maybe_remove_writeonly_store): New function
5657 factored out from ...
5658 (execute_fixup_cfg): Here. Call maybe_remove_writeonly_store.
5660 2021-10-21 Andrew Pinski <apinski@marvell.com>
5662 * tree-cfg.c (execute_fixup_cfg): Remove comment
5663 about standalone pass.
5665 2021-10-21 Andrew Pinski <apinski@marvell.com>
5667 * tree-cfg.c (execute_fixup_cfg): Output when the statement
5668 is removed when it is a write only var.
5670 2021-10-21 Aldy Hernandez <aldyh@redhat.com>
5672 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
5673 Avoid threading circular paths.
5675 2021-10-20 Alex Coplan <alex.coplan@arm.com>
5677 * calls.c (initialize_argument_information): Remove some dead
5678 code, remove handling for function_arg returning const_int.
5679 * doc/tm.texi: Delete documentation for unused target hooks.
5680 * doc/tm.texi.in: Likewise.
5681 * target.def (load_bounds_for_arg): Delete.
5682 (store_bounds_for_arg): Delete.
5683 (load_returned_bounds): Delete.
5684 (store_returned_bounds): Delete.
5685 * targhooks.c (default_load_bounds_for_arg): Delete.
5686 (default_store_bounds_for_arg): Delete.
5687 (default_load_returned_bounds): Delete.
5688 (default_store_returned_bounds): Delete.
5689 * targhooks.h (default_load_bounds_for_arg): Delete.
5690 (default_store_bounds_for_arg): Delete.
5691 (default_load_returned_bounds): Delete.
5692 (default_store_returned_bounds): Delete.
5694 2021-10-20 Jonathan Wakely <jwakely@redhat.com>
5696 * doc/extend.texi (Basic Asm): Clarify that asm is not an
5698 * doc/invoke.texi (-fno-asm): Fix description for C++.
5700 2021-10-20 Jonathan Wakely <jwakely@redhat.com>
5702 * doc/install.texi: Remove link to old.html
5704 2021-10-20 Tamar Christina <tamar.christina@arm.com>
5706 * config/aarch64/aarch64-simd.md (*aarch64_cmtst_same_<mode>): New.
5708 2021-10-20 Tamar Christina <tamar.christina@arm.com>
5710 * config/aarch64/aarch64-simd.md (*aarch64_narrow_trunc<mode>): New.
5712 2021-10-20 Tamar Christina <tamar.christina@arm.com>
5714 * config/aarch64/aarch64-simd.md (aarch64_simd_ashr<mode>): Add case cmp
5716 * config/aarch64/constraints.md (D1): New.
5718 2021-10-20 Tamar Christina <tamar.christina@arm.com>
5720 * config/aarch64/aarch64-simd.md
5721 (*aarch64_<srn_op>topbits_shuffle<mode>_le): New.
5722 (*aarch64_topbits_shuffle<mode>_le): New.
5723 (*aarch64_<srn_op>topbits_shuffle<mode>_be): New.
5724 (*aarch64_topbits_shuffle<mode>_be): New.
5725 * config/aarch64/predicates.md
5726 (aarch64_simd_shift_imm_vec_exact_top): New.
5728 2021-10-20 Tamar Christina <tamar.christina@arm.com>
5730 * config/aarch64/aarch64-simd.md (*aarch64_<srn_op>shrn<mode>_vect,
5731 *aarch64_<srn_op>shrn<mode>2_vect_le,
5732 *aarch64_<srn_op>shrn<mode>2_vect_be): New.
5733 * config/aarch64/iterators.md (srn_op): New.
5735 2021-10-20 Chung-Lin Tang <cltang@codesourcery.com>
5737 * omp-low.c (omp_copy_decl_2): For !ctx, use record_vars to add new copy
5739 (scan_sharing_clauses): Place copy of OMP_CLAUSE_IN_REDUCTION decl in
5740 ctx->outer instead of ctx.
5742 2021-10-20 Martin Liska <mliska@suse.cz>
5745 2021-10-19 Martin Liska <mliska@suse.cz>
5748 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p): Strip whitespaces.
5749 * system.h (strip_whilespaces): New function.
5751 2021-10-20 Martin Liska <mliska@suse.cz>
5754 2021-10-19 Martin Liska <mliska@suse.cz>
5757 * config/aarch64/aarch64.c (aarch64_process_one_target_attr):
5760 2021-10-20 Wilco Dijkstra <wdijkstr@arm.com>
5762 * config/aarch64/aarch64.c (aarch64_case_values_threshold):
5763 Change to 8 with -Os, 11 otherwise.
5765 2021-10-20 Wilco Dijkstra <wdijkstr@arm.com>
5767 * config/aarch64/aarch64.c (neoversev1_tunings):
5768 Enable AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
5769 (neoversen2_tunings): Likewise.
5771 2021-10-20 Wilco Dijkstra <wdijkstr@arm.com>
5774 * config/aarch64/aarch64.md (frint_pattern): Update comment.
5775 * config/aarch64/aarch64-simd-builtins.def: Change frintn to roundeven.
5776 * config/aarch64/arm_fp16.h: Change frintn to roundeven.
5777 * config/aarch64/arm_neon.h: Likewise.
5778 * config/aarch64/iterators.md (frint_pattern): Use roundeven for FRINTN.
5780 2021-10-20 Martin Liska <mliska@suse.cz>
5782 * config/arm/arm.c (arm_unwind_emit_sequence): Do not declare
5783 already declared global variable.
5784 (arm_unwind_emit_set): Use out_file as function argument.
5785 (arm_unwind_emit): Likewise.
5786 * config/darwin.c (machopic_output_data_section_indirection): Likewise.
5787 (machopic_output_stub_indirection): Likewise.
5788 (machopic_output_indirection): Likewise.
5789 (machopic_finish): Likewise.
5790 * config/i386/i386.c (ix86_asm_output_function_label): Likewise.
5791 * config/i386/winnt.c (i386_pe_seh_unwind_emit): Likewise.
5792 * config/ia64/ia64.c (process_epilogue): Likewise.
5793 (process_cfa_adjust_cfa): Likewise.
5794 (process_cfa_register): Likewise.
5795 (process_cfa_offset): Likewise.
5796 (ia64_asm_unwind_emit): Likewise.
5797 * config/s390/s390.c (s390_asm_output_function_label): Likewise.
5799 2021-10-20 Andre Simoes Dias Vieira <andre.simoesdiasvieira@arm.com>
5801 * config/aarch64/aarch64-builtins.c
5802 (aarch64_general_gimple_fold_builtin):
5803 lower vld1 and vst1 variants of the neon builtins
5804 * config/aarch64/aarch64-protos.h:
5805 (aarch64_general_gimple_fold_builtin): Add gsi parameter.
5806 * config/aarch64/aarch64.c (aarch64_general_gimple_fold_builtin):
5809 2021-10-20 Andre Simoes Dias Vieira <andre.simoesdiasvieira@arm.com>
5811 * match.pd: Generate IFN_TRUNC.
5813 2021-10-20 Richard Biener <rguenther@suse.de>
5815 PR tree-optimization/102853
5816 * tree-data-ref.c (split_constant_offset_1): Bail out
5817 immediately if the expression traps on overflow.
5819 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
5821 * tree-ssa-threadbackward.c (back_threader::~back_threader): Remove.
5823 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
5825 * tree-ssa-threadupdate.c (back_jt_path_registry::adjust_paths_after_duplication):
5826 Remove superflous debugging message.
5827 (back_jt_path_registry::duplicate_thread_path): Same.
5829 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
5831 * tree-ssa-threadbackward.c (back_threader_registry::back_threader_registry):
5833 (back_threader_registry::register_path): Remove m_threaded_paths.
5835 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
5837 PR tree-optimization/102814
5838 * doc/invoke.texi: Document --param=max-fsm-thread-length.
5839 * params.opt: Add --param=max-fsm-thread-length.
5840 * tree-ssa-threadbackward.c
5841 (back_threader_profitability::profitable_path_p): Fail on paths
5842 longer than max-fsm-thread-length.
5844 2021-10-20 Eric Botcazou <ebotcazou@adacore.com>
5846 PR middle-end/102764
5847 * cfgexpand.c (expand_gimple_basic_block): Disregard a final debug
5848 statement to reset the current location for the outgoing edges.
5850 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
5851 Richard Biener <rguenther@suse.de>
5853 * tree-ssa-threadupdate.c (cancel_thread): Dump threading reason
5854 on the same line as the threading cancellation.
5855 (jt_path_registry::cancel_invalid_paths): Avoid rotating loops.
5856 Avoid threading through loop headers where the path remains in the
5859 2021-10-20 Siddhesh Poyarekar <siddhesh@gotplt.org>
5861 * tree-object-size.c (unknown): Make into a function. Adjust
5863 (unknown_object_size): Simplify implementation.
5865 2021-10-20 Hongtao Liu <hongtao.liu@intel.com>
5866 Kewen Lin <linkw@linux.ibm.com>
5868 * doc/sourcebuild.texi (Effective-Target Keywords): Document
5869 vect_slp_v2qi_store, vect_slp_v4qi_store, vect_slp_v8qi_store,
5870 vect_slp_v16qi_store, vect_slp_v2hi_store,
5871 vect_slp_v4hi_store, vect_slp_v2si_store, vect_slp_v4si_store.
5873 2021-10-19 Jonathan Wakely <jwakely@redhat.com>
5875 * doc/extend.texi (Basic PowerPC Built-in Functions): Fix typo.
5877 2021-10-19 Paul A. Clarke <pc@us.ibm.com>
5881 * config/rs6000/emmintrin.h: Guard POWER8 intrinsics.
5882 * config/rs6000/pmmintrin.h: Same.
5883 * config/rs6000/smmintrin.h: Same.
5884 * config/rs6000/tmmintrin.h: Same.
5886 2021-10-19 Paul A. Clarke <pc@us.ibm.com>
5888 * config.gcc (extra_headers): Add nmmintrin.h.
5890 2021-10-19 Richard Biener <rguenther@suse.de>
5892 * tree-vectorizer.h (vect_supportable_dr_alignment): Add
5893 misalignment parameter.
5894 * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs):
5895 Do not change DR_MISALIGNMENT in place, instead pass the
5896 adjusted misalignment to vect_supportable_dr_alignment.
5897 (vect_peeling_supportable): Likewise.
5898 (vect_peeling_hash_get_lowest_cost): Adjust.
5899 (vect_enhance_data_refs_alignment): Likewise.
5900 (vect_vfa_access_size): Likewise.
5901 (vect_supportable_dr_alignment): Add misalignment
5902 parameter and simplify.
5903 * tree-vect-stmts.c (get_negative_load_store_type): Adjust.
5904 (get_group_load_store_type): Likewise.
5905 (get_load_store_type): Likewise.
5907 2021-10-19 Clément Chigot <clement.chigot@atos.net>
5909 * config/rs6000/rs6000.c (rs6000_xcoff_file_end): Move
5910 __tls_get_addr reference to .text csect.
5912 2021-10-19 Martin Liska <mliska@suse.cz>
5915 * config/aarch64/aarch64.c (aarch64_process_one_target_attr):
5918 2021-10-19 Richard Biener <rguenther@suse.de>
5920 * tree-vectorizer.h (vect_get_store_cost): Adjust signature.
5921 (vect_get_load_cost): Likewise.
5922 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
5923 alignment support scheme and misalignment as arguments
5925 (vect_get_peeling_costs_all_drs): Compute that info here
5926 and note that we shouldn't need to.
5927 * tree-vect-stmts.c (vect_model_store_cost): Get
5928 alignment support scheme and misalignment as arguments.
5929 (vect_get_store_cost): Likewise.
5930 (vect_model_load_cost): Likewise.
5931 (vect_get_load_cost): Likewise.
5932 (vectorizable_store): Pass down alignment support scheme
5933 and misalignment to costing.
5934 (vectorizable_load): Likewise.
5936 2021-10-19 Richard Biener <rguenther@suse.de>
5938 * tree-vect-stmts.c (get_negative_load_store_type): Add
5939 offset output parameter and initialize it.
5940 (get_group_load_store_type): Likewise.
5941 (get_load_store_type): Likewise.
5942 (vectorizable_store): Use offset as computed by
5943 get_load_store_type.
5944 (vectorizable_load): Likewise.
5946 2021-10-19 Richard Biener <rguenther@suse.de>
5948 PR tree-optimization/102827
5949 * tree-if-conv.c (predicate_statements): Add pe parameter
5950 and use that edge to insert invariant stmts on.
5951 (combine_blocks): Pass through pe.
5952 (tree_if_conversion): Compute the edge to insert invariant
5953 stmts on and pass it along.
5955 2021-10-19 Roger Sayle <roger@nextmovesoftware.com>
5958 * config/bfin/bfin.md (addsubv2hi3, subaddv2hi3, ssaddsubv2hi3,
5959 sssubaddv2hi3): Swap the order of operators in vec_concat.
5961 2021-10-19 Xionghu Luo <luoxhu@linux.ibm.com>
5963 * config/rs6000/altivec.md (*altivec_vmrghb_internal): Delete.
5964 (altivec_vmrghb_direct): New.
5965 (*altivec_vmrghh_internal): Delete.
5966 (altivec_vmrghh_direct): New.
5967 (*altivec_vmrghw_internal): Delete.
5968 (altivec_vmrghw_direct_<mode>): New.
5969 (altivec_vmrghw_direct): Delete.
5970 (*altivec_vmrglb_internal): Delete.
5971 (altivec_vmrglb_direct): New.
5972 (*altivec_vmrglh_internal): Delete.
5973 (altivec_vmrglh_direct): New.
5974 (*altivec_vmrglw_internal): Delete.
5975 (altivec_vmrglw_direct_<mode>): New.
5976 (altivec_vmrglw_direct): Delete.
5977 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
5978 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const):
5980 * config/rs6000/vsx.md (vsx_xxmrghw_<mode>): Adjust.
5981 (vsx_xxmrglw_<mode>): Adjust.
5983 2021-10-19 Aldy Hernandez <aldyh@redhat.com>
5985 * passes.def: Change threading comment before pass_ccp pass.
5987 2021-10-19 Haochen Gui <guihaoc@gcc.gnu.org>
5989 * config/rs6000/rs6000-call.c (altivec_expand_lxvr_builtin):
5990 Modify the expansion for sign extension. All extensions are done
5991 within VSX registers.
5993 2021-10-19 Richard Biener <rguenther@suse.de>
5995 * tree-vect-stmts.c (get_group_load_store_type): Add
5996 misalignment output parameter and initialize it.
5997 (get_group_load_store_type): Likewise.
5998 (vectorizable_store): Remove now redundant queries.
5999 (vectorizable_load): Likewise.
6001 2021-10-19 Richard Biener <rguenther@suse.de>
6003 * tree-vectorizer.h (vect_supportable_dr_alignment): Remove
6004 check_aligned argument.
6005 * tree-vect-data-refs.c (vect_supportable_dr_alignment):
6007 (vect_peeling_hash_insert): Add supportable_if_not_aligned
6008 argument and do not call vect_supportable_dr_alignment here.
6009 (vect_peeling_supportable): Adjust.
6010 (vect_enhance_data_refs_alignment): Compute whether the
6011 access is supported with different alignment here and
6012 pass that down to vect_peeling_hash_insert.
6013 (vect_vfa_access_size): Adjust.
6014 * tree-vect-stmts.c (vect_get_store_cost): Likewise.
6015 (vect_get_load_cost): Likewise.
6016 (get_negative_load_store_type): Likewise.
6017 (get_group_load_store_type): Likewise.
6018 (get_load_store_type): Likewise.
6020 2021-10-19 Martin Liska <mliska@suse.cz>
6023 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p): Strip whitespaces.
6024 * system.h (strip_whilespaces): New function.
6026 2021-10-19 dianhong xu <dianhong.xu@intel.com>
6028 * config/i386/avx512fp16intrin.h:
6029 (_mm512_set1_pch): New intrinsic.
6030 * config/i386/avx512fp16vlintrin.h:
6031 (_mm256_set1_pch): New intrinsic.
6032 (_mm_set1_pch): Ditto.
6034 2021-10-18 Andrew MacLeod <amacleod@redhat.com>
6036 PR tree-optimization/102796
6037 * gimple-range.cc (gimple_ranger::range_on_edge): Process EH edges
6038 normally. Return get_tree_range for non gimple_range_ssa_p names.
6039 (gimple_ranger::range_of_stmt): Use get_tree_range for non
6040 gimple_range_ssa_p names.
6042 2021-10-18 Uroš Bizjak <ubizjak@gmail.com>
6045 * config/i386/i386.c (ix86_print_operand_address):
6046 Error out for non-address_operand asm operands.
6048 2021-10-18 Richard Biener <rguenther@suse.de>
6050 * tree-vect-data-refs.c (vect_peeling_hash_insert): Do
6051 not auto-convert dr_alignment_support to bool.
6052 (vect_peeling_supportable): Likewise.
6053 (vect_enhance_data_refs_alignment): Likewise.
6054 (vect_supportable_dr_alignment): Commonize read/write case.
6055 * tree-vect-stmts.c (vect_get_store_cost): Use
6056 dr_alignment_support, not int, for the vect_supportable_dr_alignment
6058 (vect_get_load_cost): Likewise.
6060 2021-10-18 Siddhesh Poyarekar <siddhesh@gotplt.org>
6062 * tree-object-size.c (object_sizes_execute): Consolidate LHS
6063 null check and do it early.
6065 2021-10-18 Richard Biener <rguenther@suse.de>
6067 * tree-vect-stmts.c (vectorizable_store): Use the
6068 computed alignment scheme instead of querying
6071 2021-10-18 Richard Biener <rguenther@suse.de>
6073 * tree-vect-stmts.c (vectorizable_store): Do not recompute
6074 alignment scheme already determined by get_load_store_type.
6076 2021-10-18 Aldy Hernandez <aldyh@redhat.com>
6078 * tree-ssa-threadbackward.c (class pass_thread_jumps_full):
6079 Clone corresponding pass.
6081 2021-10-18 Roger Sayle <roger@nextmovesoftware.com>
6083 * combine.c (recog_for_combine): For an unrecognized move/set of
6084 a constant, try force_const_mem to place it in the constant pool.
6085 * cse.c (constant_pool_entries_cost, constant_pool_entries_regcost):
6086 Delete global variables (that are no longer assigned a cost value).
6087 (cse_insn): Simplify logic for deciding whether to place a folded
6088 constant in the constant pool using force_const_mem.
6089 (cse_main): Remove zero initialization of constant_pool_entries_cost
6090 and constant_pool_entries_regcost.
6091 * config/i386/i386.c (ix86_rtx_costs): Make memory accesses
6092 fractionally more expensive, when optimizing for speed.
6094 2021-10-18 Martin Liska <mliska@suse.cz>
6096 PR gcov-profile/102746
6097 PR gcov-profile/102747
6098 * gcov.c (main): Return return_code.
6099 (output_gcov_file): Mark return_code when error happens.
6100 (generate_results): Likewise.
6101 (read_graph_file): Likewise.
6102 (read_count_file): Likewise.
6104 2021-10-18 Roger Sayle <roger@nextmovesoftware.com>
6106 * config/bfin/bfin.md (define_constants): Remove UNSPEC_ONES.
6107 (define_insn "ones"): Replace UNSPEC_ONES with a truncate of
6108 a popcount, allowing compile-time evaluation/simplification.
6109 (popcountsi2, popcounthi2): New expanders using a "ones" insn.
6111 2021-10-18 Richard Biener <rguenther@suse.de>
6113 PR tree-optimization/102788
6114 * tree-vect-patterns.c (vect_init_pattern_stmt): Allow
6116 (vect_pattern_recog_1): Likewise.
6117 (vect_recog_bool_pattern): Continue matching the pattern
6118 even if we do not have a vector type for a conversion
6121 2021-10-18 Roger Sayle <roger@nextmovesoftware.com>
6123 * simplify-rtx.c (simplify_const_unary_operation) [SS_NEG, SS_ABS]:
6124 Evalute SS_NEG and SS_ABS of a constant argument.
6126 2021-10-18 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
6129 * gimple-match-head.c (try_conditional_simplification): Add case for single operand.
6130 * internal-fn.def: Add entry for COND_NEG internal function.
6131 * internal-fn.c (FOR_EACH_CODE_MAPPING): Add entry for
6132 NEGATE_EXPR, COND_NEG mapping.
6133 * optabs.def: Add entry for cond_neg_optab.
6134 * match.pd (UNCOND_UNARY, COND_UNARY): New operator lists.
6135 (vec_cond COND (foo A) B) -> (IFN_COND_FOO COND A B): New pattern.
6136 (vec_cond COND B (foo A)) -> (IFN_COND_FOO ~COND A B): Likewise.
6138 2021-10-18 Aldy Hernandez <aldyh@redhat.com>
6140 * tree-ssa-strlen.c (class strlen_pass): Rename from
6142 (handle_builtin_stxncpy_strncat): Move to strlen_pass.
6143 (handle_assign): Same.
6144 (adjust_last_stmt): Same.
6145 (maybe_warn_overflow): Same.
6146 (handle_builtin_strlen): Same.
6147 (handle_builtin_strchr): Same.
6148 (handle_builtin_strcpy): Same.
6149 (handle_builtin_strncat): Same.
6150 (handle_builtin_stxncpy_strncat): Same.
6151 (handle_builtin_memcpy): Same.
6152 (handle_builtin_strcat): Same.
6153 (handle_alloc_call): Same.
6154 (handle_builtin_memset): Same.
6155 (handle_builtin_memcmp): Same.
6156 (get_len_or_size): Same.
6157 (strxcmp_eqz_result): Same.
6158 (handle_builtin_string_cmp): Same.
6159 (handle_pointer_plus): Same.
6160 (count_nonzero_bytes_addr): Same.
6161 (count_nonzero_bytes): Same.
6162 (handle_store): Same.
6163 (strlen_check_and_optimize_call): Same.
6164 (handle_integral_assign): Same.
6165 (check_and_optimize_stmt): Same.
6166 (printf_strlen_execute): Rename strlen_dom_walker to strlen_pass.
6168 2021-10-18 Richard Biener <rguenther@suse.de>
6170 PR tree-optimization/102798
6171 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref):
6172 Only copy points-to info to newly generated SSA names.
6174 2021-10-18 Martin Liska <mliska@suse.cz>
6176 * dbgcnt.c (dbg_cnt_process_opt): Remove unused but set variable.
6177 * gcov.c (get_cycles_count): Likewise.
6178 * lto-compress.c (lto_compression_zlib): Likewise.
6179 (lto_uncompression_zlib): Likewise.
6180 * targhooks.c (default_pch_valid_p): Likewise.
6182 2021-10-17 Aldy Hernandez <aldyh@redhat.com>
6184 * tree-pass.h (make_pass_thread_jumps_full): New.
6185 * tree-ssa-threadbackward.c (pass_thread_jumps::gate): Inline.
6186 (try_thread_blocks): Add resolve and speed arguments.
6187 (pass_thread_jumps::execute): Inline.
6188 (do_early_thread_jumps): New.
6189 (do_thread_jumps): New.
6190 (make_pass_thread_jumps): Move.
6191 (pass_early_thread_jumps::gate): Inline.
6192 (pass_early_thread_jumps::execute): Inline.
6193 (class pass_thread_jumps_full): New.
6195 2021-10-16 Piotr Kubaj <pkubaj@FreeBSD.org>
6197 * configure.ac: Treat powerpc64*-*-freebsd* the same as
6198 powerpc64-*-freebsd*.
6199 * configure: Regenerate.
6201 2021-10-16 H.J. Lu <hjl.tools@gmail.com>
6203 * value-query.cc (get_ssa_name_ptr_info_nonnull): Change
6204 set_ptr_nonull to set_ptr_nonnull in comments.
6206 2021-10-16 Jan Hubicka <hubicka@ucw.cz>
6208 PR tree-optimization/102720
6209 * tree-ssa-structalias.c (compute_points_to_sets): Fix producing
6210 of call used and clobbered sets.
6212 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
6214 * config/darwin.h (LINK_COMMAND_SPEC_A): Update 'r' handling to
6215 skip gomp and itm when r or nodefaultlibs is given.
6216 (DSYMUTIL_SPEC): Do not call dsymutil for '-r' link lines.
6217 Update ordering of exclusions, remove duplicate 'v' addition
6218 (collect2 will add this from the main command line).
6220 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
6222 * config/darwin-driver.c (darwin_driver_init): Revise comments, handle
6223 filelist and framework options in specs instead of code.
6224 * config/darwin.h (SUBTARGET_DRIVER_SELF_SPECS): Update to handle link
6225 specs that are really driver ones.
6226 (DARWIN_CC1_SPEC): Likewise.
6227 (CPP_SPEC): Likewise.
6228 (SYSROOT_SPEC): Append space.
6229 (LINK_SYSROOT_SPEC): Remove most driver link specs.
6230 (STANDARD_STARTFILE_PREFIX_2): Update link-related specs.
6231 (STARTFILE_SPEC): Likewise.
6232 (ASM_MMACOSX_VERSION_MIN_SPEC): Fix line wrap.
6233 (ASM_SPEC): Update driver-related specs.
6234 (ASM_FINAL_SPEC): Likewise.
6235 * config/darwin.opt: Remove now unused option aliases.
6236 * config/i386/darwin.h (EXTRA_ASM_OPTS): Ensure space after opt.
6237 (ASM_SPEC): Update driver-related specs.
6239 2021-10-15 Roger Sayle <roger@nextmovesoftware.com>
6241 * config/i386/i386.c (ix86_hardreg_mov_ok): For vector modes,
6242 allow standard_sse_constant_p immediate constants.
6244 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
6246 * config.gcc: Add tm-dwarf2.h to tm_d-file.
6248 2021-10-15 Andrew MacLeod <amacleod@redhat.com>
6250 * gimple-range-fold.h (gimple_range_ssa_p): Don't process names
6251 that occur in abnormal phis.
6252 * gimple-range.cc (gimple_ranger::range_on_edge): Return false for
6253 abnormal and EH edges.
6254 * gimple-ssa-evrp.c (rvrp_folder::value_of_expr): Ditto.
6255 (rvrp_folder::value_on_edge): Ditto.
6256 (rvrp_folder::value_of_stmt): Ditto.
6257 (hybrid_folder::value_of_expr): Ditto for ranger queries.
6258 (hybrid_folder::value_on_edge): Ditto.
6259 (hybrid_folder::value_of_stmt): Ditto.
6260 * value-query.cc (gimple_range_global): Always return a range if
6261 the type is supported.
6263 2021-10-15 John David Anglin <danglin@gcc.gnu.org>
6265 * config/pa/pa.md: Consistently use "rG" constraint for copy
6266 instruction in move patterns.
6268 2021-10-15 Andrew Stubbs <ams@codesourcery.com>
6270 * config/gcn/gcn-hsa.h (S_FIJI): Set unconditionally.
6273 * config/gcn/gcn.c: Hard code SRAM ECC settings for old architectures.
6274 * config/gcn/mkoffload.c (ELFABIVERSION_AMDGPU_HSA): Rename to ...
6275 (ELFABIVERSION_AMDGPU_HSA_V3): ... this.
6276 (ELFABIVERSION_AMDGPU_HSA_V4): New.
6277 (SET_SRAM_ECC_UNSUPPORTED): New.
6278 (copy_early_debug_info): Create elf flags to match the other objects.
6279 (main): Just let the attribute flags pass through.
6281 2021-10-15 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
6283 * tree-loop-distribution.c (reduction_var_overflows_first):
6284 Pass the type of reduction_var as first argument as it is also
6285 done for the load type.
6286 (loop_distribution::transform_reduction_loop): Add missing
6287 TREE_TYPE while determining precission of reduction_var.
6289 2021-10-15 Richard Biener <rguenther@suse.de>
6291 * defaults.h (PREFERRED_DEBUGGING_TYPE): Choose DWARF2_DEBUG
6293 * toplev.c (process_options): Warn when STABS debugging is
6294 enabled but not the preferred format.
6295 * config/pa/som.h (PREFERRED_DEBUGGING_TYPE): Define to
6297 * config/pdp11/pdp11.h (PREFERRED_DEBUGGING_TYPE): Likewise.
6299 2021-10-15 Richard Biener <rguenther@suse.de>
6302 * tree-inline.c (copy_bb): Avoid underflowing nargs.
6304 2021-10-15 Hongyu Wang <hongyu.wang@intel.com>
6306 * config/i386/i386-expand.c (ix86_expand_vec_perm): Convert
6307 HFmode input operand to HImode.
6308 (ix86_vectorize_vec_perm_const): Likewise.
6309 * config/i386/sse.md (*avx512bw_permvar_truncv16siv16hi_1_hf):
6311 (*avx512f_permvar_truncv8siv8hi_1_hf):
6314 2021-10-15 Richard Biener <rguenther@suse.de>
6316 PR middle-end/102682
6317 * expmed.c (store_bit_field_1): Ensure a LHS subreg would
6318 not create a paradoxical subreg.
6320 2021-10-15 Hongyu Wang <hongyu.wang@intel.com>
6322 * config/i386/i386-expand.c (ix86_expand_vector_init):
6323 For half_vector concat for HFmode, handle them like HImode.
6325 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
6327 * tree-ssa-threadbackward.c (class back_threader): Add m_resolve.
6328 (back_threader::back_threader): Same.
6329 (back_threader::resolve_phi): Try to solve without looking back if
6331 (back_threader::find_paths_to_names): Same.
6332 (try_thread_blocks): Pass resolve argument to back threader.
6333 (pass_early_thread_jumps::execute): Same.
6335 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
6337 * doc/invoke.texi: Remove max-fsm-thread-length,
6338 max-fsm-thread-paths, and fsm-maximum-phi-arguments.
6340 * tree-ssa-threadbackward.c (back_threader::back_threader): Remove
6342 (back_threader_registry::back_threader_registry): Same.
6343 (back_threader_profitability::profitable_path_p): Remove
6344 param_max_fsm_thread-length.
6345 (back_threader_registry::register_path): Remove
6346 m_max_allowable_paths.
6348 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
6350 * tree-ssa-threadbackward.c (class back_threader): Make m_imports
6352 (back_threader::~back_threader): Do not release m_path.
6354 2021-10-14 Andrew MacLeod <amacleod@redhat.com>
6356 PR tree-optimization/102738
6357 * vr-values.c (simplify_using_ranges::simplify): Handle RSHIFT_EXPR.
6359 2021-10-14 Kwok Cheung Yeung <kcy@codesourcery.com>
6361 * omp-general.c (omp_check_context_selector): Move from c-omp.c.
6362 (omp_mark_declare_variant): Move from c-omp.c.
6363 (omp_context_name_list_prop): Update for Fortran strings.
6364 * omp-general.h (omp_check_context_selector): New prototype.
6365 (omp_mark_declare_variant): New prototype.
6367 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
6369 * config/rs6000/rs6000.c (rs6000_density_test): Move early
6370 exit test further up the function.
6372 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
6374 * config/arm/arm.c (arm_add_stmt_cost): Delete.
6375 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
6377 2021-10-14 Martin Jambor <mjambor@suse.cz>
6379 * doc/invoke.texi (Optimize Options): Add entry for
6380 ipa-cp-recursive-freq-factor.
6382 2021-10-14 Tamar Christina <tamar.christina@arm.com>
6384 * match.pd: New rule.
6386 2021-10-14 Jan Hubicka <hubicka@ucw.cz>
6389 * ipa-modref-tree.h (modref_access_node::update2):
6390 Also check that parm_offset is unchanged.
6391 (modref_ref_node::insert_access): Fix updating of
6394 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
6396 * tree-ssa-threadbackward.c (back_threader::resolve_phi): Add
6399 2021-10-14 Richard Biener <rguenther@suse.de>
6401 PR tree-optimization/102659
6402 * tree-if-conv.c (if_convertible_gimple_assign_stmt_p): Also
6403 rewrite pointer typed undefined overflow operations.
6404 (predicate_statements): Likewise. Make sure to emit invariant
6405 conversions in the preheader.
6406 * tree-vectorizer.c (vect_loop_vectorized_call): Look through
6407 non-empty preheaders.
6408 * tree-data-ref.c (dr_analyze_indices): Strip useless
6409 conversions to the MEM_REF base type.
6411 2021-10-14 Martin Liska <mliska@suse.cz>
6413 * common.opt: Stop using AUTODETECT_VALUE
6414 and use EnabledBy where possible.
6415 * opts.c: Enable OPT_fvar_tracking with optimize >= 1.
6416 * toplev.c (AUTODETECT_VALUE): Remove macro.
6417 (process_options): Simplify by using EnabledBy and
6418 OPT_fvar_tracking. Use OPTION_SET_P macro instead of
6421 2021-10-14 Jonathan Wright <jonathan.wright@arm.com>
6423 * config/aarch64/arm_neon.h (vld1_s8_x3): Use signed type for
6425 (vld1_s32_x3): Likewise.
6427 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
6429 PR tree-optimization/102736
6430 PR tree-optimization/102736
6431 * gimple-range-path.cc (path_range_query::range_on_path_entry):
6432 Assert that the requested range is defined outside the path.
6433 (path_range_query::ssa_range_in_phi): Do not call
6434 range_on_path_entry for SSA names that are defined within the
6437 2021-10-14 Iain Sandoe <iain@sandoe.co.uk>
6439 * config/darwin-driver.c (darwin_find_version_from_kernel):
6440 Quote internal identifiers and avoid contractions in
6442 (darwin_default_min_version): Likewise.
6443 (darwin_driver_init): Likewise.
6445 2021-10-14 Martin Jambor <mjambor@suse.cz>
6447 * params.opt (ipa-cp-recursive-freq-factor): New.
6448 * ipa-cp.c (ipcp_value): Switch to inline initialization. New members
6449 scc_no, self_recursion_generated_level, same_scc and
6450 self_recursion_generated_p.
6451 (ipcp_lattice::add_value): Replaced parameter unlimited with
6452 same_lat_gen_level, usit it determine limit of values and store it to
6454 (ipcp_lattice<valtype>::print): Dump the new fileds.
6455 (allocate_and_init_ipcp_value): Take same_lat_gen_level as a new
6456 parameter and store it to the new value.
6457 (self_recursively_generated_p): Removed.
6458 (propagate_vals_across_arith_jfunc): Use self_recursion_generated_p
6459 instead of self_recursively_generated_p, store self generation level
6461 (value_topo_info<valtype>::add_val): Set scc_no.
6462 (value_topo_info<valtype>::propagate_effects): Multiply frequencies of
6463 recursively feeding values and self generated values by appropriate
6466 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
6468 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Remove
6469 redundant test for flag_vect_cost_model.
6471 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
6473 * bitmap.c (debug): New overloaded function for auto_bitmaps.
6474 * bitmap.h (debug): Same.
6476 2021-10-14 Hongyu Wang <hongyu.wang@intel.com>
6478 * config/i386/avx512fp16intrin.h (_mm512_mask_fcmadd_pch):
6479 Adjust builtin call.
6480 (_mm512_mask3_fcmadd_pch): Likewise.
6481 (_mm512_mask_fmadd_pch): Likewise
6482 (_mm512_mask3_fmadd_pch): Likewise
6483 (_mm512_mask_fcmadd_round_pch): Likewise
6484 (_mm512_mask3_fcmadd_round_pch): Likewise
6485 (_mm512_mask_fmadd_round_pch): Likewise
6486 (_mm512_mask3_fmadd_round_pch): Likewise
6487 (_mm_mask_fcmadd_sch): Likewise
6488 (_mm_mask3_fcmadd_sch): Likewise
6489 (_mm_mask_fmadd_sch): Likewise
6490 (_mm_mask3_fmadd_sch): Likewise
6491 (_mm_mask_fcmadd_round_sch): Likewise
6492 (_mm_mask3_fcmadd_round_sch): Likewise
6493 (_mm_mask_fmadd_round_sch): Likewise
6494 (_mm_mask3_fmadd_round_sch): Likewise
6495 (_mm_fcmadd_round_sch): Likewise
6496 * config/i386/avx512fp16vlintrin.h (_mm_mask_fmadd_pch):
6497 Adjust builtin call.
6498 (_mm_mask3_fmadd_pch): Likewise
6499 (_mm256_mask_fmadd_pch): Likewise
6500 (_mm256_mask3_fmadd_pch): Likewise
6501 (_mm_mask_fcmadd_pch): Likewise
6502 (_mm_mask3_fcmadd_pch): Likewise
6503 (_mm256_mask_fcmadd_pch): Likewise
6504 (_mm256_mask3_fcmadd_pch): Likewise
6505 * config/i386/i386-builtin.def: Add mask3 builtin for complex
6506 fma, and adjust mask_builtin to corresponding expander.
6507 * config/i386/i386-expand.c (ix86_expand_round_builtin):
6508 Skip eraseing embedded rounding for expanders that emits
6510 * config/i386/sse.md (complexmove): New mode_attr.
6511 (<avx512>_fmaddc_<mode>_mask1<round_expand_name>): New expander.
6512 (<avx512>_fcmaddc_<mode>_mask1<round_expand_name>): Likewise.
6513 (avx512fp16_fmaddcsh_v8hf_mask1<round_expand_name>): Likewise.
6514 (avx512fp16_fcmaddcsh_v8hf_mask1<round_expand_name>): Likewise.
6515 (avx512fp16_fcmaddcsh_v8hf_mask3<round_expand_name>): Likewise.
6516 (avx512fp16_fmaddcsh_v8hf_mask3<round_expand_name>): Likewise.
6517 * config/i386/subst.md (round_embedded_complex): New subst.
6519 2021-10-14 John David Anglin <danglin@gcc.gnu.org>
6521 * config/pa/pa.md (cbranchsf4): Disable if TARGET_SOFT_FLOAT.
6522 (cbranchdf4): Likewise.
6523 Add missing move patterns for TARGET_SOFT_FLOAT.
6525 2021-10-13 Roger Sayle <roger@nextmovesoftware.com>
6527 * config/i386/i386-expand.c (ix86_expand_vector_move): Use a
6528 pseudo intermediate when moving a SUBREG into a hard register,
6529 by checking ix86_hardreg_mov_ok.
6530 (ix86_expand_vector_extract): Store zero-extended SImode
6531 intermediate in a pseudo, then set target using a SUBREG_PROMOTED
6533 * config/i386/sse.md (mov<VMOVE>_internal): Prevent CSE creating
6534 complex (SUBREG) sets of (vector) hard registers before reload, by
6535 checking ix86_hardreg_mov_ok.
6537 2021-10-13 Indu Bhagat <indu.bhagat@oracle.com>
6539 * ctfc.h (enum ctf_dtu_d_union_enum): Remove redundant comma.
6541 2021-10-13 Indu Bhagat <indu.bhagat@oracle.com>
6543 * dwarf2ctf.c (gen_ctf_array_type): Fix typo in comment.
6545 2021-10-13 Martin Sebor <msebor@redhat.com>
6547 PR middle-end/102630
6548 * pointer-query.cc (compute_objsize_r): Handle named address spaces.
6550 2021-10-13 Iain Sandoe <iain@sandoe.co.uk>
6552 * collect2.c (is_lto_object_file): Release simple-object
6553 resources, close files.
6555 2021-10-13 Andre Vieira <andre.simoesdiasvieira@arm.com>
6557 * config/arm/arm.c (thumb2_legitimate_address_p): Use VALID_MVE_MODE
6558 when checking mve addressing modes.
6559 (mve_vector_mem_operand): Fix the way we handle pre, post and offset
6561 (arm_print_operand): Fix printing of POST_ and PRE_MODIFY.
6562 * config/arm/mve.md: Use mve_memory_operand predicate everywhere where
6563 there is a single Ux constraint.
6565 2021-10-13 John David Anglin <danglin@gcc.gnu.org>
6567 * config/pa/pa.md (muldi3): Add support for inlining 64-bit
6568 multiplication on 32-bit PA 1.1 and 2.0 targets.
6570 2021-10-13 Martin Liska <mliska@suse.cz>
6572 PR gcov-profile/90364
6573 * coverage.c (build_info): Emit checksum to the global variable.
6574 (build_info_type): Add new field for checksum.
6575 (coverage_obj_finish): Pass object_checksum.
6576 (coverage_init): Use 0 as checksum for .gcno files.
6577 * gcov-dump.c (dump_gcov_file): Dump also new checksum field.
6578 * gcov.c (read_graph_file): Read also checksum.
6579 * doc/invoke.texi: Document the behaviour change.
6581 2021-10-13 Richard Biener <rguenther@suse.de>
6583 * gimple-iterator.h (gsi_iterator_update): Add GSI_LAST_NEW_STMT,
6584 start at integer value 2.
6585 * gimple-iterator.c (gsi_insert_seq_nodes_before): Update
6586 the iterator for GSI_LAST_NEW_STMT.
6587 (gsi_insert_seq_nodes_after): Likewise.
6588 * tree-if-conv.c (predicate_statements): Use GSI_LAST_NEW_STMT.
6589 * tree-ssa.c (execute_update_addresses_taken): Correct bogus
6590 arguments to gsi_replace.
6592 2021-10-13 Martin Liska <mliska@suse.cz>
6595 * common.opt: Use EnabledBy instead of detection in
6596 finish_options and process_options.
6597 * opts.c (finish_options): Remove handling of
6598 x_flag_unroll_all_loops.
6599 * toplev.c (process_options): Likewise for flag_web and
6600 flag_rename_registers.
6602 2021-10-13 Richard Biener <rguenther@suse.de>
6604 PR tree-optimization/102659
6605 * tree-if-conv.c (need_to_rewrite_undefined): New flag.
6606 (if_convertible_gimple_assign_stmt_p): Mark the loop for
6607 rewrite when stmts with undefined behavior on integer
6609 (combine_blocks): Predicate also when we need to rewrite stmts.
6610 (predicate_statements): Rewrite affected stmts to something
6611 with well-defined behavior on overflow.
6612 (tree_if_conversion): Initialize need_to_rewrite_undefined.
6614 2021-10-13 Richard Biener <rguenther@suse.de>
6617 * ipa-sra.c (ptr_parm_has_nonarg_uses): Fix volatileness
6620 2021-10-13 Jose E. Marchesi <jose.marchesi@oracle.com>
6622 * dwarf2ctf.c: Fix typo in comment.
6624 2021-10-12 Bill Schmidt <wschmidt@linux.ibm.com>
6627 * config/rs6000/altivec.h (vec_cpsgn): Swap operand order.
6628 * config/rs6000/rs6000-overload.def (VEC_COPYSIGN): Use SKIP to
6629 avoid generating an automatic #define of vec_cpsgn. Use the
6630 correct built-in for V4SFmode that doesn't depend on VSX.
6632 2021-10-12 Uroš Bizjak <ubizjak@gmail.com>
6636 * config/i386/i386.md (*add<mode>_1_slp): Rewrite as
6637 define_insn_and_split pattern. Add alternative 1 and split it
6638 post reload to insert operand 1 into the low part of operand 0.
6639 (*sub<mode>_1_slp): Ditto.
6640 (*and<mode>_1_slp): Ditto.
6641 (*<any_or:code><mode>_1_slp): Ditto.
6642 (*ashl<mode>3_1_slp): Ditto.
6643 (*<any_shiftrt:insn><mode>3_1_slp): Ditto.
6644 (*<any_rotate:insn><mode>3_1_slp): Ditto.
6645 (*neg<mode>_1_slp): New insn_and_split pattern.
6646 (*one_cmpl<mode>_1_slp): Ditto.
6648 2021-10-12 David Edelsohn <dje.gcc@gmail.com>
6650 * doc/install.texi: Update MinGW and mingw-64 Binaries
6653 2021-10-12 Daniel Le Duc Khoi Nguyen <greenrecyclebin@gmail.com>
6655 * doc/extend.texi (Common Variable Attributes): Fix typos in
6656 alloc_size documentation.
6658 2021-10-12 Richard Biener <rguenther@suse.de>
6660 PR tree-optimization/102696
6661 * tree-vect-slp.c (vect_build_slp_tree_2): Properly mark
6662 the tree fatally failed when we reject a BIT_FIELD_REF.
6664 2021-10-12 Richard Biener <rguenther@suse.de>
6666 PR tree-optimization/102572
6667 * tree-vect-stmts.c (vect_build_gather_load_calls): When
6668 gathering the vectorized defs for the mask pass in the
6669 desired mask vector type so invariants will be handled
6672 2021-10-12 Tamar Christina <tamar.christina@arm.com>
6674 * config/aarch64/aarch64-sve.md (*fcm<cmp_op><mode>_bic_combine,
6675 *fcm<cmp_op><mode>_nor_combine, *fcmuo<mode>_bic_combine,
6676 *fcmuo<mode>_nor_combine): New.
6678 2021-10-12 Eric Botcazou <ebotcazou@adacore.com>
6681 * config/sparc/sparc-modes.def (OI): New integer mode.
6683 2021-10-12 Jakub Jelinek <jakub@redhat.com>
6685 * gimple-fold.h (clear_padding_type_may_have_padding_p): Declare.
6686 * gimple-fold.c (clear_padding_type_may_have_padding_p): No longer
6689 2021-10-12 Jakub Jelinek <jakub@redhat.com>
6691 * tree-vectorizer.h (loop_cost_model): New function.
6692 (unlimited_cost_model): Use it.
6693 * tree-vect-loop.c (vect_analyze_loop_costing): Use loop_cost_model
6694 call instead of flag_vect_cost_model.
6695 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
6696 (vect_prune_runtime_alias_test_list): Likewise. Also use it instead
6697 of flag_simd_cost_model.
6699 2021-10-12 liuhongt <hongtao.liu@intel.com>
6702 * config/i386/i386-expand.c (emit_reduc_half): Handle
6704 * config/i386/mmx.md (reduc_<code>_scal_v4qi): New expander.
6705 (reduc_plus_scal_v4qi): Ditto.
6707 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
6709 * config/rs6000/smmintrin.h (_mm_cmpeq_epi64, _mm_cmpgt_epi64,
6710 _mm_mullo_epi32, _mm_mul_epi32, _mm_packus_epi32): New.
6711 * config/rs6000/nmmintrin.h: Copy from i386, tweak to suit.
6713 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
6715 * config/rs6000/smmintrin.h (_mm_cvtepi8_epi16, _mm_cvtepi8_epi32,
6716 _mm_cvtepi8_epi64, _mm_cvtepi16_epi32, _mm_cvtepi16_epi64,
6717 _mm_cvtepi32_epi64, _mm_cvtepu8_epi16, _mm_cvtepu8_epi32,
6718 _mm_cvtepu8_epi64, _mm_cvtepu16_epi32, _mm_cvtepu16_epi64,
6719 _mm_cvtepu32_epi64): New.
6721 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
6723 * config/rs6000/smmintrin.h (_mm_test_all_zeros,
6724 _mm_test_all_ones, _mm_test_mix_ones_zeros): Rewrite as macro.
6726 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
6728 * config/rs6000/smmintrin.h (_mm_min_epi8, _mm_min_epu16,
6729 _mm_min_epi32, _mm_min_epu32, _mm_max_epi8, _mm_max_epu16,
6730 _mm_max_epi32, _mm_max_epu32): New.
6732 2021-10-11 Jan Hubicka <hubicka@ucw.cz>
6734 * ipa-modref-tree.h (struct modref_access_node): Revert
6736 (struct modref_ref_node): Likewise.
6738 2021-10-11 Jan Hubicka <hubicka@ucw.cz>
6740 * ipa-modref-tree.h (modref_tree::global_access_p): New member
6743 (implicint_const_eaf_flags,implicit_pure_eaf_flags,
6744 ignore_stores_eaf_flags): Move to ipa-modref.h
6745 (remove_useless_eaf_flags): Remove early exit on NOCLOBBER.
6746 (modref_summary::global_memory_read_p): New member function.
6747 (modref_summary::global_memory_written_p): New member function.
6748 * ipa-modref.h (modref_summary::global_memory_read_p,
6749 modref_summary::global_memory_written_p): Declare.
6750 (implicint_const_eaf_flags,implicit_pure_eaf_flags,
6751 ignore_stores_eaf_flags): move here.
6752 * tree-ssa-structalias.c: Include ipa-modref-tree.h, ipa-modref.h
6754 (handle_rhs_call): Rewrite.
6755 (handle_call_arg): New function.
6756 (determine_global_memory_access): New function.
6757 (handle_const_call): Remove
6758 (handle_pure_call): Remove
6759 (find_func_aliases_for_call): Update use of handle_rhs_call.
6760 (compute_points_to_sets): Handle global memory acccesses
6763 2021-10-11 Diane Meirowitz <diane.meirowitz@oracle.com>
6765 * doc/invoke.texi: Add link to UndefinedBehaviorSanitizer
6766 documentation, mention UBSAN_OPTIONS, similar to what is done
6767 for AddressSanitizer.
6769 2021-10-11 Richard Biener <rguenther@suse.de>
6771 PR middle-end/102683
6772 * internal-fn.c (expand_DEFERRED_INIT): Check for mode
6773 availability before building an integer type for storage
6776 2021-10-11 Richard Biener <rguenther@suse.de>
6778 PR middle-end/101480
6779 * gimple.c (gimple_call_fnspec): Do not mark operator new/delete
6782 2021-10-11 Martin Liska <mliska@suse.cz>
6784 * common.opt: Remove Init(2) for some options.
6785 * toplev.c (process_options): Do not use AUTODETECT_VALUE, but
6786 use rather OPTION_SET_P.
6788 2021-10-11 Martin Liska <mliska@suse.cz>
6790 * common.opt: Remove usage of IRA_REGION_AUTODETECT.
6791 * flag-types.h (enum ira_region): Likewise.
6792 * toplev.c (process_options): Use OPTION_SET_P instead of
6793 IRA_REGION_AUTODETECT.
6795 2021-10-11 Jakub Jelinek <jakub@redhat.com>
6797 * omp-low.c (omp_runtime_api_call): Handle omp_get_max_teams,
6798 omp_[sg]et_teams_thread_limit and omp_set_num_teams.
6800 2021-10-11 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
6802 * config/s390/s390-protos.h (s390_rawmemchr): Add prototype.
6803 * config/s390/s390.c (s390_rawmemchr): New function.
6804 * config/s390/s390.md (rawmemchr<SINT:mode>): New expander.
6805 * config/s390/vector.md (@vec_vfees<mode>): Basically a copy of
6806 the pattern vfees<mode> from vx-builtins.md.
6807 * config/s390/vx-builtins.md (*vfees<mode>): Remove.
6809 2021-10-11 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
6811 * builtins.c (get_memory_rtx): Change to external linkage.
6812 * builtins.h (get_memory_rtx): Add function prototype.
6813 * doc/md.texi (rawmemchr<mode>): Document.
6814 * internal-fn.c (expand_RAWMEMCHR): Define.
6815 * internal-fn.def (RAWMEMCHR): Add.
6816 * optabs.def (rawmemchr_optab): Add.
6817 * tree-loop-distribution.c (find_single_drs): Change return code
6818 behaviour by also returning true if no single store was found
6820 (loop_distribution::classify_partition): Respect the new return
6821 code behaviour of function find_single_drs.
6822 (loop_distribution::execute): Call new function
6823 transform_reduction_loop in order to replace rawmemchr or strlen
6824 like loops by calls into builtins.
6825 (generate_reduction_builtin_1): New function.
6826 (generate_rawmemchr_builtin): New function.
6827 (generate_strlen_builtin_1): New function.
6828 (generate_strlen_builtin): New function.
6829 (generate_strlen_builtin_using_rawmemchr): New function.
6830 (reduction_var_overflows_first): New function.
6831 (determine_reduction_stmt_1): New function.
6832 (determine_reduction_stmt): New function.
6833 (loop_distribution::transform_reduction_loop): New function.
6835 2021-10-11 Martin Liska <mliska@suse.cz>
6837 * tree.c (cl_option_hasher::hash): Use cl_optimization_hash
6838 and remove legacy hashing code.
6840 2021-10-11 Kito Cheng <kito.cheng@sifive.com>
6843 * builtins.c (maybe_emit_call_builtin___clear_cache): Allow
6844 CONST_INT for BEGIN and END, and use gcc_assert rather than
6847 2021-10-10 Jakub Jelinek <jakub@redhat.com>
6850 * var-tracking.c (add_stores): For cselib_sp_derived_value_p values
6851 use MO_VAL_SET if loc is not sp.
6853 2021-10-10 Andrew Pinski <apinski@marvell.com>
6855 PR tree-optimization/102622
6856 * match.pd: Swap the order of a?pow2cst:0 and a?-1:0 transformations.
6857 Swap the order of a?0:pow2cst and a?0:-1 transformations.
6859 2021-10-09 liuhongt <hongtao.liu@intel.com>
6862 * config/i386/i386-expand.c (ix86_valid_mask_cmp_mode): Handle
6864 (ix86_use_mask_cmp_p): Ditto.
6865 (ix86_expand_sse_movcc): Ditto.
6866 * config/i386/i386.md (setcc_hf_mask): New define_insn.
6867 (movhf_mask): Ditto.
6868 (UNSPEC_MOVCC_MASK): New unspec.
6869 * config/i386/sse.md (UNSPEC_PCMP): Move to i386.md.
6871 2021-10-08 Vladimir N. Makarov <vmakarov@redhat.com>
6873 PR rtl-optimization/102627
6874 * lra-constraints.c (split_reg): Use at least natural mode of hard reg.
6876 2021-10-08 Aldy Hernandez <aldyh@redhat.com>
6878 * gimple-range-cache.cc (non_null_ref::non_null_deref_p): Grow
6881 2021-10-08 Aldy Hernandez <aldyh@redhat.com>
6883 * value-range.cc (irange::debug): New.
6884 * value-range.h (irange::debug): New.
6886 2021-10-08 Richard Sandiford <richard.sandiford@arm.com>
6888 PR tree-optimization/102385
6889 * predict.h (change_edge_frequency): Declare.
6890 * predict.c (change_edge_frequency): New function.
6891 * tree-ssa-loop-manip.h (tree_transform_and_unroll_loop): Remove
6893 (tree_unroll_loop): Likewise.
6894 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Update accordingly.
6895 * tree-predcom.c (pcom_worker::tree_predictive_commoning_loop):
6897 * tree-ssa-loop-prefetch.c (loop_prefetch_arrays): Likewise.
6898 * tree-ssa-loop-manip.c (tree_unroll_loop): Likewise.
6899 (tree_transform_and_unroll_loop): Likewise. Use single_dom_exit
6900 to retrieve the exit edges. Make all the old profile update code
6901 conditional on !single_loop_p -- the case it was written for --
6902 and use a different approach for the single-loop case.
6904 2021-10-08 Martin Liska <mliska@suse.cz>
6906 * config/alpha/alpha.c (alpha_option_override): Use new macro
6908 * config/arc/arc.c (arc_override_options): Likewise.
6909 * config/arm/arm.c (arm_option_override): Likewise.
6910 * config/bfin/bfin.c (bfin_load_pic_reg): Likewise.
6911 * config/c6x/c6x.c (c6x_option_override): Likewise.
6912 * config/csky/csky.c: Likewise.
6913 * config/darwin.c (darwin_override_options): Likewise.
6914 * config/frv/frv.c (frv_option_override): Likewise.
6915 * config/i386/djgpp.h: Likewise.
6916 * config/i386/i386.c (ix86_stack_protect_guard): Likewise.
6917 (ix86_max_noce_ifcvt_seq_cost): Likewise.
6918 * config/ia64/ia64.c (ia64_option_override): Likewise.
6919 (ia64_override_options_after_change): Likewise.
6920 * config/m32c/m32c.c (m32c_option_override): Likewise.
6921 * config/m32r/m32r.c (m32r_init): Likewise.
6922 * config/m68k/m68k.c (m68k_option_override): Likewise.
6923 * config/microblaze/microblaze.c (microblaze_option_override): Likewise.
6924 * config/mips/mips.c (mips_option_override): Likewise.
6925 * config/nios2/nios2.c (nios2_option_override): Likewise.
6926 * config/nvptx/nvptx.c (nvptx_option_override): Likewise.
6927 * config/pa/pa.c (pa_option_override): Likewise.
6928 * config/riscv/riscv.c (riscv_option_override): Likewise.
6929 * config/rs6000/aix71.h: Likewise.
6930 * config/rs6000/aix72.h: Likewise.
6931 * config/rs6000/aix73.h: Likewise.
6932 * config/rs6000/rs6000.c (darwin_rs6000_override_options): Likewise.
6933 (rs6000_override_options_after_change): Likewise.
6934 (rs6000_linux64_override_options): Likewise.
6935 (glibc_supports_ieee_128bit): Likewise.
6936 (rs6000_option_override_internal): Likewise.
6937 (rs6000_file_start): Likewise.
6938 (rs6000_darwin_file_start): Likewise.
6939 * config/rs6000/rtems.h: Likewise.
6940 * config/rs6000/sysv4.h: Likewise.
6941 * config/rs6000/vxworks.h (SUB3TARGET_OVERRIDE_OPTIONS): Likewise.
6942 * config/s390/s390.c (s390_option_override): Likewise.
6943 * config/sh/linux.h: Likewise.
6944 * config/sh/netbsd-elf.h (while): Likewise.
6945 * config/sh/sh.c (sh_option_override): Likewise.
6946 * config/sol2.c (solaris_override_options): Likewise.
6947 * config/sparc/sparc.c (sparc_option_override): Likewise.
6948 * config/tilegx/tilegx.c (tilegx_option_override): Likewise.
6949 * config/visium/visium.c (visium_option_override): Likewise.
6950 * config/vxworks.c (vxworks_override_options): Likewise.
6951 * lto-opts.c (lto_write_options): Likewise.
6952 * omp-expand.c (expand_omp_simd): Likewise.
6953 * omp-general.c (omp_max_vf): Likewise.
6954 * omp-offload.c (oacc_xform_loop): Likewise.
6955 * opts.h (OPTION_SET_P): Likewise.
6956 * targhooks.c (default_max_noce_ifcvt_seq_cost): Likewise.
6957 * toplev.c (process_options): Likewise.
6958 * tree-predcom.c: Likewise.
6959 * tree-sra.c (analyze_all_variable_accesses): Likewise.
6961 2021-10-08 liuhongt <hongtao.liu@intel.com>
6964 * config/i386/i386.c (ix86_optab_supported_p):
6965 Return true for HFmode.
6966 * match.pd: Simplify (_Float16) ceil ((double) x) to
6967 __builtin_ceilf16 (a) when a is _Float16 type and
6968 direct_internal_fn_supported_p.
6970 2021-10-08 liuhongt <hongtao.liu@intel.com>
6973 * config/i386/i386-expand.c (emit_reduc_half): Hanlde V4HImode.
6974 * config/i386/mmx.md (reduc_plus_scal_v4hi): New.
6975 (reduc_<code>_scal_v4hi): New.
6977 2021-10-08 liuhongt <hongtao.liu@intel.com>
6979 * common.opt (ftree-vectorize): Add Var(flag_tree_vectorize).
6980 * doc/invoke.texi (Options That Control Optimization): Update
6982 * opts.c (default_options_table): Enable auto-vectorization at
6983 O2 with very-cheap cost model.
6984 (finish_options): Use cheap cost model for
6985 explicit -ftree{,-loop}-vectorize.
6987 2021-10-07 Indu Bhagat <indu.bhagat@oracle.com>
6989 * ctfc.c (ctfc_delete_container): Free hash table contents.
6991 2021-10-07 Indu Bhagat <indu.bhagat@oracle.com>
6993 * toplev.c (process_options): Do not warn for GNU GIMPLE.
6995 2021-10-07 Siddhesh Poyarekar <siddhesh@gotplt.org>
6997 * tree-object-size.c (addr_object_size,
6998 compute_builtin_object_size): Drop PDECL and POFF arguments.
6999 (addr_object_size): Adjust calls.
7000 * tree-object-size.h (compute_builtin_object_size): Drop PDECL
7003 2021-10-07 Roger Sayle <roger@nextmovesoftware.com>
7005 * rtl.def (SMUL_HIGHPART, UMUL_HIGHPART): New RTX codes for
7006 representing signed and unsigned high-part multiplication resp.
7007 * simplify-rtx.c (simplify_binary_operation_1) [SMUL_HIGHPART,
7008 UMUL_HIGHPART]: Simplify high-part multiplications by zero.
7009 [SS_PLUS, US_PLUS, SS_MINUS, US_MINUS, SS_MULT, US_MULT,
7010 SS_DIV, US_DIV]: Similar simplifications for saturating
7012 (simplify_const_binary_operation) [SS_PLUS, US_PLUS, SS_MINUS,
7013 US_MINUS, SS_MULT, US_MULT, SMUL_HIGHPART, UMUL_HIGHPART]:
7014 Implement compile-time evaluation for constant operands.
7015 * dwarf2out.c (mem_loc_descriptor): Skip SMUL_HIGHPART and
7017 * doc/rtl.texi (smul_highpart, umul_highpart): Document RTX codes.
7018 * doc/md.texi (smul@var{m}3_highpart, umul@var{m3}_highpart):
7019 Mention the new smul_highpart and umul_highpart RTX codes.
7020 * doc/invoke.texi: Silence @xref "compilation" warnings.
7022 2021-10-07 Martin Jambor <mjambor@suse.cz>
7025 * ipa-prop.c (ipa_edge_args_sum_t::duplicate): Also handle the
7026 case when the source reference description corresponds to a
7027 referance taken in a function src->caller is inlined to.
7029 2021-10-07 Jan Hubicka <hubicka@ucw.cz>
7032 * ipa-modref-tree.h (modref_access_node::contains_p): Handle offsets
7034 (modref_access_node::try_merge_with): Add sanity check that there
7035 are no redundant entries in the list.
7037 2021-10-07 Richard Biener <rguenther@suse.de>
7039 PR tree-optimization/102608
7040 * tree-ssa-sccvn.c (visit_stmt): Drop .DEFERRED_INIT to
7043 2021-10-07 Martin Liska <mliska@suse.cz>
7045 * toplev.c (toplev::main): Make
7046 save_opt_decoded_options a pointer type
7047 * toplev.h: Likewise.
7049 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
7051 * config/gcn/gcn-valu.md (gather<mode>_insn_2offsets<exec>): Apply
7052 HAVE_GCN_ASM_GLOBAL_LOAD_FIXED.
7053 (scatter<mode>_insn_2offsets<exec_scatter>): Likewise.
7055 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
7057 * config/gcn/gcn-hsa.h (SRAMOPT): Include the whole option string.
7058 Adjust for new -msram-ecc=any behaviour.
7059 (ASM_SPEC): Adjust -mxnack and -msram-ecc usage.
7060 * config/gcn/gcn.c (output_file_start): Implement -msram-ecc=any.
7061 * config/gcn/mkoffload.c (EF_AMDGPU_XNACK): Rename to ...
7062 (EF_AMDGPU_XNACK_V3): ... this.
7063 (EF_AMDGPU_SRAM_ECC): Rename to ...
7064 (EF_AMDGPU_SRAM_ECC_V3): ... this.
7065 (EF_AMDGPU_FEATURE_XNACK_V4): New.
7066 (EF_AMDGPU_FEATURE_XNACK_UNSUPPORTED_V4): New.
7067 (EF_AMDGPU_FEATURE_XNACK_ANY_V4): New.
7068 (EF_AMDGPU_FEATURE_XNACK_OFF_V4): New.
7069 (EF_AMDGPU_FEATURE_XNACK_ON_V4): New.
7070 (EF_AMDGPU_FEATURE_SRAMECC_V4): New.
7071 (EF_AMDGPU_FEATURE_SRAMECC_UNSUPPORTED_V4): New.
7072 (EF_AMDGPU_FEATURE_SRAMECC_ANY_V4): New.
7073 (EF_AMDGPU_FEATURE_SRAMECC_OFF_V4): New.
7074 (EF_AMDGPU_FEATURE_SRAMECC_ON_V4): New.
7075 (SET_XNACK_ON): New.
7076 (SET_XNACK_OFF): New.
7078 (SET_SRAM_ECC_ON): New.
7079 (SET_SRAM_ECC_ANY): New.
7080 (SET_SRAM_ECC_OFF): New.
7081 (TEST_SRAM_ECC_ANY): New.
7082 (TEST_SRAM_ECC_ON): New.
7083 (main): Implement HSACOv4 and -msram-ecc=any.
7085 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
7087 * config.in: Regenerate.
7088 * config/gcn/gcn-hsa.h (X_FIJI): New macro.
7092 (A_FIJI): Rename to ...
7094 (A_900): Rename to ...
7096 (A_906): Rename to ...
7098 (A_908): Rename to ...
7100 (SRAMOPT): New macro.
7101 (ASM_SPEC): Adjust xnack option usage.
7102 * config/gcn/gcn.c (output_file_start): Adjust amdgcn_target usage.
7103 * configure: Regenerate.
7104 * configure.ac: Detect LLVM assembler dialect.
7106 2021-10-07 Richard Biener <rguenther@suse.de>
7108 * tree-pretty-print.c (dump_generic_node): Do not elide
7109 printing '&' when dumping with -gimple.
7111 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
7113 * gimple-range-cache.cc (non_null_ref::adjust_range): Call new
7115 * gimple-range-fold.cc (adjust_pointer_diff_expr): Ditto.
7116 (adjust_imagpart_expr): Ditto.
7117 * value-range.cc (irange::irange_intersect): Call new routine if
7118 RHS is a single pair.
7119 (irange::intersect): New wide_int version.
7120 * value-range.h (class irange): New prototype.
7122 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
7124 * gimple-range-edge.cc (gimple_outgoing_range::gimple_outgoing_range):
7125 Add parameter to limit size when recognizing switches.
7126 (gimple_outgoing_range::edge_range_p): Check size limit.
7127 * gimple-range-edge.h (gimple_outgoing_range): Add size field.
7128 * gimple-range-gori.cc (gori_map::calculate_gori): Ignore switches
7129 that exceed the size limit.
7130 (gori_compute::gori_compute): Add initializer.
7131 * params.opt (evrp-switch-limit): New.
7132 * doc/invoke.texi: Update docs.
7134 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
7136 * value-range.h (irange::set_varying): Use TYPE_MIN_VALUE and
7137 TYPE_MAX_VALUE instead of creating new trees when possible.
7139 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
7141 * gimple-range-cache.cc (non_null_ref::adjust_range): Check for
7142 zero and non-zero more efficently.
7144 2021-10-06 Richard Biener <rguenther@suse.de>
7147 * dumpfile.h (TDF_GIMPLE_VAL): New.
7148 (dump_flag): Re-order and adjust TDF_* flags. Make
7149 the enum uint32_t. Use std::underlying_type in the
7151 (optgroup_flag): Likewise for the operator overloads.
7152 * tree-pretty-print.c (dump_generic_node): Wrap ADDR_EXPR
7153 in _Literal if TDF_GIMPLE_VAL.
7154 * gimple-pretty-print.c (dump_gimple_assign): Add
7155 TDF_GIMPLE_VAL to flags when dumping operands where only
7156 is_gimple_val are allowed.
7157 (dump_gimple_cond): Likewise.
7159 2021-10-06 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
7161 * gimple-isel.cc (gimple_expand_vec_cond_expr): Remove redundant if
7164 2021-10-05 qing zhao <qing.zhao@oracle.com>
7166 PR middle-end/102359
7167 * gimplify.c (gimplify_decl_expr): Not add initialization for an
7168 auto variable when it has been initialized by frontend.
7170 2021-10-05 Aldy Hernandez <aldyh@redhat.com>
7172 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
7175 2021-10-05 Jan-Benedict Glaw <jbglaw@lug-owl.de>
7177 * common/config/avr/avr-common.c (avr_handle_option): Mark
7178 argument as ATTRIBUTE_UNUSED.
7180 2021-10-05 Jan-Benedict Glaw <jbglaw@lug-owl.de>
7182 * config/lm32/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC):
7183 Undefine before redefinition.
7185 2021-10-05 Richard Biener <rguenther@suse.de>
7187 * toplev.c (no_backend): Remove global var.
7188 (process_options): Pass in no_backend, move post_options
7189 langhook call to toplev::main.
7190 (do_compile): Pass in no_backend, move process_options call
7192 (toplev::run_self_tests): Check no_backend at the caller.
7193 (toplev::main): Call post_options and process_options
7194 split out from do_compile, do self-tests only if
7195 no_backend is initialized.
7197 2021-10-05 Richard Biener <rguenther@suse.de>
7199 * tree-cfg.c (dump_function_to_file): Dump the UID of the
7200 function as part of the name when requested.
7201 * tree-pretty-print.c (dump_function_name): Dump the UID when
7202 requested and the langhook produced the actual name.
7204 2021-10-05 Richard Biener <rguenther@suse.de>
7206 PR middle-end/102587
7207 PR middle-end/102285
7208 * internal-fn.c (expand_DEFERRED_INIT): Fall back to
7209 zero-initialization as last resort, use the constant
7210 size as given by the DEFERRED_INIT argument to build
7213 2021-10-04 Marek Polacek <polacek@redhat.com>
7216 * doc/invoke.texi: Document -Warray-compare.
7218 2021-10-04 Richard Biener <rguenther@suse.de>
7220 * gimplify.c (is_var_need_auto_init): DECL_HARD_REGISTER
7221 variables are not to be initialized.
7223 2021-10-04 Richard Biener <rguenther@suse.de>
7225 * expr.h (non_mem_decl_p): Declare.
7226 (mem_ref_refers_to_non_mem_p): Likewise.
7227 * expr.c (non_mem_decl_p): Export.
7228 (mem_ref_refers_to_non_mem_p): Likewise.
7229 * internal-fn.c (expand_DEFERRED_INIT): Do not expand the LHS
7230 but check the base with mem_ref_refers_to_non_mem_p
7233 2021-10-04 Richard Biener <rguenther@suse.de>
7235 PR tree-optimization/102570
7236 * tree-ssa-sccvn.h (vn_reference_op_struct): Document
7237 we are using clique for the internal function code.
7238 * tree-ssa-sccvn.c (vn_reference_op_eq): Compare the
7239 internal function code.
7240 (print_vn_reference_ops): Print the internal function code.
7241 (vn_reference_op_compute_hash): Hash it.
7242 (copy_reference_ops_from_call): Record it.
7243 (visit_stmt): Remove the restriction around internal function
7245 (fully_constant_vn_reference_p): Use fold_const_call and handle
7247 (vn_reference_eq): Compare call return types.
7248 * tree-ssa-pre.c (create_expression_by_pieces): Handle
7249 generating calls to internal functions.
7250 (compute_avail): Remove the restriction around internal function
7253 2021-10-04 Aldy Hernandez <aldyh@redhat.com>
7255 PR tree-optimization/102560
7256 * gimple-ssa-warn-alloca.c (alloca_call_type): Remove static
7257 marker for invalid_range.
7259 2021-10-04 Richard Biener <rguenther@suse.de>
7261 PR middle-end/102587
7262 * internal-fn.c (expand_DEFERRED_INIT): Guard register
7263 initialization path an avoid initializing VLA registers
7266 2021-10-04 Eric Botcazou <ebotcazou@adacore.com>
7268 * config/rs6000/vxworks.h (TARGET_INIT_LIBFUNCS): Delete.
7270 2021-10-03 Martin Liska <mliska@suse.cz>
7272 * toplev.c (toplev::main): Check opt_index if it is a part
7275 2021-10-02 Aldy Hernandez <aldyh@redhat.com>
7277 PR tree-optimization/102563
7278 * range-op.cc (operator_lshift::op1_range): Do not clobber
7281 2021-10-02 Martin Liska <mliska@suse.cz>
7283 * toplev.c (toplev::main): save_decoded_options[0] is program
7284 name and so it should be skipped.
7286 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
7288 PR tree-optimization/102546
7289 * range-op.cc (operator_lshift::op1_range): Teach range-ops that
7290 X << Y is non-zero implies X is also non-zero.
7292 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
7294 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
7296 * config/aarch64/aarch64-tune.md: Regenerate.
7297 * doc/invoke.texi: Update docs.
7299 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
7301 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
7303 * config/aarch64/aarch64-tune.md: Regenerate.
7304 * doc/invoke.texi: Update docs.
7306 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
7308 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
7310 * config/aarch64/aarch64-tune.md: Regenerate.
7311 * doc/invoke.texi: Update docs.
7313 2021-10-01 Martin Sebor <msebor@redhat.com>
7316 * doc/invoke.texi (-Waddress): Update.
7317 * gengtype.c (write_types): Avoid -Waddress.
7318 * poly-int.h (POLY_SET_COEFF): Avoid using null.
7320 2021-10-01 John David Anglin <danglin@gcc.gnu.org>
7323 * config/pa/pa.c (pa_option_override): Default to dwarf version 4
7326 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
7328 * config/aarch64/aarch64.h (AARCH64_FL_V9): Update value.
7330 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
7332 * gimple-range-path.cc (path_range_query::compute_ranges): Use
7334 * gimple-range-path.h (class path_range_query): Remove shadowed
7336 (path_range_query::get_path_oracle): New.
7338 2021-10-01 Jakub Jelinek <jakub@redhat.com>
7339 Richard Biener <rguenther@suse.de>
7342 * doc/invoke.texi (-fsanitize=integer-divide-by-zero): Remove
7343 INT_MIN / -1 division detection from here ...
7344 (-fsanitize=signed-integer-overflow): ... and add it here.
7346 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
7348 * config/aarch64/aarch64-arches.def (AARCH64_ARCH): Added
7350 * config/aarch64/aarch64.h (AARCH64_FL_V9): New.
7351 (AARCH64_FL_FOR_ARCH9): New flags for Armv9-A.
7352 (AARCH64_ISA_V9): New ISA flag.
7353 * doc/invoke.texi: Update docs.
7355 2021-10-01 Martin Liska <mliska@suse.cz>
7357 * toplev.c (toplev::main): Save decoded optimization options.
7358 * toplev.h (save_opt_decoded_options): New.
7359 * doc/extend.texi: Be more clear about optimize and target
7362 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
7364 * explow.c: Include langhooks.h.
7365 (set_stack_check_libfunc): Build a proper function type.
7367 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
7370 * config/i386/i386.c (legitimate_pic_address_disp_p): For PE-COFF do
7371 not return true for external weak function symbols in medium model.
7373 2021-10-01 Jakub Jelinek <jakub@redhat.com>
7375 * tree.h (OMP_CLAUSE_ORDER_REPRODUCIBLE): Define.
7376 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_ORDER>: Print
7377 reproducible: for OMP_CLAUSE_ORDER_REPRODUCIBLE.
7378 * omp-general.c (omp_extract_for_data): If OMP_CLAUSE_ORDER is seen
7379 without OMP_CLAUSE_ORDER_UNCONSTRAINED, overwrite sched_kind to
7380 OMP_CLAUSE_SCHEDULE_STATIC.
7382 2021-10-01 Richard Biener <rguenther@suse.de>
7384 PR middle-end/102518
7385 * tree-inline.c (setup_one_parameter): Avoid substituting
7386 an invariant into contexts where a GIMPLE register is not valid.
7388 2021-09-30 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
7390 * config/arm/arm-cpus.in: Add Cortex-R52+ CPU.
7391 * config/arm/arm-tables.opt: Regenerate.
7392 * config/arm/arm-tune.md: Regenerate.
7393 * doc/invoke.texi: Update docs.
7395 2021-09-30 Uroš Bizjak <ubizjak@gmail.com>
7398 * config/i386/i386.md
7399 (sign_extend:WIDE (any_logic:NARROW (memory, immediate)) splitters):
7402 2021-09-30 Tobias Burnus <tobias@codesourcery.com>
7404 * omp-low.c (omp_runtime_api_call): Add omp_aligned_{,c}alloc and
7405 omp_{c,re}alloc, fix omp_alloc/omp_free.
7407 2021-09-30 Martin Liska <mliska@suse.cz>
7409 * defaults.h (ASM_OUTPUT_ASCII): Do not hide global variable
7410 asm_out_file and stream directly to MYFILE.
7412 2021-09-30 Richard Biener <rguenther@suse.de>
7414 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
7415 Restore and fix condition under which we apply npeel to
7416 the DRs misalignment value.
7418 2021-09-30 Richard Biener <rguenther@suse.de>
7420 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
7421 Fix npeel check for variable amount of peeling.
7423 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
7425 * lto-wrapper.c (run_gcc): Plug snprintf overflow.
7427 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
7429 * gimple-range.cc (gimple_ranger::debug): New.
7430 * gimple-range.h (class gimple_ranger): Add debug.
7432 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
7434 PR middle-end/102519
7435 * tree-vrp.c (hybrid_threader::~hybrid_threader): Free m_query.
7437 2021-09-29 Indu Bhagat <indu.bhagat@oracle.com>
7440 * btfout.c (GTY): Add GTY (()) albeit for cosmetic only purpose.
7441 (btf_finalize): Empty the hash_map btf_var_ids.
7443 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
7445 * tree-vrp.c (thread_through_all_blocks): Return bool.
7446 (execute_vrp_threader): Return TODO_* flags.
7447 (pass_data_vrp_threader): Set todo_flags_finish to 0.
7449 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
7451 * timevar.def (TV_TREE_VRP_THREADER): New.
7452 * tree-vrp.c: Use TV_TREE_VRP_THREADER for VRP threader pass.
7454 2021-09-29 David Faust <david.faust@oracle.com>
7456 * config.gcc (bpf-*-*): Do not overwrite extra_headers.
7458 2021-09-29 Jonathan Wright <jonathan.wright@arm.com>
7460 * config/aarch64/aarch64-builtins.c (TYPES_BINOP_PPU): Define
7461 new type qualifier enum.
7462 (TYPES_TERNOP_SSSU): Likewise.
7463 (TYPES_TERNOP_PPPU): Likewise.
7464 * config/aarch64/aarch64-simd-builtins.def: Define PPU, SSU,
7465 PPPU and SSSU builtin generator macros for qtbl1 and qtbx1
7467 * config/aarch64/arm_neon.h (vqtbl1_p8): Use type-qualified
7468 builtin and remove casts.
7469 (vqtbl1_s8): Likewise.
7470 (vqtbl1q_p8): Likewise.
7471 (vqtbl1q_s8): Likewise.
7472 (vqtbx1_s8): Likewise.
7473 (vqtbx1_p8): Likewise.
7474 (vqtbx1q_s8): Likewise.
7475 (vqtbx1q_p8): Likewise.
7476 (vtbl1_p8): Likewise.
7477 (vtbl2_p8): Likewise.
7478 (vtbx2_p8): Likewise.
7480 2021-09-29 Richard Biener <rguenther@suse.de>
7482 * tree-vect-data-refs.c (vect_dr_misalign_for_aligned_access):
7484 (vect_update_misalignment_for_peel): Use it to update
7485 misaligned to the value necessary for an aligned access.
7486 (vect_get_peeling_costs_all_drs): Likewise.
7487 (vect_enhance_data_refs_alignment): Likewise.
7489 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7491 * config/aarch64/aarch64.c (aarch64_expand_cpymem): Count number of
7492 emitted operations and adjust heuristic for code size.
7494 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7496 * config/aarch64/aarch64.c (aarch64_expand_setmem): Count number of
7497 emitted operations and adjust heuristic for code size.
7499 2021-09-29 Jakub Jelinek <jakub@redhat.com>
7501 PR middle-end/102504
7502 * gimplify.c (gimplify_scan_omp_clauses): Use omp_check_private even
7503 in OMP_SCOPE clauses, not just on worksharing construct clauses.
7505 2021-09-28 Geng Qi <gengqi@linux.alibaba.com>
7507 * config/riscv/riscv.md (mulv<mode>4): Call gen_smul<mode>3_highpart.
7508 (<u>mulditi3): Call <su>muldi3_highpart.
7509 (<u>muldi3_highpart): Rename to <su>muldi3_highpart.
7510 (<u>mulsidi3): Call <su>mulsi3_highpart.
7511 (<u>mulsi3_highpart): Rename to <su>mulsi3_highpart.
7513 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
7515 * config/darwin.h (DSYMUTIL_SPEC): Recognize D sources.
7517 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
7519 * config/rs6000/darwin.h (FIXED_R13): Add for PPC64.
7520 (FIRST_SAVED_GP_REGNO): Save from R13 even when it is one
7523 2021-09-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7525 * config/aarch64/aarch64.h (AARCH64_FL_LS64): Define
7526 (AARCH64_FL_V8_7): Likewise.
7527 (AARCH64_FL_FOR_ARCH8_7): Likewise.
7528 * config/aarch64/aarch64-arches.def (armv8.7-a): Define.
7529 * config/aarch64/aarch64-option-extensions.def (ls64): Define.
7530 * doc/invoke.texi: Document the above.
7532 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
7534 * dbgcnt.c (dbg_cnt_counter): New.
7535 * dbgcnt.h (dbg_cnt_counter): New.
7536 * dumpfile.c (dump_options): Add entry for TDF_THREADING.
7537 * dumpfile.h (enum dump_flag): Add TDF_THREADING.
7538 * gimple-range-path.cc (DEBUG_SOLVER): Use TDF_THREADING.
7539 * tree-ssa-threadupdate.c (dump_jump_thread_path): Dump out
7542 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
7544 * cfgcleanup.c (pass_jump::execute): Check
7545 flag_expensive_optimizations.
7546 (pass_jump_after_combine::gate): Same.
7547 * doc/invoke.texi (-fthread-jumps): Enable for -O1.
7548 * opts.c (default_options_table): Enable -fthread-jumps at -O1.
7549 * tree-ssa-threadupdate.c
7550 (fwd_jt_path_registry::remove_jump_threads_including): Bail unless
7553 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
7555 * tree-ssa-reassoc.c (biased_names): New global.
7556 (propagate_bias_p): New function.
7557 (loop_carried_phi): Remove.
7558 (propagate_rank): Propagate bias along single uses.
7559 (get_rank): Update biased_names when needed.
7561 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
7563 * passes.def (pass_reassoc): Rename parameter to early_p.
7564 * tree-ssa-reassoc.c (reassoc_bias_loop_carried_phi_ranks_p):
7566 (phi_rank): Don't bias loop-carried phi ranks
7567 before vectorization pass.
7568 (execute_reassoc): Add bias_loop_carried_phi_ranks_p parameter.
7569 (pass_reassoc::pass_reassoc): Add bias_loop_carried_phi_ranks_p
7571 (pass_reassoc::set_param): Set bias_loop_carried_phi_ranks_p
7573 (pass_reassoc::execute): Pass bias_loop_carried_phi_ranks_p to
7575 (pass_reassoc::bias_loop_carried_phi_ranks_p): New member.
7577 2021-09-28 Jakub Jelinek <jakub@redhat.com>
7580 * config/i386/i386.c (standard_80387_constant_p): Don't recognize
7581 special 80387 instruction XFmode constants if flag_rounding_math.
7583 2021-09-28 Richard Biener <rguenther@suse.de>
7585 PR tree-optimization/100112
7586 * tree-ssa-sccvn.c (visit_reference_op_load): Record the
7587 referece into the hashtable twice in case last_vuse is
7588 different from the original vuse on the stmt.
7590 2021-09-28 Jakub Jelinek <jakub@redhat.com>
7592 PR middle-end/102492
7593 * gimplify.c (gimplify_adjust_omp_clauses_1): Don't call the
7594 omp_finish_clause langhook on implicitly added OMP_CLAUSE_PRIVATE
7595 clauses on SIMD constructs.
7597 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
7599 PR tree-optimization/102511
7600 * gimple-range-path.cc (path_range_query::range_on_path_entry):
7601 Return VARYING when nothing found.
7603 2021-09-28 Hongyu Wang <hongyu.wang@intel.com>
7606 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): Add
7608 (VALID_SSE2_REG_VHF_MODE): Add V4HFmode and V2HFmode.
7609 (VALID_MMX_REG_MODE): Add V4HFmode.
7610 (SSE_REG_MODE_P): Replace VALID_AVX512FP16_REG_MODE with
7611 vector mode condition.
7612 * config/i386/i386.c (classify_argument): Parse V4HF/V2HF
7614 (function_arg_32): Add V4HFmode.
7615 (function_arg_advance_32): Likewise.
7616 * config/i386/i386.md (mode): Add V4HF/V2HF.
7617 (MODE_SIZE): Likewise.
7618 * config/i386/mmx.md (MMXMODE): Add V4HF mode.
7619 (V_32): Add V2HF mode.
7620 (VHF_32_64): New mode iterator.
7621 (*mov<mode>_internal): Adjust sse alternatives to support
7623 (*mov<mode>_internal): Adjust sse alternatives to support
7625 (<insn><mode>3): New define_insn for add/sub/mul/div.
7627 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
7629 * tree-ssa-threadbackward.c (pass_thread_jumps::gate): Check
7631 (pass_early_thread_jumps::gate): Same.
7632 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
7633 Return if !flag_thread_jumps.
7634 * tree-ssa-threadupdate.c
7635 (jt_path_registry::register_jump_thread): Assert that
7636 flag_thread_jumps is true.
7638 2021-09-28 liuhongt <hongtao.liu@intel.com>
7641 (simplify_context::simplify_binary_operation_1): Relax
7642 condition of simplifying (vec_concat:M (vec_select op0
7643 index0)(vec_select op1 index1)) to allow different modes
7644 between op0 and M, but have same inner mode.
7646 2021-09-28 liuhongt <hongtao.liu@intel.com>
7648 * config/i386/i386-expand.c (emit_reduc_half): Handle
7649 V8HF/V16HF/V32HFmode.
7650 * config/i386/sse.md (REDUC_SSE_PLUS_MODE): Add V8HF.
7651 (REDUC_SSE_SMINMAX_MODE): Ditto.
7652 (REDUC_PLUS_MODE): Add V16HF and V32HF.
7653 (REDUC_SMINMAX_MODE): Ditto.
7655 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
7657 * gimple-range-path.cc
7658 (path_range_query::precompute_ranges_in_block): Rename to...
7659 (path_range_query::compute_ranges_in_block): ...this.
7660 (path_range_query::precompute_ranges): Rename to...
7661 (path_range_query::compute_ranges): ...this.
7662 (path_range_query::precompute_relations): Rename to...
7663 (path_range_query::compute_relations): ...this.
7664 (path_range_query::precompute_phi_relations): Rename to...
7665 (path_range_query::compute_phi_relations): ...this.
7666 * gimple-range-path.h: Rename precompute* to compute*.
7667 * tree-ssa-threadbackward.c
7668 (back_threader::find_taken_edge_switch): Same.
7669 (back_threader::find_taken_edge_cond): Same.
7670 * tree-ssa-threadedge.c
7671 (hybrid_jt_simplifier::compute_ranges_from_state): Same.
7672 (hybrid_jt_state::register_equivs_stmt): Inline...
7673 * tree-ssa-threadedge.h: ...here.
7675 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
7677 * tree-vrp.c (lhs_of_dominating_assert): Remove.
7678 (class vrp_jt_state): Remove.
7679 (class vrp_jt_simplifier): Remove.
7680 (vrp_jt_simplifier::simplify): Remove.
7681 (class vrp_jump_threader): Remove.
7682 (vrp_jump_threader::vrp_jump_threader): Remove.
7683 (vrp_jump_threader::~vrp_jump_threader): Remove.
7684 (vrp_jump_threader::before_dom_children): Remove.
7685 (vrp_jump_threader::after_dom_children): Remove.
7687 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
7689 * passes.def (pass_vrp_threader): New.
7690 * tree-pass.h (make_pass_vrp_threader): Add make_pass_vrp_threader.
7691 * tree-ssa-threadedge.c (hybrid_jt_state::register_equivs_stmt): New.
7692 (hybrid_jt_simplifier::hybrid_jt_simplifier): New.
7693 (hybrid_jt_simplifier::simplify): New.
7694 (hybrid_jt_simplifier::compute_ranges_from_state): New.
7695 * tree-ssa-threadedge.h (class hybrid_jt_state): New.
7696 (class hybrid_jt_simplifier): New.
7697 * tree-vrp.c (execute_vrp): Remove ASSERT_EXPR based jump
7699 (class hybrid_threader): New.
7700 (hybrid_threader::hybrid_threader): New.
7701 (hybrid_threader::~hybrid_threader): New.
7702 (hybrid_threader::before_dom_children): New.
7703 (hybrid_threader::after_dom_children): New.
7704 (execute_vrp_threader): New.
7705 (class pass_vrp_threader): New.
7706 (make_pass_vrp_threader): New.
7708 2021-09-27 Martin Liska <mliska@suse.cz>
7710 * output.h (enum section_flag): New.
7711 (SECTION_FORGET): Remove.
7712 (SECTION_ENTSIZE): Make it (1UL << 8) - 1.
7713 (SECTION_STYLE_MASK): Define it based on other enum
7715 * varasm.c (switch_to_section): Remove unused handling of
7718 2021-09-27 Martin Liska <mliska@suse.cz>
7720 * common.opt: Add new variable flag_default_complex_method.
7721 * opts.c (finish_options): Handle flags related to
7722 x_flag_complex_method.
7723 * toplev.c (process_options): Remove option handling related
7724 to flag_complex_method.
7726 2021-09-27 Richard Biener <rguenther@suse.de>
7728 PR middle-end/102450
7729 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid using
7730 type_for_size, instead use int_mode_for_size.
7732 2021-09-27 Andrew Pinski <apinski@marvell.com>
7735 * gimplify.c (gimplify_save_expr): Return early
7736 if the type of val is error_mark_node.
7738 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
7740 * tree-ssanames.c (ssa_name_has_boolean_range): Use
7743 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
7745 * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Remove
7746 vrp_visit_cond_stmt.
7747 * tree-ssa-dom.c (cprop_operand): Convert to range_query API.
7748 (cprop_into_stmt): Same.
7749 (dom_opt_dom_walker::optimize_stmt): Same.
7751 2021-09-27 Richard Biener <rguenther@suse.de>
7753 PR tree-optimization/97351
7754 PR tree-optimization/97352
7755 PR tree-optimization/82426
7756 * tree-vectorizer.h (dr_misalignment): Add vector type
7758 (aligned_access_p): Likewise.
7759 (known_alignment_for_access_p): Likewise.
7760 (vect_supportable_dr_alignment): Likewise.
7761 (vect_known_alignment_in_bytes): Likewise. Refactor.
7762 (DR_MISALIGNMENT): Remove.
7763 (vect_update_shared_vectype): Likewise.
7764 * tree-vect-data-refs.c (dr_misalignment): Refactor, handle
7765 a vector type with larger alignment requirement and apply
7766 the negative step adjustment here.
7767 (vect_calculate_target_alignment): Remove.
7768 (vect_compute_data_ref_alignment): Get explicit vector type
7769 argument, do not apply a negative step alignment adjustment
7771 (vect_slp_analyze_node_alignment): Re-analyze alignment
7772 when we re-visit the DR with a bigger desired alignment but
7773 keep more precise results from smaller alignments.
7774 * tree-vect-slp.c (vect_update_shared_vectype): Remove.
7775 (vect_slp_analyze_node_operations_1): Do not update the
7776 shared vector type on stmts.
7777 * tree-vect-stmts.c (vect_analyze_stmt): Push/pop the
7778 vector type of an SLP node to the representative stmt-info.
7779 (vect_transform_stmt): Likewise.
7781 2021-09-27 liuhongt <hongtao.liu@intel.com>
7784 2021-09-09 liuhongt <hongtao.liu@intel.com>
7787 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
7788 (reduc_plus_scal_v4sf): .. this, New define_expand.
7789 (reduc_plus_scal_v2df): .. and this, New define_expand.
7791 2021-09-26 liuhongt <hongtao.liu@intel.com>
7793 * doc/extend.texi (Half-Precision): Remove storage only
7794 description for _Float16 w/o avx512fp16.
7796 2021-09-25 Dimitar Dimitrov <dimitar@dinux.eu>
7798 * config/pru/constraints.md (Rrio): New constraint.
7799 * config/pru/predicates.md (regio_operand): New predicate.
7800 * config/pru/pru-pragma.c (pru_register_pragmas): Register
7801 the __regio_symbol address space.
7802 * config/pru/pru-protos.h (pru_symref2ioregno): Declaration.
7803 * config/pru/pru.c (pru_symref2ioregno): New helper function.
7804 (pru_legitimate_address_p): Remove.
7805 (pru_addr_space_legitimate_address_p): Use the address space
7807 (pru_nongeneric_pointer_addrspace): New helper function.
7808 (pru_insert_attributes): New function to validate __regio_symbol
7810 (TARGET_INSERT_ATTRIBUTES): New macro.
7811 (TARGET_LEGITIMATE_ADDRESS_P): Remove.
7812 (TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P): New macro.
7813 * config/pru/pru.h (enum reg_class): Add REGIO_REGS class.
7814 * config/pru/pru.md (*regio_readsi): New pattern to read I/O
7816 (*regio_nozext_writesi): New pattern to write to I/O registers.
7817 (*regio_zext_write_r30<EQS0:mode>): Ditto.
7818 * doc/extend.texi: Document the new PRU Named Address Space.
7820 2021-09-24 Patrick Palka <ppalka@redhat.com>
7824 * real.c (encode_ieee_double): Avoid unwanted sign extension.
7825 (encode_ieee_quad): Likewise.
7827 2021-09-24 Vladimir Makarov <vmakarov@redhat.com>
7829 PR rtl-optimization/102147
7830 * ira-build.c (ira_conflict_vector_profitable_p): Make
7831 profitability calculation independent of host compiler pointer and
7834 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
7836 * gimple-range-path.cc (path_range_query::path_range_query):
7837 Move debugging header...
7838 (path_range_query::precompute_ranges): ...here.
7839 (path_range_query::internal_range_of_expr): Do not call
7840 range_on_path_entry if NAME is defined in the current block.
7842 2021-09-24 Richard Biener <rguenther@suse.de>
7844 * cfghooks.c (verify_flow_info): Verify unallocated BB and
7845 edge flags are not set.
7847 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
7849 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
7851 (jt_path_registry::register_jump_thread): Call
7852 cancel_invalid_paths.
7853 * tree-ssa-threadupdate.h (class jt_path_registry): Add
7854 cancel_invalid_paths.
7856 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
7858 PR tree-optimization/102400
7859 * tree-ssa-sccvn.c (vn_reference_insert_pieces): Initialize
7860 result_vdef to zero value.
7862 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
7864 PR tree-optimization/102451
7865 * tree-ssa-dse.c (delete_dead_or_redundant_call): Record bb of stmt
7868 2021-09-24 Hongyu Wang <hongyu.wang@intel.com>
7870 * config/i386/sse.md (cond_<insn><mode>): Extend to support
7872 (cond_mul<mode>): Likewise.
7873 (cond_div<mode>): Likewise.
7874 (cond_<code><mode>): Likewise.
7875 (cond_fma<mode>): Likewise.
7876 (cond_fms<mode>): Likewise.
7877 (cond_fnma<mode>): Likewise.
7878 (cond_fnms<mode>): Likewise.
7880 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
7882 PR tree-optimization/102463
7883 * gimple-range-fold.cc (fold_using_range::relation_fold_and_or): If
7884 there is no range-ops handler, don't look for a relation.
7886 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
7888 * gimple-range-cache.cc (ranger_cache::ranger_cache): Take
7889 non-executable_edge flag as parameter.
7890 * gimple-range-cache.h (ranger_cache): Adjust prototype.
7891 * gimple-range-gori.cc (gori_compute::gori_compute): Take
7892 non-executable_edge flag as parameter.
7893 (gori_compute::outgoing_edge_range_p): Check new flag.
7894 * gimple-range-gori.h (gori_compute): Adjust prototype.
7895 * gimple-range.cc (gimple_ranger::gimple_ranger): Create new flag.
7896 (gimple_ranger::range_on_edge): Check new flag.
7897 * gimple-range.h (gimple_ranger::non_executable_edge_flag): New.
7898 * gimple-ssa-evrp.c (rvrp_folder): Pass ranger flag to simplifer.
7899 (hybrid_folder::hybrid_folder): Set ranger non-executable flag value.
7900 (hybrid_folder::fold_stmt): Set flag value in the simplifer.
7901 * vr-values.c (simplify_using_ranges::set_and_propagate_unexecutable):
7902 Use not_executable flag if provided inmstead of EDGE_EXECUTABLE.
7903 (simplify_using_ranges::simplify_switch_using_ranges): Clear
7904 EDGE_EXECUTABLE like it originally did.
7905 (simplify_using_ranges::cleanup_edges_and_switches): Clear any
7906 NON_EXECUTABLE flags.
7907 (simplify_using_ranges::simplify_using_ranges): Adjust.
7908 * vr-values.h (class simplify_using_ranges): Adjust.
7909 (simplify_using_ranges::set_range_query): Add non-executable flag param.
7911 2021-09-23 Bill Schmidt <wschmidt@linux.ibm.com>
7914 * config/rs6000/rs6000-call.c (rs6000_aggregate_candidate): Detect
7915 zero-width bit fields and return indicator.
7916 (rs6000_discover_homogeneous_aggregate): Diagnose when the
7917 presence of a zero-width bit field changes parameter passing in
7920 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
7922 * gimple-range-fold.cc (fold_using_range::range_of_phi):
7923 Remove dominator check.
7925 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
7927 * gimple-range-path.cc (path_range_query::precompute_relations):
7928 Hoist edge calculations before using EDGE_SUCC.
7930 2021-09-23 Jonathan Wakely <jwakely@redhat.com>
7932 * configure.ac: Fix --with-multilib-list description.
7933 * configure: Regenerate.
7935 2021-09-23 Richard Biener <rguenther@suse.de>
7937 PR tree-optimization/102448
7938 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
7939 Clear alignment info copied from DR_PTR_INFO.
7941 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
7943 * config/i386/i386-expand.c (ix86_use_mask_cmp_p): Enable
7945 * config/i386/sse.md (sseintvecmodelower): Add HF vector modes.
7946 (<avx512>_store<mode>_mask): Extend to support HF vector modes.
7947 (vec_cmp<mode><avx512fmaskmodelower>): Likewise.
7948 (vcond_mask_<mode><avx512fmaskmodelower>): Likewise.
7949 (vcond<mode><mode>): New expander.
7950 (vcond<mode><sseintvecmodelower>): Likewise.
7951 (vcond<sseintvecmodelower><mode>): Likewise.
7952 (vcondu<mode><sseintvecmodelower>): Likewise.
7954 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
7956 * config/i386/sse.md (extend<ssePHmodelower><mode>2):
7958 (extendv4hf<mode>2): Likewise.
7959 (extendv2hfv2df2): Likewise.
7960 (trunc<mode><ssePHmodelower>2): Likewise.
7961 (avx512fp16_vcvt<castmode>2ph_<mode>): Rename to ...
7962 (trunc<mode>v4hf2): ... this, and drop constraints.
7963 (avx512fp16_vcvtpd2ph_v2df): Rename to ...
7964 (truncv2dfv2hf2): ... this, and likewise.
7966 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
7968 * config/i386/sse.md (float<floatunssuffix><mode><ssePHmodelower>2):
7970 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>):
7972 (float<floatunssuffix><mode>v4hf2): ... this, and drop constraints.
7973 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Rename to ...
7974 (float<floatunssuffix>v2div2hf2): ... this, and likewise.
7976 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
7978 * config/i386/i386.md (fix<fixunssuffix>_trunchf<mode>2): New expander.
7979 (fixuns_trunchfhi2): Likewise.
7980 (*fixuns_trunchfsi2zext): New define_insn.
7981 * config/i386/sse.md (ssePHmodelower): New mode_attr.
7982 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
7983 New expander for same element vector fix_truncate.
7984 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
7985 Likewise for V4HF to V4SI/V4DI fix_truncate.
7986 (fix<fixunssuffix>_truncv2hfv2di2):
7987 Likeise for V2HF to V2DI fix_truncate.
7989 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
7991 * config/i386/i386.md (<code>hf3): New expander.
7993 2021-09-23 liuhongt <hongtao.liu@intel.com>
7995 * config/i386/sse.md (FMAMODEM): extend to handle FP16.
7996 (VFH_SF_AVX512VL): Extend to handle HFmode.
7997 (VF_SF_AVX512VL): Deleted.
7999 2021-09-23 liuhongt <hongtao.liu@intel.com>
8001 * config/i386/i386.md (rinthf2): New expander.
8002 (nearbyinthf2): New expander.
8004 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
8006 * tree-ssa-dom.c (class dom_jump_threader_simplifier): Rename...
8007 (class dom_jt_state): ...this and provide virtual overrides.
8008 (dom_jt_state::register_equiv): New.
8009 (class dom_jt_simplifier): Rename from
8010 dom_jump_threader_simplifier.
8011 (dom_jump_threader_simplifier::simplify): Rename...
8012 (dom_jt_simplifier::simplify): ...to this.
8013 (pass_dominator::execute): Use dom_jt_simplifier and
8015 * tree-ssa-threadedge.c (jump_threader::jump_threader):
8017 (jt_state::register_equivs_stmt): Abstract out...
8018 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
8020 (jump_threader::thread_around_empty_blocks): Update state.
8021 (jump_threader::thread_through_normal_block): Same.
8022 (jt_state::jt_state): Remove.
8023 (jt_state::push): Remove pass specific bits. Keep block vector
8025 (jt_state::append_path): New.
8026 (jt_state::pop): Remove pass specific bits.
8027 (jt_state::register_equiv): Same.
8028 (jt_state::record_ranges_from_stmt): Same.
8029 (jt_state::register_equivs_on_edge): Same. Rename...
8030 (jt_state::register_equivs_edge): ...to this.
8031 (jt_state::dump): New.
8032 (jt_state::debug): New.
8033 (jump_threader_simplifier::simplify): Remove.
8034 (jt_state::get_path): New.
8035 * tree-ssa-threadedge.h (class jt_simplifier): Make into a base
8036 class. Expose common functionality as virtual methods.
8037 (class jump_threader_simplifier): Same. Rename...
8038 (class jt_simplifier): ...to this.
8039 * tree-vrp.c (class vrp_jump_threader_simplifier): Rename...
8040 (class vrp_jt_simplifier): ...to this. Provide pass specific
8042 (class vrp_jt_state): New.
8043 (vrp_jump_threader_simplifier::simplify): Rename...
8044 (vrp_jt_simplifier::simplify): ...to this. Inline code from
8045 what used to be the base class.
8046 (vrp_jump_threader::vrp_jump_threader): Use vrp_jt_state and
8049 2021-09-22 Tobias Burnus <tobias@codesourcery.com>
8052 * doc/invoke.texi (-Wno-missing-include-dirs.): Document Fortran
8055 2021-09-22 Roger Sayle <roger@nextmovesoftware.com>
8056 Richard Biener <rguenther@suse.de>
8058 * match.pd (negation simplifications): Implement some negation
8059 folding transformations from fold-const.c's fold_negate_expr.
8060 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Add a SIMPLIFY
8061 argument, to control whether the op should be simplified prior
8062 to looking up/assigning a value number.
8063 (vn_nary_build_or_lookup): Update call to vn_nary_build_or_lookup_1.
8064 (vn_nary_simplify): Likewise.
8065 (visit_nary_op): Likewise, but when constructing a NEGATE_EXPR
8066 now call vn_nary_build_or_lookup_1 disabling simplification.
8068 2021-09-22 Jiufu Guo <guojiufu@linux.ibm.com>
8070 PR tree-optimization/102087
8071 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
8072 Update bound/cmp/control for niter.
8074 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
8076 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
8077 Move check for non-empty BB here.
8078 (fur_source::register_outgoing_edges): ...from here.
8080 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
8082 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
8083 Remove call to improve_range_with_equivs.
8084 (path_range_query::improve_range_with_equivs): Remove
8085 * gimple-range-path.h: Remove improve_range_with_equivs.
8087 2021-09-22 dianhong xu <dianhong.xu@intel.com>
8089 * config/i386/avx512fp16intrin.h:
8090 (_mm512_mask_blend_ph): New intrinsic.
8091 (_mm512_permutex2var_ph): Ditto.
8092 (_mm512_permutexvar_ph): Ditto.
8093 * config/i386/avx512fp16vlintrin.h:
8094 (_mm256_mask_blend_ph): New intrinsic.
8095 (_mm256_permutex2var_ph): Ditto.
8096 (_mm256_permutexvar_ph): Ditto.
8097 (_mm_mask_blend_ph): Ditto.
8098 (_mm_permutex2var_ph): Ditto.
8099 (_mm_permutexvar_ph): Ditto.
8101 2021-09-22 dianhong xu <dianhong.xu@intel.com>
8103 * config/i386/avx512fp16intrin.h: Add new intrinsics.
8104 (_mm512_conj_pch): New intrinsic.
8105 (_mm512_mask_conj_pch): Ditto.
8106 (_mm512_maskz_conj_pch): Ditto.
8107 * config/i386/avx512fp16vlintrin.h: Add new intrinsics.
8108 (_mm256_conj_pch): New intrinsic.
8109 (_mm256_mask_conj_pch): Ditto.
8110 (_mm256_maskz_conj_pch): Ditto.
8111 (_mm_conj_pch): Ditto.
8112 (_mm_mask_conj_pch): Ditto.
8113 (_mm_maskz_conj_pch): Ditto.
8115 2021-09-22 dianhong xu <dianhong.xu@intel.com>
8117 * config/i386/avx512fp16intrin.h (_MM512_REDUCE_OP): New macro
8118 (_mm512_reduce_add_ph): New intrinsic.
8119 (_mm512_reduce_mul_ph): Ditto.
8120 (_mm512_reduce_min_ph): Ditto.
8121 (_mm512_reduce_max_ph): Ditto.
8122 * config/i386/avx512fp16vlintrin.h
8123 (_MM256_REDUCE_OP/_MM_REDUCE_OP): New macro.
8124 (_mm256_reduce_add_ph): New intrinsic.
8125 (_mm256_reduce_mul_ph): Ditto.
8126 (_mm256_reduce_min_ph): Ditto.
8127 (_mm256_reduce_max_ph): Ditto.
8128 (_mm_reduce_add_ph): Ditto.
8129 (_mm_reduce_mul_ph): Ditto.
8130 (_mm_reduce_min_ph): Ditto.
8131 (_mm_reduce_max_ph): Ditto.
8133 2021-09-22 dianhong xu <dianhong.xu@intel.com>
8135 * config/i386/avx512fp16intrin.h (__m512h_u, __m256h_u,
8136 __m128h_u): New typedef.
8137 (_mm512_load_ph): New intrinsic.
8138 (_mm256_load_ph): Ditto.
8139 (_mm_load_ph): Ditto.
8140 (_mm512_loadu_ph): Ditto.
8141 (_mm256_loadu_ph): Ditto.
8142 (_mm_loadu_ph): Ditto.
8143 (_mm512_store_ph): Ditto.
8144 (_mm256_store_ph): Ditto.
8145 (_mm_store_ph): Ditto.
8146 (_mm512_storeu_ph): Ditto.
8147 (_mm256_storeu_ph): Ditto.
8148 (_mm_storeu_ph): Ditto.
8149 (_mm512_abs_ph): Ditto.
8150 * config/i386/avx512fp16vlintrin.h
8151 (_mm_abs_ph): Ditto.
8152 (_mm256_abs_ph): Ditto.
8154 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
8156 * config/s390/tpf.md (prologue_tpf, epilogue_tpf): Add cc clobber.
8158 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
8161 * config/s390/s390.c (s390_expand_insv): Emit a normal move if it
8162 is actually a full copy of the source operand into the target.
8163 Don't emit a strict low part move if source and target mode match.
8165 2021-09-22 Jakub Jelinek <jakub@redhat.com>
8167 PR middle-end/102415
8168 * omp-expand.c (expand_omp_single): If region->exit is NULL,
8169 assert region->entry is GIMPLE_OMP_SCOPE region and return.
8171 2021-09-22 Jakub Jelinek <jakub@redhat.com>
8173 * tree.h (OMP_CLAUSE_ALLOCATE_ALIGN): Define.
8174 * tree.c (omp_clause_num_ops): Change number of OMP_CLAUSE_ALLOCATE
8175 arguments from 2 to 3.
8176 * tree-pretty-print.c (dump_omp_clause): Print allocator() around
8177 allocate clause allocator and print align if present.
8178 * omp-low.c (scan_sharing_clauses): Force allocate_map entry even
8179 for omp_default_mem_alloc if align modifier is present. If align
8180 modifier is present, use TREE_LIST to encode both allocator and
8182 (lower_private_allocate, lower_rec_input_clauses, create_task_copyfn):
8183 Handle align modifier on allocator clause if present.
8185 2021-09-22 liuhongt <hongtao.liu@intel.com>
8187 * config/i386/i386.md (define_attr "isa"): Add
8189 (define_attr "enabled"): Correspond fma_or_avx512vl to
8190 TARGET_FMA || TARGET_AVX512VL.
8191 * config/i386/mmx.md (fmav2sf4): Extend to AVX512 fma.
8196 2021-09-22 liuhongt <hongtao.liu@intel.com>
8198 * config/i386/i386.md (cstorehf3): New define_expand.
8200 2021-09-22 liuhongt <hongtao.liu@intel.com>
8202 * config/i386/i386.md (<rounding_insn>hf2): New expander.
8203 (sse4_1_round<mode>2): Extend from MODEF to MODEFH.
8204 * config/i386/sse.md (*sse4_1_round<ssescalarmodesuffix>):
8205 Extend from VF_128 to VFH_128.
8207 2021-09-22 liuhongt <hongtao.liu@intel.com>
8209 * config/i386/i386-features.c (i386-features.c): Handle
8211 * config/i386/i386.md (sqrthf2): New expander.
8212 (*sqrthf2): New define_insn.
8213 * config/i386/sse.md
8214 (*<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
8217 2021-09-22 liuhongt <hongtao.liu@intel.com>
8219 * config/i386/avx512fp16intrin.h (_mm_mask_fcmadd_sch):
8221 (_mm_mask3_fcmadd_sch): Likewise.
8222 (_mm_maskz_fcmadd_sch): Likewise.
8223 (_mm_fcmadd_sch): Likewise.
8224 (_mm_mask_fmadd_sch): Likewise.
8225 (_mm_mask3_fmadd_sch): Likewise.
8226 (_mm_maskz_fmadd_sch): Likewise.
8227 (_mm_fmadd_sch): Likewise.
8228 (_mm_mask_fcmadd_round_sch): Likewise.
8229 (_mm_mask3_fcmadd_round_sch): Likewise.
8230 (_mm_maskz_fcmadd_round_sch): Likewise.
8231 (_mm_fcmadd_round_sch): Likewise.
8232 (_mm_mask_fmadd_round_sch): Likewise.
8233 (_mm_mask3_fmadd_round_sch): Likewise.
8234 (_mm_maskz_fmadd_round_sch): Likewise.
8235 (_mm_fmadd_round_sch): Likewise.
8236 (_mm_fcmul_sch): Likewise.
8237 (_mm_mask_fcmul_sch): Likewise.
8238 (_mm_maskz_fcmul_sch): Likewise.
8239 (_mm_fmul_sch): Likewise.
8240 (_mm_mask_fmul_sch): Likewise.
8241 (_mm_maskz_fmul_sch): Likewise.
8242 (_mm_fcmul_round_sch): Likewise.
8243 (_mm_mask_fcmul_round_sch): Likewise.
8244 (_mm_maskz_fcmul_round_sch): Likewise.
8245 (_mm_fmul_round_sch): Likewise.
8246 (_mm_mask_fmul_round_sch): Likewise.
8247 (_mm_maskz_fmul_round_sch): Likewise.
8248 * config/i386/i386-builtin.def: Add corresponding new builtins.
8249 * config/i386/sse.md
8250 (avx512fp16_fmaddcsh_v8hf_maskz<round_expand_name>): New expander.
8251 (avx512fp16_fcmaddcsh_v8hf_maskz<round_expand_name>): Ditto.
8252 (avx512fp16_fma_<complexopname>sh_v8hf<mask_scalarcz_name><round_scalarcz_name>):
8254 (avx512fp16_<complexopname>sh_v8hf_mask<round_name>): Ditto.
8255 (avx512fp16_<complexopname>sh_v8hf<mask_scalarc_name><round_scalarcz_name>):
8257 * config/i386/subst.md (mask_scalarcz_name): New.
8258 (mask_scalarc_name): Ditto.
8259 (mask_scalarc_operand3): Ditto.
8260 (mask_scalarcz_operand4): Ditto.
8261 (round_scalarcz_name): Ditto.
8262 (round_scalarc_mask_operand3): Ditto.
8263 (round_scalarcz_mask_operand4): Ditto.
8264 (round_scalarc_mask_op3): Ditto.
8265 (round_scalarcz_mask_op4): Ditto.
8266 (round_scalarcz_constraint): Ditto.
8267 (round_scalarcz_nimm_predicate): Ditto.
8268 (mask_scalarcz): Ditto.
8269 (mask_scalarc): Ditto.
8270 (round_scalarcz): Ditto.
8272 2021-09-22 liuhongt <hongtao.liu@intel.com>
8274 * config/i386/avx512fp16intrin.h (_mm512_fcmadd_pch):
8276 (_mm512_mask_fcmadd_pch): Likewise.
8277 (_mm512_mask3_fcmadd_pch): Likewise.
8278 (_mm512_maskz_fcmadd_pch): Likewise.
8279 (_mm512_fmadd_pch): Likewise.
8280 (_mm512_mask_fmadd_pch): Likewise.
8281 (_mm512_mask3_fmadd_pch): Likewise.
8282 (_mm512_maskz_fmadd_pch): Likewise.
8283 (_mm512_fcmadd_round_pch): Likewise.
8284 (_mm512_mask_fcmadd_round_pch): Likewise.
8285 (_mm512_mask3_fcmadd_round_pch): Likewise.
8286 (_mm512_maskz_fcmadd_round_pch): Likewise.
8287 (_mm512_fmadd_round_pch): Likewise.
8288 (_mm512_mask_fmadd_round_pch): Likewise.
8289 (_mm512_mask3_fmadd_round_pch): Likewise.
8290 (_mm512_maskz_fmadd_round_pch): Likewise.
8291 (_mm512_fcmul_pch): Likewise.
8292 (_mm512_mask_fcmul_pch): Likewise.
8293 (_mm512_maskz_fcmul_pch): Likewise.
8294 (_mm512_fmul_pch): Likewise.
8295 (_mm512_mask_fmul_pch): Likewise.
8296 (_mm512_maskz_fmul_pch): Likewise.
8297 (_mm512_fcmul_round_pch): Likewise.
8298 (_mm512_mask_fcmul_round_pch): Likewise.
8299 (_mm512_maskz_fcmul_round_pch): Likewise.
8300 (_mm512_fmul_round_pch): Likewise.
8301 (_mm512_mask_fmul_round_pch): Likewise.
8302 (_mm512_maskz_fmul_round_pch): Likewise.
8303 * config/i386/avx512fp16vlintrin.h (_mm_fmadd_pch):
8305 (_mm_mask_fmadd_pch): Likewise.
8306 (_mm_mask3_fmadd_pch): Likewise.
8307 (_mm_maskz_fmadd_pch): Likewise.
8308 (_mm256_fmadd_pch): Likewise.
8309 (_mm256_mask_fmadd_pch): Likewise.
8310 (_mm256_mask3_fmadd_pch): Likewise.
8311 (_mm256_maskz_fmadd_pch): Likewise.
8312 (_mm_fcmadd_pch): Likewise.
8313 (_mm_mask_fcmadd_pch): Likewise.
8314 (_mm_mask3_fcmadd_pch): Likewise.
8315 (_mm_maskz_fcmadd_pch): Likewise.
8316 (_mm256_fcmadd_pch): Likewise.
8317 (_mm256_mask_fcmadd_pch): Likewise.
8318 (_mm256_mask3_fcmadd_pch): Likewise.
8319 (_mm256_maskz_fcmadd_pch): Likewise.
8320 (_mm_fmul_pch): Likewise.
8321 (_mm_mask_fmul_pch): Likewise.
8322 (_mm_maskz_fmul_pch): Likewise.
8323 (_mm256_fmul_pch): Likewise.
8324 (_mm256_mask_fmul_pch): Likewise.
8325 (_mm256_maskz_fmul_pch): Likewise.
8326 (_mm_fcmul_pch): Likewise.
8327 (_mm_mask_fcmul_pch): Likewise.
8328 (_mm_maskz_fcmul_pch): Likewise.
8329 (_mm256_fcmul_pch): Likewise.
8330 (_mm256_mask_fcmul_pch): Likewise.
8331 (_mm256_maskz_fcmul_pch): Likewise.
8332 * config/i386/i386-builtin-types.def (V8HF_FTYPE_V8HF_V8HF_V8HF,
8333 V8HF_FTYPE_V16HF_V16HF_V16HF, V16HF_FTYPE_V16HF_V16HF_V16HF_UQI,
8334 V32HF_FTYPE_V32HF_V32HF_V32HF_INT,
8335 V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT): Add new builtin types.
8336 * config/i386/i386-builtin.def: Add new builtins.
8337 * config/i386/i386-expand.c: Handle new builtin types.
8338 * config/i386/subst.md (SUBST_CV): New.
8339 (maskc_name): Ditto.
8340 (maskc_operand3): Ditto.
8342 (sdc_maskz_name): Ditto.
8343 (sdc_mask_op4): Ditto.
8344 (sdc_mask_op5): Ditto.
8345 (sdc_mask_mode512bit_condition): Ditto.
8347 (round_maskc_operand3): Ditto.
8348 (round_sdc_mask_operand4): Ditto.
8349 (round_maskc_op3): Ditto.
8350 (round_sdc_mask_op4): Ditto.
8351 (round_saeonly_sdc_mask_operand5): Ditto.
8352 * config/i386/sse.md (unspec): Add complex fma unspecs.
8353 (avx512fmaskcmode): New.
8354 (UNSPEC_COMPLEX_F_C_MA): Ditto.
8355 (UNSPEC_COMPLEX_F_C_MUL): Ditto.
8356 (complexopname): Ditto.
8357 (<avx512>_fmaddc_<mode>_maskz<round_expand_name>): New expander.
8358 (<avx512>_fcmaddc_<mode>_maskz<round_expand_name>): Ditto.
8359 (fma_<complexopname>_<mode><sdc_maskz_name><round_name>): New
8361 (<avx512>_<complexopname>_<mode>_mask<round_name>): Ditto.
8362 (<avx512>_<complexopname>_<mode><maskc_name><round_name>): Ditto.
8364 2021-09-22 Kewen Lin <linkw@linux.ibm.com>
8366 * config/rs6000/rs6000.opt (rs6000-density-pct-threshold,
8367 rs6000-density-size-threshold, rs6000-density-penalty,
8368 rs6000-density-load-pct-threshold,
8369 rs6000-density-load-num-threshold): New parameter.
8370 * config/rs6000/rs6000.c (rs6000_density_test): Adjust with
8371 corresponding parameters.
8373 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
8375 * gimple-range-path.cc (path_range_query::defined_outside_path):
8377 (path_range_query::range_on_path_entry): New.
8378 (path_range_query::internal_range_of_expr): Resolve unknowns
8380 (path_range_query::improve_range_with_equivs): New.
8381 (path_range_query::ssa_range_in_phi): Resolve unknowns with
8383 * gimple-range-path.h (class path_range_query): Add
8384 defined_outside_path, range_on_path_entry, and
8385 improve_range_with_equivs.
8387 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
8389 * gimple-range-path.cc (path_range_query::add_to_imports): New.
8390 (path_range_query::add_copies_to_imports): New.
8391 (path_range_query::precompute_ranges): Call
8392 add_copies_to_imports.
8393 * gimple-range-path.h (class path_range_query): Add prototypes
8394 for add_copies_to_imports and add_to_imports.
8396 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
8398 * gimple-range-path.cc (path_range_query::range_defined_in_block):
8399 Remove useless code.
8401 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
8403 * gimple-range-fold.h (class fur_source): Make oracle protected.
8404 * gimple-range-path.cc (path_range_query::path_range_query): Add
8405 resolve argument. Initialize oracle.
8406 (path_range_query::~path_range_query): Delete oracle.
8407 (path_range_query::range_of_stmt): Adapt to use relations.
8408 (path_range_query::precompute_ranges): Pre-compute relations.
8409 (class jt_fur_source): New
8410 (jt_fur_source::jt_fur_source): New.
8411 (jt_fur_source::register_relation): New.
8412 (jt_fur_source::query_relation): New.
8413 (path_range_query::precompute_relations): New.
8414 (path_range_query::precompute_phi_relations): New.
8415 * gimple-range-path.h (path_range_query): Add resolve argument.
8416 Add oracle, precompute_relations, precompute_phi_relations.
8417 * tree-ssa-threadbackward.c (back_threader::back_threader): Pass
8418 resolve argument to solver.
8420 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
8422 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
8423 Rename postfold_gcond_edges to register_outgoing_edges and
8425 (fold_using_range::postfold_gcond_edges): Rename...
8426 (fur_source::register_outgoing_edges): ...to this.
8427 * gimple-range-fold.h (postfold_gcond_edges): Rename to
8428 register_outgoing_edges and move to fur_source.
8430 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
8432 * gimple-range-fold.cc (fold_using_range::range_of_phi): Check
8433 dom_info_available_p.
8435 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
8437 * gimple-range-cache.cc (non_null_ref::non_null_ref): Use create
8438 and quick_grow_cleared instead of safe_grow_cleared.
8440 2021-09-21 Thomas Schwinge <thomas@codesourcery.com>
8443 * omp-oacc-neuter-broadcast.cc (oacc_do_neutering): Evaluate
8446 2021-09-21 Richard Earnshaw <rearnsha@arm.com>
8448 * configure.ac: Detect when the assembler supports new-style
8449 architecture extensions.
8450 * common/config/arm/arm-common.c (arm_rewrite_mcpu): Return
8451 the full CPU string if the assembler can grok it.
8452 (arm_rewrite_march): Likewise but for the architecture.
8453 * config.in: Regenerate.
8454 * configure: Regenerate.
8456 2021-09-21 Richard Biener <rguenther@suse.de>
8458 PR tree-optimization/102421
8459 * tree-vect-loop.c (vect_dissolve_slp_only_groups): Copy and
8460 adjust alignment info.
8462 2021-09-21 Kewen Lin <linkw@linux.ibm.com>
8464 * ipa-fnsummary.c (ipa_fn_summary_write): Remove inconsistent
8465 bitfield stream out.
8467 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
8469 * gimple-range-fold.cc (fold_using_range::range_of_phi): Ignore
8470 undefined edges, apply an equivalence if appropriate.
8471 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Return
8472 UNDEFINED if EDGE_EXECUTABLE is not set.
8473 * gimple-range.cc (gimple_ranger::gimple_ranger): Set all edges
8474 as EXECUTABLE upon startup.
8475 (gimple_ranger::range_on_edge): Return UNDEFINED for edges without
8476 EDGE_EXECUTABLE set.
8477 * vr-values.c (set_and_propagate_unexecutable): New.
8478 (simplify_using_ranges::fold_cond): Call set_and_propagate.
8479 (simplify_using_ranges::simplify_switch_using_ranges): Ditto.
8480 * vr-values.h: Add prototype.
8482 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
8484 * value-relation.cc (equiv_oracle::register_initial_def): New.
8485 (equiv_oracle::register_relation): Call register_initial_def.
8486 (equiv_oracle::add_equiv_to_block): New. Split register_relation.
8487 (relation_oracle::register_stmt): Check def block of PHI arguments.
8488 * value-relation.h (equiv_oracle): Add new prototypes.
8490 2021-09-20 Matthias Kretz <m.kretz@gsi.de>
8492 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
8493 Define __RECIPROCAL_MATH__, __NO_SIGNED_ZEROS__,
8494 __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__, and
8495 __ROUNDING_MATH__ according to their corresponding flags.
8496 * doc/cpp.texi: Document __RECIPROCAL_MATH__,
8497 __NO_SIGNED_ZEROS__, __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__,
8498 and __ROUNDING_MATH__.
8500 2021-09-20 Richard Biener <rguenther@suse.de>
8502 * tree-vect-stmts.c (vectorizable_load): Use the vectype
8505 2021-09-20 Richard Biener <rguenther@suse.de>
8507 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
8508 Do not compute alignment of the vectorized access here.
8510 2021-09-20 Richard Biener <rguenther@suse.de>
8512 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
8513 Store -1 for runtime alias peeling iterations.
8515 2021-09-20 Richard Biener <rguenther@suse.de>
8517 * config.gcc: Obsolete hppa[12]*-*-hpux10* and hppa[12]*-*-hpux11*.
8519 2021-09-20 Thomas Schwinge <thomas@codesourcery.com>
8521 * input.c (string_concat_db::record_string_concatenation)
8522 (string_concat_db::get_string_concatenation): Skip for
8523 'RESERVED_LOCATION_P'.
8525 2021-09-20 Richard Biener <rguenther@suse.de>
8527 PR tree-optimization/65206
8528 * tree-data-ref.h (struct data_reference): Add alt_indices,
8530 * tree-data-ref.c (free_data_ref): Release alt_indices.
8531 (dr_analyze_indices): Work on struct indices and get DR_REF as tree.
8532 (create_data_ref): Adjust.
8533 (initialize_data_dependence_relation): Split into head
8534 and tail. When the base objects fail to match up try
8535 again with pointer-based analysis of indices.
8536 * tree-vectorizer.c (vec_info_shared::check_datarefs): Do
8537 not compare the lazily computed alternate set of indices.
8539 2021-09-20 Iain Sandoe <iain@sandoe.co.uk>
8541 * gcc.c: Test for execute OK when we find the
8542 programs for assembler linker and dsymutil and those
8543 were specified at configure-time.
8545 2021-09-19 Martin Sebor <msebor@redhat.com>
8547 PR middle-end/102403
8548 * gimple-predicate-analysis.cc (predicate::init_from_control_deps):
8549 Correct a function pre/postcondition.
8551 2021-09-19 Martin Sebor <msebor@redhat.com>
8553 PR middle-end/102243
8554 * tree-ssa-strlen.c (get_range): Handle null cfun.
8556 2021-09-19 Iain Sandoe <iain@sandoe.co.uk>
8558 * config/darwin.h (LINK_COMMAND_SPEC_A): Use Darwin10
8559 unwinder shim as a convenience library.
8561 2021-09-19 Andrew Pinski <apinski@marvell.com>
8563 * doc/install.texi: Add note about
8564 binutils 2.35 is required for LTO usage.
8566 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
8568 * tree-ssa-threadbackward.c
8569 (back_threader_registry::register_path): Use push_edge.
8570 * tree-ssa-threadedge.c
8571 (jump_threader::thread_around_empty_blocks): Same.
8572 (jump_threader::thread_through_normal_block): Same.
8573 (jump_threader::thread_across_edge): Same. Also, use auto_bitmap.
8575 * tree-ssa-threadupdate.c
8576 (jt_path_registry::allocate_thread_edge): Remove.
8577 (jt_path_registry::push_edge): New.
8578 (dump_jump_thread_path): Make static.
8579 * tree-ssa-threadupdate.h (allocate_thread_edge): Remove.
8582 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
8584 * gimple-range-path.cc (path_range_query::path_range_query): Add
8586 (path_range_query::dump): Remove extern declaration of dump_ranger.
8587 * gimple-range-trace.cc (dump_ranger): Add DEBUG_FUNCTION marker.
8588 * gimple-range-trace.h (dump_ranger): Add prototype.
8590 2021-09-19 John Ericson <git@JohnEricson.me>
8592 * gcc.c (find_a_program): New function, factored out of...
8593 (find_a_file): Here.
8594 (execute): Use find_a_program when looking for programs rather
8597 2021-09-19 Matwey V. Kornilov <matwey.kornilov@gmail.com>
8599 * config/avr/avr-mcus.def: Add atmega324pb.
8600 * doc/avr-mmcu.texi: Corresponding changes.
8602 2021-09-19 Roger Sayle <roger@nextmovesoftware.com>
8605 * match.pd (cmp @0 REAL_CST@1): When @0 is also REAL_CST, apply
8606 the same transformations as to @1. For comparisons against NaN,
8607 don't check HONOR_SNANS but confirm that neither operand is a
8610 2021-09-19 Benjamin Peterson <benjamin@locrian.net>
8612 * attribs.c (make_unique_name): Delete.
8613 * attribs.h (make_unique_name): Delete.
8615 2021-09-19 Andrew Pinski <apinski@marvell.com>
8617 * lra-constraints.c (check_and_process_move): Assert
8618 that dclass and sclass are greater than or equal to NO_REGS.
8620 2021-09-18 Jakub Jelinek <jakub@redhat.com>
8622 * tree.h (OMP_CLAUSE_ORDER_UNCONSTRAINED): Define.
8623 * tree-pretty-print.c (dump_omp_clause): Print unconstrained:
8624 for OMP_CLAUSE_ORDER_UNCONSTRAINED.
8626 2021-09-18 liuhongt <hongtao.liu@intel.com>
8628 * config/i386/i386-features.c (remove_partial_avx_dependency):
8629 Restrict TARGET_USE_VECTOR_FP_CONVERTS and
8630 TARGET_USE_VECTOR_CONVERTS to conversion instructions only.
8632 2021-09-18 Jakub Jelinek <jakub@redhat.com>
8634 * gimplify.c (omp_default_clause): For C/C++ default({,first}private),
8635 if file/namespace scope variable doesn't have predetermined sharing,
8636 treat it as if there was default(none).
8638 2021-09-18 liuhongt <hongtao.liu@intel.com>
8640 * config/i386/avx512fp16intrin.h (_mm_fmadd_sh):
8642 (_mm_mask_fmadd_sh): Likewise.
8643 (_mm_mask3_fmadd_sh): Likewise.
8644 (_mm_maskz_fmadd_sh): Likewise.
8645 (_mm_fmadd_round_sh): Likewise.
8646 (_mm_mask_fmadd_round_sh): Likewise.
8647 (_mm_mask3_fmadd_round_sh): Likewise.
8648 (_mm_maskz_fmadd_round_sh): Likewise.
8649 (_mm_fnmadd_sh): Likewise.
8650 (_mm_mask_fnmadd_sh): Likewise.
8651 (_mm_mask3_fnmadd_sh): Likewise.
8652 (_mm_maskz_fnmadd_sh): Likewise.
8653 (_mm_fnmadd_round_sh): Likewise.
8654 (_mm_mask_fnmadd_round_sh): Likewise.
8655 (_mm_mask3_fnmadd_round_sh): Likewise.
8656 (_mm_maskz_fnmadd_round_sh): Likewise.
8657 (_mm_fmsub_sh): Likewise.
8658 (_mm_mask_fmsub_sh): Likewise.
8659 (_mm_mask3_fmsub_sh): Likewise.
8660 (_mm_maskz_fmsub_sh): Likewise.
8661 (_mm_fmsub_round_sh): Likewise.
8662 (_mm_mask_fmsub_round_sh): Likewise.
8663 (_mm_mask3_fmsub_round_sh): Likewise.
8664 (_mm_maskz_fmsub_round_sh): Likewise.
8665 (_mm_fnmsub_sh): Likewise.
8666 (_mm_mask_fnmsub_sh): Likewise.
8667 (_mm_mask3_fnmsub_sh): Likewise.
8668 (_mm_maskz_fnmsub_sh): Likewise.
8669 (_mm_fnmsub_round_sh): Likewise.
8670 (_mm_mask_fnmsub_round_sh): Likewise.
8671 (_mm_mask3_fnmsub_round_sh): Likewise.
8672 (_mm_maskz_fnmsub_round_sh): Likewise.
8673 * config/i386/i386-builtin-types.def
8674 (V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT): New builtin type.
8675 * config/i386/i386-builtin.def: Add new builtins.
8676 * config/i386/i386-expand.c: Handle new builtin type.
8677 * config/i386/sse.md (fmai_vmfmadd_<mode><round_name>):
8678 Ajdust to support FP16.
8679 (fmai_vmfmsub_<mode><round_name>): Ditto.
8680 (fmai_vmfnmadd_<mode><round_name>): Ditto.
8681 (fmai_vmfnmsub_<mode><round_name>): Ditto.
8682 (*fmai_fmadd_<mode>): Ditto.
8683 (*fmai_fmsub_<mode>): Ditto.
8684 (*fmai_fnmadd_<mode><round_name>): Ditto.
8685 (*fmai_fnmsub_<mode><round_name>): Ditto.
8686 (avx512f_vmfmadd_<mode>_mask<round_name>): Ditto.
8687 (avx512f_vmfmadd_<mode>_mask3<round_name>): Ditto.
8688 (avx512f_vmfmadd_<mode>_maskz<round_expand_name>): Ditto.
8689 (avx512f_vmfmadd_<mode>_maskz_1<round_name>): Ditto.
8690 (*avx512f_vmfmsub_<mode>_mask<round_name>): Ditto.
8691 (avx512f_vmfmsub_<mode>_mask3<round_name>): Ditto.
8692 (*avx512f_vmfmsub_<mode>_maskz_1<round_name>): Ditto.
8693 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
8694 (*avx512f_vmfnmsub_<mode>_mask3<round_name>): Ditto.
8695 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
8696 (*avx512f_vmfnmadd_<mode>_mask<round_name>): Renamed to ...
8697 (avx512f_vmfnmadd_<mode>_mask<round_name>) ... this, and
8698 adjust to support FP16.
8699 (avx512f_vmfnmadd_<mode>_mask3<round_name>): Ditto.
8700 (avx512f_vmfnmadd_<mode>_maskz_1<round_name>): Ditto.
8701 (avx512f_vmfnmadd_<mode>_maskz<round_expand_name>): New
8704 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
8706 * config/i386/sse.md (avx512fmaskmodelower): Extend to support
8708 (maskload<mode><avx512fmaskmodelower>): Ditto.
8709 (maskstore<mode><avx512fmaskmodelower>): Ditto.
8711 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
8713 * config/i386/i386-expand.c (ix86_expand_fp_absneg_operator):
8715 (ix86_expand_copysign): Ditto.
8716 (ix86_expand_xorsign): Ditto.
8717 * config/i386/i386.c (ix86_build_const_vector): Handle HF vector
8719 (ix86_build_signbit_mask): Ditto.
8720 (ix86_can_change_mode_class): Ditto.
8721 * config/i386/i386.md
8722 (SSEMODEF): Add HFmode.
8723 (ssevecmodef): Ditto.
8724 (<code>hf2): New define_expand.
8725 (*<code>hf2_1): New define_insn_and_split.
8726 (copysign<mode>): Extend to support HFmode under AVX512FP16.
8727 (xorsign<mode>): Ditto.
8728 * config/i386/sse.md (VFB): New mode iterator.
8729 (VFB_128_256): Ditto.
8731 (sseintvecmode2): Support HF vector mode.
8732 (<code><mode>2): Use new mode iterator.
8733 (*<code><mode>2): Ditto.
8734 (copysign<mode>3): Ditto.
8735 (xorsign<mode>3): Ditto.
8736 (<code><mode>3<mask_name>): Ditto.
8737 (<code><mode>3<mask_name>): Ditto.
8738 (<sse>_andnot<mode>3<mask_name>): Adjust for HF vector mode.
8739 (<sse>_andnot<mode>3<mask_name>): Ditto.
8740 (*<code><mode>3<mask_name>): Ditto.
8741 (*<code><mode>3<mask_name>): Ditto.
8743 2021-09-18 liuhongt <hongtao.liu@intel.com>
8745 * config/i386/avx512fp16intrin.h (_mm512_mask_fmadd_ph):
8747 (_mm512_mask3_fmadd_ph): Likewise.
8748 (_mm512_maskz_fmadd_ph): Likewise.
8749 (_mm512_fmadd_round_ph): Likewise.
8750 (_mm512_mask_fmadd_round_ph): Likewise.
8751 (_mm512_mask3_fmadd_round_ph): Likewise.
8752 (_mm512_maskz_fmadd_round_ph): Likewise.
8753 (_mm512_fnmadd_ph): Likewise.
8754 (_mm512_mask_fnmadd_ph): Likewise.
8755 (_mm512_mask3_fnmadd_ph): Likewise.
8756 (_mm512_maskz_fnmadd_ph): Likewise.
8757 (_mm512_fnmadd_round_ph): Likewise.
8758 (_mm512_mask_fnmadd_round_ph): Likewise.
8759 (_mm512_mask3_fnmadd_round_ph): Likewise.
8760 (_mm512_maskz_fnmadd_round_ph): Likewise.
8761 (_mm512_fmsub_ph): Likewise.
8762 (_mm512_mask_fmsub_ph): Likewise.
8763 (_mm512_mask3_fmsub_ph): Likewise.
8764 (_mm512_maskz_fmsub_ph): Likewise.
8765 (_mm512_fmsub_round_ph): Likewise.
8766 (_mm512_mask_fmsub_round_ph): Likewise.
8767 (_mm512_mask3_fmsub_round_ph): Likewise.
8768 (_mm512_maskz_fmsub_round_ph): Likewise.
8769 (_mm512_fnmsub_ph): Likewise.
8770 (_mm512_mask_fnmsub_ph): Likewise.
8771 (_mm512_mask3_fnmsub_ph): Likewise.
8772 (_mm512_maskz_fnmsub_ph): Likewise.
8773 (_mm512_fnmsub_round_ph): Likewise.
8774 (_mm512_mask_fnmsub_round_ph): Likewise.
8775 (_mm512_mask3_fnmsub_round_ph): Likewise.
8776 (_mm512_maskz_fnmsub_round_ph): Likewise.
8777 * config/i386/avx512fp16vlintrin.h (_mm256_fmadd_ph):
8779 (_mm256_mask_fmadd_ph): Likewise.
8780 (_mm256_mask3_fmadd_ph): Likewise.
8781 (_mm256_maskz_fmadd_ph): Likewise.
8782 (_mm_fmadd_ph): Likewise.
8783 (_mm_mask_fmadd_ph): Likewise.
8784 (_mm_mask3_fmadd_ph): Likewise.
8785 (_mm_maskz_fmadd_ph): Likewise.
8786 (_mm256_fnmadd_ph): Likewise.
8787 (_mm256_mask_fnmadd_ph): Likewise.
8788 (_mm256_mask3_fnmadd_ph): Likewise.
8789 (_mm256_maskz_fnmadd_ph): Likewise.
8790 (_mm_fnmadd_ph): Likewise.
8791 (_mm_mask_fnmadd_ph): Likewise.
8792 (_mm_mask3_fnmadd_ph): Likewise.
8793 (_mm_maskz_fnmadd_ph): Likewise.
8794 (_mm256_fmsub_ph): Likewise.
8795 (_mm256_mask_fmsub_ph): Likewise.
8796 (_mm256_mask3_fmsub_ph): Likewise.
8797 (_mm256_maskz_fmsub_ph): Likewise.
8798 (_mm_fmsub_ph): Likewise.
8799 (_mm_mask_fmsub_ph): Likewise.
8800 (_mm_mask3_fmsub_ph): Likewise.
8801 (_mm_maskz_fmsub_ph): Likewise.
8802 (_mm256_fnmsub_ph): Likewise.
8803 (_mm256_mask_fnmsub_ph): Likewise.
8804 (_mm256_mask3_fnmsub_ph): Likewise.
8805 (_mm256_maskz_fnmsub_ph): Likewise.
8806 (_mm_fnmsub_ph): Likewise.
8807 (_mm_mask_fnmsub_ph): Likewise.
8808 (_mm_mask3_fnmsub_ph): Likewise.
8809 (_mm_maskz_fnmsub_ph): Likewise.
8810 * config/i386/i386-builtin.def: Add corresponding new builtins.
8811 * config/i386/sse.md
8812 (<avx512>_fmadd_<mode>_maskz<round_expand_name>): Adjust to
8813 support HF vector modes.
8814 (<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name><round_name>):
8816 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
8817 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
8818 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
8819 (<avx512>_fmadd_<mode>_mask<round_name>): Ditto.
8820 (<avx512>_fmadd_<mode>_mask3<round_name>): Ditto.
8821 (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Ditto.
8822 (<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name><round_name>):
8824 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
8825 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
8826 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
8827 (<avx512>_fmsub_<mode>_mask<round_name>): Ditto.
8828 (<avx512>_fmsub_<mode>_mask3<round_name>): Ditto.
8829 (<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name><round_name>):
8831 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
8832 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
8833 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
8834 (<avx512>_fnmadd_<mode>_mask<round_name>): Ditto.
8835 (<avx512>_fnmadd_<mode>_mask3<round_name>): Ditto.
8836 (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Ditto.
8837 (<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name><round_name>):
8839 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
8840 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
8841 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
8842 (<avx512>_fnmsub_<mode>_mask<round_name>): Ditto.
8843 (<avx512>_fnmsub_<mode>_mask3<round_name>): Ditto.
8845 2021-09-18 liuhongt <hongtao.liu@intel.com>
8847 * config/i386/avx512fp16intrin.h (_mm512_fmaddsub_ph):
8849 (_mm512_mask_fmaddsub_ph): Likewise.
8850 (_mm512_mask3_fmaddsub_ph): Likewise.
8851 (_mm512_maskz_fmaddsub_ph): Likewise.
8852 (_mm512_fmaddsub_round_ph): Likewise.
8853 (_mm512_mask_fmaddsub_round_ph): Likewise.
8854 (_mm512_mask3_fmaddsub_round_ph): Likewise.
8855 (_mm512_maskz_fmaddsub_round_ph): Likewise.
8856 (_mm512_mask_fmsubadd_ph): Likewise.
8857 (_mm512_mask3_fmsubadd_ph): Likewise.
8858 (_mm512_maskz_fmsubadd_ph): Likewise.
8859 (_mm512_fmsubadd_round_ph): Likewise.
8860 (_mm512_mask_fmsubadd_round_ph): Likewise.
8861 (_mm512_mask3_fmsubadd_round_ph): Likewise.
8862 (_mm512_maskz_fmsubadd_round_ph): Likewise.
8863 * config/i386/avx512fp16vlintrin.h (_mm256_fmaddsub_ph):
8865 (_mm256_mask_fmaddsub_ph): Likewise.
8866 (_mm256_mask3_fmaddsub_ph): Likewise.
8867 (_mm256_maskz_fmaddsub_ph): Likewise.
8868 (_mm_fmaddsub_ph): Likewise.
8869 (_mm_mask_fmaddsub_ph): Likewise.
8870 (_mm_mask3_fmaddsub_ph): Likewise.
8871 (_mm_maskz_fmaddsub_ph): Likewise.
8872 (_mm256_fmsubadd_ph): Likewise.
8873 (_mm256_mask_fmsubadd_ph): Likewise.
8874 (_mm256_mask3_fmsubadd_ph): Likewise.
8875 (_mm256_maskz_fmsubadd_ph): Likewise.
8876 (_mm_fmsubadd_ph): Likewise.
8877 (_mm_mask_fmsubadd_ph): Likewise.
8878 (_mm_mask3_fmsubadd_ph): Likewise.
8879 (_mm_maskz_fmsubadd_ph): Likewise.
8880 * config/i386/i386-builtin.def: Add corresponding new builtins.
8881 * config/i386/sse.md (VFH_SF_AVX512VL): New mode iterator.
8882 * (<avx512>_fmsubadd_<mode>_maskz<round_expand_name>): New expander.
8883 * (<avx512>_fmaddsub_<mode>_maskz<round_expand_name>): Use
8885 * (<sd_mask_codefor>fma_fmaddsub_<mode><sd_maskz_name><round_name>):
8887 * (<avx512>_fmaddsub_<mode>_mask<round_name>): Ditto.
8888 * (<avx512>_fmaddsub_<mode>_mask3<round_name>): Ditto.
8889 * (<sd_mask_codefor>fma_fmsubadd_<mode><sd_maskz_name><round_name>):
8891 * (<avx512>_fmsubadd_<mode>_mask<round_name>): Ditto.
8892 * (<avx512>_fmsubadd_<mode>_mask3<round_name>): Ditto.
8894 2021-09-18 liuhongt <hongtao.liu@intel.com>
8897 * config/i386/i386.c (ix86_print_operand): Handle
8898 V8HF/V16HF/V32HFmode.
8899 * config/i386/i386.h (VALID_BCST_MODE_P): Add HFmode.
8900 * config/i386/sse.md (avx512bcst): Remove.
8902 2021-09-17 Martin Sebor <msebor@redhat.com>
8904 * Makefile.in (OBJS): Add gimple-predicate-analysis.o.
8905 * tree-ssa-uninit.c (max_phi_args): Move to gimple-predicate-analysis.
8906 (MASK_SET_BIT, MASK_TEST_BIT, MASK_EMPTY): Same.
8907 (check_defs): Add comment.
8908 (can_skip_redundant_opnd): Update comment.
8909 (compute_uninit_opnds_pos): Adjust to namespace change.
8910 (find_pdom): Move to gimple-predicate-analysis.cc.
8912 (struct uninit_undef_val_t): New.
8913 (is_non_loop_exit_postdominating): Move to gimple-predicate-analysis.cc.
8914 (find_control_equiv_block): Same.
8915 (MAX_NUM_CHAINS, MAX_CHAIN_LEN, MAX_POSTDOM_CHECK): Same.
8916 (MAX_SWITCH_CASES): Same.
8917 (compute_control_dep_chain): Same.
8918 (find_uninit_use): Use predicate analyzer.
8919 (struct pred_info): Move to gimple-predicate-analysis.
8920 (convert_control_dep_chain_into_preds): Same.
8921 (find_predicates): Same.
8922 (collect_phi_def_edges): Same.
8923 (warn_uninitialized_phi): Use predicate analyzer.
8924 (find_def_preds): Move to gimple-predicate-analysis.
8925 (dump_pred_info): Same.
8926 (dump_pred_chain): Same.
8927 (dump_predicates): Same.
8928 (destroy_predicate_vecs): Remove.
8929 (execute_late_warn_uninitialized): New.
8930 (get_cmp_code): Move to gimple-predicate-analysis.
8931 (is_value_included_in): Same.
8932 (value_sat_pred_p): Same.
8933 (find_matching_predicate_in_rest_chains): Same.
8934 (is_use_properly_guarded): Same.
8935 (prune_uninit_phi_opnds): Same.
8936 (find_var_cmp_const): Same.
8937 (use_pred_not_overlap_with_undef_path_pred): Same.
8938 (pred_equal_p): Same.
8939 (is_neq_relop_p): Same.
8940 (is_neq_zero_form_p): Same.
8941 (pred_expr_equal_p): Same.
8942 (is_pred_expr_subset_of): Same.
8943 (is_pred_chain_subset_of): Same.
8944 (is_included_in): Same.
8945 (is_superset_of): Same.
8947 (simplify_pred): Same.
8948 (simplify_preds_2): Same.
8949 (simplify_preds_3): Same.
8950 (simplify_preds_4): Same.
8951 (simplify_preds): Same.
8953 (push_to_worklist): Same.
8954 (get_pred_info_from_cmp): Same.
8955 (is_degenerated_phi): Same.
8956 (normalize_one_pred_1): Same.
8957 (normalize_one_pred): Same.
8958 (normalize_one_pred_chain): Same.
8959 (normalize_preds): Same.
8960 (can_one_predicate_be_invalidated_p): Same.
8961 (can_chain_union_be_invalidated_p): Same.
8962 (uninit_uses_cannot_happen): Same.
8963 (pass_late_warn_uninitialized::execute): Define.
8964 * gimple-predicate-analysis.cc: New file.
8965 * gimple-predicate-analysis.h: New file.
8967 2021-09-17 Julian Brown <julian@codesourcery.com>
8969 * config/gcn/gcn.c (gimple.h): Include.
8970 (gcn_fork_join): Emit barrier for worker-level joins.
8971 * omp-oacc-neuter-broadcast.cc (find_local_vars_to_propagate): Add
8972 writes_gang_private bitmap parameter. Set bit for blocks
8973 containing gang-private variable writes.
8974 (worker_single_simple): Don't emit barrier after predicated block.
8975 (worker_single_copy): Don't emit barrier if we're not broadcasting
8976 anything and the block contains no gang-private writes.
8977 (neuter_worker_single): Don't predicate blocks that only contain
8978 NOPs or internal marker functions. Pass has_gang_private_write
8979 argument to worker_single_copy.
8980 (oacc_do_neutering): Add writes_gang_private bitmap handling.
8982 2021-09-17 Julian Brown <julian@codesourcery.com>
8984 * config/gcn/gcn-protos.h
8985 (gcn_goacc_create_worker_broadcast_record): Update prototype.
8986 * config/gcn/gcn-tree.c (gcn_goacc_get_worker_red_decl): Use
8987 preallocated block of LDS memory. Do not cache/share decls for
8988 reduction temporaries between invocations.
8989 (gcn_goacc_reduction_teardown): Unshare VAR on second use.
8990 (gcn_goacc_create_worker_broadcast_record): Add OFFSET parameter
8991 and return temporary LDS space at that offset. Return pointer in
8993 * config/gcn/gcn.c (acc_lds_size, gang_private_hwm, lds_allocs):
8995 (ACC_LDS_SIZE): Define as acc_lds_size.
8996 (gcn_init_machine_status): Don't initialise lds_allocated,
8997 lds_allocs, reduc_decls fields of machine function struct.
8998 (gcn_option_override): Handle default size for gang-private
8999 variables and -mgang-private-size option.
9000 (gcn_expand_prologue): Use LDS_SIZE instead of LDS_SIZE-1 when
9001 initialising M0_REG.
9002 (gcn_shared_mem_layout): New function.
9003 (gcn_print_lds_decl): Update comment. Use global lds_allocs map and
9004 gang_private_hwm variable.
9005 (TARGET_GOACC_SHARED_MEM_LAYOUT): Define target hook.
9006 * config/gcn/gcn.h (machine_function): Remove lds_allocated,
9007 lds_allocs, reduc_decls. Add reduction_base, reduction_limit.
9008 * config/gcn/gcn.opt (gang_private_size_opt): New global.
9009 (mgang-private-size=): New option.
9010 * doc/tm.texi.in (TARGET_GOACC_SHARED_MEM_LAYOUT): Place
9012 * doc/tm.texi: Regenerate.
9013 * omp-oacc-neuter-broadcast.cc (targhooks.h, diagnostic-core.h):
9015 (build_sender_ref): Handle sender_decl being pointer.
9016 (worker_single_copy): Add PLACEMENT and ISOLATE_BROADCASTS
9017 parameters. Pass placement argument to
9018 create_worker_broadcast_record hook invocations. Handle
9019 sender_decl being pointer and isolate_broadcasts inserting extra
9021 (blk_offset_map_t): Add typedef.
9022 (neuter_worker_single): Add BLK_OFFSET_MAP parameter. Pass
9023 preallocated range to worker_single_copy call.
9024 (dfs_broadcast_reachable_1): New function.
9025 (idx_decl_pair_t, used_range_vec_t): New typedefs.
9026 (sort_size_descending): New function.
9027 (addr_range): New class.
9028 (splay_tree_compare_addr_range, splay_tree_free_key)
9029 (first_fit_range, merge_ranges_1, merge_ranges): New functions.
9030 (execute_omp_oacc_neuter_broadcast): Rename to...
9031 (oacc_do_neutering): ... this. Add BOUNDS_LO, BOUNDS_HI
9032 parameters. Arrange layout of shared memory for broadcast
9034 (execute_omp_oacc_neuter_broadcast): New function.
9035 (pass_omp_oacc_neuter_broadcast::gate): Remove num_workers==1
9036 handling from here. Enable pass for all OpenACC routines in order
9037 to call shared memory-layout hook.
9038 * target.def (create_worker_broadcast_record): Add OFFSET
9040 (shared_mem_layout): New hook.
9042 2021-09-17 Julian Brown <julian@codesourcery.com>
9043 Thomas Schwinge <thomas@codesourcery.com>
9045 * omp-oacc-neuter-broadcast.cc
9046 (pass_omp_oacc_neuter_broadcast::gate): Disable if num_workers is
9048 (execute_omp_oacc_neuter_broadcast): Adjust.
9050 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
9052 * value-relation.cc (class equiv_chain): Move to header file.
9053 (path_oracle::path_oracle): New.
9054 (path_oracle::~path_oracle): New.
9055 (path_oracle::register_relation): New.
9056 (path_oracle::query_relation): New.
9057 (path_oracle::reset_path): New.
9058 (path_oracle::dump): New.
9059 * value-relation.h (class equiv_chain): Move to here.
9060 (class path_oracle): New.
9062 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
9064 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a DOM
9066 * gimple-range-fold.cc (fur_depend::register_relation): Use
9067 register_stmt/edge routines.
9068 * value-relation.cc (equiv_chain::find): Relocate from equiv_oracle.
9069 (equiv_oracle::equiv_oracle): Create self equivalence cache.
9070 (equiv_oracle::~equiv_oracle): Release same.
9071 (equiv_oracle::equiv_set): Return entry from self equiv cache if there
9072 are no equivalences.
9073 (equiv_oracle::find_equiv_block): Move list find to equiv_chain.
9074 (equiv_oracle::register_relation): Rename from register_equiv.
9075 (relation_chain_head::find_relation): Relocate from dom_oracle.
9076 (relation_oracle::register_stmt): New.
9077 (relation_oracle::register_edge): New.
9078 (dom_oracle::*): Rename from relation_oracle.
9079 (dom_oracle::register_relation): Adjust to call equiv_oracle.
9080 (dom_oracle::set_one_relation): Split from register_relation.
9081 (dom_oracle::register_transitives): Consolidate 2 methods.
9082 (dom_oracle::find_relation_block): Move core to relation_chain.
9083 (dom_oracle::query_relation): Rename from find_relation_dom and adjust.
9084 * value-relation.h (class relation_oracle): New pure virtual base.
9085 (class equiv_oracle): Inherit from relation_oracle and adjust.
9086 (class dom_oracle): Rename from old relation_oracle and adjust.
9088 2021-09-17 Martin Sebor <msebor@redhat.com>
9090 PR middle-end/102200
9091 * pointer-query.cc (access_ref::inform_access): Handle MIN/MAX_EXPR.
9092 (handle_min_max_size): Change argument. Store original SSA_NAME for
9093 operands to potentially distinct (sub)objects.
9094 (compute_objsize_r): Adjust call to the above.
9096 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
9098 * config/rs6000/rs6000.c (rs6000-builtins.h): New include.
9099 (rs6000_new_builtin_vectorized_function): New function.
9100 (rs6000_new_builtin_md_vectorized_function): Likewise.
9101 (rs6000_builtin_vectorized_function): Call
9102 rs6000_new_builtin_vectorized_function.
9103 (rs6000_builtin_md_vectorized_function): Call
9104 rs6000_new_builtin_md_vectorized_function.
9106 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
9108 * config/rs6000/rs6000-builtin-new.def (ASSEMBLE_ACC): Add mmaint flag.
9109 (ASSEMBLE_PAIR): Likewise.
9110 (BUILD_ACC): Likewise.
9111 (DISASSEMBLE_ACC): Likewise.
9112 (DISASSEMBLE_PAIR): Likewise.
9113 (PMXVBF16GER2): Likewise.
9114 (PMXVBF16GER2NN): Likewise.
9115 (PMXVBF16GER2NP): Likewise.
9116 (PMXVBF16GER2PN): Likewise.
9117 (PMXVBF16GER2PP): Likewise.
9118 (PMXVF16GER2): Likewise.
9119 (PMXVF16GER2NN): Likewise.
9120 (PMXVF16GER2NP): Likewise.
9121 (PMXVF16GER2PN): Likewise.
9122 (PMXVF16GER2PP): Likewise.
9123 (PMXVF32GER): Likewise.
9124 (PMXVF32GERNN): Likewise.
9125 (PMXVF32GERNP): Likewise.
9126 (PMXVF32GERPN): Likewise.
9127 (PMXVF32GERPP): Likewise.
9128 (PMXVF64GER): Likewise.
9129 (PMXVF64GERNN): Likewise.
9130 (PMXVF64GERNP): Likewise.
9131 (PMXVF64GERPN): Likewise.
9132 (PMXVF64GERPP): Likewise.
9133 (PMXVI16GER2): Likewise.
9134 (PMXVI16GER2PP): Likewise.
9135 (PMXVI16GER2S): Likewise.
9136 (PMXVI16GER2SPP): Likewise.
9137 (PMXVI4GER8): Likewise.
9138 (PMXVI4GER8PP): Likewise.
9139 (PMXVI8GER4): Likewise.
9140 (PMXVI8GER4PP): Likewise.
9141 (PMXVI8GER4SPP): Likewise.
9142 (XVBF16GER2): Likewise.
9143 (XVBF16GER2NN): Likewise.
9144 (XVBF16GER2NP): Likewise.
9145 (XVBF16GER2PN): Likewise.
9146 (XVBF16GER2PP): Likewise.
9147 (XVF16GER2): Likewise.
9148 (XVF16GER2NN): Likewise.
9149 (XVF16GER2NP): Likewise.
9150 (XVF16GER2PN): Likewise.
9151 (XVF16GER2PP): Likewise.
9152 (XVF32GER): Likewise.
9153 (XVF32GERNN): Likewise.
9154 (XVF32GERNP): Likewise.
9155 (XVF32GERPN): Likewise.
9156 (XVF32GERPP): Likewise.
9157 (XVF64GER): Likewise.
9158 (XVF64GERNN): Likewise.
9159 (XVF64GERNP): Likewise.
9160 (XVF64GERPN): Likewise.
9161 (XVF64GERPP): Likewise.
9162 (XVI16GER2): Likewise.
9163 (XVI16GER2PP): Likewise.
9164 (XVI16GER2S): Likewise.
9165 (XVI16GER2SPP): Likewise.
9166 (XVI4GER8): Likewise.
9167 (XVI4GER8PP): Likewise.
9168 (XVI8GER4): Likewise.
9169 (XVI8GER4PP): Likewise.
9170 (XVI8GER4SPP): Likewise.
9171 (XXMFACC): Likewise.
9172 (XXMTACC): Likewise.
9173 (XXSETACCZ): Likewise.
9174 (ASSEMBLE_PAIR_V): Likewise.
9175 (BUILD_PAIR): Likewise.
9176 (DISASSEMBLE_PAIR_V): Likewise.
9179 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_mma_builtin):
9180 Handle RS6000_BIF_LXVP and RS6000_BIF_STXVP.
9181 * config/rs6000/rs6000-gen-builtins.c (attrinfo): Add ismmaint.
9182 (parse_bif_attrs): Handle ismmaint.
9183 (write_decls): Add bif_mmaint_bit and bif_is_mmaint.
9184 (write_bif_static_init): Handle ismmaint.
9186 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
9188 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_builtin): New
9190 (rs6000_gimple_fold_builtin): Call rs6000_gimple_fold_new_builtin.
9191 (rs6000_new_builtin_valid_without_lhs): New function.
9192 (rs6000_gimple_fold_new_mma_builtin): Likewise.
9193 (rs6000_gimple_fold_new_builtin): Likewise.
9195 2021-09-17 Thomas Schwinge <thomas@codesourcery.com>
9197 * hash-table.h (hash_table<Descriptor, Lazy, Allocator>::expand):
9198 Destruct stale Value objects.
9199 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
9202 2021-09-17 Roger Sayle <roger@nextmovesoftware.com>
9205 * match.pd (shift optimizations): Disable recent sign-changing
9206 optimization for shifts by zero, these will be folded later.
9208 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
9210 * config/rs6000/rs6000-builtin-new.def (__builtin_mffsl): Move from
9211 [power9] to [always].
9213 2021-09-17 Richard Biener <rguenther@suse.de>
9215 * tree-vect-stmts.c (vectorizable_load): Do not frob
9218 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
9220 * config/i386/i386-features.c (remove_partial_avx_dependency):
9221 Also check TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY and
9222 and TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY before generating
9224 * config/i386/i386.h (TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY):
9226 (TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
9227 * config/i386/i386.md (SSE FP to FP splitters): Replace
9228 TARGET_SSE_PARTIAL_REG_DEPENDENCY with
9229 TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY.
9230 (SSE INT to FP splitter): Replace TARGET_SSE_PARTIAL_REG_DEPENDENCY
9231 with TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY.
9232 * config/i386/x86-tune.def
9233 (X86_TUNE_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY): New.
9234 (X86_TUNE_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
9236 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
9239 * config/i386/i386-features.c (remove_partial_avx_dependency):
9240 Check TARGET_USE_VECTOR_FP_CONVERTS and TARGET_USE_VECTOR_CONVERTS
9241 before generating vxorps.
9243 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
9245 * config/i386/i386-options.c (processor_cost_table): Use
9246 tremont_cost for Tremont.
9247 * config/i386/x86-tune-costs.h (tremont_memcpy): New.
9248 (tremont_memset): Likewise.
9249 (tremont_cost): Likewise.
9250 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
9253 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
9255 * common/config/i386/i386-common.c: Use Haswell scheduling model
9257 * config/i386/i386.c (ix86_sched_init_global): Prepare for Tremont
9259 * config/i386/x86-tune-sched.c (ix86_issue_rate): Change Tremont
9261 (ix86_adjust_cost): Handle Tremont.
9262 * config/i386/x86-tune.def (X86_TUNE_SSE_PARTIAL_REG_DEPENDENCY):
9264 (X86_TUNE_USE_LEAVE): Likewise.
9265 (X86_TUNE_PUSH_MEMORY): Likewise.
9266 (X86_TUNE_MISALIGNED_MOVE_STRING_PRO_EPILOGUES): Likewise.
9267 (X86_TUNE_USE_CLTD): Likewise.
9268 (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI): Likewise.
9269 (X86_TUNE_AVOID_MFENCE): Likewise.
9270 (X86_TUNE_SSE_TYPELESS_STORES): Likewise.
9271 (X86_TUNE_SSE_LOAD0_BY_PXOR): Likewise.
9272 (X86_TUNE_ACCUMULATE_OUTGOING_ARGS): Disable for Tremont.
9273 (X86_TUNE_FOUR_JUMP_LIMIT): Likewise.
9274 (X86_TUNE_OPT_AGU): Likewise.
9275 (X86_TUNE_AVOID_LEA_FOR_ADDR): Likewise.
9276 (X86_TUNE_AVOID_MEM_OPND_FOR_CMOVE): Likewise.
9277 (X86_TUNE_EXPAND_ABS): Likewise.
9278 (X86_TUNE_SPLIT_MEM_OPND_FOR_FP_CONVERTS): Likewise.
9279 (X86_TUNE_SLOW_PSHUFB): Likewise.
9281 2021-09-17 Eric Botcazou <ebotcazou@adacore.com>
9283 PR rtl-optimization/102306
9284 * combine.c (try_combine): Abort the combination if we are about to
9285 duplicate volatile references.
9287 2021-09-17 liuhongt <hongtao.liu@intel.com>
9289 * config/i386/avx512fp16intrin.h (_mm_undefined_ph):
9291 (_mm256_undefined_ph): Likewise.
9292 (_mm512_undefined_ph): Likewise.
9293 (_mm_cvtsh_h): Likewise.
9294 (_mm256_cvtsh_h): Likewise.
9295 (_mm512_cvtsh_h): Likewise.
9296 (_mm512_castph_ps): Likewise.
9297 (_mm512_castph_pd): Likewise.
9298 (_mm512_castph_si512): Likewise.
9299 (_mm512_castph512_ph128): Likewise.
9300 (_mm512_castph512_ph256): Likewise.
9301 (_mm512_castph128_ph512): Likewise.
9302 (_mm512_castph256_ph512): Likewise.
9303 (_mm512_zextph128_ph512): Likewise.
9304 (_mm512_zextph256_ph512): Likewise.
9305 (_mm512_castps_ph): Likewise.
9306 (_mm512_castpd_ph): Likewise.
9307 (_mm512_castsi512_ph): Likewise.
9308 * config/i386/avx512fp16vlintrin.h (_mm_castph_ps):
9310 (_mm256_castph_ps): Likewise.
9311 (_mm_castph_pd): Likewise.
9312 (_mm256_castph_pd): Likewise.
9313 (_mm_castph_si128): Likewise.
9314 (_mm256_castph_si256): Likewise.
9315 (_mm_castps_ph): Likewise.
9316 (_mm256_castps_ph): Likewise.
9317 (_mm_castpd_ph): Likewise.
9318 (_mm256_castpd_ph): Likewise.
9319 (_mm_castsi128_ph): Likewise.
9320 (_mm256_castsi256_ph): Likewise.
9321 (_mm256_castph256_ph128): Likewise.
9322 (_mm256_castph128_ph256): Likewise.
9323 (_mm256_zextph128_ph256): Likewise.
9325 2021-09-17 liuhongt <hongtao.liu@intel.com>
9327 * config/i386/avx512fp16intrin.h (_mm_cvtsh_ss):
9329 (_mm_mask_cvtsh_ss): Likewise.
9330 (_mm_maskz_cvtsh_ss): Likewise.
9331 (_mm_cvtsh_sd): Likewise.
9332 (_mm_mask_cvtsh_sd): Likewise.
9333 (_mm_maskz_cvtsh_sd): Likewise.
9334 (_mm_cvt_roundsh_ss): Likewise.
9335 (_mm_mask_cvt_roundsh_ss): Likewise.
9336 (_mm_maskz_cvt_roundsh_ss): Likewise.
9337 (_mm_cvt_roundsh_sd): Likewise.
9338 (_mm_mask_cvt_roundsh_sd): Likewise.
9339 (_mm_maskz_cvt_roundsh_sd): Likewise.
9340 (_mm_cvtss_sh): Likewise.
9341 (_mm_mask_cvtss_sh): Likewise.
9342 (_mm_maskz_cvtss_sh): Likewise.
9343 (_mm_cvtsd_sh): Likewise.
9344 (_mm_mask_cvtsd_sh): Likewise.
9345 (_mm_maskz_cvtsd_sh): Likewise.
9346 (_mm_cvt_roundss_sh): Likewise.
9347 (_mm_mask_cvt_roundss_sh): Likewise.
9348 (_mm_maskz_cvt_roundss_sh): Likewise.
9349 (_mm_cvt_roundsd_sh): Likewise.
9350 (_mm_mask_cvt_roundsd_sh): Likewise.
9351 (_mm_maskz_cvt_roundsd_sh): Likewise.
9352 * config/i386/i386-builtin-types.def
9353 (V8HF_FTYPE_V2DF_V8HF_V8HF_UQI_INT,
9354 V8HF_FTYPE_V4SF_V8HF_V8HF_UQI_INT,
9355 V2DF_FTYPE_V8HF_V2DF_V2DF_UQI_INT,
9356 V4SF_FTYPE_V8HF_V4SF_V4SF_UQI_INT): Add new builtin types.
9357 * config/i386/i386-builtin.def: Add corrresponding new builtins.
9358 * config/i386/i386-expand.c: Handle new builtin types.
9359 * config/i386/sse.md (VF48_128): New mode iterator.
9360 (avx512fp16_vcvtsh2<ssescalarmodesuffix><mask_scalar_name><round_saeonly_scalar_name>):
9362 (avx512fp16_vcvt<ssescalarmodesuffix>2sh<mask_scalar_name><round_scalar_name>):
9365 2021-09-17 liuhongt <hongtao.liu@intel.com>
9367 * config/i386/avx512fp16intrin.h (_mm512_cvtph_pd):
9369 (_mm512_mask_cvtph_pd): Likewise.
9370 (_mm512_maskz_cvtph_pd): Likewise.
9371 (_mm512_cvt_roundph_pd): Likewise.
9372 (_mm512_mask_cvt_roundph_pd): Likewise.
9373 (_mm512_maskz_cvt_roundph_pd): Likewise.
9374 (_mm512_cvtxph_ps): Likewise.
9375 (_mm512_mask_cvtxph_ps): Likewise.
9376 (_mm512_maskz_cvtxph_ps): Likewise.
9377 (_mm512_cvtx_roundph_ps): Likewise.
9378 (_mm512_mask_cvtx_roundph_ps): Likewise.
9379 (_mm512_maskz_cvtx_roundph_ps): Likewise.
9380 (_mm512_cvtxps_ph): Likewise.
9381 (_mm512_mask_cvtxps_ph): Likewise.
9382 (_mm512_maskz_cvtxps_ph): Likewise.
9383 (_mm512_cvtx_roundps_ph): Likewise.
9384 (_mm512_mask_cvtx_roundps_ph): Likewise.
9385 (_mm512_maskz_cvtx_roundps_ph): Likewise.
9386 (_mm512_cvtpd_ph): Likewise.
9387 (_mm512_mask_cvtpd_ph): Likewise.
9388 (_mm512_maskz_cvtpd_ph): Likewise.
9389 (_mm512_cvt_roundpd_ph): Likewise.
9390 (_mm512_mask_cvt_roundpd_ph): Likewise.
9391 (_mm512_maskz_cvt_roundpd_ph): Likewise.
9392 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_pd):
9394 (_mm_mask_cvtph_pd): Likewise.
9395 (_mm_maskz_cvtph_pd): Likewise.
9396 (_mm256_cvtph_pd): Likewise.
9397 (_mm256_mask_cvtph_pd): Likewise.
9398 (_mm256_maskz_cvtph_pd): Likewise.
9399 (_mm_cvtxph_ps): Likewise.
9400 (_mm_mask_cvtxph_ps): Likewise.
9401 (_mm_maskz_cvtxph_ps): Likewise.
9402 (_mm256_cvtxph_ps): Likewise.
9403 (_mm256_mask_cvtxph_ps): Likewise.
9404 (_mm256_maskz_cvtxph_ps): Likewise.
9405 (_mm_cvtxps_ph): Likewise.
9406 (_mm_mask_cvtxps_ph): Likewise.
9407 (_mm_maskz_cvtxps_ph): Likewise.
9408 (_mm256_cvtxps_ph): Likewise.
9409 (_mm256_mask_cvtxps_ph): Likewise.
9410 (_mm256_maskz_cvtxps_ph): Likewise.
9411 (_mm_cvtpd_ph): Likewise.
9412 (_mm_mask_cvtpd_ph): Likewise.
9413 (_mm_maskz_cvtpd_ph): Likewise.
9414 (_mm256_cvtpd_ph): Likewise.
9415 (_mm256_mask_cvtpd_ph): Likewise.
9416 (_mm256_maskz_cvtpd_ph): Likewise.
9417 * config/i386/i386-builtin.def: Add corresponding new builtins.
9418 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
9419 * config/i386/i386-expand.c: Handle new builtin types.
9420 * config/i386/sse.md
9421 (VF4_128_8_256): New.
9422 (VF48H_AVX512VL): Ditto.
9423 (ssePHmode): Add HF vector modes.
9424 (castmode): Add new convertable modes.
9427 (avx512fp16_vcvt<castmode>2ph_<mode><mask_name><round_name>): Ditto.
9428 (avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
9429 (*avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
9430 (avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
9431 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
9432 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask_1): Ditto.
9433 (avx512fp16_float_extend_ph<mode>2<mask_name><round_saeonly_name>):
9435 (avx512fp16_float_extend_ph<mode>2<mask_name>): Ditto.
9436 (*avx512fp16_float_extend_ph<mode>2_load<mask_name>): Ditto.
9437 (avx512fp16_float_extend_phv2df2<mask_name>): Ditto.
9438 (*avx512fp16_float_extend_phv2df2_load<mask_name>): Ditto.
9440 2021-09-17 liuhongt <hongtao.liu@intel.com>
9442 * config/i386/avx512fp16intrin.h (_mm_cvttsh_i32):
9444 (_mm_cvttsh_u32): Likewise.
9445 (_mm_cvtt_roundsh_i32): Likewise.
9446 (_mm_cvtt_roundsh_u32): Likewise.
9447 (_mm_cvttsh_i64): Likewise.
9448 (_mm_cvttsh_u64): Likewise.
9449 (_mm_cvtt_roundsh_i64): Likewise.
9450 (_mm_cvtt_roundsh_u64): Likewise.
9451 * config/i386/i386-builtin.def: Add corresponding new builtins.
9452 * config/i386/sse.md
9453 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<round_saeonly_name>):
9456 2021-09-17 liuhongt <hongtao.liu@intel.com>
9458 * config/i386/avx512fp16intrin.h (_mm512_cvttph_epi32):
9460 (_mm512_mask_cvttph_epi32): Likewise.
9461 (_mm512_maskz_cvttph_epi32): Likewise.
9462 (_mm512_cvtt_roundph_epi32): Likewise.
9463 (_mm512_mask_cvtt_roundph_epi32): Likewise.
9464 (_mm512_maskz_cvtt_roundph_epi32): Likewise.
9465 (_mm512_cvttph_epu32): Likewise.
9466 (_mm512_mask_cvttph_epu32): Likewise.
9467 (_mm512_maskz_cvttph_epu32): Likewise.
9468 (_mm512_cvtt_roundph_epu32): Likewise.
9469 (_mm512_mask_cvtt_roundph_epu32): Likewise.
9470 (_mm512_maskz_cvtt_roundph_epu32): Likewise.
9471 (_mm512_cvttph_epi64): Likewise.
9472 (_mm512_mask_cvttph_epi64): Likewise.
9473 (_mm512_maskz_cvttph_epi64): Likewise.
9474 (_mm512_cvtt_roundph_epi64): Likewise.
9475 (_mm512_mask_cvtt_roundph_epi64): Likewise.
9476 (_mm512_maskz_cvtt_roundph_epi64): Likewise.
9477 (_mm512_cvttph_epu64): Likewise.
9478 (_mm512_mask_cvttph_epu64): Likewise.
9479 (_mm512_maskz_cvttph_epu64): Likewise.
9480 (_mm512_cvtt_roundph_epu64): Likewise.
9481 (_mm512_mask_cvtt_roundph_epu64): Likewise.
9482 (_mm512_maskz_cvtt_roundph_epu64): Likewise.
9483 (_mm512_cvttph_epi16): Likewise.
9484 (_mm512_mask_cvttph_epi16): Likewise.
9485 (_mm512_maskz_cvttph_epi16): Likewise.
9486 (_mm512_cvtt_roundph_epi16): Likewise.
9487 (_mm512_mask_cvtt_roundph_epi16): Likewise.
9488 (_mm512_maskz_cvtt_roundph_epi16): Likewise.
9489 (_mm512_cvttph_epu16): Likewise.
9490 (_mm512_mask_cvttph_epu16): Likewise.
9491 (_mm512_maskz_cvttph_epu16): Likewise.
9492 (_mm512_cvtt_roundph_epu16): Likewise.
9493 (_mm512_mask_cvtt_roundph_epu16): Likewise.
9494 (_mm512_maskz_cvtt_roundph_epu16): Likewise.
9495 * config/i386/avx512fp16vlintrin.h (_mm_cvttph_epi32):
9497 (_mm_mask_cvttph_epi32): Likewise.
9498 (_mm_maskz_cvttph_epi32): Likewise.
9499 (_mm256_cvttph_epi32): Likewise.
9500 (_mm256_mask_cvttph_epi32): Likewise.
9501 (_mm256_maskz_cvttph_epi32): Likewise.
9502 (_mm_cvttph_epu32): Likewise.
9503 (_mm_mask_cvttph_epu32): Likewise.
9504 (_mm_maskz_cvttph_epu32): Likewise.
9505 (_mm256_cvttph_epu32): Likewise.
9506 (_mm256_mask_cvttph_epu32): Likewise.
9507 (_mm256_maskz_cvttph_epu32): Likewise.
9508 (_mm_cvttph_epi64): Likewise.
9509 (_mm_mask_cvttph_epi64): Likewise.
9510 (_mm_maskz_cvttph_epi64): Likewise.
9511 (_mm256_cvttph_epi64): Likewise.
9512 (_mm256_mask_cvttph_epi64): Likewise.
9513 (_mm256_maskz_cvttph_epi64): Likewise.
9514 (_mm_cvttph_epu64): Likewise.
9515 (_mm_mask_cvttph_epu64): Likewise.
9516 (_mm_maskz_cvttph_epu64): Likewise.
9517 (_mm256_cvttph_epu64): Likewise.
9518 (_mm256_mask_cvttph_epu64): Likewise.
9519 (_mm256_maskz_cvttph_epu64): Likewise.
9520 (_mm_cvttph_epi16): Likewise.
9521 (_mm_mask_cvttph_epi16): Likewise.
9522 (_mm_maskz_cvttph_epi16): Likewise.
9523 (_mm256_cvttph_epi16): Likewise.
9524 (_mm256_mask_cvttph_epi16): Likewise.
9525 (_mm256_maskz_cvttph_epi16): Likewise.
9526 (_mm_cvttph_epu16): Likewise.
9527 (_mm_mask_cvttph_epu16): Likewise.
9528 (_mm_maskz_cvttph_epu16): Likewise.
9529 (_mm256_cvttph_epu16): Likewise.
9530 (_mm256_mask_cvttph_epu16): Likewise.
9531 (_mm256_maskz_cvttph_epu16): Likewise.
9532 * config/i386/i386-builtin.def: Add new builtins.
9533 * config/i386/sse.md
9534 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name><round_saeonly_name>):
9536 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name>): Ditto.
9537 (*avx512fp16_fix<fixunssuffix>_trunc<mode>2_load<mask_name>): Ditto.
9538 (avx512fp16_fix<fixunssuffix>_truncv2di2<mask_name>): Ditto.
9539 (avx512fp16_fix<fixunssuffix>_truncv2di2_load<mask_name>): Ditto.
9541 2021-09-17 liuhongt <hongtao.liu@intel.com>
9543 * config/i386/avx512fp16intrin.h (_mm_cvtsh_i32): New intrinsic.
9544 (_mm_cvtsh_u32): Likewise.
9545 (_mm_cvt_roundsh_i32): Likewise.
9546 (_mm_cvt_roundsh_u32): Likewise.
9547 (_mm_cvtsh_i64): Likewise.
9548 (_mm_cvtsh_u64): Likewise.
9549 (_mm_cvt_roundsh_i64): Likewise.
9550 (_mm_cvt_roundsh_u64): Likewise.
9551 (_mm_cvti32_sh): Likewise.
9552 (_mm_cvtu32_sh): Likewise.
9553 (_mm_cvt_roundi32_sh): Likewise.
9554 (_mm_cvt_roundu32_sh): Likewise.
9555 (_mm_cvti64_sh): Likewise.
9556 (_mm_cvtu64_sh): Likewise.
9557 (_mm_cvt_roundi64_sh): Likewise.
9558 (_mm_cvt_roundu64_sh): Likewise.
9559 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
9560 * config/i386/i386-builtin.def: Add corresponding new builtins.
9561 * config/i386/i386-expand.c (ix86_expand_round_builtin):
9562 Handle new builtin types.
9563 * config/i386/sse.md
9564 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix><round_name>):
9566 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix>_2): Likewise.
9567 (avx512fp16_vcvt<floatsuffix>si2sh<rex64namesuffix><round_name>): Likewise.
9569 2021-09-16 Bill Schmidt <wschmidt@linux.ibm.com>
9571 * config/rs6000/rs6000-c.c (rs6000-builtins.h): New include.
9572 (altivec_resolve_new_overloaded_builtin): New forward decl.
9573 (rs6000_new_builtin_type_compatible): New function.
9574 (altivec_resolve_overloaded_builtin): Call
9575 altivec_resolve_new_overloaded_builtin.
9576 (altivec_build_new_resolved_builtin): New function.
9577 (altivec_resolve_new_overloaded_builtin): Likewise.
9578 * config/rs6000/rs6000-call.c (rs6000_new_builtin_is_supported):
9580 * config/rs6000/rs6000-gen-builtins.c (write_decls): Remove _p from
9581 name of rs6000_new_builtin_is_supported.
9583 2021-09-16 Uroš Bizjak <ubizjak@gmail.com>
9585 * config/i386/i386-protos.h (ix86_decompose_address):
9586 Change return type to bool.
9587 * config/i386/i386.c (ix86_decompose_address): Ditto.
9589 2021-09-16 Tobias Burnus <tobias@codesourcery.com>
9592 * config/rs6000/t-rs6000 (build/rs6000-gen-builtins.o, build/rbtree.o):
9593 Added 'build/' to target, use build/%.o rule.
9594 (build/rs6000-gen-builtins$(build_exeext)): Add 'build/' and
9595 '$(build_exeext)' to target and 'build/' for the *.o files.
9596 (rs6000-builtins.c): Update for those changes; run rs6000-gen-builtins
9599 2021-09-16 Martin Jambor <mjambor@suse.cz>
9601 * cgraph.c (cgraph_node::dump): Do not check caller count sums if
9602 the body has been removed. Remove trailing whitespace.
9604 2021-09-16 Richard Biener <rguenther@suse.de>
9606 PR middle-end/102360
9607 * internal-fn.c (expand_DEFERRED_INIT): Make pattern-init
9608 of non-memory more robust.
9610 2021-09-16 Daniel Cederman <cederman@gaisler.com>
9612 * config/sparc/sparc-opts.h (enum sparc_processor_type): Add LEON5
9613 * config/sparc/sparc.c (struct processor_costs): Add LEON5 costs
9614 (leon5_adjust_cost): Increase cost of store with data dependency
9615 on ALU instruction and FPU anti-dependencies.
9616 (sparc_option_override): Add LEON5 costs
9617 (sparc_adjust_cost): Add LEON5 cost adjustments
9618 * config/sparc/sparc.h: Add LEON5
9619 * config/sparc/sparc.md: Include LEON5 scheduling information
9620 * config/sparc/sparc.opt: Add LEON5
9621 * doc/invoke.texi: Add LEON5
9622 * config/sparc/leon5.md: New file.
9624 2021-09-16 Daniel Cederman <cederman@gaisler.com>
9626 * config/sparc/sparc.md (stack_protect_set32): Add NOP to prevent
9627 sensitive sequence for B2BST errata workaround.
9629 2021-09-16 Daniel Cederman <cederman@gaisler.com>
9631 * config/sparc/sparc.c (sparc_do_work_around_errata): Do not begin
9632 functions with atomic instruction in the UT700 errata workaround.
9634 2021-09-16 Daniel Cederman <cederman@gaisler.com>
9636 * config/sparc/sparc.c (next_active_non_empty_insn): New function
9637 that returns next active non empty assembly instruction.
9638 (sparc_do_work_around_errata): Use new function.
9640 2021-09-16 Daniel Cederman <cederman@gaisler.com>
9642 * config/sparc/sparc.c (store_insn_p): Add predicate for store
9644 (load_insn_p): Add predicate for load attributes.
9645 (sparc_do_work_around_errata): Use new predicates.
9647 2021-09-16 Andreas Larsson <andreas@gaisler.com>
9649 * config/sparc/sparc.c (dump_target_flag_bits): Print bit names for
9652 2021-09-16 Martin Liska <mliska@suse.cz>
9654 * config/mips/netbsd.h: Fix typo in name of a macro.
9656 2021-09-16 liuhongt <hongtao.liu@intel.com>
9658 PR middle-end/102080
9659 * match.pd: Check mask type when doing cond_op related gimple
9661 * tree.c (is_truth_type_for): New function.
9662 * tree.h (is_truth_type_for): New declaration.
9664 2021-09-16 liuhongt <hongtao.liu@intel.com>
9666 * config/i386/avx512fp16intrin.h (_mm512_cvtepi32_ph): New
9668 (_mm512_mask_cvtepi32_ph): Likewise.
9669 (_mm512_maskz_cvtepi32_ph): Likewise.
9670 (_mm512_cvt_roundepi32_ph): Likewise.
9671 (_mm512_mask_cvt_roundepi32_ph): Likewise.
9672 (_mm512_maskz_cvt_roundepi32_ph): Likewise.
9673 (_mm512_cvtepu32_ph): Likewise.
9674 (_mm512_mask_cvtepu32_ph): Likewise.
9675 (_mm512_maskz_cvtepu32_ph): Likewise.
9676 (_mm512_cvt_roundepu32_ph): Likewise.
9677 (_mm512_mask_cvt_roundepu32_ph): Likewise.
9678 (_mm512_maskz_cvt_roundepu32_ph): Likewise.
9679 (_mm512_cvtepi64_ph): Likewise.
9680 (_mm512_mask_cvtepi64_ph): Likewise.
9681 (_mm512_maskz_cvtepi64_ph): Likewise.
9682 (_mm512_cvt_roundepi64_ph): Likewise.
9683 (_mm512_mask_cvt_roundepi64_ph): Likewise.
9684 (_mm512_maskz_cvt_roundepi64_ph): Likewise.
9685 (_mm512_cvtepu64_ph): Likewise.
9686 (_mm512_mask_cvtepu64_ph): Likewise.
9687 (_mm512_maskz_cvtepu64_ph): Likewise.
9688 (_mm512_cvt_roundepu64_ph): Likewise.
9689 (_mm512_mask_cvt_roundepu64_ph): Likewise.
9690 (_mm512_maskz_cvt_roundepu64_ph): Likewise.
9691 (_mm512_cvtepi16_ph): Likewise.
9692 (_mm512_mask_cvtepi16_ph): Likewise.
9693 (_mm512_maskz_cvtepi16_ph): Likewise.
9694 (_mm512_cvt_roundepi16_ph): Likewise.
9695 (_mm512_mask_cvt_roundepi16_ph): Likewise.
9696 (_mm512_maskz_cvt_roundepi16_ph): Likewise.
9697 (_mm512_cvtepu16_ph): Likewise.
9698 (_mm512_mask_cvtepu16_ph): Likewise.
9699 (_mm512_maskz_cvtepu16_ph): Likewise.
9700 (_mm512_cvt_roundepu16_ph): Likewise.
9701 (_mm512_mask_cvt_roundepu16_ph): Likewise.
9702 (_mm512_maskz_cvt_roundepu16_ph): Likewise.
9703 * config/i386/avx512fp16vlintrin.h (_mm_cvtepi32_ph): New
9705 (_mm_mask_cvtepi32_ph): Likewise.
9706 (_mm_maskz_cvtepi32_ph): Likewise.
9707 (_mm256_cvtepi32_ph): Likewise.
9708 (_mm256_mask_cvtepi32_ph): Likewise.
9709 (_mm256_maskz_cvtepi32_ph): Likewise.
9710 (_mm_cvtepu32_ph): Likewise.
9711 (_mm_mask_cvtepu32_ph): Likewise.
9712 (_mm_maskz_cvtepu32_ph): Likewise.
9713 (_mm256_cvtepu32_ph): Likewise.
9714 (_mm256_mask_cvtepu32_ph): Likewise.
9715 (_mm256_maskz_cvtepu32_ph): Likewise.
9716 (_mm_cvtepi64_ph): Likewise.
9717 (_mm_mask_cvtepi64_ph): Likewise.
9718 (_mm_maskz_cvtepi64_ph): Likewise.
9719 (_mm256_cvtepi64_ph): Likewise.
9720 (_mm256_mask_cvtepi64_ph): Likewise.
9721 (_mm256_maskz_cvtepi64_ph): Likewise.
9722 (_mm_cvtepu64_ph): Likewise.
9723 (_mm_mask_cvtepu64_ph): Likewise.
9724 (_mm_maskz_cvtepu64_ph): Likewise.
9725 (_mm256_cvtepu64_ph): Likewise.
9726 (_mm256_mask_cvtepu64_ph): Likewise.
9727 (_mm256_maskz_cvtepu64_ph): Likewise.
9728 (_mm_cvtepi16_ph): Likewise.
9729 (_mm_mask_cvtepi16_ph): Likewise.
9730 (_mm_maskz_cvtepi16_ph): Likewise.
9731 (_mm256_cvtepi16_ph): Likewise.
9732 (_mm256_mask_cvtepi16_ph): Likewise.
9733 (_mm256_maskz_cvtepi16_ph): Likewise.
9734 (_mm_cvtepu16_ph): Likewise.
9735 (_mm_mask_cvtepu16_ph): Likewise.
9736 (_mm_maskz_cvtepu16_ph): Likewise.
9737 (_mm256_cvtepu16_ph): Likewise.
9738 (_mm256_mask_cvtepu16_ph): Likewise.
9739 (_mm256_maskz_cvtepu16_ph): Likewise.
9740 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
9741 * config/i386/i386-builtin.def: Add corresponding new builtins.
9742 * config/i386/i386-expand.c
9743 (ix86_expand_args_builtin): Handle new builtin types.
9744 (ix86_expand_round_builtin): Ditto.
9745 * config/i386/i386-modes.def: Declare V2HF and V6HF.
9746 * config/i386/sse.md (VI2H_AVX512VL): New.
9748 (sseintvecmode): Add HF vector modes.
9749 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode><mask_name><round_name>):
9751 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
9752 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
9753 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
9754 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
9755 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask_1): Ditto.
9756 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
9757 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
9758 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
9759 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
9760 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask_1): Ditto.
9761 * config/i386/subst.md (round_qq2phsuff): New subst_attr.
9763 2021-09-16 liuhongt <hongtao.liu@intel.com>
9765 * config/i386/avx512fp16intrin.h (_mm512_cvtph_epi32):
9767 (_mm512_mask_cvtph_epi32): Likewise.
9768 (_mm512_maskz_cvtph_epi32): Likewise.
9769 (_mm512_cvt_roundph_epi32): Likewise.
9770 (_mm512_mask_cvt_roundph_epi32): Likewise.
9771 (_mm512_maskz_cvt_roundph_epi32): Likewise.
9772 (_mm512_cvtph_epu32): Likewise.
9773 (_mm512_mask_cvtph_epu32): Likewise.
9774 (_mm512_maskz_cvtph_epu32): Likewise.
9775 (_mm512_cvt_roundph_epu32): Likewise.
9776 (_mm512_mask_cvt_roundph_epu32): Likewise.
9777 (_mm512_maskz_cvt_roundph_epu32): Likewise.
9778 (_mm512_cvtph_epi64): Likewise.
9779 (_mm512_mask_cvtph_epi64): Likewise.
9780 (_mm512_maskz_cvtph_epi64): Likewise.
9781 (_mm512_cvt_roundph_epi64): Likewise.
9782 (_mm512_mask_cvt_roundph_epi64): Likewise.
9783 (_mm512_maskz_cvt_roundph_epi64): Likewise.
9784 (_mm512_cvtph_epu64): Likewise.
9785 (_mm512_mask_cvtph_epu64): Likewise.
9786 (_mm512_maskz_cvtph_epu64): Likewise.
9787 (_mm512_cvt_roundph_epu64): Likewise.
9788 (_mm512_mask_cvt_roundph_epu64): Likewise.
9789 (_mm512_maskz_cvt_roundph_epu64): Likewise.
9790 (_mm512_cvtph_epi16): Likewise.
9791 (_mm512_mask_cvtph_epi16): Likewise.
9792 (_mm512_maskz_cvtph_epi16): Likewise.
9793 (_mm512_cvt_roundph_epi16): Likewise.
9794 (_mm512_mask_cvt_roundph_epi16): Likewise.
9795 (_mm512_maskz_cvt_roundph_epi16): Likewise.
9796 (_mm512_cvtph_epu16): Likewise.
9797 (_mm512_mask_cvtph_epu16): Likewise.
9798 (_mm512_maskz_cvtph_epu16): Likewise.
9799 (_mm512_cvt_roundph_epu16): Likewise.
9800 (_mm512_mask_cvt_roundph_epu16): Likewise.
9801 (_mm512_maskz_cvt_roundph_epu16): Likewise.
9802 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_epi32):
9804 (_mm_mask_cvtph_epi32): Likewise.
9805 (_mm_maskz_cvtph_epi32): Likewise.
9806 (_mm256_cvtph_epi32): Likewise.
9807 (_mm256_mask_cvtph_epi32): Likewise.
9808 (_mm256_maskz_cvtph_epi32): Likewise.
9809 (_mm_cvtph_epu32): Likewise.
9810 (_mm_mask_cvtph_epu32): Likewise.
9811 (_mm_maskz_cvtph_epu32): Likewise.
9812 (_mm256_cvtph_epu32): Likewise.
9813 (_mm256_mask_cvtph_epu32): Likewise.
9814 (_mm256_maskz_cvtph_epu32): Likewise.
9815 (_mm_cvtph_epi64): Likewise.
9816 (_mm_mask_cvtph_epi64): Likewise.
9817 (_mm_maskz_cvtph_epi64): Likewise.
9818 (_mm256_cvtph_epi64): Likewise.
9819 (_mm256_mask_cvtph_epi64): Likewise.
9820 (_mm256_maskz_cvtph_epi64): Likewise.
9821 (_mm_cvtph_epu64): Likewise.
9822 (_mm_mask_cvtph_epu64): Likewise.
9823 (_mm_maskz_cvtph_epu64): Likewise.
9824 (_mm256_cvtph_epu64): Likewise.
9825 (_mm256_mask_cvtph_epu64): Likewise.
9826 (_mm256_maskz_cvtph_epu64): Likewise.
9827 (_mm_cvtph_epi16): Likewise.
9828 (_mm_mask_cvtph_epi16): Likewise.
9829 (_mm_maskz_cvtph_epi16): Likewise.
9830 (_mm256_cvtph_epi16): Likewise.
9831 (_mm256_mask_cvtph_epi16): Likewise.
9832 (_mm256_maskz_cvtph_epi16): Likewise.
9833 (_mm_cvtph_epu16): Likewise.
9834 (_mm_mask_cvtph_epu16): Likewise.
9835 (_mm_maskz_cvtph_epu16): Likewise.
9836 (_mm256_cvtph_epu16): Likewise.
9837 (_mm256_mask_cvtph_epu16): Likewise.
9838 (_mm256_maskz_cvtph_epu16): Likewise.
9839 * config/i386/i386-builtin-types.def: Add new builtin types.
9840 * config/i386/i386-builtin.def: Add new builtins.
9841 * config/i386/i386-expand.c
9842 (ix86_expand_args_builtin): Handle new builtin types.
9843 (ix86_expand_round_builtin): Ditto.
9844 * config/i386/sse.md (sseintconvert): New.
9846 (UNSPEC_US_FIX_NOTRUNC): Ditto.
9847 (sseintconvertsignprefix): Ditto.
9848 (avx512fp16_vcvtph2<sseintconvertsignprefix><sseintconvert>_<mode><mask_name><round_name>):
9851 2021-09-16 liuhongt <hongtao.liu@intel.com>
9853 * config/i386/avx512fp16intrin.h: (_mm_cvtsi16_si128):
9855 (_mm_cvtsi128_si16): Likewise.
9856 (_mm_mask_load_sh): Likewise.
9857 (_mm_maskz_load_sh): Likewise.
9858 (_mm_mask_store_sh): Likewise.
9859 (_mm_move_sh): Likewise.
9860 (_mm_mask_move_sh): Likewise.
9861 (_mm_maskz_move_sh): Likewise.
9862 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
9863 * config/i386/i386-builtin.def: Add corresponding new builtins.
9864 * config/i386/i386-expand.c
9865 (ix86_expand_special_args_builtin): Handle new builtin types.
9866 (ix86_expand_vector_init_one_nonzero): Adjust for FP16 target.
9867 * config/i386/sse.md (VI2F): New mode iterator.
9868 (vec_set<mode>_0): Use new mode iterator.
9869 (avx512f_mov<ssescalarmodelower>_mask): Adjust for HF vector mode.
9870 (avx512f_store<mode>_mask): Ditto.
9872 2021-09-16 Kewen Lin <linkw@linux.ibm.com>
9874 * config/rs6000/rs6000.opt (-mtoc-fusion): Remove.
9876 2021-09-15 David Edelsohn <dje.gcc@gmail.com>
9878 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
9879 Proceed if no symbol summary or the symbol alias flag is false.
9881 2021-09-15 Jakub Jelinek <jakub@redhat.com>
9885 * varasm.c (output_constructor_regular_field): Instead of assertion
9886 that array_size_for_constructor result is equal to size of
9887 TREE_TYPE (local->val) in bytes, assert that the type size is greater
9888 or equal to array_size_for_constructor result and use type size as
9891 2021-09-15 Martin Liska <mliska@suse.cz>
9894 * config/i386/vxworks.h: Use new macro TARGET_CPU_P.
9896 2021-09-15 Martin Liska <mliska@suse.cz>
9899 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
9900 Check that we have a symbol summary for a symbol.
9902 2021-09-15 Richard Biener <rguenther@suse.de>
9905 * config/rs6000/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
9906 to inherit from elfos.h
9908 2021-09-15 liuhongt <hongtao.liu@intel.com>
9911 * config/i386/i386-expand.c
9912 (ix86_expand_vector_init_interleave): Use puncklwd to pack 2
9914 (ix86_expand_vector_set): Use blendw instead of pinsrw.
9915 * config/i386/i386.c (ix86_can_change_mode_class): Adjust for
9916 AVX512FP16 which supports 16bit vector load.
9917 * config/i386/sse.md (avx512bw_interleave_highv32hi<mask_name>):
9919 (avx512bw_interleave_high<mode><mask_name>): .. this, and
9920 extend to V32HFmode.
9921 (avx2_interleave_highv16hi<mask_name>): Rename to ..
9922 (avx2_interleave_high<mode><mask_name>): .. this, and extend
9924 (vec_interleave_highv8hi<mask_name>): Rename to ..
9925 (vec_interleave_high<mode><mask_name>): .. this, and extend to V8HFmode.
9926 (<mask_codefor>avx512bw_interleave_lowv32hi<mask_name>):
9928 (<mask_codefor>avx512bw_interleave_low<mode><mask_name>):
9929 this, and extend to V32HFmode.
9930 (avx2_interleave_lowv16hi<mask_name>): Rename to ..
9931 (avx2_interleave_low<mode><mask_name>): .. this, and extend to V16HFmode.
9932 (vec_interleave_lowv8hi<mask_name>): Rename to ..
9933 (vec_interleave_low<mode><mask_name>): .. this, and extend to V8HFmode.
9934 (sse4_1_pblendw): Rename to ..
9935 (sse4_1_pblend<blendsuf>): .. this, and extend to V8HFmode.
9936 (avx2_pblendph): New define_expand.
9937 (<sse2p4_1>_pinsr<ssemodesuffix>): Refactor, use
9938 sseintmodesuffix instead of ssemodesuffix.
9939 (blendsuf): New mode attr.
9941 2021-09-15 Richard Biener <rguenther@suse.de>
9943 * tree-vectorizer.h (dr_misalignment): Move out of line.
9944 (dr_target_alignment): New.
9945 (DR_TARGET_ALIGNMENT): Wrap dr_target_alignment.
9946 (set_dr_target_alignment): New.
9947 (SET_DR_TARGET_ALIGNMENT): Wrap set_dr_target_alignment.
9948 * tree-vect-data-refs.c (dr_misalignment): Compute and
9949 return the group members misalignment.
9950 (vect_compute_data_ref_alignment): Use SET_DR_TARGET_ALIGNMENT.
9951 (vect_analyze_data_refs_alignment): Compute alignment only
9952 for the first element of a DR group.
9953 (vect_slp_analyze_node_alignment): Likewise.
9955 2021-09-15 Hongyu Wang <hongyu.wang@intel.com>
9957 * config/i386/avx512fp16intrin.h: Adjust all builtin calls.
9958 * config/i386/avx512fp16vlintrin.h: Likewise.
9959 * config/i386/i386-builtin.def: Adjust builtin name and
9960 enumeration to match AVX512F style.
9962 2021-09-15 Richard Biener <rguenther@suse.de>
9964 PR tree-optimization/102318
9965 * tree-vect-loop.c (vect_transform_cycle_phi): Revert
9966 previous change and do the mode conversion separately from
9967 the sign conversion.
9969 2021-09-15 Hongtao Liu <hongtao.liu@intel.com>
9970 Peter Cordes <peter@cordes.ca>
9973 * config/i386/sse.md (extract_suf): Add V8SF/V8SI/V4DF/V4DI.
9974 (*vec_extract<mode><ssescalarmodelower>_valign): Output
9975 vextract{i,f}{32x4,64x2} instruction when byte_offset % 16 ==
9978 2021-09-15 Richard Biener <rguenther@suse.de>
9980 * config.gcc: Remove vax-*-openbsd* configuration.
9982 2021-09-15 Richard Biener <rguenther@suse.de>
9984 * config.gcc: Remove m68k-openbsd.
9986 2021-09-15 Max Filippov <jcmvbkbc@gmail.com>
9989 * config/xtensa/t-xtensa (TM_H): Add include/xtensa-config.h.
9991 2021-09-14 Peter Bergner <bergner@linux.ibm.com>
9993 * config/rs6000/mma.md (unspec): Delete UNSPEC_MMA_XXSETACCZ.
9994 (unspecv): Add UNSPECV_MMA_XXSETACCZ.
9995 (*mma_xxsetaccz): Delete.
9996 (mma_xxsetaccz): Change to define_insn. Remove operand 1.
9997 Use UNSPECV_MMA_XXSETACCZ. Update comment.
9998 * config/rs6000/rs6000.c (rs6000_rtx_costs): Use UNSPECV_MMA_XXSETACCZ.
10000 2021-09-14 Iain Sandoe <iain@sandoe.co.uk>
10002 * Makefile.in: Remove variables related to applying no-PIE
10003 to the exes on $build.
10004 * configure: Regenerate.
10005 * configure.ac: Remove configuration related to applying
10006 no-PIE to the exes on $build.
10008 2021-09-14 Claudiu Zissulescu <claziss@synopsys.com>
10010 * config/arc/arc.md (doloop_end): Add missing mode.
10011 (loop_end): Likewise.
10013 2021-09-14 Jakub Jelinek <jakub@redhat.com>
10015 * gimplify.c (goa_stabilize_expr): Add depth argument, propagate
10016 it to recursive calls, for depth above 7 just gimplify or return.
10017 Perform a test even for MODIFY_EXPR, ADDR_EXPR, COMPOUND_EXPR with
10018 __builtin_clear_padding and TARGET_EXPR.
10019 (gimplify_omp_atomic): Adjust goa_stabilize_expr callers.
10021 2021-09-14 liuhongt <hongtao.liu@intel.com>
10023 * config/i386/avx512fp16intrin.h (_mm_fpclass_sh_mask):
10025 (_mm_mask_fpclass_sh_mask): Likewise.
10026 (_mm512_mask_fpclass_ph_mask): Likewise.
10027 (_mm512_fpclass_ph_mask): Likewise.
10028 (_mm_getexp_sh): Likewise.
10029 (_mm_mask_getexp_sh): Likewise.
10030 (_mm_maskz_getexp_sh): Likewise.
10031 (_mm512_getexp_ph): Likewise.
10032 (_mm512_mask_getexp_ph): Likewise.
10033 (_mm512_maskz_getexp_ph): Likewise.
10034 (_mm_getexp_round_sh): Likewise.
10035 (_mm_mask_getexp_round_sh): Likewise.
10036 (_mm_maskz_getexp_round_sh): Likewise.
10037 (_mm512_getexp_round_ph): Likewise.
10038 (_mm512_mask_getexp_round_ph): Likewise.
10039 (_mm512_maskz_getexp_round_ph): Likewise.
10040 (_mm_getmant_sh): Likewise.
10041 (_mm_mask_getmant_sh): Likewise.
10042 (_mm_maskz_getmant_sh): Likewise.
10043 (_mm512_getmant_ph): Likewise.
10044 (_mm512_mask_getmant_ph): Likewise.
10045 (_mm512_maskz_getmant_ph): Likewise.
10046 (_mm_getmant_round_sh): Likewise.
10047 (_mm_mask_getmant_round_sh): Likewise.
10048 (_mm_maskz_getmant_round_sh): Likewise.
10049 (_mm512_getmant_round_ph): Likewise.
10050 (_mm512_mask_getmant_round_ph): Likewise.
10051 (_mm512_maskz_getmant_round_ph): Likewise.
10052 * config/i386/avx512fp16vlintrin.h (_mm_mask_fpclass_ph_mask):
10054 (_mm_fpclass_ph_mask): Likewise.
10055 (_mm256_mask_fpclass_ph_mask): Likewise.
10056 (_mm256_fpclass_ph_mask): Likewise.
10057 (_mm256_getexp_ph): Likewise.
10058 (_mm256_mask_getexp_ph): Likewise.
10059 (_mm256_maskz_getexp_ph): Likewise.
10060 (_mm_getexp_ph): Likewise.
10061 (_mm_mask_getexp_ph): Likewise.
10062 (_mm_maskz_getexp_ph): Likewise.
10063 (_mm256_getmant_ph): Likewise.
10064 (_mm256_mask_getmant_ph): Likewise.
10065 (_mm256_maskz_getmant_ph): Likewise.
10066 (_mm_getmant_ph): Likewise.
10067 (_mm_mask_getmant_ph): Likewise.
10068 (_mm_maskz_getmant_ph): Likewise.
10069 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
10070 * config/i386/i386-builtin.def: Add corresponding new builtins.
10071 * config/i386/i386-expand.c
10072 (ix86_expand_args_builtin): Handle new builtin types.
10073 (ix86_expand_round_builtin): Ditto.
10074 * config/i386/sse.md (vecmemsuffix): Add HF vector modes.
10075 (<avx512>_getexp<mode><mask_name><round_saeonly_name>): Adjust
10076 to support HF vector modes.
10077 (avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name):
10079 (avx512dq_fpclass<mode><mask_scalar_merge_name>): Ditto.
10080 (avx512dq_vmfpclass<mode><mask_scalar_merge_name>): Ditto.
10081 (<avx512>_getmant<mode><mask_name><round_saeonly_name>): Ditto.
10082 (avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
10085 2021-09-14 liuhongt <hongtao.liu@intel.com>
10087 * config/i386/avx512fp16intrin.h (_mm512_reduce_ph):
10089 (_mm512_mask_reduce_ph): Likewise.
10090 (_mm512_maskz_reduce_ph): Likewise.
10091 (_mm512_reduce_round_ph): Likewise.
10092 (_mm512_mask_reduce_round_ph): Likewise.
10093 (_mm512_maskz_reduce_round_ph): Likewise.
10094 (_mm_reduce_sh): Likewise.
10095 (_mm_mask_reduce_sh): Likewise.
10096 (_mm_maskz_reduce_sh): Likewise.
10097 (_mm_reduce_round_sh): Likewise.
10098 (_mm_mask_reduce_round_sh): Likewise.
10099 (_mm_maskz_reduce_round_sh): Likewise.
10100 (_mm512_roundscale_ph): Likewise.
10101 (_mm512_mask_roundscale_ph): Likewise.
10102 (_mm512_maskz_roundscale_ph): Likewise.
10103 (_mm512_roundscale_round_ph): Likewise.
10104 (_mm512_mask_roundscale_round_ph): Likewise.
10105 (_mm512_maskz_roundscale_round_ph): Likewise.
10106 (_mm_roundscale_sh): Likewise.
10107 (_mm_mask_roundscale_sh): Likewise.
10108 (_mm_maskz_roundscale_sh): Likewise.
10109 (_mm_roundscale_round_sh): Likewise.
10110 (_mm_mask_roundscale_round_sh): Likewise.
10111 (_mm_maskz_roundscale_round_sh): Likewise.
10112 * config/i386/avx512fp16vlintrin.h: (_mm_reduce_ph):
10114 (_mm_mask_reduce_ph): Likewise.
10115 (_mm_maskz_reduce_ph): Likewise.
10116 (_mm256_reduce_ph): Likewise.
10117 (_mm256_mask_reduce_ph): Likewise.
10118 (_mm256_maskz_reduce_ph): Likewise.
10119 (_mm_roundscale_ph): Likewise.
10120 (_mm_mask_roundscale_ph): Likewise.
10121 (_mm_maskz_roundscale_ph): Likewise.
10122 (_mm256_roundscale_ph): Likewise.
10123 (_mm256_mask_roundscale_ph): Likewise.
10124 (_mm256_maskz_roundscale_ph): Likewise.
10125 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
10126 * config/i386/i386-builtin.def: Add corresponding new builtins.
10127 * config/i386/i386-expand.c
10128 (ix86_expand_args_builtin): Handle new builtin types.
10129 (ix86_expand_round_builtin): Ditto.
10130 * config/i386/sse.md (<mask_codefor>reducep<mode><mask_name>):
10132 (<mask_codefor>reducep<mode><mask_name><round_saeonly_name>):
10133 ... this, and adjust for round operands.
10134 (reduces<mode><mask_scalar_name>): Likewise, with ...
10135 (reduces<mode><mask_scalar_name><round_saeonly_scalar_name):
10137 (<avx512>_rndscale<mode><mask_name><round_saeonly_name>):
10138 Adjust for HF vector modes.
10139 (avx512f_rndscale<mode><mask_scalar_name><round_saeonly_scalar_name>):
10141 (*avx512f_rndscale<mode><round_saeonly_name>): Ditto.
10143 2021-09-14 liuhongt <hongtao.liu@intel.com>
10145 * config/i386/avx512fp16intrin.h: (_mm512_rcp_ph):
10147 (_mm512_mask_rcp_ph): Likewise.
10148 (_mm512_maskz_rcp_ph): Likewise.
10149 (_mm_rcp_sh): Likewise.
10150 (_mm_mask_rcp_sh): Likewise.
10151 (_mm_maskz_rcp_sh): Likewise.
10152 (_mm512_scalef_ph): Likewise.
10153 (_mm512_mask_scalef_ph): Likewise.
10154 (_mm512_maskz_scalef_ph): Likewise.
10155 (_mm512_scalef_round_ph): Likewise.
10156 (_mm512_mask_scalef_round_ph): Likewise.
10157 (_mm512_maskz_scalef_round_ph): Likewise.
10158 (_mm_scalef_sh): Likewise.
10159 (_mm_mask_scalef_sh): Likewise.
10160 (_mm_maskz_scalef_sh): Likewise.
10161 (_mm_scalef_round_sh): Likewise.
10162 (_mm_mask_scalef_round_sh): Likewise.
10163 (_mm_maskz_scalef_round_sh): Likewise.
10164 * config/i386/avx512fp16vlintrin.h (_mm_rcp_ph):
10166 (_mm256_rcp_ph): Likewise.
10167 (_mm_mask_rcp_ph): Likewise.
10168 (_mm256_mask_rcp_ph): Likewise.
10169 (_mm_maskz_rcp_ph): Likewise.
10170 (_mm256_maskz_rcp_ph): Likewise.
10171 (_mm_scalef_ph): Likewise.
10172 (_mm256_scalef_ph): Likewise.
10173 (_mm_mask_scalef_ph): Likewise.
10174 (_mm256_mask_scalef_ph): Likewise.
10175 (_mm_maskz_scalef_ph): Likewise.
10176 (_mm256_maskz_scalef_ph): Likewise.
10177 * config/i386/i386-builtin.def: Add new builtins.
10178 * config/i386/sse.md (VFH_AVX512VL): New.
10179 (avx512fp16_rcp<mode>2<mask_name>): Ditto.
10180 (avx512fp16_vmrcpv8hf2<mask_scalar_name>): Ditto.
10181 (avx512f_vmscalef<mode><mask_scalar_name><round_scalar_name>):
10182 Adjust to support HF vector modes.
10183 (<avx512>_scalef<mode><mask_name><round_name>): Ditto.
10185 2021-09-14 liuhongt <hongtao.liu@intel.com>
10187 * config/i386/avx512fp16intrin.h: (_mm512_sqrt_ph):
10189 (_mm512_mask_sqrt_ph): Likewise.
10190 (_mm512_maskz_sqrt_ph): Likewise.
10191 (_mm512_sqrt_round_ph): Likewise.
10192 (_mm512_mask_sqrt_round_ph): Likewise.
10193 (_mm512_maskz_sqrt_round_ph): Likewise.
10194 (_mm512_rsqrt_ph): Likewise.
10195 (_mm512_mask_rsqrt_ph): Likewise.
10196 (_mm512_maskz_rsqrt_ph): Likewise.
10197 (_mm_rsqrt_sh): Likewise.
10198 (_mm_mask_rsqrt_sh): Likewise.
10199 (_mm_maskz_rsqrt_sh): Likewise.
10200 (_mm_sqrt_sh): Likewise.
10201 (_mm_mask_sqrt_sh): Likewise.
10202 (_mm_maskz_sqrt_sh): Likewise.
10203 (_mm_sqrt_round_sh): Likewise.
10204 (_mm_mask_sqrt_round_sh): Likewise.
10205 (_mm_maskz_sqrt_round_sh): Likewise.
10206 * config/i386/avx512fp16vlintrin.h (_mm_sqrt_ph): New intrinsic.
10207 (_mm256_sqrt_ph): Likewise.
10208 (_mm_mask_sqrt_ph): Likewise.
10209 (_mm256_mask_sqrt_ph): Likewise.
10210 (_mm_maskz_sqrt_ph): Likewise.
10211 (_mm256_maskz_sqrt_ph): Likewise.
10212 (_mm_rsqrt_ph): Likewise.
10213 (_mm256_rsqrt_ph): Likewise.
10214 (_mm_mask_rsqrt_ph): Likewise.
10215 (_mm256_mask_rsqrt_ph): Likewise.
10216 (_mm_maskz_rsqrt_ph): Likewise.
10217 (_mm256_maskz_rsqrt_ph): Likewise.
10218 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
10219 * config/i386/i386-builtin.def: Add corresponding new builtins.
10220 * config/i386/i386-expand.c
10221 (ix86_expand_args_builtin): Handle new builtins.
10222 (ix86_expand_round_builtin): Ditto.
10223 * config/i386/sse.md (VF_AVX512FP16VL): New.
10224 (sqrt<mode>2): Adjust for HF vector modes.
10225 (<sse>_sqrt<mode>2<mask_name><round_name>): Likewise.
10226 (<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
10228 (<sse>_rsqrt<mode>2<mask_name>): New.
10229 (avx512fp16_vmrsqrtv8hf2<mask_scalar_name>): Likewise.
10231 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
10233 PR bootstrap/101574
10234 * diagnostic-spec.c (warning_suppressed_at, copy_warning): Handle
10235 'RESERVED_LOCATION_P' locations.
10236 * warning-control.cc (get_nowarn_spec, suppress_warning)
10237 (copy_warning): Likewise.
10239 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
10241 * diagnostic-spec.h (typedef xint_hash_t): Use 'location_t' instead of...
10242 (typedef key_type_t): ... this. Remove.
10243 (nowarn_map): Document.
10244 * diagnostic-spec.c (nowarn_map): Likewise.
10245 * warning-control.cc (convert_to_key): Evolve functions into...
10246 (get_location): ... these. Adjust all users.
10248 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
10250 * warning-control.cc (copy_warning): Remove 'nowarn_map' setup.
10252 2021-09-13 Jason Merrill <jason@redhat.com>
10254 * params.opt: Add destructive-interference-size and
10255 constructive-interference-size.
10256 * doc/invoke.texi: Document them.
10257 * config/aarch64/aarch64.c (aarch64_override_options_internal):
10259 * config/arm/arm.c (arm_option_override): Set them.
10260 * config/i386/i386-options.c (ix86_option_override_internal):
10263 2021-09-13 Martin Liska <mliska@suse.cz>
10264 H.J. Lu <hjl.tools@gmail.com>
10267 * common/config/i386/cpuinfo.h (cpu_indicator_init): Add support
10268 for x86-64 micro levels for __builtin_cpu_supports.
10269 * common/config/i386/i386-cpuinfo.h (enum feature_priority):
10270 Add priorities for the micro-arch levels.
10271 (enum processor_features): Add new features.
10272 * common/config/i386/i386-isas.h: Add micro-arch features.
10273 * config/i386/i386-builtins.c (get_builtin_code_for_version):
10274 Support the micro-arch levels by callsing
10275 __builtin_cpu_supports.
10276 * doc/extend.texi: Document that the levels are support by
10277 __builtin_cpu_supports.
10279 2021-09-13 Andrew Pinski <apinski@marvell.com>
10282 * config/aarch64/aarch64-builtins.c (aarch64_fold_builtin_lane_check):
10284 (aarch64_general_fold_builtin): Handle AARCH64_SIMD_BUILTIN_LANE_CHECK.
10285 (aarch64_general_gimple_fold_builtin): Likewise.
10287 2021-09-13 Andrew Pinski <apinski@marvell.com>
10289 * config.gcc: Add m32r-*-linux* and m32rle-*-linux*
10290 to the Unsupported targets list.
10291 Remove support for m32r-*-linux* and m32rle-*-linux*.
10292 * config/m32r/linux.h: Removed.
10293 * config/m32r/t-linux: Removed.
10295 2021-09-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10298 * config/aarch64/aarch64.c (aarch64_classify_address): Don't allow
10299 register index for SVE predicate modes.
10301 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
10303 * tree-ssa-threadbackward.c
10304 (back_threader_profitability::profitable_path_p): Remove FSM
10306 (back_threader_registry::register_path): Same.
10307 * tree-ssa-threadedge.c
10308 (jump_threader::simplify_control_stmt_condition): Same.
10309 * tree-ssa-threadupdate.c (jt_path_registry::jt_path_registry):
10310 Add backedge_threads argument.
10311 (fwd_jt_path_registry::fwd_jt_path_registry): Pass
10312 backedge_threads argument.
10313 (back_jt_path_registry::back_jt_path_registry): Same.
10314 (dump_jump_thread_path): Adjust for FSM removal.
10315 (back_jt_path_registry::rewire_first_differing_edge): Same.
10316 (back_jt_path_registry::adjust_paths_after_duplication): Same.
10317 (back_jt_path_registry::update_cfg): Same.
10318 (jt_path_registry::register_jump_thread): Same.
10319 * tree-ssa-threadupdate.h (enum jump_thread_edge_type): Remove
10321 (class back_jt_path_registry): Add backedge_threads to
10324 2021-09-13 Martin Liska <mliska@suse.cz>
10327 * asan.h (sanitize_coverage_p): Handle when fn == NULL.
10329 2021-09-13 H.J. Lu <hjl.tools@gmail.com>
10332 * config/i386/i386.h (TARGET_AVX256_MOVE_BY_PIECES): New.
10333 (TARGET_AVX256_STORE_BY_PIECES): Likewise.
10334 (MOVE_MAX): Check TARGET_AVX256_MOVE_BY_PIECES and
10335 TARGET_AVX256_STORE_BY_PIECES instead of
10336 TARGET_AVX256_SPLIT_UNALIGNED_LOAD and
10337 TARGET_AVX256_SPLIT_UNALIGNED_STORE.
10338 (STORE_MAX_PIECES): Check TARGET_AVX256_STORE_BY_PIECES instead
10339 of TARGET_AVX256_SPLIT_UNALIGNED_STORE.
10340 * config/i386/x86-tune.def (X86_TUNE_AVX256_MOVE_BY_PIECES): New.
10341 (X86_TUNE_AVX256_STORE_BY_PIECES): Likewise.
10343 2021-09-13 liuhongt <hongtao.liu@intel.com>
10345 PR bootstrap/102302
10346 * expmed.c (extract_bit_field_using_extv): Use
10347 gen_lowpart_if_possible instead of gen_lowpart to avoid ICE.
10349 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
10351 * Makefile.in (OBJS): Add value-pointer-equiv.o.
10352 * gimple-ssa-evrp.c (class ssa_equiv_stack): Move to
10353 value-pointer-equiv.*.
10354 (ssa_equiv_stack::ssa_equiv_stack): Same.
10355 (ssa_equiv_stack::enter): Same.
10356 (ssa_equiv_stack::leave): Same.
10357 (ssa_equiv_stack::push_replacement): Same.
10358 (ssa_equiv_stack::get_replacement): Same.
10359 (is_pointer_ssa): Same.
10360 (class pointer_equiv_analyzer): Same.
10361 (pointer_equiv_analyzer::pointer_equiv_analyzer): Same.
10362 (pointer_equiv_analyzer::~pointer_equiv_analyzer): Same.
10363 (pointer_equiv_analyzer::set_global_equiv): Same.
10364 (pointer_equiv_analyzer::set_cond_equiv): Same.
10365 (pointer_equiv_analyzer::get_equiv): Same.
10366 (pointer_equiv_analyzer::enter): Same.
10367 (pointer_equiv_analyzer::leave): Same.
10368 (pointer_equiv_analyzer::get_equiv_expr): Same.
10369 (pta_valueize): Same.
10370 (pointer_equiv_analyzer::visit_stmt): Same.
10371 (pointer_equiv_analyzer::visit_edge): Same.
10372 (hybrid_folder::value_of_expr): Same.
10373 (hybrid_folder::value_on_edge): Same.
10374 * value-pointer-equiv.cc: New file.
10375 * value-pointer-equiv.h: New file.
10377 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
10380 * gimple-fold.c (gimple_fold_builtin_memory_op): Allow folding
10381 memcpy if the size is not more than MOVE_MAX * MOVE_RATIO.
10383 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
10386 * config/arm/arm.md (movmisaligndi): New define_expand.
10387 * config/arm/vec-common.md (movmisalign<mode>): Iterate over VDQ mode.
10389 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
10392 * emit-rtl.c (gen_highpart): Use adjust_address to handle
10393 MEM rather than calling simplify_gen_subreg.
10395 2021-09-13 Jan-Benedict Glaw <jbglaw@ług-owl.de>
10397 * config/alpha/vms.h (INIT_CUMULATIVE_ARGS): Wrap multi-statment
10398 define into a block.
10400 2021-09-13 Richard Biener <rguenther@suse.de>
10402 * config/darwin.h (DARWIN_PREFER_DWARF): Do not define.
10403 * config/i386/darwin.h (PREFERRED_DEBUGGING_TYPE): Do not
10404 change based on DARWIN_PREFER_DWARF not being defined.
10406 2021-09-13 Richard Biener <rguenther@suse.de>
10408 * config/i386/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
10409 to inherit from elfos.h
10411 2021-09-13 Richard Biener <rguenther@suse.de>
10413 * config.gcc: Add cr16-*-* to the list of obsoleted targets.
10415 2021-09-13 Richard Biener <rguenther@suse.de>
10417 * config/avr/elf.h (PREFERRED_DEBUGGING_TYPE): Remove
10418 override, pick up DWARF2_DEBUG define from elfos.h
10420 2021-09-13 Richard Biener <rguenther@suse.de>
10422 * config/rx/rx.h (PREFERRED_DEBUGGING_TYPE): Always define to
10425 2021-09-13 Richard Biener <rguenther@suse.de>
10427 * config/alpha/vms.h (PREFERRED_DEBUGGING_TYPE): Define to
10430 2021-09-13 Richard Biener <rguenther@suse.de>
10432 * config/i386/cygming.h: Always default to DWARF2 debugging.
10433 Do not define DBX_DEBUGGING_INFO, that's done via dbxcoff.h
10435 * doc/install.texi: Document binutils 2.16 as minimum
10436 requirement for mingw.
10438 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
10440 * config/rs6000/rs6000.c (struct rs6000_cost_data): New members
10441 nstmts, nloads and extra_ctor_cost.
10442 (rs6000_density_test): Add load density related heuristics. Do
10443 extra costing on vector construction statements if need.
10444 (rs6000_init_cost): Init new members.
10445 (rs6000_update_target_cost_per_stmt): New function.
10446 (rs6000_add_stmt_cost): Factor vect_nonmem hunk out to function
10447 rs6000_update_target_cost_per_stmt and call it.
10449 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
10451 * config/rs6000/rs6000.c (struct rs6000_cost_data): Remove typedef.
10452 (rs6000_init_cost): Adjust.
10454 2021-09-13 liuhongt <hongtao.liu@intel.com>
10456 * config/i386/i386.md: (UNSPEC_COPYSIGN): Remove.
10457 (UNSPEC_XORSIGN): Ditto.
10459 2021-09-12 Roger Sayle <roger@nextmovesoftware.com>
10461 * expr.c (convert_move): Preserve SUBREG_PROMOTED_VAR_P when
10462 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
10465 2021-09-11 Aldy Hernandez <aldyh@redhat.com>
10467 * tree-ssa-threadbackward.c (class back_threader_registry): Use
10468 back_jt_path_registry.
10469 * tree-ssa-threadedge.c (jump_threader::jump_threader): Use
10470 fwd_jt_path_registry.
10471 * tree-ssa-threadedge.h (class jump_threader): Same..
10472 * tree-ssa-threadupdate.c
10473 (jump_thread_path_registry::jump_thread_path_registry): Rename...
10474 (jt_path_registry::jt_path_registry): ...to this.
10475 (jump_thread_path_registry::~jump_thread_path_registry): Rename...
10476 (jt_path_registry::~jt_path_registry): ...this.
10477 (fwd_jt_path_registry::fwd_jt_path_registry): New.
10478 (fwd_jt_path_registry::~fwd_jt_path_registry): New.
10479 (jump_thread_path_registry::allocate_thread_edge): Rename...
10480 (jt_path_registry::allocate_thread_edge): ...to this.
10481 (jump_thread_path_registry::allocate_thread_path): Rename...
10482 (jt_path_registry::allocate_thread_path): ...to this.
10483 (jump_thread_path_registry::lookup_redirection_data): Rename...
10484 (fwd_jt_path_registry::lookup_redirection_data): ...to this.
10485 (jump_thread_path_registry::thread_block_1): Rename...
10486 (fwd_jt_path_registry::thread_block_1): ...to this.
10487 (jump_thread_path_registry::thread_block): Rename...
10488 (fwd_jt_path_registry::thread_block): ...to this.
10489 (jt_path_registry::thread_through_loop_header): Rename...
10490 (fwd_jt_path_registry::thread_through_loop_header): ...to this.
10491 (jump_thread_path_registry::mark_threaded_blocks): Rename...
10492 (fwd_jt_path_registry::mark_threaded_blocks): ...to this.
10493 (jump_thread_path_registry::debug_path): Rename...
10494 (jt_path_registry::debug_path): ...to this.
10495 (jump_thread_path_registry::dump): Rename...
10496 (jt_path_registry::debug): ...to this.
10497 (jump_thread_path_registry::rewire_first_differing_edge): Rename...
10498 (back_jt_path_registry::rewire_first_differing_edge): ...to this.
10499 (jump_thread_path_registry::adjust_paths_after_duplication): Rename...
10500 (back_jt_path_registry::adjust_paths_after_duplication): ...to this.
10501 (jump_thread_path_registry::duplicate_thread_path): Rename...
10502 (back_jt_path_registry::duplicate_thread_path): ...to this. Also,
10503 drop ill-formed candidates.
10504 (jump_thread_path_registry::remove_jump_threads_including): Rename...
10505 (fwd_jt_path_registry::remove_jump_threads_including): ...to this.
10506 (jt_path_registry::thread_through_all_blocks): New.
10507 (back_jt_path_registry::update_cfg): New.
10508 (fwd_jt_path_registry::update_cfg): New.
10509 (jump_thread_path_registry::register_jump_thread): Rename...
10510 (jt_path_registry::register_jump_thread): ...to this.
10511 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
10513 (class jt_path_registry): ...here.
10514 (class fwd_jt_path_registry): New.
10515 (class back_jt_path_registry): New.
10517 2021-09-10 liuhongt <hongtao.liu@intel.com>
10520 2021-09-01 liuhongt <hongtao.liu@intel.com>
10522 * emit-rtl.c (validate_subreg): Get rid of all float-int
10525 2021-09-10 Jakub Jelinek <jakub@redhat.com>
10527 * tree-core.h (enum omp_memory_order): Add OMP_MEMORY_ORDER_MASK,
10528 OMP_FAIL_MEMORY_ORDER_UNSPECIFIED, OMP_FAIL_MEMORY_ORDER_RELAXED,
10529 OMP_FAIL_MEMORY_ORDER_ACQUIRE, OMP_FAIL_MEMORY_ORDER_RELEASE,
10530 OMP_FAIL_MEMORY_ORDER_ACQ_REL, OMP_FAIL_MEMORY_ORDER_SEQ_CST and
10531 OMP_FAIL_MEMORY_ORDER_MASK enumerators.
10532 (OMP_FAIL_MEMORY_ORDER_SHIFT): Define.
10533 * gimple-pretty-print.c (dump_gimple_omp_atomic_load,
10534 dump_gimple_omp_atomic_store): Print [weak] for weak atomic
10536 * gimple.h (enum gf_mask): Change GF_OMP_ATOMIC_MEMORY_ORDER
10537 to 6-bit mask, adjust GF_OMP_ATOMIC_NEED_VALUE value and add
10538 GF_OMP_ATOMIC_WEAK.
10539 (gimple_omp_atomic_weak_p, gimple_omp_atomic_set_weak): New inline
10541 * tree.h (OMP_ATOMIC_WEAK): Define.
10542 * tree-pretty-print.c (dump_omp_atomic_memory_order): Adjust for
10543 fail memory order being encoded in the same enum and also print
10544 fail clause if present.
10545 (dump_generic_node): Print weak clause if OMP_ATOMIC_WEAK.
10546 * gimplify.c (goa_stabilize_expr): Add target_expr and rhs arguments,
10547 handle pre_p == NULL case as a test mode that only returns value
10548 but doesn't change gimplify nor change anything otherwise, adjust
10549 recursive calls, add MODIFY_EXPR, ADDR_EXPR, COND_EXPR, TARGET_EXPR
10550 and CALL_EXPR handling, adjust COMPOUND_EXPR handling for
10551 __builtin_clear_padding calls, for !rhs gimplify as lvalue rather
10553 (gimplify_omp_atomic): Adjust goa_stabilize_expr caller. Handle
10554 COND_EXPR rhs. Set weak flag on gimple load/store for
10556 * omp-expand.c (omp_memory_order_to_fail_memmodel): New function.
10557 (omp_memory_order_to_memmodel): Adjust for fail clause encoded
10559 (expand_omp_atomic_cas): New function.
10560 (expand_omp_atomic_pipeline): Use omp_memory_order_to_fail_memmodel
10562 (expand_omp_atomic): Attempt to optimize atomic compare and exchange
10563 using expand_omp_atomic_cas.
10565 2021-09-10 Aldy Hernandez <aldyh@redhat.com>
10566 Michael Matz <matz@suse.de>
10568 * tree-pass.h (PROP_loop_opts_done): New.
10569 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
10570 Intersect with global range.
10571 * tree-ssa-loop.c (tree_ssa_loop_done): Set PROP_loop_opts_done.
10572 * tree-ssa-threadbackward.c
10573 (back_threader_profitability::profitable_path_p): Disable
10574 threading through latches until after loop optimizations have run.
10576 2021-09-10 David Faust <david.faust@oracle.com>
10578 * doc/invoke.texi: Document BPF -mcpu, -mjmpext, -mjmp32 and -malu32
10581 2021-09-10 David Faust <david.faust@oracle.com>
10583 * config/bpf/bpf-opts.h (bpf_isa_version): New enum.
10584 * config/bpf/bpf-protos.h (bpf_expand_cbranch): New.
10585 * config/bpf/bpf.c (bpf_option_override): Handle -mcpu option.
10586 (bpf_expand_cbranch): New function.
10587 * config/bpf/bpf.md (AM mode iterator): Conditionalize support for SI
10589 (zero_extendsidi2): Only use mov32 instruction if it is available.
10590 (SIM mode iterator): Conditionalize support for SI mode.
10591 (JM mode iterator): New.
10592 (cbranchdi4): Update name, use new JM iterator. Use bpf_expand_cbranch.
10593 (*branch_on_di): Update name, use new JM iterator.
10594 * config/bpf/bpf.opt: (mjmpext): New option.
10595 (malu32): Likewise.
10596 (mjmp32): Likewise.
10598 (bpf_isa): New enum.
10600 2021-09-10 David Faust <david.faust@oracle.com>
10602 * config/bpf/bpf.md (zero_extendhidi2): Add new output template
10603 for register-to-register extensions.
10604 (zero_extendqidi2): Likewise.
10606 2021-09-10 Richard Biener <rguenther@suse.de>
10608 PR middle-end/102273
10609 * internal-fn.c (expand_DEFERRED_INIT): Always expand non-SSA vars.
10611 2021-09-10 Richard Biener <rguenther@suse.de>
10613 PR middle-end/102269
10614 * gimplify.c (is_var_need_auto_init): Empty types do not need
10617 2021-09-10 Richard Biener <rguenther@suse.de>
10619 * configure.ac (--with-stabs): Remove.
10620 * configure: Regenerate.
10621 * doc/install.texi: Remove --with-stabs documentation.
10623 2021-09-10 liuhongt <hongtao.liu@intel.com>
10625 * config/i386/avx512fp16intrin.h: (_mm512_cmp_ph_mask):
10627 (_mm512_mask_cmp_ph_mask): Likewise.
10628 (_mm512_cmp_round_ph_mask): Likewise.
10629 (_mm512_mask_cmp_round_ph_mask): Likewise.
10630 (_mm_cmp_sh_mask): Likewise.
10631 (_mm_mask_cmp_sh_mask): Likewise.
10632 (_mm_cmp_round_sh_mask): Likewise.
10633 (_mm_mask_cmp_round_sh_mask): Likewise.
10634 (_mm_comieq_sh): Likewise.
10635 (_mm_comilt_sh): Likewise.
10636 (_mm_comile_sh): Likewise.
10637 (_mm_comigt_sh): Likewise.
10638 (_mm_comige_sh): Likewise.
10639 (_mm_comineq_sh): Likewise.
10640 (_mm_ucomieq_sh): Likewise.
10641 (_mm_ucomilt_sh): Likewise.
10642 (_mm_ucomile_sh): Likewise.
10643 (_mm_ucomigt_sh): Likewise.
10644 (_mm_ucomige_sh): Likewise.
10645 (_mm_ucomineq_sh): Likewise.
10646 (_mm_comi_round_sh): Likewise.
10647 (_mm_comi_sh): Likewise.
10648 * config/i386/avx512fp16vlintrin.h (_mm_cmp_ph_mask): New intrinsic.
10649 (_mm_mask_cmp_ph_mask): Likewise.
10650 (_mm256_cmp_ph_mask): Likewise.
10651 (_mm256_mask_cmp_ph_mask): Likewise.
10652 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
10653 * config/i386/i386-builtin.def: Add corresponding new builtins.
10654 * config/i386/i386-expand.c
10655 (ix86_expand_args_builtin): Handle new builtin types.
10656 (ix86_expand_round_builtin): Ditto.
10657 * config/i386/i386.md (ssevecmode): Add HF mode.
10658 (MODEFH): New mode iterator.
10659 * config/i386/sse.md
10660 (V48H_AVX512VL): New mode iterator to support HF vector modes.
10661 Ajdust corresponding description.
10662 (ssecmpintprefix): New.
10663 (VI12_AVX512VL): Adjust to support HF vector modes.
10664 (cmp_imm_predicate): Likewise.
10665 (<avx512>_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
10667 (avx512f_vmcmp<mode>3<round_saeonly_name>): Likewise.
10668 (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Likewise.
10669 (<sse>_<unord>comi<round_saeonly_name>): Likewise.
10671 2021-09-10 liuhongt <hongtao.liu@intel.com>
10673 * config/i386/avx512fp16intrin.h: (_mm512_max_ph): New intrinsic.
10674 (_mm512_mask_max_ph): Likewise.
10675 (_mm512_maskz_max_ph): Likewise.
10676 (_mm512_min_ph): Likewise.
10677 (_mm512_mask_min_ph): Likewise.
10678 (_mm512_maskz_min_ph): Likewise.
10679 (_mm512_max_round_ph): Likewise.
10680 (_mm512_mask_max_round_ph): Likewise.
10681 (_mm512_maskz_max_round_ph): Likewise.
10682 (_mm512_min_round_ph): Likewise.
10683 (_mm512_mask_min_round_ph): Likewise.
10684 (_mm512_maskz_min_round_ph): Likewise.
10685 (_mm_max_sh): Likewise.
10686 (_mm_mask_max_sh): Likewise.
10687 (_mm_maskz_max_sh): Likewise.
10688 (_mm_min_sh): Likewise.
10689 (_mm_mask_min_sh): Likewise.
10690 (_mm_maskz_min_sh): Likewise.
10691 (_mm_max_round_sh): Likewise.
10692 (_mm_mask_max_round_sh): Likewise.
10693 (_mm_maskz_max_round_sh): Likewise.
10694 (_mm_min_round_sh): Likewise.
10695 (_mm_mask_min_round_sh): Likewise.
10696 (_mm_maskz_min_round_sh): Likewise.
10697 * config/i386/avx512fp16vlintrin.h (_mm_max_ph): New intrinsic.
10698 (_mm256_max_ph): Likewise.
10699 (_mm_mask_max_ph): Likewise.
10700 (_mm256_mask_max_ph): Likewise.
10701 (_mm_maskz_max_ph): Likewise.
10702 (_mm256_maskz_max_ph): Likewise.
10703 (_mm_min_ph): Likewise.
10704 (_mm256_min_ph): Likewise.
10705 (_mm_mask_min_ph): Likewise.
10706 (_mm256_mask_min_ph): Likewise.
10707 (_mm_maskz_min_ph): Likewise.
10708 (_mm256_maskz_min_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 builtin types.
10713 * config/i386/sse.md
10714 (<code><mode>3<mask_name><round_saeonly_name>): Adjust to
10715 support HF vector modes.
10716 (*<code><mode>3<mask_name><round_saeonly_name>): Likewise.
10717 (ieee_<ieee_maxmin><mode>3<mask_name><round_saeonly_name>):
10719 (<sse>_vm<code><mode>3<mask_scalar_name><round_saeonly_scalar_name>):
10721 * config/i386/subst.md (round_saeonly_mode512bit_condition):
10722 Adjust for HF vector modes.
10724 2021-09-10 Liu, Hongtao <hongtao.liu@intel.com>
10726 * config/i386/avx512fp16intrin.h (_mm_add_sh): New intrinsic.
10727 (_mm_mask_add_sh): Likewise.
10728 (_mm_maskz_add_sh): Likewise.
10729 (_mm_sub_sh): Likewise.
10730 (_mm_mask_sub_sh): Likewise.
10731 (_mm_maskz_sub_sh): Likewise.
10732 (_mm_mul_sh): Likewise.
10733 (_mm_mask_mul_sh): Likewise.
10734 (_mm_maskz_mul_sh): Likewise.
10735 (_mm_div_sh): Likewise.
10736 (_mm_mask_div_sh): Likewise.
10737 (_mm_maskz_div_sh): Likewise.
10738 (_mm_add_round_sh): Likewise.
10739 (_mm_mask_add_round_sh): Likewise.
10740 (_mm_maskz_add_round_sh): Likewise.
10741 (_mm_sub_round_sh): Likewise.
10742 (_mm_mask_sub_round_sh): Likewise.
10743 (_mm_maskz_sub_round_sh): Likewise.
10744 (_mm_mul_round_sh): Likewise.
10745 (_mm_mask_mul_round_sh): Likewise.
10746 (_mm_maskz_mul_round_sh): Likewise.
10747 (_mm_div_round_sh): Likewise.
10748 (_mm_mask_div_round_sh): Likewise.
10749 (_mm_maskz_div_round_sh): Likewise.
10750 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
10751 * config/i386/i386-builtin.def: Add corresponding new builtins.
10752 * config/i386/i386-expand.c
10753 (ix86_expand_round_builtin): Handle new builtins.
10754 * config/i386/sse.md (VF_128): Change description.
10755 (<sse>_vm<plusminus_insn><mode>3<mask_scalar_name><round_scalar_name>):
10756 Adjust to support HF vector modes.
10757 (<sse>_vm<multdiv_mnemonic><mode>3<mask_scalar_name><round_scalar_name>):
10760 2021-09-10 H.J. Lu <hjl.tools@gmail.com>
10762 * config/i386/i386-expand.c
10763 (ix86_avx256_split_vector_move_misalign): Handle V16HF mode.
10764 * config/i386/i386.c
10765 (ix86_preferred_simd_mode): Handle HF mode.
10766 * config/i386/sse.md (V_256H): New mode iterator.
10767 (avx_vextractf128<mode>): Use it.
10768 (VEC_INIT_MODE): Align vector HFmode condition to vector
10769 HImodes since there're no real HF instruction used.
10770 (VEC_INIT_HALF_MODE): Ditto.
10772 (VIHF_AVX512BW): Ditto.
10773 (*vec_extracthf): Ditto.
10774 (VEC_EXTRACT_MODE): Ditto.
10776 2021-09-10 Richard Biener <rguenther@suse.de>
10779 * config/dbx.h: Remove.
10780 * config/dbxcoff.h: Do not define PREFERRED_DEBUGGING_TYPE.
10781 * config/lynx.h: Likewise.
10783 2021-09-10 liuhongt <hongtao.liu@intel.com>
10785 * config/i386/i386-expand.c (ix86_expand_copysign): Expand
10786 right into ANDNOT + AND + IOR, using paradoxical subregs.
10787 (ix86_split_copysign_const): Remove.
10788 (ix86_split_copysign_var): Ditto.
10789 * config/i386/i386-protos.h (ix86_split_copysign_const): Dotto.
10790 (ix86_split_copysign_var): Ditto.
10791 * config/i386/i386.md (@copysign<mode>3_const): Ditto.
10792 (@copysign<mode>3_var): Ditto.
10794 2021-09-09 qing zhao <qing.zhao@oracle.com>
10796 * builtins.c (expand_builtin_memset): Make external visible.
10797 * builtins.h (expand_builtin_memset): Declare extern.
10798 * common.opt (ftrivial-auto-var-init=): New option.
10799 * doc/extend.texi: Document the uninitialized attribute.
10800 * doc/invoke.texi: Document -ftrivial-auto-var-init.
10801 * flag-types.h (enum auto_init_type): New enumerated type
10803 * gimple-fold.c (clear_padding_type): Add one new parameter.
10804 (clear_padding_union): Likewise.
10805 (clear_padding_emit_loop): Likewise.
10806 (clear_type_padding_in_mask): Likewise.
10807 (gimple_fold_builtin_clear_padding): Handle this new parameter.
10808 * gimplify.c (gimple_add_init_for_auto_var): New function.
10809 (gimple_add_padding_init_for_auto_var): New function.
10810 (is_var_need_auto_init): New function.
10811 (gimplify_decl_expr): Add initialization to automatic variables per
10813 (gimplify_call_expr): Add one new parameter for call to
10814 __builtin_clear_padding.
10815 (gimplify_init_constructor): Add padding initialization in the end.
10816 * internal-fn.c (INIT_PATTERN_VALUE): New macro.
10817 (expand_DEFERRED_INIT): New function.
10818 * internal-fn.def (DEFERRED_INIT): New internal function.
10819 * tree-cfg.c (verify_gimple_call): Verify calls to .DEFERRED_INIT.
10820 * tree-sra.c (generate_subtree_deferred_init): New function.
10821 (scan_function): Avoid setting cannot_scalarize_away_bitmap for
10822 calls to .DEFERRED_INIT.
10823 (sra_modify_deferred_init): New function.
10824 (sra_modify_function_body): Handle calls to DEFERRED_INIT specially.
10825 * tree-ssa-structalias.c (find_func_aliases_for_call): Likewise.
10826 * tree-ssa-uninit.c (warn_uninit): Handle calls to DEFERRED_INIT
10828 (check_defs): Likewise.
10829 (warn_uninitialized_vars): Likewise.
10830 * tree-ssa.c (ssa_undefined_value_p): Likewise.
10831 * tree.c (build_common_builtin_nodes): Build tree node for
10832 BUILT_IN_CLEAR_PADDING when needed.
10834 2021-09-09 Richard Biener <rguenther@suse.de>
10836 * tree-ssa-loop-im.c (fill_always_executed_in_1): Walk
10839 2021-09-09 Richard Biener <rguenther@suse.de>
10841 * tree-ssa-loop-im.c (fill_always_executed_in_1): Integrate
10842 DOM walk from get_loop_body_in_dom_order using a worklist
10845 2021-09-09 liuhongt <hongtao.liu@intel.com>
10847 * config.gcc: Add avx512fp16vlintrin.h.
10848 * config/i386/avx512fp16intrin.h: (_mm512_add_ph): New intrinsic.
10849 (_mm512_mask_add_ph): Likewise.
10850 (_mm512_maskz_add_ph): Likewise.
10851 (_mm512_sub_ph): Likewise.
10852 (_mm512_mask_sub_ph): Likewise.
10853 (_mm512_maskz_sub_ph): Likewise.
10854 (_mm512_mul_ph): Likewise.
10855 (_mm512_mask_mul_ph): Likewise.
10856 (_mm512_maskz_mul_ph): Likewise.
10857 (_mm512_div_ph): Likewise.
10858 (_mm512_mask_div_ph): Likewise.
10859 (_mm512_maskz_div_ph): Likewise.
10860 (_mm512_add_round_ph): Likewise.
10861 (_mm512_mask_add_round_ph): Likewise.
10862 (_mm512_maskz_add_round_ph): Likewise.
10863 (_mm512_sub_round_ph): Likewise.
10864 (_mm512_mask_sub_round_ph): Likewise.
10865 (_mm512_maskz_sub_round_ph): Likewise.
10866 (_mm512_mul_round_ph): Likewise.
10867 (_mm512_mask_mul_round_ph): Likewise.
10868 (_mm512_maskz_mul_round_ph): Likewise.
10869 (_mm512_div_round_ph): Likewise.
10870 (_mm512_mask_div_round_ph): Likewise.
10871 (_mm512_maskz_div_round_ph): Likewise.
10872 * config/i386/avx512fp16vlintrin.h: New header.
10873 * config/i386/i386-builtin-types.def (V16HF, V8HF, V32HF):
10874 Add new builtin types.
10875 * config/i386/i386-builtin.def: Add corresponding builtins.
10876 * config/i386/i386-expand.c
10877 (ix86_expand_args_builtin): Handle new builtin types.
10878 (ix86_expand_round_builtin): Likewise.
10879 * config/i386/immintrin.h: Include avx512fp16vlintrin.h
10880 * config/i386/sse.md (VFH): New mode_iterator.
10882 (avx512fmaskmode): Add HF vector modes.
10883 (avx512fmaskhalfmode): Likewise.
10884 (<plusminus_insn><mode>3<mask_name><round_name>): Adjust to for
10886 (*<plusminus_insn><mode>3<mask_name><round_name>): Likewise.
10887 (mul<mode>3<mask_name><round_name>): Likewise.
10888 (*mul<mode>3<mask_name><round_name>): Likewise.
10889 (div<mode>3): Likewise.
10890 (<sse>_div<mode>3<mask_name><round_name>): Likewise.
10891 * config/i386/subst.md (SUBST_V): Add HF vector modes.
10892 (SUBST_A): Likewise.
10893 (round_mode512bit_condition): Adjust for V32HFmode.
10895 2021-09-09 liuhongt <hongtao.liu@intel.com>
10898 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
10899 (reduc_plus_scal_v4sf): .. this, New define_expand.
10900 (reduc_plus_scal_v2df): .. and this, New define_expand.
10902 2021-09-09 liuhongt <hongtao.liu@intel.com>
10905 * config/i386/sse.md (*vec_extract<mode><ssescalarmodelower>_valign):
10908 2021-09-08 Jonathan Wakely <jwakely@redhat.com>
10911 * doc/trouble.texi (Copy Assignment): Fix description of
10912 behaviour and fix code in example.
10914 2021-09-08 Segher Boessenkool <segher@kernel.crashing.org>
10917 * config/rs6000/rs6000-logue.c (rs6000_emit_epilogue): For ELFv2 use
10918 r11 instead of r12 for restoring CR.
10920 2021-09-08 Jakub Jelinek <jakub@redhat.com>
10921 liuhongt <hongtao.liu@intel.com>
10924 * config/i386/i386.md (@xorsign<mode>3_1): Remove.
10925 * config/i386/i386-expand.c (ix86_expand_xorsign): Expand right away
10926 into AND with mask and XOR, using paradoxical subregs.
10927 (ix86_split_xorsign): Remove.
10928 * config/i386/i386-protos.h (ix86_split_xorsign): Remove.
10930 2021-09-08 Di Zhao <dizhao@os.amperecomputing.com>
10932 * tree-ssa-sccvn.c (vn_nary_op_insert_into): fix result compare
10934 2021-09-08 Jakub Jelinek <jakub@redhat.com>
10937 * config/i386/i386.md (xorsign<mode>3): If operands[1] is equal to
10938 operands[2], emit abs<mode>2 instead.
10939 (@xorsign<mode>3_1): Add early-clobbers for output operand, enable
10940 first alternative even for avx, add another alternative with
10941 =&Yv <- 0, Yv, Yvm constraints.
10942 * config/i386/i386-expand.c (ix86_split_xorsign): If op0 is equal
10943 to op1, emit vpandn instead.
10945 2021-09-08 liuhongt <hongtao.liu@intel.com>
10947 * config/i386/avx512fp16intrin.h (_mm_set_ph): New intrinsic.
10948 (_mm256_set_ph): Likewise.
10949 (_mm512_set_ph): Likewise.
10950 (_mm_setr_ph): Likewise.
10951 (_mm256_setr_ph): Likewise.
10952 (_mm512_setr_ph): Likewise.
10953 (_mm_set1_ph): Likewise.
10954 (_mm256_set1_ph): Likewise.
10955 (_mm512_set1_ph): Likewise.
10956 (_mm_setzero_ph): Likewise.
10957 (_mm256_setzero_ph): Likewise.
10958 (_mm512_setzero_ph): Likewise.
10959 (_mm_set_sh): Likewise.
10960 (_mm_load_sh): Likewise.
10961 (_mm_store_sh): Likewise.
10962 * config/i386/i386-builtin-types.def (V8HF): New type.
10963 (DEF_FUNCTION_TYPE (V8HF, V8HI)): New builtin function type
10964 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
10965 Support vector HFmodes.
10966 (ix86_expand_vector_init_one_nonzero): Likewise.
10967 (ix86_expand_vector_init_one_var): Likewise.
10968 (ix86_expand_vector_init_interleave): Likewise.
10969 (ix86_expand_vector_init_general): Likewise.
10970 (ix86_expand_vector_set): Likewise.
10971 (ix86_expand_vector_extract): Likewise.
10972 (ix86_expand_vector_init_concat): Likewise.
10973 (ix86_expand_sse_movcc): Handle vector HFmodes.
10974 (ix86_expand_vector_set_var): Ditto.
10975 * config/i386/i386-modes.def: Add HF vector modes in comment.
10976 * config/i386/i386.c (classify_argument): Add HF vector modes.
10977 (ix86_hard_regno_mode_ok): Allow HF vector modes for AVX512FP16.
10978 (ix86_vector_mode_supported_p): Likewise.
10979 (ix86_set_reg_reg_cost): Handle vector HFmode.
10980 (ix86_get_ssemov): Handle vector HFmode.
10981 (function_arg_advance_64): Pass unamed V16HFmode and V32HFmode
10983 (function_arg_advance_32): Pass V8HF/V16HF/V32HF by sse reg for 32bit
10985 (function_arg_advance_32): Ditto.
10986 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): New.
10987 (VALID_AVX256_REG_OR_OI_MODE): Rename to ..
10988 (VALID_AVX256_REG_OR_OI_VHF_MODE): .. this, and add V16HF.
10989 (VALID_SSE2_REG_VHF_MODE): New.
10990 (VALID_AVX512VL_128_REG_MODE): Add V8HF and TImode.
10991 (SSE_REG_MODE_P): Add vector HFmode.
10992 * config/i386/i386.md (mode): Add HF vector modes.
10993 (MODE_SIZE): Likewise.
10994 (ssemodesuffix): Add ph suffix for HF vector modes.
10995 * config/i386/sse.md (VFH_128): New mode iterator.
10996 (VMOVE): Adjust for HF vector modes.
10998 (V_256_512): Likewise.
10999 (avx512): Likewise.
11000 (avx512fmaskmode): Likewise.
11001 (shuffletype): Likewise.
11002 (sseinsnmode): Likewise.
11003 (ssedoublevecmode): Likewise.
11004 (ssehalfvecmode): Likewise.
11005 (ssehalfvecmodelower): Likewise.
11006 (ssePScmode): Likewise.
11007 (ssescalarmode): Likewise.
11008 (ssescalarmodelower): Likewise.
11009 (sseintprefix): Likewise.
11011 (bcstscalarsuff): Likewise.
11012 (xtg_mode): Likewise.
11013 (VI12HF_AVX512VL): New mode_iterator.
11014 (VF_AVX512FP16): Likewise.
11016 (VIHF_256): Likewise.
11017 (VIHF_AVX512BW): Likewise.
11018 (V16_256): Likewise.
11019 (V32_512): Likewise.
11020 (sseintmodesuffix): New mode_attr.
11021 (sse): Add scalar and vector HFmodes.
11022 (ssescalarmode): Add vector HFmode mapping.
11023 (ssescalarmodesuffix): Add sh suffix for HFmode.
11024 (*<sse>_vm<insn><mode>3): Use VFH_128.
11025 (*<sse>_vm<multdiv_mnemonic><mode>3): Likewise.
11026 (*ieee_<ieee_maxmin><mode>3): Likewise.
11027 (<avx512>_blendm<mode>): New define_insn.
11028 (vec_setv8hf): New define_expand.
11029 (vec_set<mode>_0): New define_insn for HF vector set.
11030 (*avx512fp16_movsh): Likewise.
11031 (avx512fp16_movsh): Likewise.
11032 (vec_extract_lo_v32hi): Rename to ...
11033 (vec_extract_lo_<mode>): ... this, and adjust to allow HF
11035 (vec_extract_hi_v32hi): Likewise.
11036 (vec_extract_hi_<mode>): Likewise.
11037 (vec_extract_lo_v16hi): Likewise.
11038 (vec_extract_lo_<mode>): Likewise.
11039 (vec_extract_hi_v16hi): Likewise.
11040 (vec_extract_hi_<mode>): Likewise.
11041 (vec_set_hi_v16hi): Likewise.
11042 (vec_set_hi_<mode>): Likewise.
11043 (vec_set_lo_v16hi): Likewise.
11044 (vec_set_lo_<mode>): Likewise.
11045 (*vec_extract<mode>_0): New define_insn_and_split for HF
11047 (*vec_extracthf): New define_insn.
11048 (VEC_EXTRACT_MODE): Add HF vector modes.
11049 (PINSR_MODE): Add V8HF.
11050 (sse2p4_1): Likewise.
11051 (pinsr_evex_isa): Likewise.
11052 (<sse2p4_1>_pinsr<ssemodesuffix>): Adjust to support
11053 insert for V8HFmode.
11054 (pbroadcast_evex_isa): Add HF vector modes.
11055 (AVX2_VEC_DUP_MODE): Likewise.
11056 (VEC_INIT_MODE): Likewise.
11057 (VEC_INIT_HALF_MODE): Likewise.
11058 (avx2_pbroadcast<mode>): Adjust to support HF vector mode
11060 (avx2_pbroadcast<mode>_1): Likewise.
11061 (<avx512>_vec_dup<mode>_1): Likewise.
11062 (<avx512>_vec_dup<mode><mask_name>): Likewise.
11063 (<mask_codefor><avx512>_vec_dup_gpr<mode><mask_name>):
11066 2021-09-08 Guo, Xuepeng <xuepeng.guo@intel.com>
11067 H.J. Lu <hongjiu.lu@intel.com>
11068 Liu Hongtao <hongtao.liu@intel.com>
11069 Wang Hongyu <hongyu.wang@intel.com>
11070 Xu Dianhong <dianhong.xu@intel.com>
11072 * common/config/i386/cpuinfo.h (get_available_features):
11073 Detect FEATURE_AVX512FP16.
11074 * common/config/i386/i386-common.c
11075 (OPTION_MASK_ISA_AVX512FP16_SET,
11076 OPTION_MASK_ISA_AVX512FP16_UNSET,
11077 OPTION_MASK_ISA2_AVX512FP16_SET,
11078 OPTION_MASK_ISA2_AVX512FP16_UNSET): New.
11079 (OPTION_MASK_ISA2_AVX512BW_UNSET,
11080 OPTION_MASK_ISA2_AVX512BF16_UNSET): Add AVX512FP16.
11081 (ix86_handle_option): Handle -mavx512fp16.
11082 * common/config/i386/i386-cpuinfo.h (enum processor_features):
11083 Add FEATURE_AVX512FP16.
11084 * common/config/i386/i386-isas.h: Add entry for AVX512FP16.
11085 * config.gcc: Add avx512fp16intrin.h.
11086 * config/i386/avx512fp16intrin.h: New intrinsic header.
11087 * config/i386/cpuid.h: Add bit_AVX512FP16.
11088 * config/i386/i386-builtin-types.def: (FLOAT16): New primitive type.
11089 * config/i386/i386-builtins.c: Support _Float16 type for i386
11091 (ix86_register_float16_builtin_type): New function.
11092 (ix86_float16_type_node): New.
11093 * config/i386/i386-c.c (ix86_target_macros_internal): Define
11095 * config/i386/i386-expand.c (ix86_expand_branch): Support
11097 (ix86_prepare_fp_compare_args): Adjust TARGET_SSE_MATH &&
11098 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
11099 (ix86_expand_fp_movcc): Ditto.
11100 * config/i386/i386-isa.def: Add PTA define for AVX512FP16.
11101 * config/i386/i386-options.c (isa2_opts): Add -mavx512fp16.
11102 (ix86_valid_target_attribute_inner_p): Add avx512fp16 attribute.
11103 * config/i386/i386.c (ix86_get_ssemov): Use
11104 vmovdqu16/vmovw/vmovsh for HFmode/HImode scalar or vector.
11105 (ix86_get_excess_precision): Use
11106 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when TARGET_AVX512FP16
11108 (sse_store_index): Use SFmode cost for HFmode cost.
11109 (inline_memory_move_cost): Add HFmode, and perfer SSE cost over
11110 GPR cost for HFmode.
11111 (ix86_hard_regno_mode_ok): Allow HImode in sse register.
11112 (ix86_mangle_type): Add manlging for _Float16 type.
11113 (inline_secondary_memory_needed): No memory is needed for
11114 16bit movement between gpr and sse reg under
11116 (ix86_multiplication_cost): Adjust TARGET_SSE_MATH &&
11117 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
11118 (ix86_division_cost): Ditto.
11119 (ix86_rtx_costs): Ditto.
11120 (ix86_add_stmt_cost): Ditto.
11121 (ix86_optab_supported_p): Ditto.
11122 * config/i386/i386.h (VALID_AVX512F_SCALAR_MODE): Add HFmode.
11123 (SSE_FLOAT_MODE_SSEMATH_OR_HF_P): Add HFmode.
11124 (PTA_SAPPHIRERAPIDS): Add PTA_AVX512FP16.
11125 * config/i386/i386.md (mode): Add HFmode.
11126 (MODE_SIZE): Add HFmode.
11127 (isa): Add avx512fp16.
11128 (enabled): Handle avx512fp16.
11129 (ssemodesuffix): Add sh suffix for HFmode.
11130 (comm): Add mult, div.
11131 (plusminusmultdiv): New code iterator.
11132 (insn): Add mult, div.
11133 (*movhf_internal): Adjust for avx512fp16 instruction.
11134 (*movhi_internal): Ditto.
11135 (*cmpi<unord>hf): New define_insn for HFmode.
11136 (*ieee_s<ieee_maxmin>hf3): Likewise.
11137 (extendhf<mode>2): Likewise.
11138 (trunc<mode>hf2): Likewise.
11139 (float<floatunssuffix><mode>hf2): Likewise.
11140 (*<insn>hf): Likewise.
11141 (cbranchhf4): New expander.
11142 (movhfcc): Likewise.
11143 (<insn>hf3): Likewise.
11144 (mulhf3): Likewise.
11145 (divhf3): Likewise.
11146 * config/i386/i386.opt: Add mavx512fp16.
11147 * config/i386/immintrin.h: Include avx512fp16intrin.h.
11148 * doc/invoke.texi: Add mavx512fp16.
11149 * doc/extend.texi: Add avx512fp16 Usage Notes.
11151 2021-09-08 liuhongt <hongtao.liu@intel.com>
11153 * common.opt: Support -fexcess-precision=16.
11154 * config/aarch64/aarch64.c (aarch64_excess_precision): Return
11155 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when
11156 EXCESS_PRECISION_TYPE_FLOAT16.
11157 * config/arm/arm.c (arm_excess_precision): Ditto.
11158 * config/i386/i386.c (ix86_get_excess_precision): Ditto.
11159 * config/m68k/m68k.c (m68k_excess_precision): Issue an error
11160 when EXCESS_PRECISION_TYPE_FLOAT16.
11161 * config/s390/s390.c (s390_excess_precision): Ditto.
11162 * coretypes.h (enum excess_precision_type): Add
11163 EXCESS_PRECISION_TYPE_FLOAT16.
11164 * doc/tm.texi (TARGET_C_EXCESS_PRECISION): Update documents.
11165 * doc/tm.texi.in (TARGET_C_EXCESS_PRECISION): Ditto.
11166 * doc/extend.texi (Half-Precision): Document
11167 -fexcess-precision=16.
11168 * flag-types.h (enum excess_precision): Add
11169 EXCESS_PRECISION_FLOAT16.
11170 * target.def (excess_precision): Update document.
11171 * tree.c (excess_precision_type): Set excess_precision_type to
11172 EXCESS_PRECISION_FLOAT16 when -fexcess-precision=16.
11174 2021-09-08 liuhongt <hongtao.liu@intel.com>
11176 * doc/extend.texi: (@node Floating Types): Adjust the wording.
11177 (@node Half-Precision): Ditto.
11179 2021-09-07 Takayuki 'January June' Suwa <jjsuwa_sys3175@yahoo.co.jp>
11182 * config/xtensa/xtensa.c (xtensa_emit_move_sequence): Add
11183 'CONST_INT_P (src)' to the condition of the block that tries to
11184 eliminate literal when loading integer contant.
11186 2021-09-07 David Faust <david.faust@oracle.com>
11188 * doc/extend.texi (BPF Type Attributes) New node.
11189 Document new preserve_access_index attribute.
11190 Document new preserve_access_index builtin.
11191 * doc/invoke.texi: Document -mco-re and -mno-co-re options.
11193 2021-09-07 David Faust <david.faust@oracle.com>
11195 * config/bpf/bpf.c: Adjust includes.
11196 (bpf_handle_preserve_access_index_attribute): New function.
11197 (bpf_attribute_table): Use it here.
11198 (bpf_builtins): Add BPF_BUILTIN_PRESERVE_ACCESS_INDEX.
11199 (bpf_option_override): Handle "-mco-re" option.
11200 (bpf_asm_init_sections): New.
11201 (TARGET_ASM_INIT_SECTIONS): Redefine.
11202 (bpf_file_end): New.
11203 (TARGET_ASM_FILE_END): Redefine.
11204 (bpf_init_builtins): Add "__builtin_preserve_access_index".
11205 (bpf_core_compute, bpf_core_get_index): New.
11206 (is_attr_preserve_access): New.
11207 (bpf_expand_builtin): Handle new builtins.
11208 (bpf_core_newdecl, bpf_core_is_maybe_aggregate_access): New.
11209 (bpf_core_walk): New.
11210 (bpf_resolve_overloaded_builtin): New.
11211 (TARGET_RESOLVE_OVERLOADED_BUILTIN): Redefine.
11212 (handle_attr): New.
11213 (pass_bpf_core_attr): New RTL pass.
11214 * config/bpf/bpf-passes.def: New file.
11215 * config/bpf/bpf-protos.h (make_pass_bpf_core_attr): New.
11216 * config/bpf/coreout.c: New file.
11217 * config/bpf/coreout.h: Likewise.
11218 * config/bpf/t-bpf (TM_H): Add $(srcdir)/config/bpf/coreout.h.
11219 (coreout.o): New rule.
11220 (PASSES_EXTRA): Add $(srcdir)/config/bpf/bpf-passes.def.
11221 * config.gcc (bpf): Add coreout.h to extra_headers.
11222 Add coreout.o to extra_objs.
11223 Add $(srcdir)/config/bpf/coreout.c to target_gtfiles.
11225 2021-09-07 David Faust <david.faust@oracle.com>
11227 * btfout.c (get_btf_id): Function is no longer static.
11228 * ctfc.h: Expose it here.
11230 2021-09-07 David Faust <david.faust@oracle.com>
11232 * ctfc.c (ctf_lookup_tree_type): New function.
11233 * ctfc.h: Likewise.
11235 2021-09-07 David Faust <david.faust@oracle.com>
11237 * ctfc.c (ctf_dtd_lookup): Function is no longer static.
11238 * ctfc.h: Analogous change.
11240 2021-09-07 David Faust <david.faust@oracle.com>
11242 * dwarf2out.c (lookup_type_die): Function is no longer static.
11243 * dwarf2out.h: Expose it here.
11245 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
11247 * dwarf2ctf.c (ctf_debug_finalize): Make it static.
11248 (ctf_debug_early_finish): New definition.
11249 (ctf_debug_finish): Likewise.
11250 * dwarf2ctf.h (ctf_debug_finalize): Remove declaration.
11251 (ctf_debug_early_finish): New declaration.
11252 (ctf_debug_finish): Likewise.
11253 * dwarf2out.c (dwarf2out_finish): Invoke ctf_debug_finish.
11254 (dwarf2out_early_finish): Invoke ctf_debug_early_finish.
11256 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
11258 * config/bpf/bpf.c (bpf_option_override): For BPF backend, disable LTO
11259 support when compiling for CO-RE.
11260 * config/bpf/bpf.opt: Add new command line option -mco-re.
11262 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
11264 * flag-types.h (enum debug_info_type): Add new enum
11265 DINFO_TYPE_BTF_WITH_CORE.
11266 (BTF_WITH_CORE_DEBUG): New bitmask.
11267 * flags.h (btf_with_core_debuginfo_p): New declaration.
11268 * opts.c (btf_with_core_debuginfo_p): New definition.
11270 2021-09-07 Jason Merrill <jason@redhat.com>
11272 * tree.h (error_operand_p): Change to inline function.
11274 2021-09-07 Aldy Hernandez <aldyh@redhat.com>
11276 * tree-ssa-threadedge.c (forwarder_block_p): Rename to...
11277 (empty_block_with_phis_p): ...this.
11278 (potentially_threadable_block): Same.
11279 (jump_threader::thread_through_normal_block): Same.
11281 2021-09-07 Eric Botcazou <ebotcazou@adacore.com>
11284 * dwarf2out.c (mark_base_types): New overloaded function.
11285 (dwarf2out_early_finish): Invoke it on the COMDAT type list as well
11286 as the compilation unit, and call move_marked_base_types afterward.
11288 2021-09-07 H.J. Lu <hjl.tools@gmail.com>
11291 * config/i386/i386-expand.c (ix86_expand_convert_uns_sisf_sse):
11293 (ix86_expand_vector_convert_uns_vsivsf): Likewise.
11295 2021-09-07 Richard Biener <rguenther@suse.de>
11297 PR tree-optimization/102226
11298 * tree-vect-loop.c (vect_transform_cycle_phi): Record
11299 the converted value for the epilogue PHI use.
11301 2021-09-07 Martin Liska <mliska@suse.cz>
11303 PR gcov-profile/80223
11304 * ipa-inline.c (can_inline_edge_p): Similarly to sanitizer
11305 options, do not inline when no_profile_instrument_function
11306 attributes are different in early inliner. It's fine to inline
11307 it after PGO instrumentation.
11309 2021-09-07 Richard Biener <rguenther@suse.de>
11311 PR tree-optimization/101555
11312 * tree-ssa-pre.c (translate_vuse_through_block): Do not
11313 perform an alias walk to determine the validity of the
11314 mem at the start of the block which is already guaranteed
11315 by means of prune_clobbered_mems.
11316 (phi_translate_1): Pass edge to translate_vuse_through_block.
11318 2021-09-07 Xionghu Luo <luoxhu@linux.ibm.com>
11321 * config/rs6000/rs6000.md (fmod<mode>3): New define_expand.
11322 (remainder<mode>3): Likewise.
11324 2021-09-07 YunQiang Su <yunqiang.su@cipunited.com>
11326 * config/mips/mips.c (mips_file_start): add .module for
11329 2021-09-06 Roger Sayle <roger@nextmovesoftware.com>
11331 * wide-int.cc (wi::clz): Reorder tests to ensure the result
11332 is zero for all negative values.
11334 2021-09-06 Tobias Burnus <tobias@codesourcery.com>
11336 * doc/invoke.texi (-foffload-options): Fix @opindex.
11338 2021-09-06 H.J. Lu <hjl.tools@gmail.com>
11341 * config/i386/i386-expand.c (ix86_split_xorsign): Use operands[2].
11342 * config/i386/i386.md (@xorsign<mode>3_1): Add non-destructive
11343 source alternative for AVX.
11345 2021-09-06 liuhongt <hongtao.liu@intel.com>
11347 PR middle-end/102182
11348 * optabs.c (expand_fix): Add from1 to avoid from being
11351 2021-09-06 Eric Botcazou <ebotcazou@adacore.com>
11353 * dwarf2out.c (modified_type_die): Deal with all array types earlier
11354 and use local variable consistently throughout the function.
11356 2021-09-06 Jakub Jelinek <jakub@redhat.com>
11358 PR tree-optimization/102207
11359 * match.pd: Don't demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
11360 were promoted from signed to wider unsigned type.
11362 2021-09-06 Andrew Pinski <apinski@marvell.com>
11364 PR tree-optimization/63184
11365 * match.pd: Add simplification of pointer_diff of two pointer_plus
11366 with addr_expr in the first operand of each pointer_plus.
11367 Add simplificatoin of ne/eq of two pointer_plus with addr_expr
11368 in the first operand of each pointer_plus.
11370 2021-09-06 Richard Biener <rguenther@suse.de>
11372 PR tree-optimization/102176
11373 * tree-vect-slp.c (vect_slp_gather_vectorized_scalar_stmts):
11375 (vect_bb_slp_scalar_cost): Use the computed set of
11376 vectorized scalar stmts instead of relying on the out-of-date
11377 and not accurate PURE_SLP_STMT.
11378 (vect_bb_vectorization_profitable_p): Compute the set
11379 of vectorized scalar stmts.
11381 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
11383 * gimple-range-path.cc (path_range_query::range_of_stmt): Remove
11384 GIMPLE_COND special casing.
11385 (path_range_query::range_defined_in_block): Use range_of_stmt
11386 instead of calling fold_range directly.
11388 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
11390 * gimple-range-path.cc (path_range_query::range_of_expr): Set
11391 m_undefined_path when appropriate.
11392 (path_range_query::internal_range_of_expr): Copy from range_of_expr.
11393 (path_range_query::unreachable_path_p): New.
11394 (path_range_query::precompute_ranges): Set m_undefined_path.
11395 * gimple-range-path.h (path_range_query::unreachable_path_p): New.
11396 (path_range_query::internal_range_of_expr): New.
11397 * tree-ssa-threadbackward.c (back_threader::find_taken_edge_cond):
11398 Use unreachable_path_p.
11400 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
11402 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
11403 Remove argument and call find_taken_edge.
11404 (back_threader::resolve_phi): Do not calculate taken edge before
11405 calling maybe_register_path.
11406 (back_threader::find_paths_to_names): Same.
11408 2021-09-05 Jeff Law <jlaw@localhost.localdomain>
11410 * config/h8300/h8300.md (QHSI2 mode iterator): New mode iterator.
11411 * config/h8300/testcompare.md (store_c): Update name, use new
11413 (store_neg_c, store_shifted_c): New patterns.
11415 2021-09-03 Segher Boessenkool <segher@kernel.crashing.org>
11418 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): On ELFv2 use r11
11419 instead of r12 for CR save, in all cases.
11421 2021-09-03 Andrew Pinski <apinski@marvell.com>
11423 * config/aarch64/aarch64-sve-builtins.cc (register_vector_type):
11424 Handle error_mark_node as the type of the type_decl.
11426 2021-09-03 Andrew Pinski <apinski@marvell.com>
11428 * config/aarch64/aarch64-builtins.c (struct aarch64_simd_type_info):
11430 (aarch64_simd_types): Likewise.
11431 (aarch64_simd_intOI_type_node): Likewise.
11432 (aarch64_simd_intCI_type_node): Likewise.
11433 (aarch64_simd_intXI_type_node): Likewise.
11434 * config/aarch64/aarch64.h (aarch64_fp16_type_node): Likewise.
11435 (aarch64_fp16_ptr_type_node): Likewise.
11436 (aarch64_bf16_type_node): Likewise.
11437 (aarch64_bf16_ptr_type_node): Likewise.
11439 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
11441 * range-op.cc (operator_minus::op1_op2_relation_effect): Abstract
11443 (minus_op1_op2_relation_effect): ...here.
11444 (class operator_pointer_diff): New.
11445 (operator_pointer_diff::op1_op2_relation_effect): Call
11446 minus_op1_op2_relation_effect.
11447 (integral_table::integral_table): Add entry for POINTER_DIFF_EXPR.
11449 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
11451 * tree-ssa-threadbackward.c (back_threader::thread_through_all_blocks):
11452 Add may_peel_loop_headers.
11453 (back_threader_registry::thread_through_all_blocks): Same.
11454 (try_thread_blocks): Pass may_peel_loop_headers argument.
11455 (pass_early_thread_jumps::execute): Same.
11457 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
11459 * tree-ssa-threadedge.c (has_phis_p): New.
11460 (forwarder_block_p): New.
11461 (potentially_threadable_block): Call forwarder_block_p.
11462 (jump_threader::thread_around_empty_blocks): Call has_phis_p.
11463 (jump_threader::thread_through_normal_block): Call
11466 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
11468 * tree-ssa-threadbackward.c (back_threader::dump): New.
11469 (back_threader::debug): New.
11470 (back_threader_profitability::profitable_path_p): Dump blocks
11471 even if we are bailing early.
11473 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
11475 * tree-ssa-threadupdate.c (cancel_thread): New.
11476 (jump_thread_path_registry::thread_block_1): Use cancel_thread.
11477 (jump_thread_path_registry::mark_threaded_blocks): Same.
11478 (jump_thread_path_registry::register_jump_thread): Same.
11480 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
11482 * tree-ssa-threadedge.c (jt_state::push): Only call methods for
11483 which objects are available.
11484 (jt_state::pop): Same.
11485 (jt_state::register_equiv): Same.
11486 (jt_state::register_equivs_on_edge): Same.
11488 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
11490 * tree-ssa-threadedge.c (jump_threader::thread_across_edge):
11491 Move pop until after a thread is registered.
11493 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
11495 * tree-ssa-threadupdate.c (debug): New.
11497 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
11499 * gimple-range-trace.cc (push_dump_file::push_dump_file): New.
11500 (push_dump_file::~push_dump_file): New.
11501 (dump_ranger): Change dump_file temporarily while dumping
11503 * gimple-range-trace.h (class push_dump_file): New.
11505 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
11507 * gimple-range-trace.cc (debug_seed_ranger): Remove static.
11508 (dump_ranger): Dump function name.
11510 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
11512 * gimple-range-path.cc (path_range_query::range_defined_in_block):
11513 Adjust for non-null.
11514 (path_range_query::adjust_for_non_null_uses): New.
11515 (path_range_query::precompute_ranges): Call
11516 adjust_for_non_null_uses.
11517 * gimple-range-path.h: Add m_non_null and
11518 adjust_for_non_null_uses.
11520 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
11522 * gimple-range-path.cc (path_range_query::dump): Dump path
11524 (path_range_query::precompute_ranges): Dump entire path.
11526 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
11528 * value-relation.cc (relation_oracle::debug): New.
11529 * value-relation.h (relation_oracle::debug): New.
11531 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
11533 * tree-ssa-loop-ch.c: Remove unnecessary include file.
11535 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
11537 * gimple-range-fold.cc (fold_using_range::postfold_gcond_edges):
11538 Skip statements with no defining BB.
11539 * gimple-range-path.cc (path_range_query::range_defined_in_block):
11540 Do not get confused by statements with no defining BB.
11542 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
11544 * gimple-range-fold.cc (adjust_imagpart_expr): Move from
11545 gimple_range_adjustment. Add support for constants.
11546 (adjust_realpart_expr): New.
11547 (gimple_range_adjustment): Move IMAGPART_EXPR code to
11548 adjust_imagpart_expr.
11549 * range-op.cc (integral_table::integral_table): Add entry for
11552 2021-09-03 Jakub Jelinek <jakub@redhat.com>
11554 * omp-expand.c (expand_omp_atomic_pipeline): Use
11555 IFN_ATOMIC_COMPARE_EXCHANGE instead of
11556 BUILT_IN_SYNC_VAL_COMPARE_AND_SWAP_? so that memory order
11559 2021-09-03 Jakub Jelinek <jakub@redhat.com>
11562 * tree.h (DECL_FIELD_ABI_IGNORED): Changed into rvalue only macro
11563 that is false if DECL_BIT_FIELD.
11564 (SET_DECL_FIELD_ABI_IGNORED, DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD,
11565 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD): Define.
11566 * tree-streamer-out.c (pack_ts_decl_common_value_fields): For
11567 DECL_BIT_FIELD stream DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead
11568 of DECL_FIELD_ABI_IGNORED.
11569 * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Use
11570 SET_DECL_FIELD_ABI_IGNORED instead of writing to
11571 DECL_FIELD_ABI_IGNORED and for DECL_BIT_FIELD use
11572 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead.
11573 * lto-streamer-out.c (hash_tree): For DECL_BIT_FIELD hash
11574 DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead of DECL_FIELD_ABI_IGNORED.
11576 2021-09-03 liuhongt <hongtao.liu@intel.com>
11579 * config/i386/amxbf16intrin.h : Remove macro check for __AMX_BF16__.
11580 * config/i386/amxint8intrin.h : Remove macro check for __AMX_INT8__.
11581 * config/i386/amxtileintrin.h : Remove macro check for __AMX_TILE__.
11583 2021-09-02 Martin Sebor <msebor@redhat.com>
11585 PR tree-optimization/17506
11587 * tree-ssa-uninit.c (warn_uninit): Remove conditional guarding note.
11589 2021-09-02 Richard Biener <rguenther@suse.de>
11591 * tree-ssa-loop-im.c (fill_always_executed_in_1): Refine
11592 fix for PR78185 and continue processing when leaving
11593 finite inner loops.
11595 2021-09-02 Jakub Jelinek <jakub@redhat.com>
11597 PR tree-optimization/99591
11598 * match.pd: Demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
11601 2021-09-02 Richard Biener <rguenther@suse.de>
11604 2021-09-02 Richard Biener <rguenther@suse.de>
11606 PR tree-optimization/102155
11607 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
11608 over a part of the RPO array and do not recurse here.
11609 Dump blocks marked as always executed.
11610 (fill_always_executed_in): Walk over the RPO array and
11611 process loops whose header we run into.
11612 (loop_invariant_motion_in_fun): Compute the first RPO
11613 using rev_post_order_and_mark_dfs_back_seme in iteration
11614 order and pass that to fill_always_executed_in.
11616 2021-09-02 liuhongt <hongtao.liu@intel.com>
11618 * config/i386/i386-modes.def (FLOAT_MODE): Define ieee HFmode.
11619 * config/i386/i386.c (enum x86_64_reg_class): Add
11620 X86_64_SSEHF_CLASS.
11621 (merge_classes): Handle X86_64_SSEHF_CLASS.
11622 (examine_argument): Ditto.
11623 (construct_container): Ditto.
11624 (classify_argument): Ditto, and set HFmode/HCmode to
11625 X86_64_SSEHF_CLASS.
11626 (function_value_32): Return _FLoat16/Complex Float16 by
11628 (function_value_64): Return _Float16/Complex Float16 by SSE
11630 (ix86_print_operand): Handle CONST_DOUBLE HFmode.
11631 (ix86_secondary_reload): Require gpr as intermediate register
11632 to store _Float16 from sse register when sse4 is not
11634 (ix86_libgcc_floating_mode_supported_p): Enable _FLoat16 under
11636 (ix86_scalar_mode_supported_p): Ditto.
11637 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Defined.
11638 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add HFmode.
11639 (VALID_INT_MODE_P): Add HFmode and HCmode.
11640 * config/i386/i386.md (*pushhf_rex64): New define_insn.
11642 (*movhf_internal): Ditto.
11643 * doc/extend.texi (Half-Precision Floating Point): Documemt
11646 2021-09-02 Richard Biener <rguenther@suse.de>
11648 PR tree-optimization/102155
11649 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
11650 over a part of the RPO array and do not recurse here.
11651 Dump blocks marked as always executed.
11652 (fill_always_executed_in): Walk over the RPO array and
11653 process loops whose header we run into.
11654 (loop_invariant_motion_in_fun): Compute the first RPO
11655 using rev_post_order_and_mark_dfs_back_seme in iteration
11656 order and pass that to fill_always_executed_in.
11658 2021-09-02 YunQiang Su <syq@debian.org>
11661 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
11663 * config/mips/mips.c (mips_module_isa_name): New.
11664 mips_file_start: add .module mipsREV to all asm output
11666 2021-09-01 Jeff Law <jlaw@localhost.localdomain>
11668 PR tree-optimization/102152
11669 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Reduce a vector
11670 comparison to a scalar comparison before calling
11671 update_stmt_if_modified.
11673 2021-09-01 Andrew Pinski <apinski@marvell.com>
11676 * config/aarch64/aarch64.c (aarch64_expand_setmem):
11677 Check STRICT_ALIGNMENT before creating an overlapping
11680 2021-09-01 Martin Sebor <msebor@redhat.com>
11682 * gimple-ssa-warn-access.cc (get_size_range): Add argument.
11683 (check_access): Pass additional argument.
11684 (check_memop_access): Remove template and make a member function.
11685 (maybe_check_dealloc_call): Make a pass_waccess member function.
11686 (class pass_waccess): Add, rename, and remove members.
11687 (pass_waccess::pass_waccess): Adjust to name change.
11688 (pass_waccess::~pass_waccess): Same.
11689 (check_alloca): Make a member function.
11690 (check_alloc_size_call): Same.
11691 (check_strcat): Same.
11692 (check_strncat): Same.
11693 (check_stxcpy): Same.
11694 (check_stxncpy): Same.
11695 (check_strncmp): Same.
11696 (maybe_warn_rdwr_sizes): Rename...
11697 (pass_waccess::maybe_check_access_sizes): ...to this.
11698 (pass_waccess::check_call): Adjust to name changes.
11699 (pass_waccess::maybe_check_dealloc_call): Make a pass_waccess member
11701 (pass_waccess::execute): Adjust to name changes.
11702 * gimple-ssa-warn-access.h (check_memop_access): Remove.
11703 * pointer-query.cc (access_ref::phi): Handle null pointer.
11704 (access_ref::inform_access): Same.
11705 (pointer_query::put_ref): Modify a cached value, not a copy of it.
11706 (pointer_query::dump): New function.
11707 (compute_objsize_r): Avoid overwriting access_ref::bndrng. Cache
11709 * pointer-query.h (pointer_query::dump): Declare.
11710 * tree-ssa-strlen.c (get_range): Simplify. Use function query.
11711 (dump_strlen_info): Use function query.
11712 (printf_strlen_execute): Factor code out into pointer_query::put_ref.
11714 2021-09-01 Thomas Schwinge <thomas@codesourcery.com>
11716 * tree.c (walk_tree_1) <OMP_CLAUSE>: Simplify.
11718 2021-09-01 Iain Sandoe <iain@sandoe.co.uk>
11720 * doc/extend.texi: Document unavailable attribute.
11721 * print-tree.c (print_node): Handle unavailable attribute.
11722 * tree-core.h (struct tree_base): Add a bit to carry unavailability.
11723 * tree.c (error_unavailable_use): New.
11724 * tree.h (TREE_UNAVAILABLE): New.
11725 (error_unavailable_use): New.
11727 2021-09-01 Jakub Jelinek <jakub@redhat.com>
11729 PR tree-optimization/102124
11730 * tree-vect-patterns.c (vect_recog_widen_op_pattern): For ORIG_CODE
11731 MINUS_EXPR, if itype is unsigned with smaller precision than type,
11732 add an extra cast to signed variant of itype to ensure sign-extension.
11734 2021-09-01 Martin Liska <mliska@suse.cz>
11736 * graph.c (draw_cfg_node_succ_edges): Do not color fallthru
11737 edges and rather use colors for TRUE and FALSE edges.
11739 2021-09-01 Richard Biener <rguenther@suse.de>
11741 PR tree-optimization/93491
11742 * tree-ssa-pre.c (compute_avail): Set BB_MAY_NOTRETURN
11743 after processing the stmt itself. Do not consider
11744 pure functions possibly not returning. Properly avoid
11745 adding possibly trapping calls to EXP_GEN when there's
11746 a preceeding possibly not returning call.
11747 * tree-ssa-sccvn.c (vn_reference_may_trap): Conservatively
11750 2021-09-01 Richard Biener <rguenther@suse.de>
11752 PR tree-optimization/102139
11753 * tree-vectorizer.h (vec_base_alignments): Adjust hash-map
11754 type to record a std::pair of the stmt-info and the innermost
11756 (dr_vec_info::group): New member.
11757 * tree-vect-data-refs.c (vect_record_base_alignment): Adjust.
11758 (vect_compute_data_ref_alignment): Verify the recorded
11759 base alignment can be used.
11760 (data_ref_pair): Remove.
11761 (dr_group_sort_cmp): Adjust.
11762 (vect_analyze_data_ref_accesses): Store the group-ID in the
11763 dr_vec_info and operate on a vector of dr_vec_infos.
11765 2021-09-01 YunQiang Su <yunqiang.su@cipunited.com>
11767 * read-md.c (md_reader::handle_enum): support value assignation.
11768 * doc/md.texi: record define_c_enum value assignation support.
11770 2021-09-01 Jakub Jelinek <jakub@redhat.com>
11772 PR tree-optimization/102141
11773 * gimple-ssa-store-merging.c (bswap_view_convert): Add BEFORE
11774 argument. If false, emit stmts after gsi instead of before, and
11776 (bswap_replace): Adjust callers. When converting output of bswap,
11777 emit VIEW_CONVERT prepratation stmts after a copy of gsi instead
11780 2021-09-01 liuhongt <hongtao.liu@intel.com>
11782 * emit-rtl.c (validate_subreg): Get rid of all float-int
11785 2021-09-01 liuhongt <hongtao.liu@intel.com>
11788 2021-08-30 liuhongt <hongtao.liu@intel.com>
11790 * expmed.c (extract_bit_field_1): Make sure we're playing with
11791 integral modes before call extract_integral_bit_field.
11792 (extract_integral_bit_field): Add a parameter of type
11793 scalar_int_mode which corresponds to of tmode.
11794 And call extract_and_convert_fixed_bit_field instead of
11795 extract_fixed_bit_field and convert_extracted_bit_field.
11796 (extract_and_convert_fixed_bit_field): New function, it's a
11797 combination of extract_fixed_bit_field and
11798 convert_extracted_bit_field.
11800 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
11802 * tree.c (walk_tree_1) <OMP_CLAUSE_TILE>: Handle three operands.
11804 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
11806 * omp-general.h (omp_is_reference): Rename to...
11807 (omp_privatize_by_reference): ... this. Adjust all users...
11808 * omp-general.c: ... here, ...
11809 * gimplify.c: ... here, ...
11810 * omp-expand.c: ... here, ...
11811 * omp-low.c: ... here.
11813 2021-08-31 Martin Sebor <msebor@redhat.com>
11815 * gimple-ssa-warn-access.cc (maybe_warn_alloc_args_overflow): Test
11816 pointer element for equality to zero, not that of the cotaining
11819 2021-08-31 Martin Sebor <msebor@redhat.com>
11821 * gcc-rich-location.h (gcc_rich_location): Make ctor explicit.
11823 2021-08-31 Martin Sebor <msebor@redhat.com>
11825 * function.h (function): Add comments.
11826 (get_range_query): Same. Add attribute returns nonnull.
11828 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
11830 * expr.c (convert_modes): Don't use subreg_promoted_mode on a
11831 SUBREG if it can't be guaranteed to a SUBREG_PROMOTED_VAR_P set.
11832 Instead use the standard (safer) is_a <scalar_int_mode> idiom.
11834 2021-08-31 Jeff Law <jlaw@localhost.localdomain>
11836 * config.gcc (cris-*-elf, cris-*-none): Remove dbxelf.h from
11838 (m32r-*-elf, m32rle-*-elf, m32r-*-linux): Likewise.
11839 (mn10300-*-*, am33_2.0-*-linux*): Likewise.
11840 (xtensa*-*-elf, xtensa*-*-linux, xtensa*-*-uclinux): Likewise.
11841 (m32c-*-elf*, m32c-*-rtems*): Likewise.
11842 * config/cris/cris.h (DBX_NO_XREFS): Remove.
11843 (DBX_CONTIN_LENGTH, DBX_CONTIN_CHAR): Likewise.
11844 * config/m32r/m32r.h (DBXOUT_SOURCE_LINE): Likewise.
11845 (DBX_DEBUGGING_INFO, DBX_CONTIN_LENGTH): Likewise.
11846 * config/mn10300/mn10300.h (DEFAULT_GDB_EXTENSIONS): Likewise.
11847 * config/mn10300/linux.h (DBX_REGISTER_NAMES): Likewise.
11849 2021-08-31 Marcel Vollweiler <marcel@codesourcery.com>
11851 * gimplify.c (gimplify_scan_omp_clauses): Error handling. 'ancestor' only
11852 allowed on target constructs and only with particular other clauses.
11853 * omp-expand.c (expand_omp_target): Output of 'sorry, not supported' if
11854 'ancestor' is used.
11855 * omp-low.c (check_omp_nesting_restrictions): Error handling. No nested OpenMP
11856 structs when 'ancestor' is used.
11857 (scan_omp_1_stmt): No usage of OpenMP runtime routines in a target region when
11858 'ancestor' is used.
11859 * tree-pretty-print.c (dump_omp_clause): Append 'ancestor'.
11860 * tree.h (OMP_CLAUSE_DEVICE_ANCESTOR): Define macro.
11862 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
11864 * expr.c (convert_modes): Preserve SUBREG_PROMOTED_VAR_P when
11865 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
11867 * simplify-rtx.c (simplify_unary_operation_1) [SIGN_EXTEND]:
11868 Likewise, preserve SUBREG_PROMOTED_VAR_P when creating a (wider)
11869 partial subreg from a SUBREG_PROMOTED_VAR_P subreg. Generate
11870 SIGN_EXTEND of the SUBREG_REG when a subreg would be paradoxical.
11871 [ZERO_EXTEND]: Likewise, preserve SUBREG_PROMOTED_VAR_P when
11872 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
11873 subreg. Generate ZERO_EXTEND of the SUBREG_REG when a subreg
11874 would be paradoxical.
11876 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
11878 * combine.c (combine_simplify_rtx): Avoid converting an explicit
11879 TRUNCATE into a lowpart SUBREG on !TRULY_NOOP_TRUNCATION targets.
11880 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11882 2021-08-31 Richard Biener <rguenther@suse.de>
11884 PR tree-optimization/102142
11885 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Fix
11886 condition under which to unset the visited flag.
11888 2021-08-31 Richard Biener <rguenther@suse.de>
11890 PR middle-end/102129
11891 * tree-ssa-ter.c (find_replaceable_in_bb): Do not move
11892 possibly trapping expressions across calls.
11894 2021-08-31 Jakub Jelinek <jakub@redhat.com>
11896 PR tree-optimization/102134
11897 * tree-ssa-ccp.c (bit_value_binop) <case RSHIFT_EXPR>: If sgn is
11898 UNSIGNED and r1val | r1mask has MSB set, ensure lzcount doesn't
11901 2021-08-31 Andrew Pinski <apinski@marvell.com>
11904 * collect-utils.c (setup_signals): New declaration.
11905 * collect-utils.h (setup_signals): New function.
11906 * collect2.c (handler): Delete.
11907 (main): Instead of manually setting up the signals,
11908 just call setup_signals.
11909 * lto-wrapper.c (main): Likewise.
11911 2021-08-31 Andrew Pinski <apinski@marvell.com>
11914 * config/i386/i386-protos.h (x86_output_aligned_bss):
11915 Change align argument to unsigned type.
11916 (x86_elf_aligned_decl_common): Likewise.
11917 * config/i386/i386.c (x86_elf_aligned_decl_common): Likewise.
11918 (x86_output_aligned_bss): Likewise.
11920 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
11922 * config/mips/mips.c (mips_module_isa_name): New.
11923 mips_file_start: add .module mipsREV to all asm output
11925 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
11927 * config/mips/mips.h (struct mips_cpu_info): define enum mips_isa;
11928 use enum instead of int for 'isa' member.
11929 * config.gcc, config/mips/mips.c, config/mips/mips-cpus.def,
11930 config/mips/netbsd.h: replace hardcoded numbers with enum.
11932 2021-08-31 liuhongt <hongtao.liu@intel.com>
11934 * config/i386/sse.md (*<avx512>_ucmp<mode>3_1): Change from
11935 define_split to define_insn_and_split.
11936 (*avx2_eq<mode>3): Removed.
11937 (<avx512>_eq<mode>3<mask_scalar_merge_name>): Adjust pattern
11938 (<avx512>_eq<mode>3<mask_scalar_merge_name>_1): Rename to ..
11939 (*<avx512>_eq<mode>3<mask_scalar_merge_name>_1): .. this, and
11941 (*avx2_gt<mode>3): Removed.
11942 (<avx512>_gt<mode>3<mask_scalar_merge_name>): Change from
11943 define_insn to define_expand, and adjust pattern.
11944 (UNSPEC_MASKED_EQ, UNSPEC_MASKED_GT): Removed.
11946 2021-08-30 David Malcolm <dmalcolm@redhat.com>
11949 * Makefile.in (ANALYZER_OBJS): Add analyzer/call-info.o.
11951 2021-08-30 Jason Merrill <jason@redhat.com>
11953 * doc/invoke.texi: Document -Wmissing-requires.
11955 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
11957 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Remove
11958 TARGET_EXTRA_BUILTINS guard.
11960 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
11962 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Change
11963 initialization of V2DI_type_node and unsigned_V2DI_type_node.
11965 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
11967 * config/rs6000/darwin.h (SUBTARGET_INIT_BUILTINS): Use the new
11968 decl when new_builtins_are_live.
11969 * config/rs6000/rs6000-builtin-new.def (__builtin_cfstring): New
11972 2021-08-30 Pat Haugen <pthaugen@linux.ibm.com>
11974 * config/rs6000/rs6000-cpus.def (ISA_3_1_MASKS_SERVER): Add
11975 OPTION_MASK_P10_FUSION_2STORE.
11976 (POWERPC_MASKS): Likewise.
11977 * config/rs6000/rs6000.c (rs6000_option_override_internal): Enable
11978 store fusion for Power10.
11979 (is_fusable_store): New.
11980 (power10_sched_reorder): Likewise.
11981 (rs6000_sched_reorder): Do Power10 specific reordering.
11982 (rs6000_sched_reorder2): Likewise.
11983 * config/rs6000/rs6000.opt: Add new option.
11985 2021-08-30 Richard Biener <rguenther@suse.de>
11987 PR tree-optimization/102128
11988 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
11989 Move scanning for if-converted scalar code to the caller
11990 and instead delay clearing the visited flag for profitable
11992 (vect_slp_region): Cost all subgraphs before scheduling.
11993 For if-converted BB vectorization scan for scalar COND_EXPRs
11994 and do not vectorize if any found and the cost model is
11997 2021-08-30 Richard Biener <rguenther@suse.de>
11999 * common.opt (fexceptions): Mark
12000 EnabledBy(fnon-call-exceptions).
12001 * doc/invoke.texi (fnon-call-exceptions): Document this
12002 enables -fexceptions.
12004 2021-08-30 Sebastian Huber <sebastian.huber@embedded-brains.de>
12006 * tsystem.h (abort): Define abort() if inhibit_libc is defined and it
12007 is not already defined.
12009 2021-08-30 liuhongt <hongtao.liu@intel.com>
12011 * expmed.c (extract_bit_field_1): Make sure we're playing with
12012 integral modes before call extract_integral_bit_field.
12013 (extract_integral_bit_field): Add a parameter of type
12014 scalar_int_mode which corresponds to of tmode.
12015 And call extract_and_convert_fixed_bit_field instead of
12016 extract_fixed_bit_field and convert_extracted_bit_field.
12017 (extract_and_convert_fixed_bit_field): New function, it's a
12018 combination of extract_fixed_bit_field and
12019 convert_extracted_bit_field.
12021 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
12023 * config/darwin.c (darwin_libc_has_function): Do not run
12024 the checks for x86 or modern Darwin. Make sure that there
12025 is a value set for darwin_macosx_version_min before testing.
12027 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
12029 * config/i386/darwin.h (CLEAR_INSN_CACHE): New.
12031 2021-08-28 Jan Hubicka <hubicka@ucw.cz>
12033 * ipa-modref-tree.h (modref_access_node::merge): Break out
12034 logic combining offsets and logic merging ranges to ...
12035 (modref_access_node::combined_offsets): ... here
12036 (modref_access_node::update2): ... here
12037 (modref_access_node::closer_pair_p): New member function.
12038 (modref_access_node::forced_merge): New member function.
12039 (modre_ref_node::insert): Do merging when table is full.
12041 2021-08-28 YunQiang Su <yunqiang.su@cipunited.com>
12044 * config.gcc: MIPS: use N64 ABI by default if the triple end
12045 with -gnuabi64, which is used by Debian since 2013.
12047 2021-08-28 Alexandre Oliva <oliva@adacore.com>
12049 * ipa-modref.c (analyze_function): Skip debug stmts.
12050 * tree-inline.c (estimate_num_insn): Consider builtins even
12051 without a cgraph_node.
12053 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
12055 * config/h8300/bitfield.md (cstore<mode>4): Remove expander.
12056 * config/h8300/h8300.c (h8300_expand_branch): Remove function.
12057 * config/h8300/h8300-protos.h (h8300_expadn_branch): Remove prototype.
12058 * config/h8300/h8300.md (eqne): New code iterator.
12059 (geultu, geultu_to_c): Similarly.
12060 * config/h8300/testcompare.md (cstore<mode>4): Dummy expander.
12061 (store_c_<mode>, store_c_i_<mode>): New define_insn_and_splits
12062 (cmp<mode>_c): New pattern
12064 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
12066 * tree-ssa-dom.c (reduce_vector_comparison_to_scalar_comparison): New
12068 (dom_opt_dom_walker::optimize_stmt): Use it.
12070 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
12072 * config/darwin.c (finalize_ctors): Add a section-start linker-
12074 (finalize_dtors): Likewise.
12075 * config/darwin.h (MIN_LD64_INIT_TERM_START_LABELS): New.
12077 2021-08-27 Bill Schmidt <wschmidt@linux.ibm.com>
12079 * config/rs6000/rs6000-call.c (rs6000-builtins.h): New #include.
12080 (rs6000_init_builtins): Call rs6000_init_generated_builtins. Skip the
12081 old initialization logic when new builtins are enabled.
12082 * config/rs6000/rs6000-gen-builtins.c (write_decls): Rename
12083 rs6000_autoinit_builtins to rs6000_init_generated_builtins.
12084 (write_init_file): Likewise.
12086 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
12088 * configure.ac (darwin2[[0-9]]* | darwin19*): Alter use of
12089 gcc_GAS_CHECK_FEATURE to remove an extraneous parameter.
12090 (amdgcn-* | gcn-*) Likewise.
12092 2021-08-27 Anthony Sharp <anthonysharp15@gmail.com>
12094 * symbol-summary.h: Added missing template keyword.
12096 2021-08-27 Richard Biener <rguenther@suse.de>
12098 PR tree-optimization/45178
12099 * tree-ssa-dce.c (find_obviously_necessary_stmts): For
12100 infinite loops without exit do not mark control dependent
12101 edges of the latch necessary.
12103 2021-08-27 konglin1 <lingling.kong@intel.com>
12106 * config/i386/sse.md: (<avx512>scattersi<mode>): Add mask operand to
12108 (<avx512>scattersi<mode>): Likewise.
12109 (*avx512f_scattersi<VI48F:mode>): Merge mask operand to set_dest.
12110 (*avx512f_scatterdi<VI48F:mode>): Likewise
12112 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
12114 * config/rs6000/rs6000.c (rs6000_builtin_md_vectorized_function): Add
12115 support for built-in functions MISC_BUILTIN_DIVWE, MISC_BUILTIN_DIVWEU,
12116 MISC_BUILTIN_DIVDE, MISC_BUILTIN_DIVDEU, P10_BUILTIN_CFUGED,
12117 P10_BUILTIN_CNTLZDM, P10_BUILTIN_CNTTZDM, P10_BUILTIN_PDEPD and
12118 P10_BUILTIN_PEXTD on Power10.
12120 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
12122 * config/rs6000/rs6000-call.c (builtin_function_type): Add unsigned
12123 signedness for some Power10 bifs.
12125 2021-08-27 David Edelsohn <dje.gcc@gmail.com>
12128 * config/rs6000/rs6000.c (rs6000_adjust_field_align): Use
12129 computed alignment if the entire struct has attribute packed.
12131 2021-08-27 liuhongt <hongtao.liu@intel.com>
12135 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
12136 IX86_BUILTIN_SHUFPD512, IX86_BUILTIN_SHUFPS512,
12137 IX86_BUILTIN_SHUFPD256, IX86_BUILTIN_SHUFPS,
12138 IX86_BUILTIN_SHUFPS256.
12139 (ix86_masked_all_ones): New function.
12141 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
12143 * config/i386/i386.md (*btr<mode>_1): Call force_reg unconditionally.
12144 (conditional moves with memory inputs splitters): Ditto.
12145 * config/i386/sse.md (one_cmpl<mode>2): Simplify.
12147 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
12149 * ipa-modref-tree.h (modref_access_node::try_merge_with): Restart
12150 search after merging.
12152 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
12154 * config/rs6000/rs6000-overload.def: Add remaining overloads.
12156 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
12158 * config/rs6000/rs6000-builtin-new.def: Add cell stanza.
12160 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
12162 * config/rs6000/rs6000-builtin-new.def: Add ieee128-hw, dfp,
12163 crypto, and htm stanzas.
12165 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
12167 * config/rs6000/rs6000-builtin-new.def: Add mma stanza.
12169 2021-08-26 Martin Sebor <msebor@redhat.com>
12171 * tree-ssa-uninit.c (warn_uninit): Refactor and simplify.
12172 (warn_uninit_phi_uses): Remove argument from calls to warn_uninit.
12173 (warn_uninitialized_vars): Same. Reduce visibility of locals.
12174 (warn_uninitialized_phi): Same.
12176 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
12178 * tree-ssa-ccp.c (get_individual_bits): Helper function to
12179 extract the individual bits from a widest_int constant (mask).
12180 (gray_code_bit_flips): New read-only table for effiently
12181 enumerating permutations/combinations of bits.
12182 (bit_value_binop) [LROTATE_EXPR, RROTATE_EXPR]: Handle rotates
12183 by unknown counts that are guaranteed less than the target
12184 precision and four or fewer unknown bits by enumeration.
12185 [LSHIFT_EXPR, RSHIFT_EXPR]: Likewise, also handle shifts by
12186 enumeration under the same conditions. Handle remaining
12187 shifts as a mask based upon the minimum possible shift value.
12189 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
12190 Richard Biener <rguenther@suse.de>
12192 * match.pd (shift transformations): Remove a redundant
12193 !POINTER_TYPE_P check.
12195 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
12198 * config/i386/i386.md (cmove reg-reg move elimination peephole2s):
12199 Set all_regs to true in the call to replace_rtx.
12201 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
12203 * ipa-modref-tree.c (test_insert_search_collapse): Update test.
12204 * ipa-modref-tree.h (modref_base_node::insert): Be smarter when
12205 hiting --param modref-max-refs limit.
12206 (modref_tree:insert_base): Be smarter when hitting
12207 --param modref-max-bases limit. Add new parameter REF.
12208 (modref_tree:insert): Update.
12209 (modref_tree:merge): Update.
12210 * ipa-modref.c (read_modref_records): Update.
12212 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
12214 * params.opt: (modref-max-adjustments): Add full stop.
12216 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
12218 * ipa-modref-tree.h (modref_ref_node::verify): New member
12220 (modref_ref_node::insert): Use it.
12221 (modref_ref_node::try_mere_with): Fix off by one error.
12223 2021-08-26 Martin Liska <mliska@suse.cz>
12224 Stefan Kneifel <stefan.kneifel@bluewin.ch>
12226 * cgraph.h (create_version_clone_with_body): Add new parameter.
12227 * cgraphclones.c: Likewise.
12228 * multiple_target.c (create_dispatcher_calls): Do not use
12230 (create_target_clone): Likewise here.
12232 2021-08-26 Jonathan Yong <10walls@gmail.com>
12234 * doc/extend.texi: Add note about reserved priorities
12235 to the constructor attribute.
12237 2021-08-25 Martin Sebor <msebor@redhat.com>
12239 * gimple-range-cache.cc (ssa_global_cache::dump): Avoid printing
12240 range table header alone.
12241 * gimple-range.cc (gimple_ranger::export_global_ranges): Same.
12243 2021-08-25 Jan Hubicka <hubicka@ucw.cz>
12245 * doc/invoke.texi: Document --param modref-max-adjustments.
12246 * ipa-modref-tree.c (test_insert_search_collapse): Update.
12247 (test_merge): Update.
12248 * ipa-modref-tree.h (struct modref_access_node): Add adjustments;
12249 (modref_access_node::operator==): Fix handling of access ranges.
12250 (modref_access_node::contains): Constify parameter; handle also
12251 mismatched parm offsets.
12252 (modref_access_node::update): New function.
12253 (modref_access_node::merge): New function.
12254 (unspecified_modref_access_node): Update constructor.
12255 (modref_ref_node::insert_access): Add record_adjustments parameter;
12257 (modref_ref_node::try_merge_with): New private function.
12258 (modref_tree::insert): New record_adjustments parameter.
12259 (modref_tree::merge): New record_adjustments parameter.
12260 (modref_tree::copy_from): Update.
12261 * ipa-modref.c (dump_access): Dump adjustments field.
12262 (get_access): Update constructor.
12263 (record_access): Update call of insert.
12264 (record_access_lto): Update call of insert.
12265 (merge_call_side_effects): Add record_adjustments parameter.
12266 (get_access_for_fnspec): Update.
12267 (process_fnspec): Update.
12268 (analyze_call): Update.
12269 (analyze_function): Update.
12270 (read_modref_records): Update.
12271 (ipa_merge_modref_summary_after_inlining): Update.
12272 (propagate_unknown_call): Update.
12273 (modref_propagate_in_scc): Update.
12274 * params.opt (param-max-modref-adjustments=): New.
12276 2021-08-25 Michael Meissner <meissner@linux.ibm.com>
12278 * config/rs6000/vsx.md (UNSPEC_XXSPLTIDP): Rename from
12280 (xxspltiw_v4si): Use vecperm type attribute.
12281 (xxspltiw_v4si_inst): Use vecperm type attribute.
12282 (xxspltiw_v4sf_inst): Likewise.
12283 (xxspltidp_v2df): Use vecperm type attribute. Use
12284 UNSPEC_XXSPLTIDP instead of UNSPEC_XXSPLTID.
12285 (xxspltidp_v2df_inst): Likewise.
12286 (xxsplti32dx_v4si): Use vecperm type attribute.
12287 (xxsplti32dx_v4si_inst): Likewise.
12288 (xxsplti32dx_v4sf_inst): Likewise.
12289 (xxblend_<mode>): Likewise.
12290 (xxpermx): Likewise.
12291 (xxpermx_inst): Likewise.
12292 (xxeval): Likewise.
12294 2021-08-25 Lewis Hyatt <lhyatt@gmail.com>
12297 * coretypes.h (typedef diagnostic_input_charset_callback): Declare.
12298 * diagnostic.c (diagnostic_initialize_input_context): New function.
12299 * diagnostic.h (diagnostic_initialize_input_context): Declare.
12300 * input.c (default_charset_callback): New function.
12301 (file_cache::initialize_input_context): New function.
12302 (file_cache_slot::create): Added ability to convert the input
12303 according to the input context.
12304 (file_cache::file_cache): Initialize the new input context.
12305 (class file_cache_slot): Added new m_alloc_offset member.
12306 (file_cache_slot::file_cache_slot): Initialize the new member.
12307 (file_cache_slot::~file_cache_slot): Handle potentially offset buffer.
12308 (file_cache_slot::maybe_grow): Likewise.
12309 (file_cache_slot::needs_read_p): Handle NULL fp, which is now possible.
12310 (file_cache_slot::get_next_line): Likewise.
12311 * input.h (class file_cache): Added input context member.
12313 2021-08-25 Richard Biener <rguenther@suse.de>
12315 PR tree-optimization/102046
12316 * tree-vect-slp.c (vect_build_slp_tree_2): Conservatively
12317 update ->any_pattern when swapping operands.
12319 2021-08-25 Hongyu Wang <hongyu.wang@intel.com>
12322 * config/i386/i386.c (ix86_live_on_entry): Adjust comment.
12323 (ix86_decompose_address): Remove retval check for ASHIFT,
12324 allow non-canonical zero extend if AND mask covers ASHIFT
12326 (ix86_legitimate_address_p): Adjust condition for decompose.
12327 (ix86_rtx_costs): Adjust cost for lea with non-canonical
12329 Co-Authored by: Uros Bizjak <ubizjak@gmail.com>
12331 2021-08-25 Jiufu Guo <guojiufu@linux.ibm.com>
12333 PR tree-optimization/101145
12334 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
12336 (number_of_iterations_lt): Invoke above function.
12337 (adjust_cond_for_loop_until_wrap):
12338 Merge to number_of_iterations_until_wrap.
12339 (number_of_iterations_cond): Update invokes for
12340 adjust_cond_for_loop_until_wrap and number_of_iterations_lt.
12342 2021-08-25 konglin1 <lingling.kong@intel.com>
12345 * config/i386/avx512dqintrin.h (_mm512_fpclass_ps_mask): Fix
12346 macro define in O0.
12347 (_mm512_mask_fpclass_ps_mask): Ditto.
12349 2021-08-25 Kewen Lin <linkw@linux.ibm.com>
12351 * config/rs6000/altivec.md (vec_unpacku_hi_v16qi): Remove.
12352 (vec_unpacku_hi_v8hi): Likewise.
12353 (vec_unpacku_lo_v16qi): Likewise.
12354 (vec_unpacku_lo_v8hi): Likewise.
12355 (vec_unpacku_hi_<VP_small_lc>): New define_expand.
12356 (vec_unpacku_lo_<VP_small_lc>): Likewise.
12358 2021-08-24 David Edelsohn <dje.gcc@gmail.com>
12360 * config/rs6000/aix.h (SYSTEM_IMPLICIT_EXTERN_C): Delete.
12361 * config/rs6000/aix71.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
12362 * config/rs6000/aix72.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
12363 * config/rs6000/aix73.h (TARGET_AIX_VERSION): Increase to 73.
12365 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
12367 PR middle-end/102031
12368 * simplify-rtx.c (simplify_truncation): When comparing precisions
12369 use "subreg_prec" variable, not "subreg_mode".
12371 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
12373 * config/rs6000/rs6000-builtin-new.def: Add power10 and power10-64
12376 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
12378 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Initialize
12379 various pointer type nodes.
12380 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add enum
12381 values for various pointer types.
12382 (ptr_V16QI_type_node): New macro.
12383 (ptr_V1TI_type_node): New macro.
12384 (ptr_V2DI_type_node): New macro.
12385 (ptr_V2DF_type_node): New macro.
12386 (ptr_V4SI_type_node): New macro.
12387 (ptr_V4SF_type_node): New macro.
12388 (ptr_V8HI_type_node): New macro.
12389 (ptr_unsigned_V16QI_type_node): New macro.
12390 (ptr_unsigned_V1TI_type_node): New macro.
12391 (ptr_unsigned_V8HI_type_node): New macro.
12392 (ptr_unsigned_V4SI_type_node): New macro.
12393 (ptr_unsigned_V2DI_type_node): New macro.
12394 (ptr_bool_V16QI_type_node): New macro.
12395 (ptr_bool_V8HI_type_node): New macro.
12396 (ptr_bool_V4SI_type_node): New macro.
12397 (ptr_bool_V2DI_type_node): New macro.
12398 (ptr_bool_V1TI_type_node): New macro.
12399 (ptr_pixel_type_node): New macro.
12400 (ptr_intQI_type_node): New macro.
12401 (ptr_uintQI_type_node): New macro.
12402 (ptr_intHI_type_node): New macro.
12403 (ptr_uintHI_type_node): New macro.
12404 (ptr_intSI_type_node): New macro.
12405 (ptr_uintSI_type_node): New macro.
12406 (ptr_intDI_type_node): New macro.
12407 (ptr_uintDI_type_node): New macro.
12408 (ptr_intTI_type_node): New macro.
12409 (ptr_uintTI_type_node): New macro.
12410 (ptr_long_integer_type_node): New macro.
12411 (ptr_long_unsigned_type_node): New macro.
12412 (ptr_float_type_node): New macro.
12413 (ptr_double_type_node): New macro.
12414 (ptr_long_double_type_node): New macro.
12415 (ptr_dfloat64_type_node): New macro.
12416 (ptr_dfloat128_type_node): New macro.
12417 (ptr_ieee128_type_node): New macro.
12418 (ptr_ibm128_type_node): New macro.
12419 (ptr_vector_pair_type_node): New macro.
12420 (ptr_vector_quad_type_node): New macro.
12421 (ptr_long_long_integer_type_node): New macro.
12422 (ptr_long_long_unsigned_type_node): New macro.
12424 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
12426 * config/rs6000/rs6000-builtin-new.def: Add power9-vector, power9,
12427 and power9-64 stanzas.
12429 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
12430 Tom de Vries <tdevries@suse.de>
12432 * config.gcc (nvptx-*-*): Define {c,c++}_target_objs.
12433 * config/nvptx/nvptx-protos.h (nvptx_cpu_cpp_builtins): Prototype.
12434 * config/nvptx/nvptx.h (TARGET_CPU_CPP_BUILTINS): Implement with
12435 a call to the new nvptx_cpu_cpp_builtins function in nvptx-c.c.
12436 * config/nvptx/t-nvptx (nvptx-c.o): New rule.
12437 * config/nvptx/nvptx-c.c: New source file.
12438 (nvptx_cpu_cpp_builtins): Move implementation here.
12440 2021-08-24 Martin Sebor <msebor@redhat.com>
12442 PR middle-end/101600
12443 PR middle-end/101977
12444 * gimple-ssa-warn-access.cc (maybe_warn_for_bound): Tighten up
12445 the phrasing of a warning.
12446 (check_access): Use the remaining size after subtracting any offset
12447 rather than the whole object size.
12448 * pointer-query.cc (access_ref::get_ref): Clear BASE0 flag if it's
12449 clear for any nonnull PHI argument.
12450 (compute_objsize): Clear argument.
12452 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
12454 * config/rs6000/rs6000-builtin-new.def: Add power8-vector stanza.
12456 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
12458 * config/rs6000/rs6000-builtin-new.def: Add power7 and power7-64
12461 2021-08-24 Andrew MacLeod <amacleod@redhat.com>
12463 * value-relation.cc (rr_transitive_table): New.
12464 (relation_transitive): New.
12465 (value_relation::swap): Remove.
12466 (value_relation::apply_transitive): New.
12467 (relation_oracle::relation_oracle): Allocate a new tmp bitmap.
12468 (relation_oracle::register_relation): Call register_transitives.
12469 (relation_oracle::register_transitives): New.
12470 * value-relation.h (relation_oracle): Add new temporary bitmap and
12473 2021-08-24 H.J. Lu <hjl.tools@gmail.com>
12476 * config/i386/i386-expand.c (ix86_expand_vector_move): Broadcast
12477 from integer to a pseudo vector register.
12479 2021-08-24 Richard Biener <rguenther@suse.de>
12481 PR tree-optimization/100089
12482 * tree-vectorizer.h (vect_slp_bb): Rename to ...
12483 (vect_slp_if_converted_bb): ... this and get the original
12484 loop as new argument.
12485 * tree-vectorizer.c (try_vectorize_loop_1): Revert previous fix,
12486 pass original loop to vect_slp_if_converted_bb.
12487 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
12488 If orig_loop was passed scan the not vectorized stmts
12489 for COND_EXPRs and force not profitable if found.
12490 (vect_slp_region): Pass down all SLP instances to costing
12491 if orig_loop was specified.
12492 (vect_slp_bbs): Pass through orig_loop.
12493 (vect_slp_bb): Rename to ...
12494 (vect_slp_if_converted_bb): ... this and get the original
12495 loop as new argument.
12496 (vect_slp_function): Adjust.
12498 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
12501 * config/arm/arm.md (attribute arch): Add fix_vlldm.
12502 (arch_enabled): Use it.
12503 * config/arm/vfp.md (lazy_store_multiple_insn): Add alternative to
12504 use when erratum mitigation is needed.
12506 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
12509 * config/arm/arm.opt (mfix-cmse-cve-2021-35465): New option.
12510 * doc/invoke.texi (Arm Options): Document it.
12511 * config/arm/arm-cpus.in (quirk_vlldm): New feature bit.
12512 (ALL_QUIRKS): Add quirk_vlldm.
12513 (cortex-m33): Add quirk_vlldm.
12514 (cortex-m35p, cortex-m55): Likewise.
12515 * config/arm/arm.c (arm_option_override): Enable fix_vlldm if
12516 targetting an affected CPU and not explicitly controlled on
12519 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
12521 * config/arm/vfp.md (lazy_store_multiple_insn): Rewrite as valid RTL.
12522 (lazy_load_multiple_insn): Likewise.
12524 2021-08-24 liuhongt <hongtao.liu@intel.com>
12527 * config/i386/sse.md (<avx512>_vternlog<mode><sd_maskz_name>):
12528 Enable avx512 embedded broadcast.
12529 (*<avx512>_vternlog<mode>_all): Ditto.
12530 (<avx512>_vternlog<mode>_mask): Ditto.
12532 2021-08-24 liuhongt <hongtao.liu@intel.com>
12535 * config/i386/i386.c (ix86_rtx_costs): Define cost for
12537 * config/i386/i386.h (STRIP_UNARY): New macro.
12538 * config/i386/predicates.md (reg_or_notreg_operand): New
12540 * config/i386/sse.md (*<avx512>_vternlog<mode>_all): New define_insn.
12541 (*<avx512>_vternlog<mode>_1): New pre_reload
12542 define_insn_and_split.
12543 (*<avx512>_vternlog<mode>_2): Ditto.
12544 (*<avx512>_vternlog<mode>_3): Ditto.
12545 (any_logic1,any_logic2): New code iterator.
12546 (logic_op): New code attribute.
12547 (ternlogsuffix): Extend to VNxDF and VNxSF.
12549 2021-08-24 Richard Biener <rguenther@suse.de>
12551 * doc/invoke.texi (vect-inner-loop-cost-factor): Adjust.
12552 * params.opt (--param vect-inner-loop-cost-factor): Adjust
12554 * tree-vect-loop.c (vect_analyze_loop_form): Initialize
12555 inner_loop_cost_factor to the minimum of the estimated number
12556 of iterations of the inner loop and vect-inner-loop-cost-factor.
12558 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
12559 Richard Biener <rguenther@suse.de>
12561 * config/i386/i386-features.c (compute_convert_gain): Provide
12562 more accurate values for CONST_INT, when optimizing for size.
12563 * config/i386/i386.c (COSTS_N_BYTES): Move definition from here...
12564 * config/i386/i386.h (COSTS_N_BYTES): to here.
12566 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
12567 Jakub Jelinek <jakub@redhat.com>
12569 PR middle-end/102029
12570 * match.pd (shift transformations): Add an additional check for
12571 !POINTER_TYPE_P in the recently added left shift transformation.
12573 2021-08-24 liuhongt <hongtao.liu@intel.com>
12575 PR tree-optimization/100089
12576 * tree-vectorizer.c (try_vectorize_loop_1): Disable slp in
12577 loop vectorizer when cost model is very-cheap.
12579 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
12581 * config/rs6000/rs6000-gen-builtins.c (parse_bif_entry): Don't call
12582 asprintf, which is not available on AIX.
12584 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
12586 * config.gcc (target_gtfiles): Add ./rs6000-builtins.h.
12587 * config/rs6000/t-rs6000 (EXTRA_GTYPE_DEPS): Set.
12589 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
12591 * config.gcc (powerpc*-*-*): Add rs6000-builtins.o to extra_objs.
12592 * config/rs6000/rs6000-gen-builtins.c (main): Close init_file
12594 * config/rs6000/t-rs6000 (rs6000-gen-builtins.o): New target.
12595 (rbtree.o): Likewise.
12596 (rs6000-gen-builtins): Likewise.
12597 (rs6000-builtins.c): Likewise.
12598 (rs6000-builtins.h): Likewise.
12599 (rs6000.o): Add dependency.
12600 (EXTRA_HEADERS): Add rs6000-vecdefines.h.
12601 (rs6000-vecdefines.h): New target.
12602 (rs6000-builtins.o): Likewise.
12603 (rs6000-call.o): Add rs6000-builtins.h as a dependency.
12604 (rs6000-c.o): Likewise.
12606 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
12609 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace):
12610 Diagnose buffer overrun.
12611 (safe_inc_pos): Fix overrun detection.
12612 (match_identifier): Diagnose buffer overrun.
12613 (match_integer): Likewise.
12614 (match_to_right_bracket): Likewise.
12616 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
12618 * ipa-modref-tree.h (modref_access_node::range_info_useful_p):
12619 Improve range compare.
12620 (modref_access_node::contains): New member function.
12621 (modref_access_node::search): Remove.
12622 (modref_access_node::insert): Be smarter about subaccesses.
12624 2021-08-23 Thomas Schwinge <thomas@codesourcery.com>
12626 * config/i386/i386-options.c (ix86_omp_device_kind_arch_isa)
12627 <omp_device_arch> [ACCEL_COMPILER]: Match "intel_mic".
12628 * config/i386/t-omp-device (omp-device-properties-i386) <arch>:
12631 2021-08-23 Jeff Law <jlaw@localhost.localdomain>
12633 * config/h8300/h8300-protos.h (h8300_expand_epilogue): Add new
12635 * config/h8300/jumpcall.md (call, call_value): Restrict to
12636 !SIBLING_CALL_P cases.
12637 (subcall, sibcall_value): New patterns & expanders.
12638 * config/h8300/proepi.md (epilogue): Pass new argument to
12639 h8300_expand_epilogue.
12640 (sibcall_epilogue): New expander.
12641 * config/h8300/h8300.c (h8300_expand_epilogue): Handle sibcall
12643 (h8300_ok_for_sibcall_p): New function.
12644 (TARGET_FUNCTION_OK_FOR_SIBCALL): define.
12646 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
12648 * simplify-rtx.c (simplify_unary_operation_1): [TRUNCATE]:
12649 Handle case where the operand is already the desired mode.
12651 2021-08-23 Richard Biener <rguenther@suse.de>
12654 * tree-ssa-structalias.c (ipa_pta_execute): Check in_other_partition
12655 in addition to has_gimple_body.
12657 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
12659 PR middle-end/101949
12660 * ipa-modref.c (analyze_ssa_name_flags): Fix merging of
12663 2021-08-23 Martin Liska <mliska@suse.cz>
12665 * doc/invoke.texi: Put the option out of -mxl-mode-app-model
12668 2021-08-23 Richard Biener <rguenther@suse.de>
12670 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
12671 Properly scale the inner loop cost only once.
12673 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
12675 * tree-ssa-ccp.c (bit_value_binop) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]:
12676 Provide bounds for unsigned (and signed with non-negative operands)
12677 division and modulus.
12679 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
12681 * simplify-rtx.c (simplify_truncation): Generalize simplification
12682 of (truncate:A (subreg:B X)).
12683 (simplify_unary_operation_1) [FLOAT_TRUNCATE, FLOAT_EXTEND,
12684 SIGN_EXTEND, ZERO_EXTEND]: Handle cases where the operand
12685 already has the desired machine mode.
12686 (test_scalar_int_ops): Add tests that useless extensions and
12687 truncations are optimized away.
12688 (test_scalar_int_ext_ops): New self-test function to confirm
12689 that truncations of extensions are correctly simplified.
12690 (test_scalar_int_ext_ops2): New self-test function to check
12691 truncations of truncations, extensions of extensions, and
12692 truncations of extensions.
12693 (test_scalar_ops): Call the above two functions with a
12694 representative sampling of integer machine modes.
12696 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
12698 * match.pd (shift transformations): Change the sign of an
12699 LSHIFT_EXPR if it reduces the number of explicit conversions.
12701 2021-08-23 Jakub Jelinek <jakub@redhat.com>
12703 PR tree-optimization/86723
12704 * gimple-ssa-store-merging.c (find_bswap_or_nop_finalize): Add
12705 cast64_to_32 argument, set *cast64_to_32 to false, unless n is
12706 non-memory permutation of 64-bit src which only has bytes of
12707 0 or [5..8] and n->range is 4.
12708 (find_bswap_or_nop): Add cast64_to_32 and mask arguments, adjust
12709 find_bswap_or_nop_finalize caller, support bswap with some bytes
12710 zeroed, as long as at least two bytes are not zeroed.
12711 (bswap_replace): Add mask argument and handle masking of bswap
12713 (maybe_optimize_vector_constructor): Adjust find_bswap_or_nop
12714 caller, punt if cast64_to_32 or mask is not all ones.
12715 (pass_optimize_bswap::execute): Adjust find_bswap_or_nop_finalize
12716 caller, for now punt if cast64_to_32.
12718 2021-08-23 Richard Biener <rguenther@suse.de>
12720 PR tree-optimization/79334
12721 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Record
12722 a type also for COMPONENT_REFs.
12723 (vn_reference_may_trap): Check ARRAY_REF with constant index
12724 against the array domain.
12726 2021-08-23 liuhongt <hongtao.liu@intel.com>
12729 * config/i386/sse.md (*avx512f_pshufb_truncv8hiv8qi_1): Add
12730 TARGET_AVX512BW to condition.
12732 2021-08-23 Jakub Jelinek <jakub@redhat.com>
12735 * dwarf2out.c (gen_variable_die): Add DW_AT_location for global
12736 register variables already during early_dwarf if possible.
12738 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
12740 * config/arm/arm_mve.h: Fix __arm_vctp16q return type.
12742 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
12745 * config/arm/arm.opt: Fix typo.
12746 * config/arm/t-rmprofile: Fix typo.
12748 2021-08-23 Jakub Jelinek <jakub@redhat.com>
12750 * tree.h (OMP_CLAUSE_GRAINSIZE_STRICT): Define.
12751 (OMP_CLAUSE_NUM_TASKS_STRICT): Define.
12752 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_GRAINSIZE,
12753 case OMP_CLAUSE_NUM_TASKS>: Print strict: modifier.
12754 * omp-expand.c (expand_task_call): Use GOMP_TASK_FLAG_STRICT in iflags
12755 if either grainsize or num_tasks clause has the strict modifier.
12757 2021-08-23 Martin Liska <mliska@suse.cz>
12759 * dbgcnt.def (DEBUG_COUNTER): New counter.
12760 * gimple.c (gimple_call_arg_flags): Use it in IPA PTA.
12762 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
12764 * ipa-modref.c (analyze_ssa_name_flags): Improve handling of return slot.
12766 2021-08-23 Xi Ruoyao <xry111@mengyan1223.wang>
12769 * config/mips/mips-protos.h (mips_msa_output_shift_immediate):
12771 * config/mips/mips.c (mips_msa_output_shift_immediate): New
12773 * config/mips/mips-msa.md (vashl<mode>3, vashr<mode>3,
12774 vlshr<mode>3): Call it.
12776 2021-08-22 Jan Hubicka <hubicka@ucw.cz>
12777 Martin Liska <mliska@suse.cz>
12779 PR middle-end/101949
12780 * ipa-modref.c (analyze_ssa_name_flags): Indirect call implies
12783 2021-08-21 Dragan Mladjenovic <OT_Dragan.Mladjenovic@mediatek.com>
12785 * config/mips/mips.c (mips_function_rodata_section,
12786 TARGET_ASM_FUNCTION_RODATA_SECTION): Removed.
12788 2021-08-21 John David Anglin <danglin@gcc.gnu.org>
12790 * config/pa/pa.c (pa_asm_output_aligned_common): Remove warning.
12792 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
12794 * configure.ac (thread-local storage support): Remove tls_first_major
12795 and tls_first_minor. Use "$conftest_s" to check support.
12796 * configure: Regenerate.
12798 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
12800 * configure.ac: Fixup formatting.
12802 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
12804 * acinclude.m4 (gcc_GAS_CHECK_FEATURE): Remove third argument and ...
12805 * configure.ac: ... update all callers.
12807 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
12810 * acinclude.m4 (_gcc_COMPUTE_GAS_VERSION, _gcc_GAS_VERSION_GTE_IFELSE)
12811 (gcc_GAS_VERSION_GTE_IFELSE): Remove.
12812 (gcc_GAS_CHECK_FEATURE): Do not handle in-tree case specially.
12813 * configure.ac: Remove gcc_cv_gas_major_version, gcc_cv_gas_minor_version.
12814 Remove remaining checks for in-tree assembler.
12815 * configure: Regenerate.
12817 2021-08-20 Jeff Law <jlaw@localhost.localdomain>
12819 * config/h8300/h8300.c (shift_alg_hi): Improve arithmetic shift right
12820 by 15 bits for H8/300H and H8/S. Improve logical shifts by 12
12822 (shift_alg_si): Improve arithmetic right shift by 28-30 bits for
12823 H8/300H. Improve arithmetic shift right by 15 bits for H8/S.
12824 Improve logical shifts by 27 bits for H8/S.
12825 (get_shift_alg): Corresponding changes.
12826 (h8300_option_override): Revert to loops for -Os when profitable.
12828 2021-08-20 Richard Biener <rguenther@suse.de>
12830 * tree-vect-data-refs.c (dr_group_sort_cmp): Do not compare
12832 (vect_analyze_data_ref_accesses): Likewise. Assign the BB
12833 index as group_id when dataref_groups were not computed.
12834 * tree-vect-slp.c (vect_slp_bbs): Bump current_group when
12835 we advace to the next BB.
12837 2021-08-20 Jakub Jelinek <jakub@redhat.com>
12839 * omp-builtins.def (BUILT_IN_GOMP_WARNING, BUILT_IN_GOMP_ERROR): New
12842 2021-08-20 Martin Liska <mliska@suse.cz>
12844 PR gcov-profile/89961
12845 * gcov.c (make_gcov_file_name): Rewrite using std::string.
12846 (mangle_name): Simplify, do not used the second argument.
12847 (strip_extention): New function.
12848 (get_md5sum): Likewise.
12849 (get_gcov_intermediate_filename): Handle properly -p and -x
12851 (output_gcov_file): Use string type.
12852 (generate_results): Likewise.
12853 (md5sum_to_hex): Remove.
12855 2021-08-20 Michael Meissner <meissner@linux.ibm.com>
12857 * config/rs6000/altivec.md (UNSPEC_XXEVAL): Move to vsx.md.
12858 (UNSPEC_XXSPLTIW): Move to vsx.md.
12859 (UNSPEC_XXSPLTID): Move to vsx.md.
12860 (UNSPEC_XXSPLTI32DX): Move to vsx.md.
12861 (UNSPEC_XXBLEND): Move to vsx.md.
12862 (UNSPEC_XXPERMX): Move to vsx.md.
12863 (VM3): Move to vsx.md.
12864 (VM3_char): Move to vsx.md.
12865 (xxspltiw_v4si): Move to vsx.md.
12866 (xxspltiw_v4sf): Move to vsx.md.
12867 (xxspltiw_v4sf_inst): Move to vsx.md.
12868 (xxspltidp_v2df): Move to vsx.md.
12869 (xxspltidp_v2df_inst): Move to vsx.md.
12870 (xxsplti32dx_v4si_inst): Move to vsx.md.
12871 (xxsplti32dx_v4sf): Move to vsx.md.
12872 (xxsplti32dx_v4sf_inst): Move to vsx.md.
12873 (xxblend_<mode>): Move to vsx.md.
12874 (xxpermx): Move to vsx.md.
12875 (xxpermx_inst): Move to vsx.md.
12876 * config/rs6000/vsx.md (UNSPEC_XXEVAL): Move from altivec.md.
12877 (UNSPEC_XXSPLTIW): Move from altivec.md.
12878 (UNSPEC_XXSPLTID): Move from altivec.md.
12879 (UNSPEC_XXSPLTI32DX): Move from altivec.md.
12880 (UNSPEC_XXBLEND): Move from altivec.md.
12881 (UNSPEC_XXPERMX): Move from altivec.md.
12882 (VM3): Move from altivec.md.
12883 (VM3_char): Move from altivec.md.
12884 (xxspltiw_v4si): Move from altivec.md.
12885 (xxspltiw_v4sf): Move from altivec.md.
12886 (xxspltiw_v4sf_inst): Move from altivec.md.
12887 (xxspltidp_v2df): Move from altivec.md.
12888 (xxspltidp_v2df_inst): Move from altivec.md.
12889 (xxsplti32dx_v4si_inst): Move from altivec.md.
12890 (xxsplti32dx_v4sf): Move from altivec.md.
12891 (xxsplti32dx_v4sf_inst): Move from altivec.md.
12892 (xxblend_<mode>): Move from altivec.md.
12893 (xxpermx): Move from altivec.md.
12894 (xxpermx_inst): Move from altivec.md.
12896 2021-08-19 Roger Sayle <roger@nextmovesoftware.com>
12898 * tree-vect-generic.c (expand_vector_operations_1): Use either
12899 gimplify_build1 or gimplify_build2 instead of gimple_build_assign
12900 when constructing scalar splat expressions.
12902 2021-08-19 Peter Bergner <bergner@linux.ibm.com>
12905 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Cast
12906 pointer to __vector_pair *.
12908 2021-08-19 Martin Sebor <msebor@redhat.com>
12910 * gimple-range.cc: Add comments.
12911 * gimple-range.h: Same.
12913 2021-08-19 Martin Sebor <msebor@redhat.com>
12915 PR middle-end/101984
12916 * gimple-ssa-warn-access.cc (pass_waccess::execute): Also call
12919 2021-08-19 Jeff Law <jlaw@localhost.localdomain>
12921 * config.gcc (h8300-*-elf*): Do not include dbxelf.h.
12922 (h8300-*-linux*, v850-*-rtems*, v850*-elf*): Likewise.
12923 * config/v850/v850.h (DEFAULT_GDB_EXTENSIONS): Remove.
12925 2021-08-19 Jakub Jelinek <jakub@redhat.com>
12927 PR middle-end/101950
12928 * optabs.c (expand_clrsb_using_clz): New function.
12929 (expand_unop): Use it as another clrsb expansion fallback.
12931 2021-08-19 liuhongt <hongtao.liu@intel.com>
12934 2021-07-28 liuhongt <hongtao.liu@intel.com>
12937 * config/i386/i386.h (processor_costs): Add new member
12939 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
12940 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
12941 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
12942 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
12943 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
12944 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
12945 generic_cost, core_cost): Initialize integer_to_sse same value
12947 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
12948 * config/i386/i386.c (ix86_builtin_vectorization_cost):
12949 Use integer_to_sse instead of sse_op to calculate the cost of
12952 2021-08-18 Iain Sandoe <iain@sandoe.co.uk>
12954 * config.gcc: Include rpath.opt for Darwin.
12955 * config/darwin.h (DRIVER_SELF_SPECS): Handle -rpath.
12957 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
12959 PR bootstrap/101959
12960 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
12963 2021-08-18 Jonathan Wright <jonathan.wright@arm.com>
12965 * config/aarch64/arm_neon.h (vld3_lane_f64): Use float RTL
12966 pattern and type cast.
12967 (vld4_lane_f32): Use float RTL pattern.
12968 (vld4q_lane_f64): Use float type cast.
12970 2021-08-18 Jan Hubicka <hubicka@ucw.cz>
12972 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Check also
12975 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
12977 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor): Extend.
12978 (test_map_of_type_with_ctor_and_dtor_expand): Add function.
12979 (hash_map_tests_c_tests): Call it.
12981 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
12983 * ggc.h (enum ggc_collect): New.
12984 (ggc_collect): Use it.
12985 * ggc-page.c: Adjust.
12986 * ggc-common.c: Likewise.
12987 * ggc-tests.c: Likewise.
12988 * read-rtl-function.c: Likewise.
12989 * selftest-run-tests.c: Likewise.
12990 * doc/gty.texi (Invoking the garbage collector): Likewise.
12992 2021-08-18 liuhongt <hongtao.liu@intel.com>
12995 * config/i386/i386.h (TARGET_V2DF_REDUCTION_PREFER_HADDPD):
12997 * config/i386/sse.md (*sse3_haddv2df3_low): Add
12998 TARGET_V2DF_REDUCTION_PREFER_HADDPD.
12999 (*sse3_hsubv2df3_low): Ditto.
13000 * config/i386/x86-tune.def
13001 (X86_TUNE_V2DF_REDUCTION_PREFER_HADDPD): New tune.
13003 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
13005 * gimple-range-gori.cc (gori_compute::gori_compute): Enable tracing.
13006 (gori_compute::compute_operand_range): Add tracing.
13007 (gori_compute::logical_combine): Ditto.
13008 (gori_compute::compute_logical_operands): Ditto.
13009 (gori_compute::compute_operand1_range): Ditto.
13010 (gori_compute::compute_operand2_range): Ditto.
13011 (gori_compute::outgoing_edge_range_p): Ditto.
13012 * gimple-range-gori.h (class gori_compute): Add range_tracer.
13014 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
13016 * flag-types.h (enum evrp_mode): Adjust evrp-mode values.
13017 * gimple-range-cache.cc (DEBUG_RANGE_CACHE): Relocate from.
13018 * gimple-range-trace.h (DEBUG_RANGE_CACHE): Here.
13019 * params.opt (--param=evrp-mode): Adjust options.
13021 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
13023 * Makefile.in (OBJS): Add gimple-range-trace.o.
13024 * gimple-range-cache.h (enable_new_values): Remove unused prototype.
13025 * gimple-range-fold.cc: Adjust headers.
13026 * gimple-range-trace.cc: New.
13027 * gimple-range-trace.h: New.
13028 * gimple-range.cc (gimple_ranger::gimple_ranger): Enable tracer.
13029 (gimple_ranger::range_of_expr): Add tracing.
13030 (gimple_ranger::range_on_entry): Ditto.
13031 (gimple_ranger::range_on_exit): Ditto.
13032 (gimple_ranger::range_on_edge): Ditto.
13033 (gimple_ranger::fold_range_internal): Ditto.
13034 (gimple_ranger::dump_bb): Do not calculate edge range twice.
13035 (trace_ranger::*): Remove.
13036 (enable_ranger): Never create a trace_ranger.
13037 (debug_seed_ranger): Move to gimple-range-trace.cc.
13038 (dump_ranger): Ditto.
13039 (debug_ranger): Ditto.
13040 * gimple-range.h: Include gimple-range-trace.h.
13041 (range_on_entry, range_on_exit): No longer virtual.
13042 (class trace_ranger): Remove.
13043 (DEBUG_RANGE_CACHE): Move to gimple-range-trace.h.
13045 2021-08-17 Martin Sebor <msebor@redhat.com>
13047 PR middle-end/101854
13048 * builtins.c (expand_builtin_alloca): Move warning code to check_alloca
13049 in gimple-ssa-warn-access.cc.
13050 * calls.c (alloc_max_size): Move code to check_alloca.
13051 (get_size_range): Move to pointer-query.cc.
13052 (maybe_warn_alloc_args_overflow): Move to gimple-ssa-warn-access.cc.
13053 (get_attr_nonstring_decl): Move to tree.c.
13054 (fntype_argno_type): Move to gimple-ssa-warn-access.cc.
13055 (append_attrname): Same.
13056 (maybe_warn_rdwr_sizes): Same.
13057 (initialize_argument_information): Move code to
13058 gimple-ssa-warn-access.cc.
13059 * calls.h (maybe_warn_alloc_args_overflow): Move to
13060 gimple-ssa-warn-access.h.
13061 (get_attr_nonstring_decl): Move to tree.h.
13062 (maybe_warn_nonstring_arg): Move to gimple-ssa-warn-access.h.
13063 (enum size_range_flags): Move to pointer-query.h.
13064 (get_size_range): Same.
13065 * gimple-ssa-warn-access.cc (has_location): Remove unused overload
13066 to avoid Clang -Wunused-function.
13067 (get_size_range): Declare static.
13068 (maybe_emit_free_warning): Rename...
13069 (maybe_check_dealloc_call): ...to this for consistency.
13070 (class pass_waccess): Add members.
13071 (pass_waccess::~pass_waccess): Defined.
13072 (alloc_max_size): Move here from calls.c.
13073 (maybe_warn_alloc_args_overflow): Same.
13074 (check_alloca): New function.
13075 (check_alloc_size_call): New function.
13076 (check_strncat): Handle another warning flag.
13077 (pass_waccess::check_builtin): Handle alloca.
13078 (fntype_argno_type): Move here from calls.c.
13079 (append_attrname): Same.
13080 (maybe_warn_rdwr_sizes): Same.
13081 (pass_waccess::check_call): Define.
13082 (check_nonstring_args): New function.
13083 (pass_waccess::check): Call new member functions.
13084 (pass_waccess::execute): Enable ranger.
13085 * gimple-ssa-warn-access.h (get_size_range): Move here from calls.h.
13086 (maybe_warn_nonstring_arg): Same.
13087 * gimple-ssa-warn-restrict.c: Remove #include.
13088 * pointer-query.cc (get_size_range): Move here from calls.c.
13089 * pointer-query.h (enum size_range_flags): Same.
13090 (get_size_range): Same.
13091 * tree.c (get_attr_nonstring_decl): Move here from calls.c.
13092 * tree.h (get_attr_nonstring_decl): Move here from calls.h.
13094 2021-08-17 Thomas Schwinge <thomas@codesourcery.com>
13096 * ggc.h (ggc_collect): Add 'force_collect' parameter.
13097 * ggc-page.c (ggc_collect): Use that one instead of global
13098 'ggc_force_collect'. Adjust all users.
13099 * doc/gty.texi (Invoking the garbage collector): Update.
13100 * ggc-internal.h (ggc_force_collect): Remove.
13101 * ggc-common.c (ggc_force_collect): Likewise.
13102 * selftest.h (forcibly_ggc_collect): Remove.
13103 * ggc-tests.c (selftest::forcibly_ggc_collect): Likewise.
13104 * read-rtl-function.c (test_loading_labels): Adjust.
13105 * selftest-run-tests.c (run_tests): Likewise.
13107 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
13109 * config/darwin.c (darwin_file_end): Reset and reclaim the
13110 section names table at the end of compile.
13112 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
13115 * config.in: Regenerate.
13116 * config/i386/darwin.h (EXTRA_ASM_OPTS): New
13117 (ASM_SPEC): Pass options to disable branch shortening where
13119 * configure: Regenerate.
13120 * configure.ac: Detect versions of 'as' that support the
13121 optimisation which has the bug.
13123 2021-08-17 Richard Biener <rguenther@suse.de>
13125 * optabs-query.c (supports_vec_gather_load_p): Also check
13127 (supports_vec_scatter_store_p): Likewise.
13128 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Fall
13129 back to masked variants if non-masked are not supported.
13130 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
13131 When we need to use masked gather/scatter but do not have
13132 a mask set up a constant true one.
13133 * tree-vect-stmts.c (vect_check_scalar_mask): Also allow
13134 non-SSA_NAME masks.
13136 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
13138 * tree-ssa-ccp.c (bit_value_binop) [MINUS_EXPR]: Use same
13139 algorithm as PLUS_EXPR to improve subtraction bit bounds.
13140 [POINTER_DIFF_EXPR]: Treat as synonymous with MINUS_EXPR.
13142 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
13144 * tree-ssa-ccp.c (bit_value_mult_const): New helper function to
13145 calculate the mask-value pair result of a multiplication by an
13147 (bit_value_binop) [MULT_EXPR]: Call it from here for
13148 multiplications by (sparse) non-negative constants.
13150 2021-08-17 Christophe Lyon <christophe.lyon@foss.st.com>
13153 * config.gcc (gcc_cv_initfini_array): Leave undefined for
13154 uclinuxfdpiceabi targets.
13156 2021-08-17 Alexandre Oliva <oliva@adacore.com>
13158 * tree-inline.c (maybe_move_debug_stmts_to_successors): Don't
13159 reverse debug stmts.
13161 2021-08-17 Alexandre Oliva <oliva@adacore.com>
13163 * tree-cfg.c (dump_function_to_file): Use fun, not cfun.
13165 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
13167 * config/aarch64/arm_neon.h (__LD4_LANE_FUNC): Delete.
13168 (__LD4Q_LANE_FUNC): Likewise.
13169 (vld4_lane_u8): Define without macro.
13170 (vld4_lane_u16): Likewise.
13171 (vld4_lane_u32): Likewise.
13172 (vld4_lane_u64): Likewise.
13173 (vld4_lane_s8): Likewise.
13174 (vld4_lane_s16): Likewise.
13175 (vld4_lane_s32): Likewise.
13176 (vld4_lane_s64): Likewise.
13177 (vld4_lane_f16): Likewise.
13178 (vld4_lane_f32): Likewise.
13179 (vld4_lane_f64): Likewise.
13180 (vld4_lane_p8): Likewise.
13181 (vld4_lane_p16): Likewise.
13182 (vld4_lane_p64): Likewise.
13183 (vld4q_lane_u8): Likewise.
13184 (vld4q_lane_u16): Likewise.
13185 (vld4q_lane_u32): Likewise.
13186 (vld4q_lane_u64): Likewise.
13187 (vld4q_lane_s8): Likewise.
13188 (vld4q_lane_s16): Likewise.
13189 (vld4q_lane_s32): Likewise.
13190 (vld4q_lane_s64): Likewise.
13191 (vld4q_lane_f16): Likewise.
13192 (vld4q_lane_f32): Likewise.
13193 (vld4q_lane_f64): Likewise.
13194 (vld4q_lane_p8): Likewise.
13195 (vld4q_lane_p16): Likewise.
13196 (vld4q_lane_p64): Likewise.
13197 (vld4_lane_bf16): Likewise.
13198 (vld4q_lane_bf16): Likewise.
13200 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
13202 * config/aarch64/arm_neon.h (__LD3_LANE_FUNC): Delete.
13203 (__LD3Q_LANE_FUNC): Delete.
13204 (vld3_lane_u8): Define without macro.
13205 (vld3_lane_u16): Likewise.
13206 (vld3_lane_u32): Likewise.
13207 (vld3_lane_u64): Likewise.
13208 (vld3_lane_s8): Likewise.
13209 (vld3_lane_s16): Likewise.
13210 (vld3_lane_s32): Likewise.
13211 (vld3_lane_s64): Likewise.
13212 (vld3_lane_f16): Likewise.
13213 (vld3_lane_f32): Likewise.
13214 (vld3_lane_f64): Likewise.
13215 (vld3_lane_p8): Likewise.
13216 (vld3_lane_p16): Likewise.
13217 (vld3_lane_p64): Likewise.
13218 (vld3q_lane_u8): Likewise.
13219 (vld3q_lane_u16): Likewise.
13220 (vld3q_lane_u32): Likewise.
13221 (vld3q_lane_u64): Likewise.
13222 (vld3q_lane_s8): Likewise.
13223 (vld3q_lane_s16): Likewise.
13224 (vld3q_lane_s32): Likewise.
13225 (vld3q_lane_s64): Likewise.
13226 (vld3q_lane_f16): Likewise.
13227 (vld3q_lane_f32): Likewise.
13228 (vld3q_lane_f64): Likewise.
13229 (vld3q_lane_p8): Likewise.
13230 (vld3q_lane_p16): Likewise.
13231 (vld3q_lane_p64): Likewise.
13232 (vld3_lane_bf16): Likewise.
13233 (vld3q_lane_bf16): Likewise.
13235 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
13237 * config/aarch64/arm_neon.h (__LD2_LANE_FUNC): Delete.
13238 (__LD2Q_LANE_FUNC): Likewise.
13239 (vld2_lane_u8): Define without macro.
13240 (vld2_lane_u16): Likewise.
13241 (vld2_lane_u32): Likewise.
13242 (vld2_lane_u64): Likewise.
13243 (vld2_lane_s8): Likewise.
13244 (vld2_lane_s16): Likewise.
13245 (vld2_lane_s32): Likewise.
13246 (vld2_lane_s64): Likewise.
13247 (vld2_lane_f16): Likewise.
13248 (vld2_lane_f32): Likewise.
13249 (vld2_lane_f64): Likewise.
13250 (vld2_lane_p8): Likewise.
13251 (vld2_lane_p16): Likewise.
13252 (vld2_lane_p64): Likewise.
13253 (vld2q_lane_u8): Likewise.
13254 (vld2q_lane_u16): Likewise.
13255 (vld2q_lane_u32): Likewise.
13256 (vld2q_lane_u64): Likewise.
13257 (vld2q_lane_s8): Likewise.
13258 (vld2q_lane_s16): Likewise.
13259 (vld2q_lane_s32): Likewise.
13260 (vld2q_lane_s64): Likewise.
13261 (vld2q_lane_f16): Likewise.
13262 (vld2q_lane_f32): Likewise.
13263 (vld2q_lane_f64): Likewise.
13264 (vld2q_lane_p8): Likewise.
13265 (vld2q_lane_p16): Likewise.
13266 (vld2q_lane_p64): Likewise.
13267 (vld2_lane_bf16): Likewise.
13268 (vld2q_lane_bf16): Likewise.
13270 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
13272 * haifa-sched.c (advance_one_cycle): Output more context-synchronization
13275 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
13277 * haifa-sched.c (enum rfs_decision, rfs_str): Add RFS_AUTOPREF.
13278 (rank_for_schedule): Use it.
13280 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
13282 PR rtl-optimization/91598
13283 * haifa-sched.c (autopref_rank_for_schedule): Prioritize "irrelevant"
13284 insns after memory reads and before memory writes.
13286 2021-08-17 Alistair_Lee <alistair.lee@arm.com>
13288 * rtl.h (CONST_VECTOR_P): New macro.
13289 * config/aarch64/aarch64.c (aarch64_get_sve_pred_bits): Use RTL
13290 code testing macros.
13291 (aarch64_ptrue_all_mode): Likewise.
13292 (aarch64_expand_mov_immediate): Likewise.
13293 (aarch64_const_vec_all_in_range_p): Likewise.
13294 (aarch64_rtx_costs): Likewise.
13295 (aarch64_legitimate_constant_p): Likewise.
13296 (aarch64_simd_valid_immediate): Likewise.
13297 (aarch64_simd_make_constant): Likewise.
13298 (aarch64_convert_mult_to_shift): Likewise.
13299 (aarch64_expand_sve_vec_perm): Likewise.
13300 (aarch64_vec_fpconst_pow_of_2): Likewise.
13302 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
13304 PR tree-optimization/101938
13305 * range-op.cc (operator_abs::op1_range): Special case
13306 -TYPE_MIN_VALUE for flag_wrapv.
13308 2021-08-17 Kewen Lin <linkw@linux.ibm.com>
13310 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Add the cost for
13313 2021-08-17 Jakub Jelinek <jakub@redhat.com>
13315 * tree.def (OMP_SCOPE): New tree code.
13316 * tree.h (OMP_SCOPE_BODY, OMP_SCOPE_CLAUSES): Define.
13317 * tree-nested.c (convert_nonlocal_reference_stmt,
13318 convert_local_reference_stmt, convert_gimple_call): Handle
13320 * tree-pretty-print.c (dump_generic_node): Handle OMP_SCOPE.
13321 * gimple.def (GIMPLE_OMP_SCOPE): New gimple code.
13322 * gimple.c (gimple_build_omp_scope): New function.
13323 (gimple_copy): Handle GIMPLE_OMP_SCOPE.
13324 * gimple.h (gimple_build_omp_scope): Declare.
13325 (gimple_has_substatements): Handle GIMPLE_OMP_SCOPE.
13326 (gimple_omp_scope_clauses, gimple_omp_scope_clauses_ptr,
13327 gimple_omp_scope_set_clauses): New inline functions.
13328 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_SCOPE.
13329 * gimple-pretty-print.c (dump_gimple_omp_scope): New function.
13330 (pp_gimple_stmt_1): Handle GIMPLE_OMP_SCOPE.
13331 * gimple-walk.c (walk_gimple_stmt): Likewise.
13332 * gimple-low.c (lower_stmt): Likewise.
13333 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
13334 (gimplify_scan_omp_clauses): For task reductions, handle OMP_SCOPE
13335 like ORT_WORKSHARE constructs. Adjust diagnostics for %<scope%>
13336 allowing task reductions. Reject inscan reductions on scope.
13337 (omp_find_stores_stmt): Handle GIMPLE_OMP_SCOPE.
13338 (gimplify_omp_workshare, gimplify_expr): Handle OMP_SCOPE.
13339 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_SCOPE.
13340 (estimate_num_insns): Likewise.
13341 * omp-low.c (build_outer_var_ref): Look through GIMPLE_OMP_SCOPE
13342 contexts if var isn't privatized there.
13343 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_SCOPE.
13344 (scan_omp_1_stmt): Likewise.
13345 (maybe_add_implicit_barrier_cancel): Look through outer
13347 (lower_omp_scope): New function.
13348 (lower_omp_task_reductions): Handle OMP_SCOPE.
13349 (lower_omp_1): Handle GIMPLE_OMP_SCOPE.
13350 (diagnose_sb_1, diagnose_sb_2): Likewise.
13351 * omp-expand.c (expand_omp_single): Support also GIMPLE_OMP_SCOPE.
13352 (expand_omp): Handle GIMPLE_OMP_SCOPE.
13353 (omp_make_gimple_edges): Likewise.
13354 * omp-builtins.def (BUILT_IN_GOMP_SCOPE_START): New built-in.
13356 2021-08-17 Richard Biener <rguenther@suse.de>
13358 PR tree-optimization/101925
13359 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Set
13360 reverse on COMPONENT_REF and ARRAY_REF according to
13361 what reverse_storage_order_for_component_p does.
13362 (vn_reference_eq): Compare reversed on reference ops.
13363 (reverse_storage_order_for_component_p): New overload.
13364 (vn_reference_lookup_3): Check reverse_storage_order_for_component_p
13365 on the reference looked up.
13367 2021-08-17 Jeff Law <jlaw@localhost.localdomain>
13369 * config/h8300/h8300.c (shift_alg_si): Avoid loops for most SImode
13370 shifts on the H8/S.
13371 (h8300_option_override): Use loops on H8/S more often when optimizing
13373 (get_shift_alg): Handle new "special" cases on H8/S. Simplify
13374 accordingly. Handle various arithmetic right shifts with special
13375 sequences that we couldn't handle before.
13377 2021-08-16 Jeff Law <jlaw@localhost.localdomain>
13379 * config.gcc (rl78-*-elf*): Do not include dbxelf.h.
13381 2021-08-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
13383 * config/sparc/rtemself.h (SPARC_GCOV_TYPE_SIZE): Define.
13384 * config/sparc/sparc.c (sparc_gcov_type_size): New.
13385 (TARGET_GCOV_TYPE_SIZE): Redefine if SPARC_GCOV_TYPE_SIZE is defined.
13386 * coverage.c (get_gcov_type): Use targetm.gcov_type_size().
13387 * doc/tm.texi (TARGET_GCOV_TYPE_SIZE): Add hook under "Misc".
13388 * doc/tm.texi.in: Regenerate.
13389 * target.def (gcov_type_size): New target hook.
13390 * targhooks.c (default_gcov_type_size): New.
13391 * targhooks.h (default_gcov_type_size): Declare.
13392 * tree-profile.c (gimple_gen_edge_profiler): Use precision of
13394 (gimple_gen_time_profiler): Likewise.
13396 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
13398 * dwarf2out.c (add_scalar_info): Deal with DW_AT_data_bit_offset.
13400 2021-08-16 Tobias Burnus <tobias@codesourcery.com>
13402 PR middle-end/101931
13403 * omp-low.c (omp_runtime_api_call): Update for routines
13404 added in the meanwhile.
13406 2021-08-16 Martin Liska <mliska@suse.cz>
13408 PR tree-optimization/100393
13409 * tree-switch-conversion.c (group_cluster::dump): Use
13410 get_comparison_count.
13411 (jump_table_cluster::find_jump_tables): Pre-compute number of
13412 comparisons and then decrement it. Cache also max_ratio.
13413 (jump_table_cluster::can_be_handled): Change signature.
13414 * tree-switch-conversion.h (get_comparison_count): New.
13416 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
13418 * dwarf2out.c (add_data_member_location_attribute): Use GNAT
13419 encodings only when -fgnat-encodings=all is specified.
13420 (add_bound_info): Likewise.
13421 (add_byte_size_attribute): Likewise.
13422 (gen_member_die): Likewise.
13424 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
13426 * omp-oacc-neuter-broadcast.cc
13427 (execute_omp_oacc_neuter_broadcast): Plug 'par' memory leak.
13429 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
13431 * omp-oacc-neuter-broadcast.cc
13432 (execute_omp_oacc_neuter_broadcast): Clarify memory management for
13435 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
13437 * omp-oacc-neuter-broadcast.cc (field_map): Move variable into...
13438 (execute_omp_oacc_neuter_broadcast): ... here.
13439 (install_var_field, build_receiver_ref, build_sender_ref): Take
13440 'field_map_t *' parameter. Adjust all users.
13441 (worker_single_copy, neuter_worker_single): Take a
13442 'record_field_map_t *' parameter. Adjust all users.
13444 2021-08-16 liuhongt <hongtao.liu@intel.com>
13447 * config/i386/i386.md (ldexp<mode>3): Force operands[1] to
13450 2021-08-16 Martin Liska <mliska@suse.cz>
13453 * multiple_target.c (create_dispatcher_calls): Make default
13454 function local only if it is a definition.
13456 2021-08-16 Martin Liska <mliska@suse.cz>
13459 * ipa-icf-gimple.c (func_checker::compare_ssa_name): Do not
13460 consider equal SSA_NAMEs when one is a param.
13462 2021-08-16 liuhongt <hongtao.liu@intel.com>
13465 * config/i386/i386-expand.c (ix86_expand_vec_perm_vpermt2):
13466 Support vpermi2b for V32QI/V16QImode.
13467 (ix86_extract_perm_from_pool_constant): New function.
13468 (ix86_expand_vec_one_operand_perm_avx512): Support
13469 vpermw/vpermb under TARGET_AVX512BW/TARGET_AVX512VBMI.
13470 (expand_vec_perm_1): Adjust comments for upper.
13471 * config/i386/i386-protos.h (ix86_extract_perm_from_pool_constant):
13473 * config/i386/predicates.md (permvar_truncate_operand): New predicate.
13474 (pshufb_truncv4siv4hi_operand): Ditto.
13475 (pshufb_truncv8hiv8qi_operand): Ditto.
13476 * config/i386/sse.md (*avx512bw_permvar_truncv16siv16hi_1):
13477 New pre_reload define_insn_and_split.
13478 (*avx512f_permvar_truncv8siv8hi_1): Ditto.
13479 (*avx512f_vpermvar_truncv8div8si_1): Ditto.
13480 (*avx512f_permvar_truncv32hiv32qi_1): Ditto.
13481 (*avx512f_permvar_truncv16hiv16qi_1): Ditto.
13482 (*avx512f_permvar_truncv4div4si_1): Ditto.
13483 (*avx512f_pshufb_truncv8hiv8qi_1): Ditto.
13484 (*avx512f_pshufb_truncv4siv4hi_1): Ditto.
13485 (*avx512f_pshufd_truncv2div2si_1): Ditto.
13487 2021-08-16 Kito Cheng <kito.cheng@sifive.com>
13489 * config/riscv/multilib-generator: Support code model option for
13491 * doc/install.texi: Add document of new option for
13492 --with-multilib-generator.
13494 2021-08-15 Clément Chigot <clement.chigot@atos.net>
13496 * config/rs6000/rs6000.c (xcoff_tls_exec_model_detected): New.
13497 (rs6000_legitimize_tls_address_aix): Use it.
13498 (rs6000_xcoff_file_end): Add ".ref __tls_get_addr" when
13499 xcoff_tls_exec_model_detected is true.
13501 2021-08-15 Jeff Law <jlaw@localhost.localdomain>
13503 * config/h8300/h8300.c (shift_alg_si): Retune H8/300H shifts
13504 to allow a bit more code growth, saving many dozens of cycles.
13505 (h8300_option_override): Adjus shift_alg_si if optimizing for
13507 (get_shift_alg): Use special + inline shifts for residuals
13510 2021-08-14 Stafford Horne <shorne@gmail.com>
13513 * config/or1k/or1k-opts.h: New file.
13514 * config/or1k/or1k.c (or1k_legitimize_address_1, print_reloc):
13515 Support generating gotha relocations if -mcmodel=large is
13517 * config/or1k/or1k.h (TARGET_CMODEL_SMALL, TARGET_CMODEL_LARGE):
13519 * config/or1k/or1k.opt (mcmodel=): New option.
13520 * doc/invoke.texi (OpenRISC Options): Document mcmodel.
13522 2021-08-14 Martin Sebor <msebor@redhat.com>
13524 PR middle-end/101791
13525 * gimple-ssa-warn-access.cc (new_delete_mismatch_p): Use new argument
13526 to valid_new_delete_pair_p.
13527 * tree.c (valid_new_delete_pair_p): Add argument.
13528 * tree.h (valid_new_delete_pair_p): Same.
13530 2021-08-14 Jakub Jelinek <jakub@redhat.com>
13533 * config/i386/i386-expand.c (expand_vec_perm_broadcast_1)
13534 <case E_V64QImode>: For this mode assert
13535 !TARGET_AVX512BW || d->perm[0] rather than !TARGET_AVX2 || d->perm[0].
13537 2021-08-13 Michael Meissner <meissner@linux.ibm.com>
13540 * config/rs6000/altivec.md (xxeval): Use register_predicate
13541 instead of altivec_register_predicate.
13543 2021-08-13 Martin Sebor <msebor@redhat.com>
13545 PR middle-end/101734
13546 * tree-ssa-uninit.c (maybe_warn_read_write_only): New function.
13547 (maybe_warn_operand): Call it.
13549 2021-08-13 Martin Liska <mliska@suse.cz>
13552 * attribs.c (decl_attributes): Make naked functions "noipa"
13555 2021-08-13 Martin Liska <mliska@suse.cz>
13558 * symtab.c (symtab_node::noninterposable_alias): Do not create
13559 local aliases for target_clone functions as the clonning pass
13562 2021-08-13 Martin Liska <mliska@suse.cz>
13564 * opts.c (LIVE_PATCHING_OPTION): Define.
13565 (control_options_for_live_patching): Use it in error messages.
13567 2021-08-13 Jan Hubicka <hubicka@ucw.cz>
13569 * ipa-modref.c (dump_eaf_flags): Dump EAF_NOREAD.
13570 (implicit_const_eaf_flags, implicit_pure_eaf_flags,
13571 ignore_stores_eaf_flags): New constants.
13572 (remove_useless_eaf_flags): New function.
13573 (eaf_flags_useful_p): Use it.
13574 (deref_flags): Add EAF_NOT_RETURNED if flag is unused;
13576 (modref_lattice::init): Add EAF_NOREAD.
13577 (modref_lattice::add_escape_point): Do not reacord escape point if
13579 (modref_lattice::merge): EAF_NOESCAPE implies EAF_NODIRECTESCAPE;
13580 use remove_useless_eaf_flags.
13581 (modref_lattice::merge_deref): Use ignore_stores_eaf_flags.
13582 (modref_lattice::merge_direct_load): Add EAF_NOREAD
13583 (analyze_ssa_name_flags): Fix handling EAF_NOT_RETURNED
13584 (analyze_parms): Use remove_useless_eaf_flags.
13585 (ipa_merge_modref_summary_after_inlining): Use ignore_stores_eaf_flags.
13586 (modref_merge_call_site_flags): Add caller and ecf_flags parameter;
13587 use remove_useless_eaf_flags.
13588 (modref_propagate_flags_in_scc): Update.
13589 * ipa-modref.h: Turn eaf_flags_t back to char.
13590 * tree-core.h (EAF_NOT_RETURNED): Fix.
13591 (EAF_NOREAD): New constant
13592 * tree-ssa-alias.c: (ref_maybe_used_by_call_p_1): Check for
13594 * tree-ssa-structalias.c (handle_rhs_call): Handle new flags.
13595 (handle_pure_call): Likewise.
13597 2021-08-12 Jakub Jelinek <jakub@redhat.com>
13599 * tree.def (OMP_MASKED): New tree code.
13600 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_FILTER.
13601 * tree.h (OMP_MASKED_BODY, OMP_MASKED_CLAUSES, OMP_MASKED_COMBINED,
13602 OMP_CLAUSE_FILTER_EXPR): Define.
13603 * tree.c (omp_clause_num_ops): Add OMP_CLAUSE_FILTER entry.
13604 (omp_clause_code_name): Likewise.
13605 (walk_tree_1): Handle OMP_CLAUSE_FILTER.
13606 * tree-nested.c (convert_nonlocal_omp_clauses,
13607 convert_local_omp_clauses): Handle OMP_CLAUSE_FILTER.
13608 (convert_nonlocal_reference_stmt, convert_local_reference_stmt,
13609 convert_gimple_call): Handle GIMPLE_OMP_MASTER.
13610 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_FILTER.
13611 (dump_generic_node): Handle OMP_MASTER.
13612 * gimple.def (GIMPLE_OMP_MASKED): New gimple code.
13613 * gimple.c (gimple_build_omp_masked): New function.
13614 (gimple_copy): Handle GIMPLE_OMP_MASKED.
13615 * gimple.h (gimple_build_omp_masked): Declare.
13616 (gimple_has_substatements): Handle GIMPLE_OMP_MASKED.
13617 (gimple_omp_masked_clauses, gimple_omp_masked_clauses_ptr,
13618 gimple_omp_masked_set_clauses): New inline functions.
13619 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_MASKED.
13620 * gimple-pretty-print.c (dump_gimple_omp_masked): New function.
13621 (pp_gimple_stmt_1): Handle GIMPLE_OMP_MASKED.
13622 * gimple-walk.c (walk_gimple_stmt): Likewise.
13623 * gimple-low.c (lower_stmt): Likewise.
13624 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
13625 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_FILTER. For clauses
13626 that take one expression rather than decl or constant, force
13627 gimplification of that into a SSA_NAME or temporary unless min
13629 (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_FILTER.
13630 (gimplify_expr): Handle OMP_MASKED.
13631 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_MASKED.
13632 (estimate_num_insns): Likewise.
13633 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_FILTER.
13634 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_MASKED. Adjust
13635 diagnostics for existence of masked construct.
13636 (scan_omp_1_stmt, lower_omp_master, lower_omp_1, diagnose_sb_1,
13637 diagnose_sb_2): Handle GIMPLE_OMP_MASKED.
13638 * omp-expand.c (expand_omp_synch, expand_omp, omp_make_gimple_edges):
13641 2021-08-12 Uroš Bizjak <ubizjak@gmail.com>
13644 * config/i386/i386.md (avx512f_scalef<mode>2): New insn pattern.
13645 (ldexp<mode>3): Use avx512f_scalef<mode>2.
13646 (UNSPEC_SCALEF): Move from ...
13647 * config/i386/sse.md (UNSPEC_SCALEF): ... here.
13649 2021-08-12 Jan Hubicka <hubicka@ucw.cz>
13651 * ipa-split.c (consider_split): Fix condition testing void functions.
13653 2021-08-12 Aldy Hernandez <aldyh@redhat.com>
13655 * doc/invoke.texi: Remove docs for threader-mode param.
13656 * flag-types.h (enum threader_mode): Remove.
13657 * params.opt: Remove threader-mode param.
13658 * tree-ssa-threadbackward.c (class back_threader): Remove
13659 path_is_unreachable_p.
13660 Make find_paths private.
13661 Add maybe_thread and thread_through_all_blocks.
13662 Remove reference marker for m_registry.
13663 Remove reference marker for m_profit.
13664 (back_threader::back_threader): Adjust for registry and profit not
13666 (dump_path): Move down.
13667 (debug): Move down.
13668 (class thread_jumps): Remove.
13669 (class back_threader_registry): Remove m_all_paths.
13671 (thread_jumps::thread_through_all_blocks): Move to back_threader
13673 (fsm_find_thread_path): Remove
13674 (back_threader::maybe_thread): New.
13675 (back_threader::thread_through_all_blocks): Move from
13677 (back_threader_registry::back_threader_registry): Remove
13679 (back_threader_registry::~back_threader_registry): Remove.
13680 (thread_jumps::find_taken_edge): Remove.
13681 (thread_jumps::check_subpath_and_update_thread_path): Remove.
13682 (thread_jumps::maybe_register_path): Remove.
13683 (thread_jumps::handle_phi): Remove.
13684 (handle_assignment_p): Remove.
13685 (thread_jumps::handle_assignment): Remove.
13686 (thread_jumps::fsm_find_control_statement_thread_paths): Remove.
13687 (thread_jumps::find_jump_threads_backwards): Remove.
13688 (thread_jumps::find_jump_threads_backwards_with_ranger): Remove.
13689 (try_thread_blocks): Rename find_jump_threads_backwards to
13691 (pass_early_thread_jumps::execute): Same.
13693 2021-08-12 Tobias Burnus <tobias@codesourcery.com>
13695 * tree-core.h (omp_clause_proc_bind_kind): Add
13696 OMP_CLAUSE_PROC_BIND_PRIMARY.
13697 * tree-pretty-print.c (dump_omp_clause): Add TODO comment to
13698 change 'master' to 'primary' in proc_bind for OpenMP 5.1.
13700 2021-08-12 Claudiu Zissulescu <claziss@synopsys.com>
13702 * common/config/arc/arc-common.c (arc_option_init_struct): Remove
13703 fno-common reference.
13704 * config/arc/arc.c (arc_override_options): Remove overriding of
13707 2021-08-12 Jakub Jelinek <jakub@redhat.com>
13710 * config/i386/i386-expand.c (ix86_expand_vec_one_operand_perm_avx512):
13711 If d->testing_p, return true after performing checks instead of
13712 actually expanding the insn.
13713 (expand_vec_perm_broadcast_1): Handle V32HImode - assert
13714 !TARGET_AVX512BW and return false.
13716 2021-08-12 Eric Botcazou <ebotcazou@gcc.gnu.org>
13718 * configure.ac (PE linker --disable-dynamicbase support): New check.
13719 * configure: Regenerate.
13720 * config.in: Likewise.
13721 * config/i386/mingw32.h (LINK_SPEC_DISABLE_DYNAMICBASE): New define.
13722 (LINK_SPEC): Use it.
13723 * config/i386/mingw-w64.h (LINK_SPEC_DISABLE_DYNAMICBASE): Likewise.
13724 (LINK_SPEC): Likewise.
13726 2021-08-12 liuhongt <hongtao.liu@intel.com>
13729 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_2): New
13730 post_reload define_insn_and_split.
13731 (*avx512bw_zero_extendv32qiv32hi2_2): Ditto.
13732 (*sse4_1_zero_extendv8qiv8hi2_4): Ditto.
13733 (*avx512f_zero_extendv16hiv16si2_2): Ditto.
13734 (*avx2_zero_extendv8hiv8si2_2): Ditto.
13735 (*sse4_1_zero_extendv4hiv4si2_4): Ditto.
13736 (*avx512f_zero_extendv8siv8di2_2): Ditto.
13737 (*avx2_zero_extendv4siv4di2_2): Ditto.
13738 (*sse4_1_zero_extendv2siv2di2_4): Ditto.
13739 (VI248_256, VI248_512, VI148_512, VI148_256, VI148_128): New
13742 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
13744 * config/rs6000/rs6000-builtin-new.def: Add always, power5, and
13747 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
13749 * config/rs6000/rs6000-builtin-new.def: Add vsx stanza.
13751 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
13753 * config/rs6000/rs6000-builtin-new.def: Finish altivec stanza.
13754 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Move
13755 initialization of pcvoid_type_node here...
13756 (altivec_init_builtins): ...from here.
13757 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add
13758 RS6000_BTI_const_ptr_void.
13759 (pcvoid_type_node): New macro.
13761 2021-08-11 Richard Biener <rguenther@suse.de>
13764 * tree-ssa-forwprop.c (pass_forwprop::execute): Do not decompose
13765 hard-register accesses.
13767 2021-08-11 Richard Biener <rguenther@suse.de>
13769 * tree-ssa-operands.c (operands_scanner::get_expr_operands):
13770 Do not look at COMPONENT_REF FIELD_DECLs TREE_THIS_VOLATILE
13771 to determine has_volatile_ops.
13773 2021-08-11 Eric Botcazou <ebotcazou@gcc.gnu.org>
13775 * cfgexpand.c (expand_used_vars): Reuse attribs local variable.
13777 2021-08-11 Jan Hubicka <hubicka@ucw.cz>
13778 Alexandre Oliva <oliva@adacore.com>
13780 * ipa-modref.c (modref_lattice::dump): Fix escape_point's min_flags
13782 (modref_lattice::merge_deref): Fix handling of indirect scape points.
13783 (update_escape_summary_1): Likewise.
13784 (update_escape_summary): Likewise.
13785 (ipa_merge_modref_summary_after_inlining): Likewise.
13787 2021-08-11 Richard Biener <rguenther@suse.de>
13789 PR middle-end/101858
13790 * fold-const.c (fold_binary_loc): Guard simplification
13791 of X < (cast) (1 << Y) to integer types.
13793 2021-08-11 Richard Biener <rguenther@suse.de>
13795 PR tree-optimization/101861
13796 * tree-vect-stmts.c (vectorizable_load): Fix error in
13797 previous change with regard to gather vectorization.
13799 2021-08-11 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
13802 * config/arm/arm_neon.h (vdup_n_s8): Replace call to builtin
13804 (vdup_n_s16): Likewise.
13805 (vdup_n_s32): Likewise.
13806 (vdup_n_s64): Likewise.
13807 (vdup_n_u8): Likewise.
13808 (vdup_n_u16): Likewise.
13809 (vdup_n_u32): Likewise.
13810 (vdup_n_u64): Likewise.
13811 (vdup_n_p8): Likewise.
13812 (vdup_n_p16): Likewise.
13813 (vdup_n_p64): Likewise.
13814 (vdup_n_f16): Likewise.
13815 (vdup_n_f32): Likewise.
13816 (vdupq_n_s8): Likewise.
13817 (vdupq_n_s16): Likewise.
13818 (vdupq_n_s32): Likewise.
13819 (vdupq_n_s64): Likewise.
13820 (vdupq_n_u8): Likewise.
13821 (vdupq_n_u16): Likewise.
13822 (vdupq_n_u32): Likewise.
13823 (vdupq_n_u64): Likewise.
13824 (vdupq_n_p8): Likewise.
13825 (vdupq_n_p16): Likewise.
13826 (vdupq_n_p64): Likewise.
13827 (vdupq_n_f16): Likewise.
13828 (vdupq_n_f32): Likewise.
13829 (vmov_n_s8): Replace call to builtin with call to corresponding
13831 (vmov_n_s16): Likewise.
13832 (vmov_n_s32): Likewise.
13833 (vmov_n_s64): Likewise.
13834 (vmov_n_u8): Likewise.
13835 (vmov_n_u16): Likewise.
13836 (vmov_n_u32): Likewise.
13837 (vmov_n_u64): Likewise.
13838 (vmov_n_p8): Likewise.
13839 (vmov_n_p16): Likewise.
13840 (vmov_n_f16): Likewise.
13841 (vmov_n_f32): Likewise.
13842 (vmovq_n_s8): Likewise.
13843 (vmovq_n_s16): Likewise.
13844 (vmovq_n_s32): Likewise.
13845 (vmovq_n_s64): Likewise.
13846 (vmovq_n_u8): Likewise.
13847 (vmovq_n_u16): Likewise.
13848 (vmovq_n_u32): Likewise.
13849 (vmovq_n_u64): Likewise.
13850 (vmovq_n_p8): Likewise.
13851 (vmovq_n_p16): Likewise.
13852 (vmovq_n_f16): Likewise.
13853 (vmovq_n_f32): Likewise.
13854 * config/arm/arm_neon_builtins.def: Remove entries for vdup_n.
13856 2021-08-11 liuhongt <hongtao.liu@intel.com>
13859 * config/i386/i386.md (ldexp<mode>3): Extend to vscalefs[sd]
13860 when TARGET_AVX512F and TARGET_SSE_MATH.
13862 2021-08-10 Jakub Jelinek <jakub@redhat.com>
13865 * config/i386/i386-expand.c (expand_vec_perm_even_odd): Return false
13866 for V32HImode if !TARGET_AVX512BW.
13867 (ix86_vectorize_vec_perm_const) <case E_V32HImode, case E_V64QImode>:
13868 If !TARGET_AVX512BW and TARGET_AVX512F and d.testing_p, don't fail
13869 early, but actually check the permutation.
13871 2021-08-10 Richard Biener <rguenther@suse.de>
13873 PR tree-optimization/101809
13874 * tree-vect-stmts.c (get_load_store_type): Allow emulated
13875 gathers with offset vector nunits being a constant multiple
13876 of the data vector nunits.
13877 (vect_get_gather_scatter_ops): Use the appropriate nunits
13878 for the offset vector defs.
13879 (vectorizable_store): Adjust call to
13880 vect_get_gather_scatter_ops.
13881 (vectorizable_load): Likewise. Handle the case of less
13882 offset vectors than data vectors.
13884 2021-08-10 Jakub Jelinek <jakub@redhat.com>
13887 * config/i386/sse.md (*avx512f_shuf_<shuffletype>64x2_1<mask_name>_1,
13888 *avx512f_shuf_<shuffletype>32x4_1<mask_name>_1): New define_insn
13891 2021-08-10 Richard Biener <rguenther@suse.de>
13893 PR tree-optimization/101801
13894 PR tree-optimization/101819
13895 * tree-vectorizer.h (vect_emulated_vector_p): Declare.
13896 * tree-vect-loop.c (vect_emulated_vector_p): New function.
13897 (vectorizable_reduction): Re-instantiate a check for emulated
13899 * tree-vect-stmts.c (vectorizable_shift): Likewise.
13900 (vectorizable_operation): Likewise. Cost emulated vector
13901 operations according to the scalar sequence synthesized by
13904 2021-08-10 Richard Biener <rguenther@suse.de>
13906 PR middle-end/101824
13907 * tree-nested.c (get_frame_field): Mark the COMPONENT_REF as
13908 volatile in case the variable was.
13910 2021-08-10 H.J. Lu <hjl.tools@gmail.com>
13913 * config/i386/constraints.md (BC): Document for integer SSE
13914 constant all bits set operand.
13915 (BF): New constraint for const floating-point all bits set
13917 * config/i386/i386.c (standard_sse_constant_p): Likewise.
13918 (standard_sse_constant_opcode): Likewise.
13919 * config/i386/sse.md (sseconstm1): New mode attribute.
13920 (mov<mode>_internal): Replace BC with <sseconstm1>.
13922 2021-08-10 liuhongt <hongtao.liu@intel.com>
13924 * config/i386/sse.md (cond_<insn><mode>): New expander.
13925 (VI248_AVX512VLBW): New mode iterator.
13926 * config/i386/predicates.md
13927 (nonimmediate_or_const_vec_dup_operand): New predicate.
13929 2021-08-09 Andrew MacLeod <amacleod@redhat.com>
13931 PR tree-optimization/101741
13932 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Check
13933 type of parameter for toupper/tolower.
13935 2021-08-09 Martin Jambor <mjambor@suse.cz>
13937 PR testsuite/101654
13938 * ipa-prop.c (propagate_controlled_uses): Removed a spurious space.
13940 2021-08-09 Pat Haugen <pthaugen@linux.ibm.com>
13942 * config/rs6000/rs6000.c (is_load_insn1): Verify destination is a
13944 (is_store_insn1): Verify source is a register.
13946 2021-08-09 Uroš Bizjak <ubizjak@gmail.com>
13949 * config/i386/mmx.md (<any_logic:code>v2sf3):
13950 Rename from *mmx_<any_logic:code>v2sf3
13952 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
13954 * config/nvptx/nvptx.c: Cross-reference parts adapted in
13955 'gcc/omp-oacc-neuter-broadcast.cc'.
13956 * omp-low.c: Likewise.
13957 * omp-oacc-neuter-broadcast.cc: Cross-reference parts adapted from
13960 2021-08-09 Julian Brown <julian@codesourcery.com>
13961 Kwok Cheung Yeung <kcy@codesourcery.com>
13962 Thomas Schwinge <thomas@codesourcery.com>
13964 * config/gcn/gcn.c (gcn_init_builtins): Override decls for
13965 BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START,
13966 BUILT_IN_GOACC_SINGLE_COPY_END and BUILT_IN_GOACC_BARRIER.
13967 (gcn_goacc_validate_dims): Turn on worker partitioning unconditionally.
13968 (gcn_fork_join): Update comment.
13969 * config/gcn/gcn.opt (flag_worker_partitioning): Remove.
13970 (macc_experimental_workers): Remove unused option.
13972 2021-08-09 Julian Brown <julian@codesourcery.com>
13973 Nathan Sidwell <nathan@codesourcery.com> (via 'gcc/config/nvptx/nvptx.c' master)
13974 Kwok Cheung Yeung <kcy@codesourcery.com>
13975 Thomas Schwinge <thomas@codesourcery.com>
13977 * Makefile.in (OBJS): Add omp-oacc-neuter-broadcast.o.
13978 * doc/tm.texi.in (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD):
13979 Add documentation hook.
13980 * doc/tm.texi: Regenerate.
13981 * omp-oacc-neuter-broadcast.cc: New file.
13982 * omp-builtins.def (BUILT_IN_GOACC_BARRIER)
13983 (BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START)
13984 (BUILT_IN_GOACC_SINGLE_COPY_END): New builtins.
13985 * passes.def (pass_omp_oacc_neuter_broadcast): Add pass.
13986 * target.def (goacc.create_worker_broadcast_record): Add target
13988 * tree-pass.h (make_pass_omp_oacc_neuter_broadcast): Add
13990 * config/gcn/gcn-protos.h (gcn_goacc_adjust_propagation_record):
13991 Rename prototype to...
13992 (gcn_goacc_create_worker_broadcast_record): ... this.
13993 * config/gcn/gcn-tree.c (gcn_goacc_adjust_propagation_record): Rename
13995 (gcn_goacc_create_worker_broadcast_record): ... this.
13996 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_PROPAGATION_RECORD):
13998 (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD): ... this.
14000 2021-08-09 Tejas Belagod <tejas.belagod@arm.com>
14003 * config/aarch64/aarch64-simd.md (vlshr<mode>3, vashr<mode>3): Use
14004 the right iterator.
14006 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
14008 * Makefile.in (GTFILES): Remove '$(srcdir)/omp-offload.c'.
14010 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
14012 * builtins.def (DEF_GOACC_BUILTIN, DEF_GOMP_BUILTIN): Don't
14013 consider '-foffload-abi'.
14014 * common.opt (-foffload-abi): Remove 'Var', 'Init'.
14015 * opts.c (common_handle_option) <-foffload-abi> [ACCEL_COMPILER]:
14018 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
14020 * optc-gen.awk: Sanity check that 'Init' doesn't appear without
14023 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
14025 * omp-builtins.def (BUILT_IN_ACC_GET_DEVICE_TYPE): Remove.
14027 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
14029 * doc/gty.texi (Files): Update.
14031 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
14033 * doc/gty.texi (Files): Fix GTY header file example.
14035 2021-08-09 Roger Sayle <roger@nextmovesoftware.com>
14037 * tree-ssa-ccp.c (value_mask_to_min_max): Helper function to
14038 determine the upper and lower bounds from a mask-value pair.
14039 (bit_value_unop) [ABS_EXPR, ABSU_EXPR]: Add support for
14040 absolute value and unsigned absolute value expressions.
14041 (bit_value_binop): Initialize *VAL's precision.
14042 [LT_EXPR, LE_EXPR]: Use value_mask_to_min_max to determine
14043 upper and lower bounds of operands. Add LE_EXPR/GE_EXPR
14044 support when the operands are unknown but potentially equal.
14045 [MIN_EXPR, MAX_EXPR]: Support minimum/maximum expressions.
14047 2021-08-09 Bin Cheng <bin.cheng@linux.alibaba.com>
14049 * config/aarch64/aarch64.md
14050 (*extend<SHORT:mode><GPI:mode>2_aarch64): Use %<GPI:w>0.
14052 2021-08-08 Sergei Trofimovich <siarheit@google.com>
14054 * lra-constraints.c: Fix s/otput/output/ typo.
14056 2021-08-06 Martin Sebor <msebor@redhat.com>
14058 * builtins.c (expand_builtin_memchr): Move to gimple-ssa-warn-access.cc.
14059 (expand_builtin_strcat): Same.
14060 (expand_builtin_stpncpy): Same.
14061 (expand_builtin_strncat): Same.
14062 (check_read_access): Same.
14063 (check_memop_access): Same.
14064 (expand_builtin_strlen): Move checks to gimple-ssa-warn-access.cc.
14065 (expand_builtin_strnlen): Same.
14066 (expand_builtin_memcpy): Same.
14067 (expand_builtin_memmove): Same.
14068 (expand_builtin_mempcpy): Same.
14069 (expand_builtin_strcpy): Same.
14070 (expand_builtin_strcpy_args): Same.
14071 (expand_builtin_stpcpy_1): Same.
14072 (expand_builtin_strncpy): Same.
14073 (expand_builtin_memset): Same.
14074 (expand_builtin_bzero): Same.
14075 (expand_builtin_strcmp): Same.
14076 (expand_builtin_strncmp): Same.
14077 (expand_builtin): Remove handlers.
14078 (fold_builtin_strlen): Add a comment.
14079 * builtins.h (check_access): Move to gimple-ssa-warn-access.cc.
14080 * calls.c (maybe_warn_nonstring_arg): Same.
14081 * diagnostic-spec.c (nowarn_spec_t::nowarn_spec_t): Add warning option.
14082 * gimple-fold.c (gimple_fold_builtin_strcpy): Pass argument to callee.
14083 (gimple_fold_builtin_stpcpy): Same.
14084 * gimple-ssa-warn-access.cc (has_location): New function.
14085 (get_location): Same.
14086 (get_callee_fndecl): Same.
14087 (call_nargs): Same.
14089 (warn_string_no_nul): Define.
14090 (unterminated_array): Same.
14091 (check_nul_terminated_array): Same.
14092 (maybe_warn_nonstring_arg): Same.
14093 (maybe_warn_for_bound): Same.
14094 (warn_for_access): Same.
14095 (check_access): Same.
14096 (check_memop_access): Same.
14097 (check_read_access): Same.
14098 (warn_dealloc_offset): Use helper functions.
14099 (maybe_emit_free_warning): Same.
14100 (class pass_waccess): Add members.
14101 (check_strcat): New function.
14102 (check_strncat): New function.
14103 (check_stxcpy): New function.
14104 (check_stxncpy): New function.
14105 (check_strncmp): New function.
14106 (pass_waccess::check_builtin): New function.
14107 (pass_waccess::check): Call it.
14108 * gimple-ssa-warn-access.h (warn_string_no_nul): Move here from
14110 (maybe_warn_for_bound): Same.
14111 (check_access): Same.
14112 (check_memop_access): Same.
14113 (check_read_access): Same.
14114 * pointer-query.h (struct access_data): Define a ctor overload.
14116 2021-08-06 Richard Biener <rguenther@suse.de>
14118 PR tree-optimization/101801
14119 * tree-vectorizer.h (vect_worthwhile_without_simd_p): Rename...
14120 (vect_can_vectorize_without_simd_p): ... to this.
14121 * tree-vect-loop.c (vect_worthwhile_without_simd_p): Rename...
14122 (vect_can_vectorize_without_simd_p): ... to this and fold
14123 in vect_min_worthwhile_factor.
14124 (vect_min_worthwhile_factor): Remove.
14125 (vectorizable_reduction): Adjust and remove the cost part.
14126 * tree-vect-stmts.c (vectorizable_shift): Likewise.
14127 (vectorizable_operation): Likewise.
14129 2021-08-06 Uroš Bizjak <ubizjak@gmail.com>
14132 * config/i386/i386.md (cmove reg-to-reg move elimination peephole2s):
14133 Add general_gr_operand predicate to operand 3.
14135 2021-08-06 Roger Sayle <roger@nextmovesoftware.com>
14137 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern): Use
14138 CFN_BUILT_IN_CLRSB* instead of BUILT_IN_CLRSB* for consistency.
14140 2021-08-06 Tamar Christina <tamar.christina@arm.com>
14142 * config/aarch64/aarch64-sve-builtins.cc (register_svpattern,
14143 register_svprfop): Pass vec<> by pointer.
14144 * langhooks-def.h (lhd_simulate_enum_decl): Likewise.
14145 * langhooks.c (lhd_simulate_enum_decl): Likewise.
14146 * langhooks.h (struct lang_hooks_for_types): Likewise.
14148 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
14150 * config/aarch64/arm_neon.h (vst1_bf16_x2): Use
14151 __builtin_memcpy instead of constructing an additional
14152 __builtin_aarch64_simd_oi one vector at a time.
14153 (vst1q_bf16_x2): Likewise.
14154 (vst1_bf16_x3): Use __builtin_memcpy instead of constructing
14155 an additional __builtin_aarch64_simd_ci one vector at a time.
14156 (vst1q_bf16_x3): Likewise.
14157 (vst1_bf16_x4): Use __builtin_memcpy instead of a union.
14158 (vst1q_bf16_x4): Likewise.
14159 (vst2_bf16): Use __builtin_memcpy instead of constructing an
14160 additional __builtin_aarch64_simd_oi one vector at a time.
14161 (vst2q_bf16): Likewise.
14162 (vst3_bf16): Use __builtin_memcpy instead of constructing an
14163 additional __builtin_aarch64_simd_ci mode one vector at a
14165 (vst3q_bf16): Likewise.
14166 (vst4_bf16): Use __builtin_memcpy instead of constructing an
14167 additional __builtin_aarch64_simd_xi one vector at a time.
14168 (vst4q_bf16): Likewise.
14170 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
14172 * config/aarch64/arm_neon.h (__ST2_LANE_FUNC): Delete.
14173 (__ST2Q_LANE_FUNC): Delete.
14174 (vst2_lane_f16): Use __builtin_memcpy to copy vector
14175 structure instead of constructing __builtin_aarch64_simd_oi
14176 one vector at a time.
14177 (vst2_lane_f32): Likewise.
14178 (vst2_lane_f64): Likewise.
14179 (vst2_lane_p8): Likewise.
14180 (vst2_lane_p16): Likewise.
14181 (vst2_lane_p64): Likewise.
14182 (vst2_lane_s8): Likewise.
14183 (vst2_lane_s16): Likewise.
14184 (vst2_lane_s32): Likewise.
14185 (vst2_lane_s64): Likewise.
14186 (vst2_lane_u8): Likewise.
14187 (vst2_lane_u16): Likewise.
14188 (vst2_lane_u32): Likewise.
14189 (vst2_lane_u64): Likewise.
14190 (vst2_lane_bf16): Likewise.
14191 (vst2q_lane_f16): Use __builtin_memcpy to copy vector
14192 structure instead of using a union.
14193 (vst2q_lane_f32): Likewise.
14194 (vst2q_lane_f64): Likewise.
14195 (vst2q_lane_p8): Likewise.
14196 (vst2q_lane_p16): Likewise.
14197 (vst2q_lane_p64): Likewise.
14198 (vst2q_lane_s8): Likewise.
14199 (vst2q_lane_s16): Likewise.
14200 (vst2q_lane_s32): Likewise.
14201 (vst2q_lane_s64): Likewise.
14202 (vst2q_lane_u8): Likewise.
14203 (vst2q_lane_u16): Likewise.
14204 (vst2q_lane_u32): Likewise.
14205 (vst2q_lane_u64): Likewise.
14206 (vst2q_lane_bf16): Likewise.
14208 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
14210 * config/aarch64/arm_neon.h (__ST3_LANE_FUNC): Delete.
14211 (__ST3Q_LANE_FUNC): Delete.
14212 (vst3_lane_f16): Use __builtin_memcpy to copy vector
14213 structure instead of constructing __builtin_aarch64_simd_ci
14214 one vector at a time.
14215 (vst3_lane_f32): Likewise.
14216 (vst3_lane_f64): Likewise.
14217 (vst3_lane_p8): Likewise.
14218 (vst3_lane_p16): Likewise.
14219 (vst3_lane_p64): Likewise.
14220 (vst3_lane_s8): Likewise.
14221 (vst3_lane_s16): Likewise.
14222 (vst3_lane_s32): Likewise.
14223 (vst3_lane_s64): Likewise.
14224 (vst3_lane_u8): Likewise.
14225 (vst3_lane_u16): Likewise.
14226 (vst3_lane_u32): Likewise.
14227 (vst3_lane_u64): Likewise.
14228 (vst3_lane_bf16): Likewise.
14229 (vst3q_lane_f16): Use __builtin_memcpy to copy vector
14230 structure instead of using a union.
14231 (vst3q_lane_f32): Likewise.
14232 (vst3q_lane_f64): Likewise.
14233 (vst3q_lane_p8): Likewise.
14234 (vst3q_lane_p16): Likewise.
14235 (vst3q_lane_p64): Likewise.
14236 (vst3q_lane_s8): Likewise.
14237 (vst3q_lane_s16): Likewise.
14238 (vst3q_lane_s32): Likewise.
14239 (vst3q_lane_s64): Likewise.
14240 (vst3q_lane_u8): Likewise.
14241 (vst3q_lane_u16): Likewise.
14242 (vst3q_lane_u32): Likewise.
14243 (vst3q_lane_u64): Likewise.
14244 (vst3q_lane_bf16): Likewise.
14246 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
14248 * config/aarch64/arm_neon.h (__ST4_LANE_FUNC): Delete.
14249 (__ST4Q_LANE_FUNC): Delete.
14250 (vst4_lane_f16): Use __builtin_memcpy to copy vector
14251 structure instead of constructing __builtin_aarch64_simd_xi
14252 one vector at a time.
14253 (vst4_lane_f32): Likewise.
14254 (vst4_lane_f64): Likewise.
14255 (vst4_lane_p8): Likewise.
14256 (vst4_lane_p16): Likewise.
14257 (vst4_lane_p64): Likewise.
14258 (vst4_lane_s8): Likewise.
14259 (vst4_lane_s16): Likewise.
14260 (vst4_lane_s32): Likewise.
14261 (vst4_lane_s64): Likewise.
14262 (vst4_lane_u8): Likewise.
14263 (vst4_lane_u16): Likewise.
14264 (vst4_lane_u32): Likewise.
14265 (vst4_lane_u64): Likewise.
14266 (vst4_lane_bf16): Likewise.
14267 (vst4q_lane_f16): Use __builtin_memcpy to copy vector
14268 structure instead of using a union.
14269 (vst4q_lane_f32): Likewise.
14270 (vst4q_lane_f64): Likewise.
14271 (vst4q_lane_p8): Likewise.
14272 (vst4q_lane_p16): Likewise.
14273 (vst4q_lane_p64): Likewise.
14274 (vst4q_lane_s8): Likewise.
14275 (vst4q_lane_s16): Likewise.
14276 (vst4q_lane_s32): Likewise.
14277 (vst4q_lane_s64): Likewise.
14278 (vst4q_lane_u8): Likewise.
14279 (vst4q_lane_u16): Likewise.
14280 (vst4q_lane_u32): Likewise.
14281 (vst4q_lane_u64): Likewise.
14282 (vst4q_lane_bf16): Likewise.
14284 2021-08-06 Martin Liska <mliska@suse.cz>
14286 * config/rs6000/rs6000.c (rs6000_option_override_internal): When
14287 a target option is restored, it can have
14288 rs6000_long_double_type_size set to FLOAT_PRECISION_TFmode
14289 and error should not be emitted.
14291 2021-08-06 Sebastian Huber <sebastian.huber@embedded-brains.de>
14293 * gcov-io.h (gcov_write): Declare.
14294 * gcov-io.c (gcov_write): New.
14295 (gcov_write_counter): Remove.
14296 (gcov_write_tag_length): Likewise.
14297 (gcov_write_summary): Replace gcov_write_tag_length() with calls to
14298 gcov_write_unsigned().
14299 * doc/invoke.texi (fprofile-info-section): Mention
14300 __gcov_info_to_gdca().
14302 2021-08-06 Martin Sebor <msebor@redhat.com>
14304 * dominance.c (prune_bbs_to_update_dominators): Adjust by-value vec
14305 arguments to by-reference.
14306 (iterate_fix_dominators): Same.
14307 * dominance.h (iterate_fix_dominators): Same.
14308 * ipa-prop.h: Call auto_vec::to_vec_legacy.
14309 * tree-data-ref.c (dump_data_dependence_relation): Adjust by-value vec
14310 arguments to by-reference.
14311 (debug_data_dependence_relation): Same.
14312 (dump_data_dependence_relations): Same.
14313 * tree-data-ref.h (debug_data_dependence_relation): Same.
14314 (dump_data_dependence_relations): Same.
14315 * tree-predcom.c (dump_chains): Same.
14316 (initialize_root_vars_lm): Same.
14317 (determine_unroll_factor): Same.
14318 (replace_phis_by_defined_names): Same.
14319 (insert_init_seqs): Same.
14320 (pcom_worker::tree_predictive_commoning_loop): Call
14321 auto_vec::to_vec_legacy.
14322 * tree-ssa-pre.c (insert_into_preds_of_block): Adjust by-value vec
14323 arguments to by-reference.
14324 * tree-ssa-threadbackward.c (populate_worklist): Same.
14325 (back_threader::resolve_def): Same.
14326 * tree-vect-data-refs.c (vect_check_nonzero_value): Same.
14327 (vect_enhance_data_refs_alignment): Same.
14328 (vect_check_lower_bound): Same.
14329 (vect_prune_runtime_alias_test_list): Same.
14330 (vect_permute_store_chain): Same.
14331 * tree-vect-slp-patterns.c (vect_normalize_conj_loc): Same.
14332 * tree-vect-stmts.c (vect_create_vectorized_demotion_stmts): Same.
14333 * tree-vectorizer.h (vect_permute_store_chain): Same.
14334 * vec.c (test_init): New function.
14335 (vec_c_tests): Call new function.
14336 * vec.h (vec): Declare ctors, dtor, and assignment.
14337 (auto_vec::vec_to_legacy): New function.
14338 (vec::copy): Adjust initialization.
14340 2021-08-05 H.J. Lu <hjl.tools@gmail.com>
14343 * config/i386/i386.c (ix86_can_inline_p): Ignore MASK_80387 if
14344 callee only uses GPRs.
14345 * config/i386/ia32intrin.h: Revert commit 5463cee2770.
14346 * config/i386/serializeintrin.h: Revert commit 71958f740f1.
14347 * config/i386/x86gprintrin.h: Add
14348 #pragma GCC target("general-regs-only") and #pragma GCC pop_options
14349 to disable non-GPR ISAs.
14351 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
14353 PR middle-end/101787
14354 * doc/md.texi (cond_ashl, cond_ashr, cond_lshr): Document.
14356 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
14358 * tree-vectorizer.h (vect_is_store_elt_extraction, vect_is_reduction)
14359 (vect_reduc_type, vect_embedded_comparison_type, vect_comparison_type)
14360 (vect_is_extending_load, vect_is_integer_truncation): New functions,
14361 moved from aarch64.c but given different names.
14362 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction)
14363 (aarch64_is_reduction, aarch64_reduc_type)
14364 (aarch64_embedded_comparison_type, aarch64_comparison_type)
14365 (aarch64_extending_load_p, aarch64_integer_truncation_p): Delete
14366 in favor of the above. Update callers accordingly.
14368 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
14371 * config/arm/arm-cpus.in (generic-armv7-a): Add quirk to suppress
14372 writing .cpu directive in asm output.
14373 * config/arm/arm.c (arm_identify_fpu_from_isa): New variable.
14374 (arm_last_printed_arch_string): Delete.
14375 (arm_last-printed_fpu_string): Delete.
14376 (arm_configure_build_target): If use of floating-point/SIMD is
14377 disabled, remove all fp/simd related features from the target ISA.
14378 (last_arm_targ_options): New variable.
14379 (arm_print_asm_arch_directives): Add new parameters. Change order
14380 of emitted directives and handle all cases here.
14381 (arm_file_start): Always call arm_print_asm_arch_directives, move
14382 all generation of .arch/.arch_extension here.
14383 (arm_file_end): Call arm_print_asm_arch.
14384 (arm_declare_function_name): Call arm_print_asm_arch_directives
14385 instead of printing .arch/.fpu directives directly.
14387 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
14389 * config/arm/arm.c (arm_configure_build_target): Don't call
14390 arm_option_reconfigure_globals.
14391 (arm_option_restore): Call arm_option_reconfigure_globals after
14392 reconfiguring the target.
14393 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
14395 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
14397 * config/arm/arm.c (arm_configure_build_target): Ensure the target's
14398 arch_name is always set.
14400 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
14402 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
14403 of vec_select high-half from being added into Neon subtract
14406 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
14408 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
14409 of vec_select high-half from being added into Neon add cost.
14411 2021-08-05 Kewen Lin <linkw@linux.ibm.com>
14413 * cfgloop.h (loops_list::loops_list): Add one optional argument
14414 root and adjust accordingly, update loop tree walking and factor
14416 * cfgloop.c (loops_list::walk_loop_tree): ... this. New function.
14418 2021-08-05 Eric Botcazou <ebotcazou@gcc.gnu.org>
14420 PR tree-optimization/101626
14421 * tree-sra.c (propagate_subaccesses_from_rhs): Do not set the
14422 reverse scalar storage order on a pointer or vector component.
14424 2021-08-05 liuhongt <hongtao.liu@intel.com>
14426 * config/i386/sse.md (cond_<code><mode>): New expander.
14428 2021-08-05 liuhongt <hongtao.liu@intel.com>
14430 * config/i386/sse.md (cond_<code><mode>): New expander.
14432 2021-08-05 liuhongt <hongtao.liu@intel.com>
14434 * config/i386/sse.md (cond_<code><mode>): New expander.
14436 2021-08-04 David Malcolm <dmalcolm@redhat.com>
14439 * Makefile.in (ANALYZER_OBJS): Add analyzer/region-model-asm.o.
14441 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
14444 * config/i386/i386.h (STORE_MAX_PIECES): Allow 16/32/64 bytes
14445 only if TARGET_INTER_UNIT_MOVES_TO_VEC is true.
14447 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
14450 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
14451 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
14452 data with SSE register from one memory location to another.
14454 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
14456 * config/s390/s390.c (expand_perm_with_vpdi): New function.
14457 (vectorize_vec_perm_const_1): Call expand_perm_with_vpdi.
14458 * config/s390/vector.md (*vpdi1<mode>, @vpdi1<mode>): Enable a
14459 parameterized expander.
14460 (*vpdi4<mode>, @vpdi4<mode>): Likewise.
14462 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
14464 * config/s390/s390.c (MAX_VECT_LEN): Define macro.
14465 (struct expand_vec_perm_d): Define struct.
14466 (expand_perm_with_merge): New function.
14467 (vectorize_vec_perm_const_1): New function.
14468 (s390_vectorize_vec_perm_const): New function.
14469 (TARGET_VECTORIZE_VEC_PERM_CONST): Define target macro.
14471 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
14473 * config/s390/vector.md (V_HW_64): Remove mode iterator.
14474 (*vec_load_pair<mode>): Use V_HW_2 instead of V_HW_64.
14475 * config/s390/vx-builtins.md
14476 (vec_scatter_element<V_HW_2:mode>_SI): Use V_HW_2 instead of
14479 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
14481 * config/s390/s390.md (UNSPEC_VEC_PERMI): Remove constant
14483 * config/s390/vector.md (*vpdi1<mode>, *vpdi4<mode>): New pattern
14485 * config/s390/vx-builtins.md (*vec_permi<mode>): Emit generic rtx
14486 instead of an unspec.
14488 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
14490 * config/s390/s390-modes.def: Add more vector modes to support
14491 concatenation of two vectors.
14492 * config/s390/s390-protos.h (s390_expand_merge_perm_const): Add
14494 (s390_expand_merge): Likewise.
14495 * config/s390/s390.c (s390_expand_merge_perm_const): New function.
14496 (s390_expand_merge): New function.
14497 * config/s390/s390.md (UNSPEC_VEC_MERGEH, UNSPEC_VEC_MERGEL):
14498 Remove constant definitions.
14499 * config/s390/vector.md (V_HW_2): Add mode iterators.
14500 (VI_HW_4, V_HW_4): Rename VI_HW_4 to V_HW_4.
14501 (vec_2x_nelts, vec_2x_wide): New mode attributes.
14502 (*vmrhb, *vmrlb, *vmrhh, *vmrlh, *vmrhf, *vmrlf, *vmrhg, *vmrlg):
14503 New pattern definitions.
14504 (vec_widen_umult_lo_<mode>, vec_widen_umult_hi_<mode>)
14505 (vec_widen_smult_lo_<mode>, vec_widen_smult_hi_<mode>)
14506 (vec_unpacks_lo_v4sf, vec_unpacks_hi_v4sf, vec_unpacks_lo_v2df)
14507 (vec_unpacks_hi_v2df): Adjust expanders to emit non-unspec RTX for
14509 * config/s390/vx-builtins.md (V_HW_4): Remove mode iterator. Now
14511 (vec_mergeh<mode>, vec_mergel<mode>): Use s390_expand_merge to
14512 emit vec merge pattern.
14514 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
14516 * config/aarch64/aarch64.c (aarch64_strip_extend_vec_half):
14518 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent cost of
14519 vec_select high-half from being added into Neon multiply
14521 * rtlanal.c (vec_series_highpart_p): Define.
14522 * rtlanal.h (vec_series_highpart_p): Declare.
14524 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
14526 * config/aarch64/aarch64.c (aarch64_strip_duplicate_vec_elt):
14528 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent
14529 vec_select cost from being added into Neon multiply cost.
14531 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
14533 * tree-vect-loop.c (vect_better_loop_vinfo_p): Detect cases in
14534 which old_loop_vinfo is an epilogue loop that handles a constant
14535 number of iterations.
14537 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
14539 * tree-vect-loop.c (vect_analyze_loop): Print a dump message
14540 when a reanalyzed loop fails to be cheaper than the current
14543 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
14545 * config/aarch64/aarch64.c: Fix a typo.
14547 2021-08-04 Vincent Lefèvre <vincent-gcc@vinc17.net>
14549 PR gcov-profile/101773
14550 * gcov-io.c (gcov_close): Check return code of a fclose.
14552 2021-08-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
14555 * dwarf2out.c (dwarf2out_assembly_start): Emit a dummy
14556 .file statement when needed.
14558 2021-08-04 Richard Biener <rguenther@suse.de>
14560 * tree-vect-data-refs.c (vect_check_gather_scatter):
14561 Include widening conversions only when the result is
14562 still handed by native gather or the current offset
14563 size not already matches the data size.
14564 Also succeed analysis in case there's no native support,
14565 noted by a IFN_LAST ifn and a NULL decl.
14566 (vect_analyze_data_refs): Always consider gathers.
14567 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
14568 Test for no IFN gather rather than decl gather.
14569 * tree-vect-stmts.c (vect_model_load_cost): Pass in the
14570 gather-scatter info and cost emulated gathers accordingly.
14571 (vect_truncate_gather_scatter_offset): Properly test for
14573 (vect_use_strided_gather_scatters_p): Likewise.
14574 (get_load_store_type): Handle emulated gathers and its
14576 (vectorizable_load): Likewise. Emulate them by extracting
14577 scalar offsets, doing scalar loads and a vector construct.
14579 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
14582 * expr.c (op_by_pieces_d::op_by_pieces_d): Add a max_pieces
14583 argument to set m_max_size.
14584 (move_by_pieces_d): Pass MOVE_MAX_PIECES to op_by_pieces_d.
14585 (store_by_pieces_d): Pass STORE_MAX_PIECES to op_by_pieces_d.
14586 (compare_by_pieces_d): Pass COMPARE_MAX_PIECES to op_by_pieces_d.
14588 2021-08-04 Roger Sayle <roger@nextmovesoftware.com>
14589 Marc Glisse <marc.glisse@inria.fr>
14591 * match.pd (bit_ior, bit_xor): Canonicalize (X*C1)|(X*C2) and
14592 (X*C1)^(X*C2) as X*(C1+C2), and related variants, using
14593 tree_nonzero_bits to ensure that operands are bit-wise disjoint.
14595 2021-08-04 Richard Biener <rguenther@suse.de>
14597 * tree-ssa-forwprop.c (pass_forwprop::execute): Split
14598 out code to decompose vector loads ...
14599 (optimize_vector_load): ... here. Generalize it to
14600 handle intermediate widening and TARGET_MEM_REF loads
14601 and apply it to loads with a supported vector mode as well.
14603 2021-08-04 Richard Biener <rguenther@suse.de>
14605 PR tree-optimization/101756
14606 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Make sure
14607 the result of the reduction epilogue is compatible to the original
14610 2021-08-04 liuhongt <hongtao.liu@intel.com>
14613 * config/i386/i386.md (peephole2): Refine predicate from
14614 register_operand to general_reg_operand.
14616 2021-08-04 Aldy Hernandez <aldyh@redhat.com>
14618 * gimple-range-path.h (path_range_query::dump): Mark override.
14620 2021-08-04 Richard Biener <rguenther@suse.de>
14622 PR tree-optimization/101769
14623 * tree-tailcall.c (eliminate_tail_call): Add the created loop
14624 for the first recursion and return it via the new output parameter.
14625 (optimize_tail_call): Pass through new output param.
14626 (tree_optimize_tail_calls_1): After creating all latches,
14627 add the created loop to the loop tree. Do not mark loops for fixup.
14629 2021-08-04 Martin Liska <mliska@suse.cz>
14631 * doc/invoke.texi: Document threader-mode param.
14633 2021-08-04 liuhongt <hongtao.liu@intel.com>
14635 * config/i386/sse.md (cond_fma<mode>): New expander.
14636 (cond_fms<mode>): Ditto.
14637 (cond_fnma<mode>): Ditto.
14638 (cond_fnms<mode>): Ditto.
14640 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
14642 * config/rs6000/vsx.md (*vsx_le_perm_store_<mode>): Use && instead of &.
14644 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
14646 * config/rs6000/constraints.md: Remove "e" from the list of available
14647 constraint characters.
14649 2021-08-03 Eugene Rozenfeld <erozen@microsoft.com>
14651 PR gcov-profile/71672
14652 * auto-profile.c (afdo_indirect_call): Fix setup of the historgram value for indirect calls.
14654 2021-08-03 Paul A. Clarke <pc@us.ibm.com>
14656 * config/rs6000/smmintrin.h (_mm_minpos_epu16): New.
14658 2021-08-03 H.J. Lu <hjl.tools@gmail.com>
14660 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): In 64-bit mode,
14661 try XMM31 to avoid vzeroupper.
14663 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
14665 * doc/invoke.texi: Document -mtune=neoverse-512tvb and
14666 -mcpu=neoverse-512tvb.
14667 * config/aarch64/aarch64-cores.def (neoverse-512tvb): New entry.
14668 * config/aarch64/aarch64-tune.md: Regenerate.
14669 * config/aarch64/aarch64.c (neoverse512tvb_sve_vector_cost)
14670 (neoverse512tvb_sve_issue_info, neoverse512tvb_vec_issue_info)
14671 (neoverse512tvb_vector_cost, neoverse512tvb_tunings): New structures.
14672 (aarch64_adjust_body_cost_sve): Handle -mtune=neoverse-512tvb.
14673 (aarch64_adjust_body_cost): Likewise.
14675 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
14677 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Only
14678 record issue information for operations that occur in the
14681 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
14683 * config/aarch64/aarch64.c (aarch64_multiply_add_p): Add a vec_flags
14684 parameter. Detect cases in which an Advanced SIMD MLA would almost
14685 certainly require a MOV.
14686 (aarch64_count_ops): Update accordingly.
14688 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
14690 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction): New
14691 function, split out from...
14692 (aarch64_detect_vector_stmt_subtype): ...here.
14693 (aarch64_add_stmt_cost): Treat extracting element 0 as free.
14695 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
14697 * config/aarch64/aarch64-protos.h (sve_vec_cost):
14698 Add gather_load_x32_cost and gather_load_x64_cost.
14699 * config/aarch64/aarch64.c (generic_sve_vector_cost)
14700 (a64fx_sve_vector_cost, neoversev1_sve_vector_cost): Update
14701 accordingly, using the values given by the scalar_load * number
14702 of elements calculation that we used previously.
14703 (aarch64_detect_vector_stmt_subtype): Use the new fields.
14705 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
14707 * config/aarch64/aarch64.c (aarch64_adjust_body_cost_sve): New
14708 function, split out from...
14709 (aarch64_adjust_body_cost): ...here.
14711 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
14713 * config/aarch64/fractional-cost.h: New file.
14714 * config/aarch64/aarch64.c: Include <algorithm> (indirectly)
14715 and cost_fraction.h.
14716 (vec_cost_fraction): New typedef.
14717 (aarch64_detect_scalar_stmt_subtype): Use it for statement costs.
14718 (aarch64_detect_vector_stmt_subtype): Likewise.
14719 (aarch64_sve_adjust_stmt_cost, aarch64_adjust_stmt_cost): Likewise.
14720 (aarch64_estimate_min_cycles_per_iter): Use vec_cost_fraction
14722 (aarch64_adjust_body_cost): Likewise.
14723 (aarch64_test_cost_fraction): New function.
14724 (aarch64_run_selftests): Call it.
14726 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
14728 * config/aarch64/aarch64-protos.h (tune_params::sve_width): Turn
14730 * config/aarch64/aarch64.c (aarch64_cmp_autovec_modes): Update
14732 (aarch64_estimated_poly_value): Likewise. Use the least significant
14733 set bit for the minimum and likely values. Use the most significant
14734 set bit for the maximum value.
14736 2021-08-03 liuhongt <hongtao.liu@intel.com>
14738 * config/i386/sse.md (cond_<insn><mode>): New expander.
14739 (cond_mul<mode>): Ditto.
14741 2021-08-03 Kewen Lin <linkw@linux.ibm.com>
14743 * tree-cfg.c (move_sese_region_to_fn): Fix typos on dloop.
14745 2021-08-03 liuhongt <hongtao.liu@intel.com>
14747 * config/i386/sse.md (cond_<insn><mode>):New expander.
14748 (cond_mul<mode>): Ditto.
14749 (cond_div<mode>): Ditto.
14751 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
14753 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Also
14754 check stack_realign_needed for stack realignment.
14755 (ix86_legitimate_constant_p): Always allow CONST_WIDE_INT smaller
14756 than the largest integer supported by vector register.
14757 * config/i386/i386.h (MAX_MOVE_MAX): New. Set to 64.
14758 (MOVE_MAX): Set to bytes of the largest integer supported by
14760 (STORE_MAX_PIECES): New.
14762 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
14764 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
14765 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
14766 data from one memory location to another.
14768 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
14770 PR middle-end/90773
14771 * config/i386/i386.c (TARGET_GEN_MEMSET_SCRATCH_RTX): New.
14773 2021-08-02 Aldy Hernandez <aldyh@redhat.com>
14775 PR tree-optimization/101724
14776 * params.opt: Remove --param=threader-iterative.
14777 * tree-ssa-threadbackward.c (pass_thread_jumps::execute): Remove
14780 2021-08-02 Tom de Vries <tdevries@suse.de>
14782 PR middle-end/101665
14783 * doc/extend.texi (nonnull attribute): Improve documentation.
14785 2021-08-02 Andrew Pinski <apinski@marvell.com>
14787 PR rtl-optimization/101683
14788 * rtlanal.c (may_trap_p_1): Handle UNSIGNED_FIX.
14790 2021-08-02 Roger Sayle <roger@nextmovesoftware.com>
14792 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern):
14793 Renamed from cond_removal_in_popcount_clz_ctz_pattern.
14794 Add support for BSWAP, FFS, PARITY and CLRSB builtins.
14795 (tree_ssa_phiop_worker): Update call to function above.
14797 2021-08-01 H.J. Lu <hjl.tools@gmail.com>
14800 * config/i386/i386.md (bsr_rex64_1_zext): New.
14801 (combine splitter for constant - clzll): Replace gen_bsr_rex64_1
14802 with gen_bsr_rex64_1_zext.
14804 2021-07-31 Jakub Jelinek <jakub@redhat.com>
14807 * config/i386/i386.md (bsr_rex64_1, bsr_1, bsr_zext_1): New
14808 define_insn patterns.
14809 (*bsr_rex64_2, *bsr_2): New define_insn_and_split patterns.
14810 Add combine splitters for constant - clz.
14811 (clz<mode>2): Use a temporary pseudo for bsr result.
14813 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
14815 * config/rs6000/smmintrin.h (_mm_floor_pd, _mm_floor_ps,
14816 _mm_floor_sd, _mm_floor_ss): New.
14818 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
14820 * config/rs6000/smmintrin.h (_mm_ceil_pd, _mm_ceil_ps,
14821 _mm_ceil_sd, _mm_ceil_ss): New.
14823 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
14825 * config/rs6000/smmintrin.h (_mm_blend_pd, _mm_blendv_pd,
14826 _mm_blend_ps, _mm_blendv_ps): New.
14828 2021-07-30 Roger Sayle <roger@nextmovesoftware.com>
14829 Uroš Bizjak <ubizjak@gmail.com>
14831 * config/i386/i386.md (*dec_cmov<mode>): New define_insn_and_split
14832 to generate a conditional move using the carry flag after sub $1.
14833 (peephole2): Eliminate a register-to-register move by inverting
14834 the condition of a conditional move.
14836 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
14838 * config/mmix/mmix.md ("call", "call_value", "*call_real")
14839 ("*call_value_real"): Don't generate rtx mentioning the generic
14840 operands 1 and 2 to "call", and similarly for "call_value".
14841 * config/mmix/mmix.c (mmix_print_operand_punct_valid_p)
14842 (mmix_print_operand): Use '!' instead of 'p'.
14844 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
14846 * doc/md.texi (call): Correct information about operand 2.
14847 * config/mmix/mmix.md ("call", "call_value"): Remove fixed FIXMEs.
14849 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
14851 * range-op.cc (operator_trunc_mod::wi_fold): Fold constants.
14853 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
14855 * range-op.cc (operator_div::wi_fold): Return UNDEFINED for [0, 0] divisor.
14857 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
14859 * gimple-range-cache.cc (*::set_bb_range): Change const basic_block to
14860 const_basic_block..
14861 (*::get_bb_range): Ditto.
14862 (*::bb_range_p): Ditto.
14863 * gimple-range-cache.h: Change prototypes.
14865 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
14867 PR middle-end/90773
14868 * builtins.c (builtin_memcpy_read_str): Change the mode argument
14869 from scalar_int_mode to fixed_size_mode.
14870 (builtin_strncpy_read_str): Likewise.
14871 (gen_memset_value_from_prev): New function.
14872 (builtin_memset_read_str): Change the mode argument from
14873 scalar_int_mode to fixed_size_mode. Use gen_memset_value_from_prev
14874 and support CONST_VECTOR.
14875 (builtin_memset_gen_str): Likewise.
14876 (try_store_by_multiple_pieces): Use by_pieces_constfn to declare
14878 * builtins.h (builtin_strncpy_read_str): Replace scalar_int_mode
14879 with fixed_size_mode.
14880 (builtin_memset_read_str): Likewise.
14881 * expr.c (widest_int_mode_for_size): Renamed to ...
14882 (widest_fixed_size_mode_for_size): Add a bool argument to
14883 indicate if QI vector mode can be used.
14884 (by_pieces_ninsns): Call widest_fixed_size_mode_for_size
14885 instead of widest_int_mode_for_size.
14886 (pieces_addr::adjust): Change the mode argument from
14887 scalar_int_mode to fixed_size_mode.
14888 (op_by_pieces_d): Make m_len read-only. Add a bool member,
14889 m_qi_vector_mode, to indicate that QI vector mode can be used.
14890 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
14891 initialize m_qi_vector_mode. Call widest_fixed_size_mode_for_size
14892 instead of widest_int_mode_for_size.
14893 (op_by_pieces_d::get_usable_mode): Change the mode argument from
14894 scalar_int_mode to fixed_size_mode. Call
14895 widest_fixed_size_mode_for_size instead of
14896 widest_int_mode_for_size.
14897 (op_by_pieces_d::smallest_fixed_size_mode_for_size): New member
14898 function to return the smallest integer or QI vector mode.
14899 (op_by_pieces_d::run): Call widest_fixed_size_mode_for_size
14900 instead of widest_int_mode_for_size. Call
14901 smallest_fixed_size_mode_for_size instead of
14902 smallest_int_mode_for_size.
14903 (store_by_pieces_d::store_by_pieces_d): Add a bool argument to
14904 indicate that QI vector mode can be used and pass it to
14905 op_by_pieces_d::op_by_pieces_d.
14906 (can_store_by_pieces): Call widest_fixed_size_mode_for_size
14907 instead of widest_int_mode_for_size. Pass memsetp to
14908 widest_fixed_size_mode_for_size to support QI vector mode.
14909 Allow all CONST_VECTORs for memset if vec_duplicate is supported.
14910 (store_by_pieces): Pass memsetp to
14911 store_by_pieces_d::store_by_pieces_d.
14912 (clear_by_pieces_1): Removed.
14913 (clear_by_pieces): Replace clear_by_pieces_1 with
14914 builtin_memset_read_str and pass true to store_by_pieces_d to
14915 support vector mode broadcast.
14916 (string_cst_read_str): Change the mode argument from
14917 scalar_int_mode to fixed_size_mode.
14918 * expr.h (by_pieces_constfn): Change scalar_int_mode to
14920 (by_pieces_prev): Likewise.
14921 * rtl.h (lowpart_subreg_regno): New.
14922 * rtlanal.c (lowpart_subreg_regno): New. A wrapper around
14923 simplify_subreg_regno.
14924 * target.def (gen_memset_scratch_rtx): New hook.
14925 * doc/tm.texi.in: Add TARGET_GEN_MEMSET_SCRATCH_RTX.
14926 * doc/tm.texi: Regenerated.
14928 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
14931 * config/mips/mips.c (mips_atomic_assign_expand_fenv): Use
14932 TARGET_EXPR instead of MODIFY_EXPR.
14934 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
14937 * config/mips/mips-protos.h (mips_expand_vec_cmp_expr): Declare.
14938 * config/mips/mips.c (mips_expand_vec_cmp_expr): New function.
14939 * config/mips/mips-msa.md (vec_cmp<MSA:mode><mode_i>): New
14941 (vec_cmpu<IMSA:mode><mode_i>): New expander.
14943 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
14946 * config/i386/i386-options.c (ix86_option_override_internal):
14947 Don't enable LZCNT/POPCNT if they have been disabled explicitly.
14949 2021-07-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
14952 * config/arm/arm_neon.h (vld1_p64): Replace call to builtin by
14953 explicitly dereferencing __a.
14954 (vld1_s64): Likewise.
14955 (vld1_u64): Likewise.
14956 * config/arm/arm_neon_builtins.def (vld1): Remove entry for di
14957 and change to VAR13.
14959 2021-07-30 Aldy Hernandez <aldyh@redhat.com>
14961 * gimple-loop-versioning.cc (lv_dom_walker::lv_dom_walker): Remove
14962 use of m_range_analyzer.
14963 (loop_versioning::lv_dom_walker::before_dom_children): Same.
14964 (loop_versioning::lv_dom_walker::after_dom_children): Remove.
14965 (loop_versioning::prune_loop_conditions): Replace vr_values use
14966 with range_query interface.
14967 (pass_loop_versioning::execute): Use ranger.
14969 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
14972 * ipa-devirt.c (ipa_odr_read_section): Compare the precision of
14973 enum values, and emit a warning if they mismatch.
14975 2021-07-30 Kewen Lin <linkw@linux.ibm.com>
14977 * cfgloop.h (as_const): New function.
14978 (class loop_iterator): Rename to ...
14979 (class loops_list): ... this.
14980 (loop_iterator::next): Rename to ...
14981 (loops_list::Iter::fill_curr_loop): ... this and adjust.
14982 (loop_iterator::loop_iterator): Rename to ...
14983 (loops_list::loops_list): ... this and adjust.
14984 (loops_list::Iter): New class.
14985 (loops_list::iterator): New type.
14986 (loops_list::const_iterator): New type.
14987 (loops_list::begin): New function.
14988 (loops_list::end): Likewise.
14989 (loops_list::begin const): Likewise.
14990 (loops_list::end const): Likewise.
14991 (FOR_EACH_LOOP): Remove.
14992 (FOR_EACH_LOOP_FN): Remove.
14993 * cfgloop.c (flow_loops_dump): Adjust FOR_EACH_LOOP* with range-based
14994 for loop with loops_list instance.
14995 (sort_sibling_loops): Likewise.
14996 (disambiguate_loops_with_multiple_latches): Likewise.
14997 (verify_loop_structure): Likewise.
14998 * cfgloopmanip.c (create_preheaders): Likewise.
14999 (force_single_succ_latches): Likewise.
15000 * config/aarch64/falkor-tag-collision-avoidance.c
15001 (execute_tag_collision_avoidance): Likewise.
15002 * config/mn10300/mn10300.c (mn10300_scan_for_setlb_lcc): Likewise.
15003 * config/s390/s390.c (s390_adjust_loops): Likewise.
15004 * doc/loop.texi: Likewise.
15005 * gimple-loop-interchange.cc (pass_linterchange::execute): Likewise.
15006 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
15007 * gimple-loop-versioning.cc (loop_versioning::analyze_blocks): Likewise.
15008 (loop_versioning::make_versioning_decisions): Likewise.
15009 * gimple-ssa-split-paths.c (split_paths): Likewise.
15010 * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
15011 * graphite.c (canonicalize_loop_form): Likewise.
15012 (graphite_transform_loops): Likewise.
15013 * ipa-fnsummary.c (analyze_function_body): Likewise.
15014 * ipa-pure-const.c (analyze_function): Likewise.
15015 * loop-doloop.c (doloop_optimize_loops): Likewise.
15016 * loop-init.c (loop_optimizer_finalize): Likewise.
15017 (fix_loop_structure): Likewise.
15018 * loop-invariant.c (calculate_loop_reg_pressure): Likewise.
15019 (move_loop_invariants): Likewise.
15020 * loop-unroll.c (decide_unrolling): Likewise.
15021 (unroll_loops): Likewise.
15022 * modulo-sched.c (sms_schedule): Likewise.
15023 * predict.c (predict_loops): Likewise.
15024 (pass_profile::execute): Likewise.
15025 * profile.c (branch_prob): Likewise.
15026 * sel-sched-ir.c (sel_finish_pipelining): Likewise.
15027 (sel_find_rgns): Likewise.
15028 * tree-cfg.c (replace_loop_annotate): Likewise.
15029 (replace_uses_by): Likewise.
15030 (move_sese_region_to_fn): Likewise.
15031 * tree-if-conv.c (pass_if_conversion::execute): Likewise.
15032 * tree-loop-distribution.c (loop_distribution::execute): Likewise.
15033 * tree-parloops.c (parallelize_loops): Likewise.
15034 * tree-predcom.c (tree_predictive_commoning): Likewise.
15035 * tree-scalar-evolution.c (scev_initialize): Likewise.
15036 (scev_reset): Likewise.
15037 * tree-ssa-dce.c (find_obviously_necessary_stmts): Likewise.
15038 * tree-ssa-live.c (remove_unused_locals): Likewise.
15039 * tree-ssa-loop-ch.c (ch_base::copy_headers): Likewise.
15040 * tree-ssa-loop-im.c (analyze_memory_references): Likewise.
15041 (tree_ssa_lim_initialize): Likewise.
15042 * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables): Likewise.
15043 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize): Likewise.
15044 * tree-ssa-loop-manip.c (get_loops_exits): Likewise.
15045 * tree-ssa-loop-niter.c (estimate_numbers_of_iterations): Likewise.
15046 (free_numbers_of_iterations_estimates): Likewise.
15047 * tree-ssa-loop-prefetch.c (tree_ssa_prefetch_arrays): Likewise.
15048 * tree-ssa-loop-split.c (tree_ssa_split_loops): Likewise.
15049 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Likewise.
15050 * tree-ssa-loop.c (gate_oacc_kernels): Likewise.
15051 (pass_scev_cprop::execute): Likewise.
15052 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
15053 * tree-ssa-sccvn.c (do_rpo_vn): Likewise.
15054 * tree-ssa-threadupdate.c
15055 (jump_thread_path_registry::thread_through_all_blocks): Likewise.
15056 * tree-vectorizer.c (vectorize_loops): Likewise.
15057 * tree-vrp.c (vrp_asserts::find_assert_locations): Likewise.
15059 2021-07-29 Hans-Peter Nilsson <hp@bitrange.com>
15061 * config/mmix/mmix.c (mmix_function_arg_1): Avoid
15062 generating a VOIDmode register for e.g the
15063 function_arg_info::end_marker.
15065 2021-07-29 Jeff Law <jeffreyalaw@gmail.com>
15067 * config/h8300/h8300-modes.def: Add CCZ, CCV and CCC, drop CCZNV.
15068 * config/h8300/h8300.md (H8cc mode iterator): Add CCZ.
15069 (cc mode_attr): Similarly.
15070 (ccz subst_attr): Similarly.
15071 * config/h8300/jumpcall.md: Add new patterns for branch-on-bit.
15072 * config/h8300/testcompare.md: Remove various cc0 based patterns
15073 that had been commented out. Add pattern to set CCZ from a bit
15076 2021-07-29 Thomas Schwinge <thomas@codesourcery.com>
15077 Julian Brown <julian@codesourcery.com>
15078 Kwok Cheung Yeung <kcy@codesourcery.com>
15080 * omp-offload.c (oacc_loop_xform_head_tail, oacc_loop_process):
15081 'update_stmt' after modification.
15082 (pass_oacc_loop_designation): New function, extracted out of...
15083 (pass_oacc_device_lower): ... this.
15084 (pass_data_oacc_loop_designation, pass_oacc_loop_designation)
15085 (make_pass_oacc_loop_designation): New
15086 * passes.def: Add it.
15087 * tree-parloops.c (create_parallel_loop): Adjust.
15088 * tree-pass.h (make_pass_oacc_loop_designation): New.
15090 2021-07-29 Aldy Hernandez <aldyh@redhat.com>
15092 * flag-types.h (enum threader_mode): New.
15093 * params.opt: Add entry for --param=threader-mode.
15094 * tree-ssa-threadbackward.c (THREADER_ITERATIVE_MODE): New.
15095 (class back_threader): New.
15096 (back_threader::back_threader): New.
15097 (back_threader::~back_threader): New.
15098 (back_threader::maybe_register_path): New.
15099 (back_threader::find_taken_edge): New.
15100 (back_threader::find_taken_edge_switch): New.
15101 (back_threader::find_taken_edge_cond): New.
15102 (back_threader::resolve_def): New.
15103 (back_threader::resolve_phi): New.
15104 (back_threader::find_paths_to_names): New.
15105 (back_threader::find_paths): New.
15108 (thread_jumps::find_jump_threads_backwards): Call ranger threader.
15109 (thread_jumps::find_jump_threads_backwards_with_ranger): New.
15110 (pass_thread_jumps::execute): Abstract out code...
15111 (try_thread_blocks): ...here.
15112 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
15113 Abstract out threading candidate code to...
15114 (single_succ_to_potentially_threadable_block): ...here.
15115 * tree-ssa-threadedge.h (single_succ_to_potentially_threadable_block):
15117 * tree-ssa-threadupdate.c (register_jump_thread): Return boolean.
15118 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
15119 Return bool from register_jump_thread.
15121 2021-07-29 Andreas Krebbel <krebbel@linux.ibm.com>
15123 * target.def: in0 and in1 do not need to be registers.
15124 * doc/tm.texi: Regenerate.
15126 2021-07-29 liuhongt <hongtao.liu@intel.com>
15129 * config/i386/i386.c (ix86_widen_mult_cost): New function.
15130 (ix86_add_stmt_cost): Use ix86_widen_mult_cost for
15133 2021-07-29 Jiufu Guo <guojiufu@linux.ibm.com>
15136 * config/rs6000/rs6000.c (TARGET_PREFERRED_DOLOOP_MODE): New hook.
15137 (rs6000_preferred_doloop_mode): New hook.
15138 * doc/tm.texi: Regenerate.
15139 * doc/tm.texi.in: Add hook preferred_doloop_mode.
15140 * target.def (preferred_doloop_mode): New hook.
15141 * targhooks.c (default_preferred_doloop_mode): New hook.
15142 * targhooks.h (default_preferred_doloop_mode): New hook.
15143 * tree-ssa-loop-ivopts.c (compute_doloop_base_on_mode): New function.
15144 (add_iv_candidate_for_doloop): Call targetm.preferred_doloop_mode
15145 and compute_doloop_base_on_mode.
15147 2021-07-28 Martin Sebor <msebor@redhat.com>
15149 PR middle-end/101494
15150 * tree-ssa-uninit.c (maybe_warn_operand): Correct object offset
15151 and size computation.
15153 2021-07-28 Martin Sebor <msebor@redhat.com>
15155 PR middle-end/101601
15156 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Remove
15158 Handle pointers to functions.
15160 2021-07-28 Martin Sebor <msebor@redhat.com>
15162 * Makefile.in (OBJS): Add gimple-ssa-warn-access.o and pointer-query.o.
15163 * attribs.h (fndecl_dealloc_argno): Move fndecl_dealloc_argno to tree.h.
15164 * builtins.c (compute_objsize_r): Move to pointer-query.cc.
15165 (access_ref::access_ref): Same.
15166 (access_ref::phi): Same.
15167 (access_ref::get_ref): Same.
15168 (access_ref::size_remaining): Same.
15169 (access_ref::offset_in_range): Same.
15170 (access_ref::add_offset): Same.
15171 (access_ref::inform_access): Same.
15172 (ssa_name_limit_t::visit_phi): Same.
15173 (ssa_name_limit_t::leave_phi): Same.
15174 (ssa_name_limit_t::next): Same.
15175 (ssa_name_limit_t::next_phi): Same.
15176 (ssa_name_limit_t::~ssa_name_limit_t): Same.
15177 (pointer_query::pointer_query): Same.
15178 (pointer_query::get_ref): Same.
15179 (pointer_query::put_ref): Same.
15180 (pointer_query::flush_cache): Same.
15181 (warn_string_no_nul): Move to gimple-ssa-warn-access.cc.
15182 (check_nul_terminated_array): Same.
15183 (unterminated_array): Same.
15184 (maybe_warn_for_bound): Same.
15185 (check_read_access): Same.
15186 (warn_for_access): Same.
15187 (get_size_range): Same.
15188 (check_access): Same.
15189 (gimple_call_alloc_size): Move to tree.c.
15190 (gimple_parm_array_size): Move to pointer-query.cc.
15191 (get_offset_range): Same.
15192 (gimple_call_return_array): Same.
15193 (handle_min_max_size): Same.
15194 (handle_array_ref): Same.
15195 (handle_mem_ref): Same.
15196 (compute_objsize): Same.
15197 (gimple_call_alloc_p): Move to gimple-ssa-warn-access.cc.
15198 (call_dealloc_argno): Same.
15199 (fndecl_dealloc_argno): Same.
15200 (new_delete_mismatch_p): Same.
15201 (matching_alloc_calls_p): Same.
15202 (warn_dealloc_offset): Same.
15203 (maybe_emit_free_warning): Same.
15204 * builtins.h (check_nul_terminated_array): Move to
15205 gimple-ssa-warn-access.h.
15206 (check_nul_terminated_array): Same.
15207 (warn_string_no_nul): Same.
15208 (unterminated_array): Same.
15209 (class ssa_name_limit_t): Same.
15210 (class pointer_query): Same.
15211 (struct access_ref): Same.
15212 (class range_query): Same.
15213 (struct access_data): Same.
15214 (gimple_call_alloc_size): Same.
15215 (gimple_parm_array_size): Same.
15216 (compute_objsize): Same.
15217 (class access_data): Same.
15218 (maybe_emit_free_warning): Same.
15219 * calls.c (initialize_argument_information): Remove call to
15220 maybe_emit_free_warning.
15221 * gimple-array-bounds.cc: Include new header..
15222 * gimple-fold.c: Same.
15223 * gimple-ssa-sprintf.c: Same.
15224 * gimple-ssa-warn-restrict.c: Same.
15225 * passes.def: Add pass_warn_access.
15226 * tree-pass.h (make_pass_warn_access): Declare.
15227 * tree-ssa-strlen.c: Include new headers.
15228 * tree.c (fndecl_dealloc_argno): Move here from builtins.c.
15229 * tree.h (fndecl_dealloc_argno): Move here from attribs.h.
15230 * gimple-ssa-warn-access.cc: New file.
15231 * gimple-ssa-warn-access.h: New file.
15232 * pointer-query.cc: New file.
15233 * pointer-query.h: New file.
15235 2021-07-28 Jakub Jelinek <jakub@redhat.com>
15237 PR middle-end/101624
15238 * ubsan.c (maybe_instrument_pointer_overflow,
15239 instrument_object_size): Only test DECL_REGISTER on VAR_DECLs,
15240 PARM_DECLs or RESULT_DECLs.
15241 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Likewise.
15243 2021-07-28 Jakub Jelinek <jakub@redhat.com>
15245 PR middle-end/101642
15246 * match.pd (bswap16 (x) == bswap16 (y)): Cast both operands
15247 to type of bswap16 for comparison.
15248 (bswap16 (x) == cst): Cast bswap16 operand to type of cst.
15250 2021-07-28 Richard Biener <rguenther@suse.de>
15252 PR tree-optimization/101615
15253 * tree-vect-slp.c (vect_optimize_slp): Materialize permutes
15254 at CTOR SLP graph entries.
15256 2021-07-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15258 * config/aarch64/aarch64.md (*extend<SHORT:mode><GPI:mode>2_aarch64):
15259 Add "r,w" alternative.
15261 2021-07-28 H.J. Lu <hjl.tools@gmail.com>
15264 * config/i386/i386.c (ix86_avx_u128_mode_needed): Don't set
15265 AVX_U128_DIRTY when all bits are zero.
15267 2021-07-28 Richard Biener <rguenther@suse.de>
15269 PR tree-optimization/101615
15270 * tree-vect-slp.c (vect_optimize_slp): Pre-existing vector
15271 external nodes cannot be permuted so make them perm_out 0.
15273 2021-07-28 Andrew Stubbs <ams@codesourcery.com>
15276 * config.in: Regenerate.
15277 * config/gcn/gcn-hsa.h (A_FIJI): New define.
15278 (A_900): New define.
15279 (A_906): New define.
15280 (A_908): New define.
15281 (ASM_SPEC): Use A_FIJI, A_900, A_906 and A_908.
15282 * config/gcn/gcn.c (output_file_start): Adjust attributes according
15283 to the assembler capabilities.
15284 * config/gcn/mkoffload.c (main): Likewise.
15285 * configure: Regenerate.
15286 * configure.ac: Add tests for LLVM assembler attribute features.
15288 2021-07-28 Andrew MacLeod <amacleod@redhat.com>
15290 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Check for
15291 cond_false and cond_true on branches.
15293 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
15295 * config/aarch64/aarch64.c (aarch64_gen_adjusted_ldpstp): use
15298 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
15300 * alias.c (init_alias_analysis): Don't skip prologue/epilogue.
15302 2021-07-28 Jakub Jelinek <jakub@redhat.com>
15305 * config/i386/sse.md (vashr<mode>3): Split into vashrv8di3 expander
15306 and vashrv4di3 expander, where the latter requires just TARGET_AVX2
15307 and has special !TARGET_AVX512VL expansion.
15308 (vashrv2di3<mask_name>): Rename to ...
15309 (vashrv2di3): ... this. Change condition to TARGET_XOP || TARGET_AVX2
15310 and add special !TARGET_XOP && !TARGET_AVX512VL expansion.
15312 2021-07-28 Martin Uecker <muecker@gwdg.de>
15314 * calls.c (maybe_warn_rdwr_sizes): Correct argument
15315 numbers in warning that were switched.
15317 2021-07-28 Kewen Lin <linkw@linux.ibm.com>
15319 PR tree-optimization/101596
15320 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Fix wrong check
15321 by using new_type's precision instead.
15323 2021-07-28 liuhongt <hongtao.liu@intel.com>
15326 * config/i386/i386.h (processor_costs): Add new member
15328 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
15329 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
15330 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
15331 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
15332 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
15333 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
15334 generic_cost, core_cost): Initialize integer_to_sse same value
15336 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
15337 * config/i386/i386.c (ix86_builtin_vectorization_cost):
15338 Use integer_to_sse instead of sse_op to calculate the cost of
15341 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
15343 * config/rs6000/rs6000-gen-builtins.c (write_ovld_static_init): New
15345 (write_init_file): Call write_ovld_static_init.
15347 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
15349 * config/rs6000/rs6000-gen-builtins.c (write_bif_static_init): New
15351 (write_init_file): Call write_bif_static_init.
15353 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
15355 * config/rs6000/rs6000-gen-builtins.c (typemap): New struct.
15356 (TYPE_MAP_SIZE): New macro.
15357 (type_map): New initialized variable.
15358 (typemap_cmp): New function.
15359 (write_type_node): Likewise.
15360 (write_fntype_init): Implement.
15362 2021-07-27 Martin Sebor <msebor@redhat.com>
15364 PR tree-optimization/101584
15365 * tree-ssa-uninit.c (builtin_call_nomodifying_p): New function.
15366 (check_defs): Call it.
15368 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
15370 * tree-ssa-dom.c (dom_jump_threader_simplifier):
15371 Put avail_exprs_stack in the class, instead of passing it to
15372 jump_threader_simplifier.
15373 (dom_jump_threader_simplifier::simplify): Add state argument.
15374 (dom_opt_dom_walker): Add state.
15375 (pass_dominator::execute): Pass state to threader.
15376 (dom_opt_dom_walker::before_dom_children): Use state.
15377 * tree-ssa-threadedge.c (jump_threader::jump_threader): Replace
15378 arguments by state.
15379 (jump_threader::record_temporary_equivalences_from_phis):
15380 Register equivalences through the state variable.
15381 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
15382 Record ranges in a statement through the state variable.
15383 (jump_threader::simplify_control_stmt_condition): Pass state to
15385 (jump_threader::simplify_control_stmt_condition_1): Same.
15386 (jump_threader::thread_around_empty_blocks): Remove obsolete
15388 (jump_threader::thread_through_normal_block): Record equivalences
15389 on edge through the state variable.
15390 (jump_threader::thread_across_edge): Abstract state pushing.
15391 (jt_state::jt_state): New.
15392 (jt_state::push): New.
15393 (jt_state::pop): New.
15394 (jt_state::register_equiv): New.
15395 (jt_state::record_ranges_from_stmt): New.
15396 (jt_state::register_equivs_on_edge): New.
15397 (jump_threader_simplifier::jump_threader_simplifier): Move from
15399 (jump_threader_simplifier::simplify): Add state argument.
15400 * tree-ssa-threadedge.h (class jt_state): New.
15401 (class jump_threader): Add state to constructor.
15402 (class jump_threader_simplifier): Add state to simplify. Remove
15403 avail_exprs_stack from class.
15404 * tree-vrp.c (vrp_jump_threader_simplifier::simplify): Add state
15406 (vrp_jump_threader::vrp_jump_threader): Add state.
15407 (vrp_jump_threader::~vrp_jump_threader): Cleanup state.
15409 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
15411 * Makefile.in (OBJS): Add gimple-range-path.o.
15412 * gimple-range-path.cc: New file.
15413 * gimple-range-path.h: New file.
15415 2021-07-27 Jonathan Wright <jonathan.wright@arm.com>
15417 * config/aarch64/aarch64-simd.md: Push sign/zero-extension
15418 inside vec_duplicate for all patterns.
15419 * simplify-rtx.c (simplify_context::simplify_unary_operation_1):
15420 Push sign/zero-extension inside vec_duplicate.
15422 2021-07-27 Richard Biener <rguenther@suse.de>
15424 PR tree-optimization/101573
15425 * tree-ssa-uninit.c (warn_uninit_phi_uses): New function
15426 looking at uninitialized PHI arg defs in some constrained cases.
15427 (warn_uninitialized_vars): Call it.
15428 (execute_early_warn_uninitialized): Calculate dominators.
15430 2021-07-27 Richard Biener <rguenther@suse.de>
15432 PR tree-optimization/39821
15433 * tree-vect-stmts.c (vect_model_promotion_demotion_cost): Use
15434 vector_stmt for widening arithmetic.
15435 (vectorizable_conversion): Adjust.
15437 2021-07-27 Martin Jambor <mjambor@suse.cz>
15439 * cgraph.h (ipa_replace_map): New field force_load_ref.
15440 * ipa-prop.h (ipa_param_descriptor): Reduce precision of move_cost,
15441 aded new flag load_dereferenced, adjusted comments.
15442 (ipa_get_param_dereferenced): New function.
15443 (ipa_set_param_dereferenced): Likewise.
15444 * cgraphclones.c (cgraph_node::create_virtual_clone): Follow it.
15445 * ipa-cp.c: Include gimple.h.
15446 (ipcp_discover_new_direct_edges): Take into account dereferenced flag.
15447 (get_replacement_map): New parameter force_load_ref, set the
15448 appropriate flag in ipa_replace_map if set.
15449 (struct symbol_and_index_together): New type.
15450 (adjust_refs_in_act_callers): New function.
15451 (adjust_references_in_caller): Likewise.
15452 (create_specialized_node): When appropriate, call
15453 adjust_references_in_caller and force only load references.
15454 * ipa-prop.c (load_from_dereferenced_name): New function.
15455 (ipa_analyze_controlled_uses): Also detect loads from a
15456 dereference, harden testing of call statements.
15457 (ipa_write_node_info): Stream the dereferenced flag.
15458 (ipa_read_node_info): Likewise.
15459 (ipa_set_jf_constant): Also create refdesc when jump function
15460 references a variable.
15461 (cgraph_node_for_jfunc): Rename to symtab_node_for_jfunc, work
15462 also on references of variables and return a symtab_node. Adjust
15464 (propagate_controlled_uses): Also remove references to VAR_DECLs.
15466 2021-07-27 Jakub Jelinek <jakub@redhat.com>
15468 PR middle-end/101586
15469 * gimple-fold.c (clear_padding_type): Ignore FIELD_DECLs with byte
15470 positions above or equal to sz except for diagnostics of flexible
15473 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
15475 PR tree-optimization/78888
15476 * gimple-range-fold.cc (get_letter_range): New.
15477 (fold_using_range::range_of_builtin_call): Call get_letter_range.
15479 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
15481 PR tree-optimization/78888
15482 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Add cases
15483 for CFN_BUILT_IN_TOUPPER and CFN_BUILT_IN_TOLOWER.
15485 2021-07-26 Roger Sayle <roger@nextmovesoftware.com>
15486 Marc Glisse <marc.glisse@inria.fr>
15488 * match.pd (rotate): Simplify equality/inequality of rotations.
15489 (bswap): Simplify equality/inequality tests of byte swapping.
15491 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
15493 * range-op.cc (operator_bitwise_xor::op1_op2_relation_effect):
15496 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
15498 * range-op.cc (operator_lshift::fold_range): Pass rel to
15499 base class fold_range.
15500 (operator_rshift::fold_range): Same.
15502 2021-07-26 Ashimida <ashimida@linux.alibaba.com>
15505 * toplev.h (min_align_loops_log): Remove declaration.
15506 (min_align_jumps_log, min_align_labels_log): Likewise.
15507 (min_align_functions_log): Likewise.
15509 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
15511 * tree-vrp.c (vrp_simplify_cond_using_ranges): Rename vr_values
15513 (execute_vrp): Abstract out simplification of conditionals...
15514 (simplify_casted_conds): ...here.
15516 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
15518 * gimple-array-bounds.cc (array_bounds_checker::get_value_range):
15519 Add gimple argument.
15520 (array_bounds_checker::check_array_ref): Same.
15521 (array_bounds_checker::check_addr_expr): Same.
15522 (array_bounds_checker::check_array_bounds): Pass statement to
15523 check_array_bounds and check_addr_expr.
15524 * gimple-array-bounds.h (check_array_bounds): Add gimple argument.
15525 (check_addr_expr): Same.
15526 (get_value_range): Same.
15528 2021-07-26 Tamar Christina <tamar.christina@arm.com>
15530 * config/aarch64/aarch64-simd-builtins.def (sdot, udot): Rename to..
15531 (sdot_prod, udot_prod): ... This.
15532 * config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>): Merged
15534 (<sur>dot_prod<vsi2qi>): ... this.
15535 (aarch64_<sur>dot_lane<vsi2qi>, aarch64_<sur>dot_laneq<vsi2qi>):
15536 Change operands order.
15537 (<sur>sadv16qi): Use new operands order.
15538 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32,
15539 vdotq_s32): Use new RTL ordering.
15541 2021-07-26 Tamar Christina <tamar.christina@arm.com>
15543 * config/aarch64/aarch64-builtins.c (TYPES_TERNOP_SUSS,
15544 aarch64_types_ternop_suss_qualifiers): New.
15545 * config/aarch64/aarch64-simd-builtins.def (usdot_prod): Use it.
15546 * config/aarch64/aarch64-simd.md (usdot_prod<vsi2qi>): Re-organize RTL.
15547 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Use it.
15549 2021-07-23 Jakub Jelinek <jakub@redhat.com>
15551 PR rtl-optimization/101562
15552 * expmed.c (store_integral_bit_field): Only use movstrict_optab
15553 if the operand isn't paradoxical.
15555 2021-07-23 Aldy Hernandez <aldyh@redhat.com>
15557 * gimple-array-bounds.h (class array_bounds_checker): Change
15558 ranges type to range_query.
15560 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
15562 * config/aarch64/arm_neon.h (vst1_s64_x2): Use
15563 __builtin_memcpy instead of constructing
15564 __builtin_aarch64_simd_oi one vector at a time.
15565 (vst1_u64_x2): Likewise.
15566 (vst1_f64_x2): Likewise.
15567 (vst1_s8_x2): Likewise.
15568 (vst1_p8_x2): Likewise.
15569 (vst1_s16_x2): Likewise.
15570 (vst1_p16_x2): Likewise.
15571 (vst1_s32_x2): Likewise.
15572 (vst1_u8_x2): Likewise.
15573 (vst1_u16_x2): Likewise.
15574 (vst1_u32_x2): Likewise.
15575 (vst1_f16_x2): Likewise.
15576 (vst1_f32_x2): Likewise.
15577 (vst1_p64_x2): Likewise.
15578 (vst1q_s8_x2): Likewise.
15579 (vst1q_p8_x2): Likewise.
15580 (vst1q_s16_x2): Likewise.
15581 (vst1q_p16_x2): Likewise.
15582 (vst1q_s32_x2): Likewise.
15583 (vst1q_s64_x2): Likewise.
15584 (vst1q_u8_x2): Likewise.
15585 (vst1q_u16_x2): Likewise.
15586 (vst1q_u32_x2): Likewise.
15587 (vst1q_u64_x2): Likewise.
15588 (vst1q_f16_x2): Likewise.
15589 (vst1q_f32_x2): Likewise.
15590 (vst1q_f64_x2): Likewise.
15591 (vst1q_p64_x2): Likewise.
15593 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
15595 * config/aarch64/arm_neon.h (vst1_s64_x3): Use
15596 __builtin_memcpy instead of constructing
15597 __builtin_aarch64_simd_ci one vector at a time.
15598 (vst1_u64_x3): Likewise.
15599 (vst1_f64_x3): Likewise.
15600 (vst1_s8_x3): Likewise.
15601 (vst1_p8_x3): Likewise.
15602 (vst1_s16_x3): Likewise.
15603 (vst1_p16_x3): Likewise.
15604 (vst1_s32_x3): Likewise.
15605 (vst1_u8_x3): Likewise.
15606 (vst1_u16_x3): Likewise.
15607 (vst1_u32_x3): Likewise.
15608 (vst1_f16_x3): Likewise.
15609 (vst1_f32_x3): Likewise.
15610 (vst1_p64_x3): Likewise.
15611 (vst1q_s8_x3): Likewise.
15612 (vst1q_p8_x3): Likewise.
15613 (vst1q_s16_x3): Likewise.
15614 (vst1q_p16_x3): Likewise.
15615 (vst1q_s32_x3): Likewise.
15616 (vst1q_s64_x3): Likewise.
15617 (vst1q_u8_x3): Likewise.
15618 (vst1q_u16_x3): Likewise.
15619 (vst1q_u32_x3): Likewise.
15620 (vst1q_u64_x3): Likewise.
15621 (vst1q_f16_x3): Likewise.
15622 (vst1q_f32_x3): Likewise.
15623 (vst1q_f64_x3): Likewise.
15624 (vst1q_p64_x3): Likewise.
15626 2021-07-23 H.J. Lu <hjl.tools@gmail.com>
15629 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): Don't return
15630 hard register when LRA is in progress.
15632 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
15634 * config/aarch64/arm_neon.h (vst1_s8_x4): Use
15635 __builtin_memcpy instead of using a union.
15636 (vst1q_s8_x4): Likewise.
15637 (vst1_s16_x4): Likewise.
15638 (vst1q_s16_x4): Likewise.
15639 (vst1_s32_x4): Likewise.
15640 (vst1q_s32_x4): Likewise.
15641 (vst1_u8_x4): Likewise.
15642 (vst1q_u8_x4): Likewise.
15643 (vst1_u16_x4): Likewise.
15644 (vst1q_u16_x4): Likewise.
15645 (vst1_u32_x4): Likewise.
15646 (vst1q_u32_x4): Likewise.
15647 (vst1_f16_x4): Likewise.
15648 (vst1q_f16_x4): Likewise.
15649 (vst1_f32_x4): Likewise.
15650 (vst1q_f32_x4): Likewise.
15651 (vst1_p8_x4): Likewise.
15652 (vst1q_p8_x4): Likewise.
15653 (vst1_p16_x4): Likewise.
15654 (vst1q_p16_x4): Likewise.
15655 (vst1_s64_x4): Likewise.
15656 (vst1_u64_x4): Likewise.
15657 (vst1_p64_x4): Likewise.
15658 (vst1q_s64_x4): Likewise.
15659 (vst1q_u64_x4): Likewise.
15660 (vst1q_p64_x4): Likewise.
15661 (vst1_f64_x4): Likewise.
15662 (vst1q_f64_x4): Likewise.
15664 2021-07-23 Jonathan Wrightt <jonathan.wright@arm.com>
15666 * config/aarch64/arm_neon.h (vst2_s64): Use __builtin_memcpy
15667 instead of constructing __builtin_aarch64_simd_oi one vector
15669 (vst2_u64): Likewise.
15670 (vst2_f64): Likewise.
15671 (vst2_s8): Likewise.
15672 (vst2_p8): Likewise.
15673 (vst2_s16): Likewise.
15674 (vst2_p16): Likewise.
15675 (vst2_s32): Likewise.
15676 (vst2_u8): Likewise.
15677 (vst2_u16): Likewise.
15678 (vst2_u32): Likewise.
15679 (vst2_f16): Likewise.
15680 (vst2_f32): Likewise.
15681 (vst2_p64): Likewise.
15682 (vst2q_s8): Likewise.
15683 (vst2q_p8): Likewise.
15684 (vst2q_s16): Likewise.
15685 (vst2q_p16): Likewise.
15686 (vst2q_s32): Likewise.
15687 (vst2q_s64): Likewise.
15688 (vst2q_u8): Likewise.
15689 (vst2q_u16): Likewise.
15690 (vst2q_u32): Likewise.
15691 (vst2q_u64): Likewise.
15692 (vst2q_f16): Likewise.
15693 (vst2q_f32): Likewise.
15694 (vst2q_f64): Likewise.
15695 (vst2q_p64): Likewise.
15697 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
15699 * config/aarch64/arm_neon.h (vst3_s64): Use __builtin_memcpy
15700 instead of constructing __builtin_aarch64_simd_ci one vector
15702 (vst3_u64): Likewise.
15703 (vst3_f64): Likewise.
15704 (vst3_s8): Likewise.
15705 (vst3_p8): Likewise.
15706 (vst3_s16): Likewise.
15707 (vst3_p16): Likewise.
15708 (vst3_s32): Likewise.
15709 (vst3_u8): Likewise.
15710 (vst3_u16): Likewise.
15711 (vst3_u32): Likewise.
15712 (vst3_f16): Likewise.
15713 (vst3_f32): Likewise.
15714 (vst3_p64): Likewise.
15715 (vst3q_s8): Likewise.
15716 (vst3q_p8): Likewise.
15717 (vst3q_s16): Likewise.
15718 (vst3q_p16): Likewise.
15719 (vst3q_s32): Likewise.
15720 (vst3q_s64): Likewise.
15721 (vst3q_u8): Likewise.
15722 (vst3q_u16): Likewise.
15723 (vst3q_u32): Likewise.
15724 (vst3q_u64): Likewise.
15725 (vst3q_f16): Likewise.
15726 (vst3q_f32): Likewise.
15727 (vst3q_f64): Likewise.
15728 (vst3q_p64): Likewise.
15730 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
15732 * config/aarch64/arm_neon.h (vst4_s64): Use __builtin_memcpy
15733 instead of constructing __builtin_aarch64_simd_xi one vector
15735 (vst4_u64): Likewise.
15736 (vst4_f64): Likewise.
15737 (vst4_s8): Likewise.
15738 (vst4_p8): Likewise.
15739 (vst4_s16): Likewise.
15740 (vst4_p16): Likewise.
15741 (vst4_s32): Likewise.
15742 (vst4_u8): Likewise.
15743 (vst4_u16): Likewise.
15744 (vst4_u32): Likewise.
15745 (vst4_f16): Likewise.
15746 (vst4_f32): Likewise.
15747 (vst4_p64): Likewise.
15748 (vst4q_s8): Likewise.
15749 (vst4q_p8): Likewise.
15750 (vst4q_s16): Likewise.
15751 (vst4q_p16): Likewise.
15752 (vst4q_s32): Likewise.
15753 (vst4q_s64): Likewise.
15754 (vst4q_u8): Likewise.
15755 (vst4q_u16): Likewise.
15756 (vst4q_u32): Likewise.
15757 (vst4q_u64): Likewise.
15758 (vst4q_f16): Likewise.
15759 (vst4q_f32): Likewise.
15760 (vst4q_f64): Likewise.
15761 (vst4q_p64): Likewise.
15763 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
15765 * config/aarch64/arm_neon.h (vtbx4_s8): Use __builtin_memcpy
15766 instead of constructing __builtin_aarch64_simd_oi one vector
15768 (vtbx4_u8): Likewise.
15769 (vtbx4_p8): Likewise.
15771 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
15773 * config/aarch64/arm_neon.h (vtbl3_s8): Use __builtin_memcpy
15774 instead of constructing __builtin_aarch64_simd_oi one vector
15776 (vtbl3_u8): Likewise.
15777 (vtbl3_p8): Likewise.
15778 (vtbl4_s8): Likewise.
15779 (vtbl4_u8): Likewise.
15780 (vtbl4_p8): Likewise.
15782 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
15784 * config/aarch64/arm_neon.h (vqtbx2_s8): Use __builtin_memcpy
15785 instead of constructing __builtin_aarch64_simd_oi one vector
15787 (vqtbx2_u8): Likewise.
15788 (vqtbx2_p8): Likewise.
15789 (vqtbx2q_s8): Likewise.
15790 (vqtbx2q_u8): Likewise.
15791 (vqtbx2q_p8): Likewise.
15792 (vqtbx3_s8): Use __builtin_memcpy instead of constructing
15793 __builtin_aarch64_simd_ci one vector at a time.
15794 (vqtbx3_u8): Likewise.
15795 (vqtbx3_p8): Likewise.
15796 (vqtbx3q_s8): Likewise.
15797 (vqtbx3q_u8): Likewise.
15798 (vqtbx3q_p8): Likewise.
15799 (vqtbx4_s8): Use __builtin_memcpy instead of constructing
15800 __builtin_aarch64_simd_xi one vector at a time.
15801 (vqtbx4_u8): Likewise.
15802 (vqtbx4_p8): Likewise.
15803 (vqtbx4q_s8): Likewise.
15804 (vqtbx4q_u8): Likewise.
15805 (vqtbx4q_p8): Likewise.
15807 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
15809 * config/aarch64/arm_neon.h (vqtbl2_s8): Use __builtin_memcpy
15810 instead of constructing __builtin_aarch64_simd_oi one vector
15812 (vqtbl2_u8): Likewise.
15813 (vqtbl2_p8): Likewise.
15814 (vqtbl2q_s8): Likewise.
15815 (vqtbl2q_u8): Likewise.
15816 (vqtbl2q_p8): Likewise.
15817 (vqtbl3_s8): Use __builtin_memcpy instead of constructing
15818 __builtin_aarch64_simd_ci one vector at a time.
15819 (vqtbl3_u8): Likewise.
15820 (vqtbl3_p8): Likewise.
15821 (vqtbl3q_s8): Likewise.
15822 (vqtbl3q_u8): Likewise.
15823 (vqtbl3q_p8): Likewise.
15824 (vqtbl4_s8): Use __builtin_memcpy instead of constructing
15825 __builtin_aarch64_simd_xi one vector at a time.
15826 (vqtbl4_u8): Likewise.
15827 (vqtbl4_p8): Likewise.
15828 (vqtbl4q_s8): Likewise.
15829 (vqtbl4q_u8): Likewise.
15830 (vqtbl4q_p8): Likewise.
15832 2021-07-23 Haochen Gui <guihaoc@gcc.gnu.org>
15835 * config/rs6000/rs6000.md (cstore<mode>4): Fix wrong fall through.
15837 2021-07-22 Andrew Pinski <apinski@marvell.com>
15839 PR tree-optimization/10153
15840 * tree-tailcall.c (create_tailcall_accumulator):
15841 Don't call fold_convert as the type should be correct already.
15842 (tree_optimize_tail_calls_1): Use build_{one,zero}_cst instead
15843 of integer_{one,zero}_node for the call of create_tailcall_accumulator.
15845 2021-07-22 Aldy Hernandez <aldyh@redhat.com>
15847 * gimple-range-cache.cc (non_null_ref::adjust_range): Replace
15848 varying_p check for null/non-null check.
15850 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
15852 PR tree-optimization/101511
15853 * value-relation.cc (relation_oracle::query_relation): Check if ssa1
15854 is in ssa2's equiv set, and don't trap if so.
15856 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
15858 PR tree-optimization/101497
15859 * gimple-range-fold.cc (fold_using_range::range_of_cond_expr): Check
15862 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
15864 PR tree-optimization/101496
15865 * vr-values.c (simplify_using_ranges::fold_cond): Call range_of_stmt
15866 first, then vrp_visit_cond_Stmt.
15868 2021-07-22 liuhongt <hongtao.liu@intel.com>
15870 * config/i386/i386-expand.c
15871 (ix86_broadcast_from_integer_constant): Rename to ..
15872 (ix86_broadcast_from_constant): .. this, and extend it to
15874 (ix86_expand_vector_move): Extend to float mode.
15875 * config/i386/i386-features.c
15876 (replace_constant_pool_with_broadcast): Remove.
15877 (remove_partial_avx_dependency_gate): Ditto.
15878 (constant_pool_broadcast): Ditto.
15879 (class pass_constant_pool_broadcast): Ditto.
15880 (make_pass_constant_pool_broadcast): Ditto.
15881 (remove_partial_avx_dependency): Adjust gate.
15882 * config/i386/i386-passes.def: Remove pass_constant_pool_broadcast.
15883 * config/i386/i386-protos.h
15884 (make_pass_constant_pool_broadcast): Remove.
15886 2021-07-22 liuhongt <hongtao.liu@intel.com>
15888 * config/i386/constraints.md (Wb): New constraint.
15890 * config/i386/i386.md (*ashlhi3_1): Extend to avx512 mask
15892 (*ashlqi3_1): Ditto.
15893 (*<insn><mode>3_1): Split to ..
15894 (*ashr<mode>3_1): this, ...
15895 (*lshr<mode>3_1): and this, also extend this pattern to avx512
15897 (*<insn><mode>3_1): Split to ..
15898 (*ashr<mode>3_1): this, ...
15899 (*lshrqi3_1): and this, also extend this pattern to avx512
15901 (*lshrhi3_1): And this, also extend this pattern to avx512
15903 * config/i386/sse.md (k<code><mode>): New define_split after
15904 it to convert generic shift pattern to mask shift ones.
15906 2021-07-21 Thomas Schwinge <thomas@codesourcery.com>
15907 Joseph Myers <joseph@codesourcery.com>
15908 Cesar Philippidis <cesar@codesourcery.com>
15910 * tree-core.h (omp_clause_code): Add 'OMP_CLAUSE_NOHOST'.
15911 * tree.c (omp_clause_num_ops, omp_clause_code_name, walk_tree_1):
15913 * tree-pretty-print.c (dump_omp_clause): Likewise.
15914 * omp-general.c (oacc_verify_routine_clauses): Likewise.
15915 * gimplify.c (gimplify_scan_omp_clauses)
15916 (gimplify_adjust_omp_clauses): Likewise.
15917 * tree-nested.c (convert_nonlocal_omp_clauses)
15918 (convert_local_omp_clauses): Likewise.
15919 * omp-low.c (scan_sharing_clauses): Likewise.
15920 * omp-offload.c (execute_oacc_device_lower): Update.
15922 2021-07-21 Martin Sebor <msebor@redhat.com>
15924 * tree-ssa-alias.c (walk_aliased_vdefs_1): Fix typos in a comment.
15926 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
15928 * config/rs6000/rs6000-gen-builtins.c (write_init_bif_table):
15931 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
15933 * config/rs6000/rs6000-gen-builtins.c (write_fntype): New
15935 (write_fntype_init): New stub function.
15936 (write_init_bif_table): Likewise.
15937 (write_init_ovld_table): New function.
15938 (write_init_file): Implement.
15940 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
15942 * config/rs6000/rs6000-gen-builtins.c
15943 (write_autogenerated_header): New function.
15944 (write_decls): Likewise.
15945 (write_extern_fntype): New callback function.
15946 (write_header_file): Implement.
15948 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
15950 * config/rs6000/rs6000-gen-builtins.c (write_defines_file):
15953 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
15955 * config/rs6000/rs6000-gen-builtins.c (complete_vector_type): New
15957 (complete_base_type): Likewise.
15958 (construct_fntype_id): Likewise.
15959 (parse_bif_entry): Call contruct_fntype_id.
15960 (parse_ovld_entry): Likewise.
15962 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
15964 * config/rs6000/rs6000-gen-builtins.c (ovld_stanza): New struct.
15965 (MAXOVLDSTANZAS): New macro.
15966 (ovld_stanzas): New variable.
15967 (curr_ovld_stanza): Likewise.
15968 (MAXOVLDS): New macro.
15969 (ovlddata): New struct.
15970 (ovlds): New variable.
15971 (curr_ovld): Likewise.
15972 (max_ovld_args): Likewise.
15973 (parse_ovld_entry): New function.
15974 (parse_ovld_stanza): Likewise.
15975 (parse_ovld): Implement.
15977 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
15979 * config/rs6000/rs6000-gen-builtins.c (parse_bif_attrs):
15982 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
15984 * config/rs6000/rs6000-gen-builtins.c (parse_args): New function.
15985 (parse_prototype): Implement.
15987 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
15989 * config/rs6000/rs6000-gen-builtins.c (bif_stanza): New enum.
15990 (curr_bif_stanza): New variable.
15991 (stanza_entry): New struct.
15992 (stanza_map): New initialized variable.
15993 (enable_string): Likewise.
15994 (fnkinds): New enum.
15995 (typelist): New struct.
15996 (attrinfo): Likewise.
15997 (MAXRESTROPNDS): New macro.
15998 (prototype): New struct.
15999 (MAXBIFS): New macro.
16000 (bifdata): New struct.
16001 (bifs): New variable.
16002 (curr_bif): Likewise.
16003 (bif_order): Likewise.
16004 (bif_index): Likewise.
16005 (fatal): New function.
16006 (stanza_name_to_stanza): Likewise.
16007 (parse_bif_attrs): New stub function.
16008 (parse_prototype): Likewise.
16009 (parse_bif_entry): New function.
16010 (parse_bif_stanza): Likewise.
16011 (parse_bif): Implement.
16012 (set_bif_order): New function.
16013 (create_bif_order): Implement.
16015 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
16017 * config/rs6000/rs6000-gen-builtins.c (rbtree.h): New #include.
16018 (num_bifs): New variable.
16019 (num_ovld_stanzas): Likewise.
16020 (num_ovlds): Likewise.
16021 (parse_codes): New enum.
16022 (bif_rbt): New variable.
16023 (ovld_rbt): Likewise.
16024 (fntype_rbt): Likewise.
16025 (bifo_rbt): Likewise.
16026 (parse_bif): New stub function.
16027 (create_bif_order): Likewise.
16028 (parse_ovld): Likewise.
16029 (write_header_file): Likewise.
16030 (write_init_file): Likewise.
16031 (write_defines_file): Likewise.
16032 (delete_output_files): New function.
16035 2021-07-21 H.J. Lu <hjl.tools@gmail.com>
16038 * config/i386/i386-builtin.def: Remove OPTION_MASK_ISA_SSE4_2
16039 from CRC32 _builtin functions.
16041 2021-07-21 Sebastian Huber <sebastian.huber@embedded-brains.de>
16043 * coverage.c (build_gcov_info_var_registration): Mark the object placed
16044 in the linker set as referenced so that it does not get optimized away.
16046 2021-07-21 Kito Cheng <kito.cheng@sifive.com>
16049 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
16051 * config.gcc (riscv*-*-*): Detect which python is available.
16053 2021-07-21 Jakub Jelinek <jakub@redhat.com>
16055 PR middle-end/101535
16056 * gimplify.c (omp_check_private): Properly skip ORT_TARGET_DATA
16057 contexts in which decl isn't privatized and for ORT_TARGET return
16058 false if decl is mapped.
16060 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
16062 * gimple-loop-jam.c: Include tree-ssa-sccvn.h.
16063 (tree_loop_unroll_and_jam): Run value-numbering on a loop that
16064 has been successfully unrolled.
16066 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
16068 * tree-ssa-loop-manip.c (determine_exit_conditions): Return a null
16069 exit condition if no tail loop is needed, and if the original exit
16070 condition should therefore be kept as-is.
16071 (tree_transform_and_unroll_loop): Handle that case here too.
16073 2021-07-21 Kewen Lin <linkw@linux.ibm.com>
16075 * tree-data-ref.c (free_dependence_relations): Adjust to pass vec
16077 (free_data_refs): Likewise.
16078 * tree-data-ref.h (free_dependence_relations): Likewise.
16079 (free_data_refs): Likewise.
16080 * tree-predcom.c (struct chain): Use auto_vec instead of vec for
16082 (struct component): Likewise.
16083 (pcom_worker::pcom_worker): Adjust for auto_vec and renaming changes.
16084 (pcom_worker::~pcom_worker): Likewise.
16085 (pcom_worker::release_chain): Adjust as auto_vec changes.
16086 (pcom_worker::loop): Rename to ...
16087 (pcom_worker::m_loop): ... this.
16088 (pcom_worker::datarefs): Rename to ...
16089 (pcom_worker::m_datarefs): ... this. Use auto_vec instead of vec.
16090 (pcom_worker::dependences): Rename to ...
16091 (pcom_worker::m_dependences): ... this. Use auto_vec instead of vec.
16092 (pcom_worker::chains): Rename to ...
16093 (pcom_worker::m_chains): ... this. Use auto_vec instead of vec.
16094 (pcom_worker::looparound_phis): Rename to ...
16095 (pcom_worker::m_looparound_phis): ... this. Use auto_vec instead of
16097 (pcom_worker::cache): Rename to ...
16098 (pcom_worker::m_cache): ... this. Use auto_vec instead of vec.
16099 (pcom_worker::release_chain): Adjust for auto_vec changes.
16100 (pcom_worker::release_chains): Adjust for auto_vec and renaming
16102 (release_component): Remove.
16103 (release_components): Adjust for release_component removal.
16104 (component_of): Adjust to use vec.
16105 (merge_comps): Likewise.
16106 (pcom_worker::aff_combination_dr_offset): Adjust for renaming changes.
16107 (pcom_worker::determine_offset): Likewise.
16108 (class comp_ptrs): Remove.
16109 (pcom_worker::split_data_refs_to_components): Adjust for renaming
16110 changes, for comp_ptrs removal with auto_vec.
16111 (pcom_worker::suitable_component_p): Adjust for renaming changes.
16112 (pcom_worker::filter_suitable_components): Adjust for release_component
16114 (pcom_worker::valid_initializer_p): Adjust for renaming changes.
16115 (pcom_worker::find_looparound_phi): Likewise.
16116 (pcom_worker::add_looparound_copies): Likewise.
16117 (pcom_worker::determine_roots_comp): Likewise.
16118 (pcom_worker::single_nonlooparound_use): Likewise.
16119 (pcom_worker::execute_pred_commoning_chain): Likewise.
16120 (pcom_worker::execute_pred_commoning): Likewise.
16121 (pcom_worker::try_combine_chains): Likewise.
16122 (pcom_worker::prepare_initializers_chain): Likewise.
16123 (pcom_worker::prepare_initializers): Likewise.
16124 (pcom_worker::prepare_finalizers_chain): Likewise.
16125 (pcom_worker::prepare_finalizers): Likewise.
16126 (pcom_worker::tree_predictive_commoning_loop): Likewise.
16128 2021-07-20 Martin Sebor <msebor@redhat.com>
16130 PR middle-end/101397
16131 * builtins.c (gimple_call_return_array): Add argument. Correct
16132 offsets for memchr, mempcpy, stpcpy, and stpncpy.
16133 (compute_objsize_r): Adjust offset computation for argument returning
16136 2021-07-20 Martin Sebor <msebor@redhat.com>
16138 PR middle-end/101300
16139 * tree-ssa-uninit.c (check_defs): Handle UBSAN built-ins.
16141 2021-07-20 Jeff Law <jlaw@localhost.localdomain>
16143 * function.c (assign_parm_setup_block): Use adjust_address instead
16144 of change_address to preserve MEM_EXPR and friends.
16146 2021-07-20 Martin Sebor <msebor@redhat.com>
16148 * cfgloop.h (single_likely_exit): Adjust by-value argument to
16149 by-const-reference.
16150 * cfgloopanal.c (single_likely_exit): Same.
16151 * cgraph.h (struct cgraph_node): Same.
16152 * cgraphclones.c (cgraph_node::create_virtual_clone): Same.
16153 * genautomata.c (merge_states): Same.
16154 * genextract.c (VEC_char_to_string): Same.
16155 * genmatch.c (dt_node::gen_kids_1): Same.
16156 (walk_captures): Adjust by-value argument to by-reference.
16157 * gimple-ssa-store-merging.c (check_no_overlap): Adjust by-value argument
16158 to by-const-reference.
16159 * gimple.c (gimple_build_call_vec): Same.
16160 (gimple_build_call_internal_vec): Same.
16161 (gimple_build_switch): Same.
16162 (sort_case_labels): Same.
16163 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
16165 * gimple.h (gimple_build_call_vec): Adjust by-value argument to
16166 by-const-reference.
16167 (gimple_build_call_internal_vec): Same.
16168 (gimple_build_switch): Same.
16169 (sort_case_labels): Same.
16170 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
16172 * haifa-sched.c (calc_priorities): Adjust by-value argument to
16173 by-const-reference.
16174 (sched_init_luids): Same.
16175 (haifa_init_h_i_d): Same.
16176 * ipa-cp.c (ipa_get_indirect_edge_target_1): Same.
16177 (adjust_callers_for_value_intersection): Adjust by-value argument to
16179 (find_more_scalar_values_for_callers_subset): Adjust by-value argument to
16180 by-const-reference.
16181 (find_more_contexts_for_caller_subset): Same.
16182 (find_aggregate_values_for_callers_subset): Same.
16183 (copy_useful_known_contexts): Same.
16184 * ipa-fnsummary.c (remap_edge_summaries): Same.
16185 (remap_freqcounting_predicate): Same.
16186 * ipa-inline.c (add_new_edges_to_heap): Adjust by-value argument to
16188 * ipa-predicate.c (predicate::remap_after_inlining): Adjust by-value argument
16189 to by-const-reference.
16190 * ipa-predicate.h (predicate::remap_after_inlining): Same.
16191 * ipa-prop.c (ipa_find_agg_cst_for_param): Same.
16192 * ipa-prop.h (ipa_find_agg_cst_for_param): Same.
16193 * ira-build.c (ira_loop_tree_body_rev_postorder): Same.
16194 * read-rtl.c (add_overload_instance): Same.
16195 * rtl.h (native_decode_rtx): Same.
16196 (native_decode_vector_rtx): Same.
16197 * sched-int.h (sched_init_luids): Same.
16198 (haifa_init_h_i_d): Same.
16199 * simplify-rtx.c (native_decode_vector_rtx): Same.
16200 (native_decode_rtx): Same.
16201 * tree-call-cdce.c (gen_shrink_wrap_conditions): Same.
16202 (shrink_wrap_one_built_in_call_with_conds): Same.
16203 (shrink_wrap_conditional_dead_built_in_calls): Same.
16204 * tree-data-ref.c (create_runtime_alias_checks): Same.
16205 (compute_all_dependences): Same.
16206 * tree-data-ref.h (compute_all_dependences): Same.
16207 (create_runtime_alias_checks): Same.
16208 (index_in_loop_nest): Same.
16209 * tree-if-conv.c (mask_exists): Same.
16210 * tree-loop-distribution.c (class loop_distribution): Same.
16211 (loop_distribution::create_rdg_vertices): Same.
16212 (dump_rdg_partitions): Same.
16213 (debug_rdg_partitions): Same.
16214 (partition_contains_all_rw): Same.
16215 (loop_distribution::distribute_loop): Same.
16216 * tree-parloops.c (oacc_entry_exit_ok_1): Same.
16217 (oacc_entry_exit_single_gang): Same.
16218 * tree-ssa-loop-im.c (hoist_memory_references): Same.
16219 (loop_suitable_for_sm): Same.
16220 * tree-ssa-loop-niter.c (bound_index): Same.
16221 * tree-ssa-reassoc.c (update_ops): Same.
16222 (swap_ops_for_binary_stmt): Same.
16223 (rewrite_expr_tree): Same.
16224 (rewrite_expr_tree_parallel): Same.
16225 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Same.
16226 * tree-ssa-sccvn.h (ao_ref_init_from_vn_reference): Same.
16227 * tree-ssa-structalias.c (process_all_all_constraints): Same.
16228 (make_constraints_to): Same.
16229 (handle_lhs_call): Same.
16230 (find_func_aliases_for_builtin_call): Same.
16231 (sort_fieldstack): Same.
16232 (check_for_overlaps): Same.
16233 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Same.
16234 (vect_create_cond_for_unequal_addrs): Same.
16235 (vect_create_cond_for_lower_bounds): Same.
16236 (vect_create_cond_for_alias_checks): Same.
16237 * tree-vect-slp-patterns.c (vect_validate_multiplication): Same.
16238 * tree-vect-slp.c (vect_analyze_slp_instance): Same.
16239 (vect_make_slp_decision): Same.
16240 (vect_slp_bbs): Same.
16241 (duplicate_and_interleave): Same.
16242 (vect_transform_slp_perm_load): Same.
16243 (vect_schedule_slp): Same.
16244 * tree-vectorizer.h (vect_transform_slp_perm_load): Same.
16245 (vect_schedule_slp): Same.
16246 (duplicate_and_interleave): Same.
16247 * tree.c (build_vector_from_ctor): Same.
16248 (build_vector): Same.
16249 (check_vector_cst): Same.
16250 (check_vector_cst_duplicate): Same.
16251 (check_vector_cst_fill): Same.
16252 (check_vector_cst_stepped): Same.
16253 * tree.h (build_vector_from_ctor): Same.
16255 2021-07-20 Jakub Jelinek <jakub@redhat.com>
16258 * config/rs6000/rs6000-protos.h (easy_altivec_constant): Change return
16259 type from bool to int.
16260 * config/rs6000/rs6000.c (vspltis_constant): Fix up handling the
16261 EASY_VECTOR_MSB case if either step or copies is not 1.
16262 (vspltis_shifted): Fix comment typo.
16263 (easy_altivec_constant): Change return type from bool to int, instead
16264 of returning true return byte size of the element mode that should be
16265 used to synthetize the constant.
16266 * config/rs6000/predicates.md (easy_vector_constant_msb): Require
16267 that vspltis_shifted is 0, handle the case where easy_altivec_constant
16268 assumes using different vector mode from CONST_VECTOR's mode.
16269 * config/rs6000/altivec.md (easy_vector_constant_msb splitter): Use
16270 easy_altivec_constant to determine mode in which -1 >> -1 should be
16271 performed, use rs6000_expand_vector_init instead of gen_vec_initv4sisi.
16273 2021-07-20 Richard Biener <rguenther@suse.de>
16276 * dwarf2out.h (dwarf_file_data): Add key member.
16277 * dwarf2out.c (dwarf_file_hasher::equal): Compare key.
16278 (dwarf_file_hasher::hash): Hash key.
16279 (lookup_filename): Remap the filename and store it in the
16280 filename member of dwarf_file_data when creating a new
16282 (file_name_acquire): Do not remap the filename again.
16283 (maybe_emit_file): Likewise.
16285 2021-07-20 Jonathan Wright <jonathan.wright@arm.com>
16287 * config/aarch64/aarch64-simd-builtins.def: Use two variant
16288 generators for all TBL/TBX intrinsics and rename to
16289 consistent forms: qtbl[1234] or qtbx[1234].
16290 * config/aarch64/aarch64-simd.md (aarch64_tbl1<mode>):
16292 (aarch64_qtbl1<mode>): This.
16293 (aarch64_tbx1<mode>): Rename to...
16294 (aarch64_qtbx1<mode>): This.
16295 (aarch64_tbl2v16qi): Delete.
16296 (aarch64_tbl3<mode>): Rename to...
16297 (aarch64_qtbl2<mode>): This.
16298 (aarch64_tbx4<mode>): Rename to...
16299 (aarch64_qtbx2<mode>): This.
16300 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_1): Use
16301 renamed qtbl1 and qtbl2 RTL patterns.
16302 * config/aarch64/arm_neon.h (vqtbl1_p8): Use renamed qtbl1
16304 (vqtbl1_s8): Likewise.
16305 (vqtbl1_u8): Likewise.
16306 (vqtbl1q_p8): Likewise.
16307 (vqtbl1q_s8): Likewise.
16308 (vqtbl1q_u8): Likewise.
16309 (vqtbx1_s8): Use renamed qtbx1 RTL pattern.
16310 (vqtbx1_u8): Likewise.
16311 (vqtbx1_p8): Likewise.
16312 (vqtbx1q_s8): Likewise.
16313 (vqtbx1q_u8): Likewise.
16314 (vqtbx1q_p8): Likewise.
16315 (vtbl1_s8): Use renamed qtbl1 RTL pattern.
16316 (vtbl1_u8): Likewise.
16317 (vtbl1_p8): Likewise.
16318 (vtbl2_s8): Likewise
16319 (vtbl2_u8): Likewise.
16320 (vtbl2_p8): Likewise.
16321 (vtbl3_s8): Use renamed qtbl2 RTL pattern.
16322 (vtbl3_u8): Likewise.
16323 (vtbl3_p8): Likewise.
16324 (vtbl4_s8): Likewise.
16325 (vtbl4_u8): Likewise.
16326 (vtbl4_p8): Likewise.
16327 (vtbx2_s8): Use renamed qtbx2 RTL pattern.
16328 (vtbx2_u8): Likewise.
16329 (vtbx2_p8): Likewise.
16330 (vqtbl2_s8): Use renamed qtbl2 RTL pattern.
16331 (vqtbl2_u8): Likewise.
16332 (vqtbl2_p8): Likewise.
16333 (vqtbl2q_s8): Likewise.
16334 (vqtbl2q_u8): Likewise.
16335 (vqtbl2q_p8): Likewise.
16336 (vqtbx2_s8): Use renamed qtbx2 RTL pattern.
16337 (vqtbx2_u8): Likewise.
16338 (vqtbx2_p8): Likewise.
16339 (vqtbx2q_s8): Likewise.
16340 (vqtbx2q_u8): Likewise.
16341 (vqtbx2q_p8): Likewise.
16342 (vtbx4_s8): Likewise.
16343 (vtbx4_u8): Likewise.
16344 (vtbx4_p8): Likewise.
16346 2021-07-20 Uroš Bizjak <ubizjak@gmail.com>
16349 * config/i386/sync.md (define_peephole2 atomic_storedi_fpu):
16351 (define_peephole2 atomic_loaddi_fpu): Ditto.
16353 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
16355 * config.gcc (riscv*-*-*): Detect which python is available.
16357 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
16359 * config/rs6000/vsx.md (mulhs_<mode>): Rename to...
16360 (smul<mode>3_highpart): ... this.
16361 (mulhu_<mode>): Rename to...
16362 (umul<mode>3_highpart): ... this.
16363 * config/rs6000/rs6000-builtin.def (MULHS_V2DI, MULHS_V4SI,
16364 MULHU_V2DI, MULHU_V4SI): Adjust.
16366 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
16368 PR tree-optimization/100696
16369 * internal-fn.c (first_commutative_argument): Add info for IFN_MULH.
16370 * internal-fn.def (IFN_MULH): New internal function.
16371 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Add support to
16372 recog normal multiply highpart as IFN_MULH.
16373 * config/i386/i386.c (ix86_add_stmt_cost): Adjust for combined
16376 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
16378 * config/elfos.h (CTF_DEBUGGING_INFO): New definition.
16379 (BTF_DEBUGGING_INFO): Likewise.
16380 * doc/tm.texi.in: Document the new macros.
16381 * doc/tm.texi: Regenerated.
16382 * toplev.c: Guard initialization of debug hooks.
16384 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
16386 * flags.h (ctf_debuginfo_p): New function declaration.
16387 * opts.c (ctf_debuginfo_p): New function definition.
16389 2021-07-19 Andrew Stubbs <ams@codesourcery.com>
16392 * config/gcn/gcn-hsa.h (DRIVER_SELF_SPECS): New.
16393 (ASM_SPEC): Set -mattr for xnack and sram-ecc.
16394 * config/gcn/gcn-opts.h (enum sram_ecc_type): New.
16395 * config/gcn/gcn-valu.md: Add a warning comment.
16396 * config/gcn/gcn.c (gcn_option_override): Add "sorry" for -mxnack.
16397 (output_file_start): Add xnack and sram-ecc state to ".amdgcn_target".
16398 * config/gcn/gcn.md: Add a warning comment.
16399 * config/gcn/gcn.opt: Add -mxnack and -msram-ecc.
16400 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): Remove
16402 (EF_AMDGPU_XNACK): New.
16403 (EF_AMDGPU_SRAM_ECC): New.
16405 (copy_early_debug_info): Use elf_flags.
16406 (main): Handle -mxnack and -msram-ecc options.
16407 * doc/invoke.texi: Document -mxnack and -msram-ecc.
16409 2021-07-19 Andrew Pinski <apinski@marvell.com>
16412 * config/aarch64/aarch64.md (csneg3_uxtw_insn): Rename to ...
16413 (*cs<neg_not_cs>3_uxtw_insn4): and extend to NEG_NOT.
16415 2021-07-19 Richard Biener <rguenther@suse.de>
16417 PR tree-optimization/101505
16418 * tree-vect-patterns.c (vect_determine_precisions): Walk
16419 PHIs also for loop vectorization.
16421 2021-07-19 Richard Biener <rguenther@suse.de>
16423 * gimple.h (gimple_expr_type): Remove.
16424 * doc/gimple.texi: Remove gimple_expr_type documentation.
16426 2021-07-19 Richard Biener <rguenther@suse.de>
16428 * tree-ssa-sccvn.c (vn_reference_eq): Handle NULL vr->type.
16429 (ao_ref_init_from_vn_reference): Likewise.
16430 (fully_constant_reference): Likewise.
16431 (vn_reference_lookup_call): Do not set vr->type to random
16433 * tree-ssa-pre.c (compute_avail): Do not try to PRE calls
16435 * tree-vect-generic.c (expand_vector_piecewise): Pass in
16436 whether we expanded parallel.
16437 (expand_vector_parallel): Adjust.
16438 (expand_vector_addition): Likewise.
16439 (expand_vector_comparison): Likewise.
16440 (expand_vector_operation): Likewise.
16441 (expand_vector_scalar_condition): Likewise.
16442 (expand_vector_conversion): Likewise.
16444 2021-07-19 Richard Biener <rguenther@suse.de>
16446 * tree-vrp.c (register_edge_assert_for_2): Use the
16448 (vrp_folder::fold_predicate_in): Likewise.
16449 * vr-values.c (gimple_assign_nonzero_p): Likewise.
16450 (vr_values::extract_range_from_comparison): Likewise.
16451 (vr_values::extract_range_from_ubsan_builtin): Use the
16452 type of the first operand.
16453 (vr_values::extract_range_basic): Push down type
16454 computation, use the appropriate LHS.
16455 (vr_values::extract_range_from_assignment): Use the
16458 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
16461 * common/config/i386/i386-common.c (ix86_handle_option): For
16462 -mgeneral-regs-only, enable the GPR only instructions which are
16463 enabled implicitly by SSE ISAs unless they have been disabled
16466 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
16469 * config/i386/i386.c (ix86_check_avx_upper_stores): Moved before
16470 ix86_avx_u128_mode_needed.
16471 (ix86_avx_u128_mode_needed): Return AVX_U128_DIRTY if callee
16472 returns AVX register.
16474 2021-07-17 Jan Hubicka <hubicka@ucw.cz>
16476 * tree-ssa-structalias.c (handle_rhs_call): Support EAF_NOT_RETURNED.
16477 (handle_const_call): Liekise
16478 (handle_pure_call): Liekise
16480 2021-07-17 Andrew MacLeod <amacleod@redhat.com>
16482 PR tree-optimization/96542
16483 * range-op.cc (range_operator::wi_fold_in_parts): New.
16484 (range_operator::fold_range): Call wi_fold_in_parts.
16485 (operator_lshift::wi_fold): Fix broken lshift by [0,0].
16486 * range-op.h (wi_fold_in_parts): Add prototype.
16488 2021-07-16 David Malcolm <dmalcolm@redhat.com>
16490 * doc/analyzer.texi: Add __analyzer_dump_state.
16492 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
16494 * config/rs6000/rbtree.c: New file.
16495 * config/rs6000/rbtree.h: New file.
16497 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
16499 * config/rs6000/rs6000-gen-builtins.c (restriction): New enum.
16500 (typeinfo): Add restr field.
16501 (match_bracketed_pair): New function.
16502 (match_const_restriction): Implement.
16504 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
16506 * config/rs6000/rs6000-gen-builtins.c (match_basetype): Implement.
16508 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
16510 * config/rs6000/rs6000-gen-builtins.c (void_status): New enum.
16511 (basetype): Likewise.
16512 (typeinfo): Likewise.
16513 (handle_pointer): New function.
16514 (match_basetype): New stub function.
16515 (match_const_restriction): Likewise.
16516 (match_type): New function.
16518 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
16520 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace): New
16522 (advance_line): Likewise.
16523 (safe_inc_pos): Likewise.
16524 (match_identifier): Likewise.
16525 (match_integer): Likewise.
16526 (match_to_right_bracket): Likewise.
16528 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
16530 * config/rs6000/rs6000-gen-builtins.c (bif_file): New variable.
16531 (ovld_file): Likewise.
16532 (header_file): Likewise.
16533 (init_file): Likewise.
16534 (defines_file): Likewise.
16535 (pgm_path): Likewise.
16536 (bif_path): Likewise.
16537 (ovld_path): Likewise.
16538 (header_path): Likewise.
16539 (init_path): Likewise.
16540 (defines_path): Likewise.
16541 (LINELEN): New macro.
16542 (linebuf): New variable.
16546 (bif_diag): New function.
16547 (ovld_diag): Likewise.
16549 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
16551 * config/rs6000/rs6000-builtin-new.def: New.
16552 * config/rs6000/rs6000-overload.def: New.
16554 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
16556 * config/rs6000/rs6000-gen-builtins.c: New.
16558 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
16560 * Makefile.in (EXTRA_GTYPE_DEPS): New variable.
16561 (s-gtype): Depend on EXTRA_GTYPE_DEPS.
16562 * gengtype-state.c (state_writer::write_state_file_list): Add a
16563 parameter to the fileslist expression for the number of build
16565 (read_state_files_list): Detect build headers and strip the
16566 initial "./" or ".\" from their names.
16567 * gengtype.c (build_headers): New global variable.
16568 (num_build_headers): Likewise.
16569 (open_base_files): Emit #include for each build header.
16570 (main): Detect and count build headers.
16571 * gengtype.h (build_headers): New extern variable.
16572 (num_build_headers): Likewise.
16574 2021-07-16 Richard Biener <rguenther@suse.de>
16576 * gimple-ssa-store-merging.c (verify_symbolic_number_p): Use
16577 the type of the LHS.
16578 (find_bswap_or_nop_1): Likewise.
16579 (find_bswap_or_nop): Likewise.
16580 * tree-vectorizer.h (vect_get_smallest_scalar_type): Adjust
16582 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
16583 Remove unused parameters, pass in the scalar type. Fix
16584 internal store function handling.
16585 * tree-vect-stmts.c (vect_analyze_stmt): Remove assert.
16586 (vect_get_vector_types_for_stmt): Move down check for
16587 existing vector stmt after we've determined a scalar type.
16588 Pass down the used scalar type to vect_get_smallest_scalar_type.
16589 * tree-vect-generic.c (expand_vector_condition): Use
16590 the type of the LHS.
16591 (expand_vector_scalar_condition): Likewise.
16592 (expand_vector_operations_1): Likewise.
16593 * tree-vect-patterns.c (vect_widened_op_tree): Likewise.
16594 (vect_recog_dot_prod_pattern): Likewise.
16595 (vect_recog_sad_pattern): Likewise.
16596 (vect_recog_widen_op_pattern): Likewise.
16597 (vect_recog_widen_sum_pattern): Likewise.
16598 (vect_recog_mixed_size_cond_pattern): Likewise.
16600 2021-07-16 Jan Hubicka <hubicka@ucw.cz>
16602 * ipa-modref.c (struct escape_entry): Use eaf_fleags_t.
16603 (dump_eaf_flags): Dump EAF_NOT_RETURNED
16604 (eaf_flags_useful_p): Use eaf_fleags_t; handle const functions
16605 and EAF_NOT_RETURNED.
16606 (modref_summary::useful_p): Likewise.
16607 (modref_summary_lto::useful_p): Likewise.
16608 (struct) modref_summary_lto: Use eaf_fleags_t.
16609 (deref_flags): Handle EAF_NOT_RETURNED.
16610 (struct escape_point): Use min_flags.
16611 (modref_lattice::init): Add EAF_NOT_RETURNED.
16612 (merge_call_lhs_flags): Ignore EAF_NOT_RETURNED functions
16613 (analyze_ssa_name_flags): Clear EAF_NOT_RETURNED on return;
16615 (analyze_parms): Also analyze const functions; update conition on
16617 (modref_write): Update streaming.
16618 (read_section): Update streaming.
16619 (remap_arg_flags): Use eaf_flags_t.
16620 (modref_merge_call_site_flags): Hanlde EAF_NOT_RETURNED.
16621 * ipa-modref.h: (eaf_flags_t): New typedef.
16622 (struct modref_summary): Use eaf_flags_t.
16623 * tree-core.h (EAF_NOT_RETURNED): New constant.
16625 2021-07-16 Richard Biener <rguenther@suse.de>
16627 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Use
16628 the type of the LHS.
16629 (gimple_assign_nonnegative_warnv_p): Likewise.
16630 (gimple_call_nonnegative_warnv_p): Likewise. Return false
16631 if the call has no LHS.
16632 * gimple.c (gimple_could_trap_p_1): Use the type of the LHS.
16633 * tree-eh.c (stmt_could_throw_1_p): Likewise.
16634 * tree-inline.c (insert_init_stmt): Likewise.
16635 * tree-ssa-loop-niter.c (get_val_for): Likewise.
16636 * tree-outof-ssa.c (ssa_is_replaceable_p): Use the type of
16638 * tree-ssa-sccvn.c (init_vn_nary_op_from_stmt): Take a
16639 gassign *. Use the type of the lhs.
16640 (vn_nary_op_lookup_stmt): Adjust.
16641 (vn_nary_op_insert_stmt): Likewise.
16643 2021-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
16645 * config/s390/predicates.md (bras_sym_operand): Accept all
16646 functions in 64-bit mode, use UNSPEC_PLT31.
16647 (larl_operand): Use UNSPEC_PLT31.
16648 * config/s390/s390.c (s390_loadrelative_operand_p): Likewise.
16649 (legitimize_pic_address): Likewise.
16650 (s390_emit_tls_call_insn): Mark __tls_get_offset as function,
16652 (s390_delegitimize_address): Use UNSPEC_PLT31.
16653 (s390_output_addr_const_extra): Likewise.
16654 (print_operand): Add @PLT to TLS calls, handle %K.
16655 (s390_function_profiler): Mark __fentry__/_mcount as function,
16656 use %K, use UNSPEC_PLT31.
16657 (s390_output_mi_thunk): Use only UNSPEC_GOT, use %K.
16658 (s390_emit_call): Use UNSPEC_PLT31.
16659 (s390_emit_tpf_eh_return): Mark __tpf_eh_return as function.
16660 * config/s390/s390.md (UNSPEC_PLT31): Rename from UNSPEC_PLT.
16661 (*movdi_64): Use %K.
16662 (reload_base_64): Likewise.
16663 (*sibcall_brc): Likewise.
16664 (*sibcall_brcl): Likewise.
16665 (*sibcall_value_brc): Likewise.
16666 (*sibcall_value_brcl): Likewise.
16668 (*brasl): Likewise.
16669 (*bras_r): Likewise.
16670 (*brasl_r): Likewise.
16671 (*bras_tls): Likewise.
16672 (*brasl_tls): Likewise.
16673 (main_base_64): Likewise.
16674 (reload_base_64): Likewise.
16675 (@split_stack_call<mode>): Likewise.
16677 2021-07-16 Richard Biener <rguenther@suse.de>
16679 PR tree-optimization/101467
16680 * tree-vect-stmts.c (vect_gen_while): Properly guard
16681 make_temp_ssa_name usage.
16683 2021-07-16 Cooper Qu <cooper.qu@linux.alibaba.com>
16685 * config.gcc: Don't use forked print-sysroot-suffix.sh and
16686 t-sysroot-suffix for C-SKY.
16687 * config/csky/print-sysroot-suffix.sh: Delete.
16688 * config/csky/t-csky-linux: Delete.
16689 * config/csky/t-sysroot-suffix: Define MULTILIB_DIRNAMES
16690 instead of CSKY_MULTILIB_DIRNAMES.
16692 2021-07-16 Richard Biener <rguenther@suse.de>
16694 * tree-vect-loop.c (vect_transform_cycle_phi): Correct sign
16695 conversion issues with the partial reduction of the reused
16696 vector accumulator.
16698 2021-07-16 Richard Biener <rguenther@suse.de>
16700 * config/i386/i386-options.c (ix86_option_override_internal): Set
16701 param_vect_partial_vector_usage to zero if not set.
16703 2021-07-15 Uroš Bizjak <ubizjak@gmail.com>
16706 * config/i386/i386.h (VALID_SSE_REG_MODE): Add TDmode.
16707 (VALID_INT_MODE_P): Add SDmode and DDmode.
16708 Add TDmode for TARGET_64BIT.
16709 (VALID_DFP_MODE_P): Remove.
16710 * config/i386/i386.c (ix86_hard_regno_mode_ok):
16711 Do not use VALID_DFP_MODE_P.
16713 2021-07-15 Andrew MacLeod <amacleod@redhat.com>
16715 * gimple-range-fold.cc (adjust_pointer_diff_expr): Use
16717 (fold_using_range::fold_stmt): Ditto.
16718 (fold_using_range::range_of_range_op): Ditto.
16719 (fold_using_range::range_of_phi): Ditto.
16720 (fold_using_range::range_of_call): Ditto.
16721 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
16722 (fold_using_range::range_of_builtin_call): Ditto.
16723 (fold_using_range::range_of_cond_expr): Ditto.
16724 * gimple-range-fold.h (gimple_range_type): New.
16726 2021-07-15 Martin Sebor <msebor@redhat.com>
16728 PR middle-end/97027
16729 * tree-ssa-strlen.c (handle_assign): New function.
16730 (maybe_warn_overflow): Add argument.
16731 (nonzero_bytes_for_type): New function.
16732 (count_nonzero_bytes): Handle more tree types. Call
16733 nonzero_bytes_for_tye.
16734 (count_nonzero_bytes): Handle types.
16735 (handle_store): Handle stores from function calls.
16736 (strlen_check_and_optimize_call): Move code to handle_assign. Call
16737 it for assignments from function calls.
16739 2021-07-15 David Malcolm <dmalcolm@redhat.com>
16744 * doc/invoke.texi: Add -Wanalyzer-use-of-uninitialized-value.
16746 2021-07-15 David Malcolm <dmalcolm@redhat.com>
16748 * doc/invoke.texi (-fdump-analyzer-exploded-paths): New.
16750 2021-07-15 Martin Sebor <msebor@redhat.com>
16754 * fold-const.c (operand_compare::operand_equal_p): Handle OEP_DECL_NAME.
16755 (operand_compare::verify_hash_value): Same.
16756 * tree-core.h (OEP_DECL_NAME): New.
16758 2021-07-15 Martin Jambor <mjambor@suse.cz>
16760 * profile-count.h (profile_count::value): Change the return type to
16762 * gimple-pretty-print.c (dump_gimple_bb_header): Adjust print
16764 * tree-cfg.c (dump_function_to_file): Likewise.
16766 2021-07-15 Bill Schmidt <wschmidt@linux.ibm.com>
16769 * config/rs6000/rs6000-p8swap.c (has_part_mult): New.
16770 (rs6000_analyze_swaps): Insns containing a subreg of a mult are
16773 2021-07-15 Richard Biener <rguenther@suse.de>
16775 * tree-vectorizer.h (vect_gen_while): Match up with
16776 vect_gen_while_not.
16777 * tree-vect-stmts.c (vect_gen_while): Adjust API to that
16778 of vect_gen_while_not.
16779 (vect_gen_while_not): Adjust.
16780 * tree-vect-loop-manip.c (vect_set_loop_controls_directly): Likewise.
16782 2021-07-15 Aldy Hernandez <aldyh@redhat.com>
16784 * gimple-range-cache.cc (non_null_ref::adjust_range): New.
16785 (ranger_cache::range_of_def): Call adjust_range.
16786 (ranger_cache::entry_range): Same.
16787 * gimple-range-cache.h (non_null_ref::adjust_range): New.
16788 * gimple-range.cc (gimple_ranger::range_of_expr): Call
16790 (gimple_ranger::range_on_entry): Same.
16792 2021-07-15 Tamar Christina <tamar.christina@arm.com>
16795 2021-07-14 Tamar Christina <tamar.christina@arm.com>
16797 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
16799 2021-07-15 Tamar Christina <tamar.christina@arm.com>
16802 2021-07-14 Tamar Christina <tamar.christina@arm.com>
16804 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
16805 (sdot_prod, udot_prod): ...These.
16806 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
16807 (aarch64_<sur>dot<vsi2qi>): Rename to...
16808 (<sur>dot_prod<vsi2qi>): ...This.
16809 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
16812 2021-07-15 Jakub Jelinek <jakub@redhat.com>
16814 PR middle-end/101437
16815 * gimplify.c (gimplify_expr): Throw away volatile reads from empty
16816 types even if they have non-BLKmode TYPE_MODE.
16818 2021-07-15 Richard Biener <rguenther@suse.de>
16821 * gcc.c (process_command): Process -gtoggle like process_options
16822 would after parsing options.
16824 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
16826 * cfgexpand.c (expand_asm_loc): Adjust.
16827 (expand_asm_stmt): Likewise.
16828 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Likewise.
16829 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
16830 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
16831 * config/avr/avr.c (avr_md_asm_adjust): Likewise.
16832 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
16833 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
16834 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
16835 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
16836 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
16837 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
16838 * config/s390/s390.c (s390_md_asm_adjust): Likewise.
16839 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
16840 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
16841 * doc/tm.texi: Regenerate.
16842 * target.def: Add location argument to md_asm_adjust.
16844 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
16846 * tree-diagnostic.c (diagnostic_report_current_function): Use the
16847 diagnostic's location, not input_location.
16849 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
16851 * cfgexpand.c (tree_conflicts_with_clobbers_p): Pass location to
16853 (expand_asm_stmt): Likewise.
16855 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
16857 * config/rs6000/rs6000.c (adjacent_mem_locations): Return the lower
16858 addressed memory rtx, if any.
16859 (rs6000_split_multireg_move): Fix code formatting.
16860 Handle MMA build built-ins with operands in adjacent memory locations.
16862 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
16864 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Move to later
16867 2021-07-14 Jason Merrill <jason@redhat.com>
16869 * sel-sched-ir.h (get_all_loop_exits): Use auto_vec.
16871 2021-07-14 Jason Merrill <jason@redhat.com>
16873 * doc/invoke.texi: -fdelete-dead-exceptions is on by default for
16876 2021-07-14 Tamar Christina <tamar.christina@arm.com>
16878 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
16879 Remove erroneous line.
16881 2021-07-14 Andrew MacLeod <amacleod@redhat.com>
16883 * params.opt (param_evrp_mode): Change default.
16885 2021-07-14 Tamar Christina <tamar.christina@arm.com>
16887 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
16888 (sdot_prod, udot_prod): ...These.
16889 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
16890 (aarch64_<sur>dot<vsi2qi>): Rename to...
16891 (<sur>dot_prod<vsi2qi>): ...This.
16892 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
16895 2021-07-14 Tamar Christina <tamar.christina@arm.com>
16897 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
16899 2021-07-14 Tamar Christina <tamar.christina@arm.com>
16901 * doc/sourcebuild.texi (arm_v8_2a_i8mm_neon_hw): Document.
16903 2021-07-14 Tamar Christina <tamar.christina@arm.com>
16905 * config/arm/neon.md (usdot_prod<vsi2qi>): New.
16907 2021-07-14 Tamar Christina <tamar.christina@arm.com>
16909 * config/aarch64/aarch64-simd.md (aarch64_usdot<vsi2qi>): Rename to...
16910 (usdot_prod<vsi2qi>): ... This.
16911 * config/aarch64/aarch64-simd-builtins.def (usdot): Rename to...
16912 (usdot_prod): ...This.
16913 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Likewise.
16914 * config/aarch64/aarch64-sve.md (@aarch64_<sur>dot_prod<vsi2qi>):
16916 (@<sur>dot_prod<vsi2qi>): ...This.
16917 * config/aarch64/aarch64-sve-builtins-base.cc
16918 (svusdot_impl::expand): Use it.
16920 2021-07-14 Tamar Christina <tamar.christina@arm.com>
16922 * optabs.def (usdot_prod_optab): New.
16923 * doc/md.texi: Document it and clarify other dot prod optabs.
16924 * optabs-tree.h (enum optab_subtype): Add optab_vector_mixed_sign.
16925 * optabs-tree.c (optab_for_tree_code): Support usdot_prod_optab.
16926 * optabs.c (expand_widen_pattern_expr): Likewise.
16927 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
16928 * tree-vect-loop.c (vectorizable_reduction): Query dot-product kind.
16929 * tree-vect-patterns.c (vect_supportable_direct_optab_p): Take optional
16931 (vect_widened_op_tree): Optionally ignore
16933 (vect_recog_dot_prod_pattern): Support usdot_prod_optab.
16935 2021-07-14 H.J. Lu <hjl.tools@gmail.com>
16938 * config/i386/driver-i386.c (host_detect_local_cpu): Check
16939 "arch [32|64]" and "tune [32|64]" for 32-bit and 64-bit codegen.
16940 Enable UINTR only for 64-bit codegen.
16941 * config/i386/i386-options.c
16942 (ix86_option_override_internal::DEF_PTA): Skip PTA_UINTR if not
16944 * config/i386/i386.h (ARCH_ARG): New.
16945 (CC1_CPU_SPEC): Pass "[arch|tune] 32" for 32-bit codegen and
16946 "[arch|tune] 64" for 64-bit codegen.
16948 2021-07-14 Richard Biener <rguenther@suse.de>
16950 PR tree-optimization/101445
16951 * tree-vect-stmts.c (vectorizable_load): Do the gap adjustment
16952 of the IV in the correct direction for negative stride
16955 2021-07-14 Jakub Jelinek <jakub@redhat.com>
16958 * godump.c (godump_str_hash): New type.
16959 (godump_container::pot_dummy_types): Use string_hash instead of
16960 ptr_hash in the hash_set.
16962 2021-07-14 Richard Biener <rguenther@suse.de>
16964 * tree-vect-loop.c (vect_find_reusable_accumulator): Handle
16965 vector types where the old vector type has a multiple of
16966 the new vector type elements.
16967 (vect_create_partial_epilog): New function, split out from...
16968 (vect_create_epilog_for_reduction): ... here.
16969 (vect_transform_cycle_phi): Reduce the re-used accumulator
16970 to the new vector type.
16972 2021-07-14 Alexandre Oliva <oliva@adacore.com>
16974 * tree-ssa-alias.c (attr_fnspec::verify): Fix index in
16975 non-'t'-sized arg check.
16977 2021-07-14 Alexandre Oliva <oliva@adacore.com>
16979 * tree-cfg.c (cleanup_dead_labels_eh): Update
16980 post_landing_pad label upon change of landing pad block's
16982 (cleanup_dead_labels): Check that a removed label is not that
16985 2021-07-13 Jonathan Wright <jonathan.wright@arm.com>
16987 * combine.c (combine_simplify_rtx): Add vec_select -> subreg
16989 * config/aarch64/aarch64.md (*zero_extend<SHORT:mode><GPI:mode>2_aarch64):
16990 Add Neon to general purpose register case for zero-extend
16992 * config/arm/vfp.md (*arm_movsi_vfp): Remove "*" from *t -> r
16993 case to prevent some cases opting to go through memory.
16994 * cse.c (fold_rtx): Add vec_select -> subreg simplification.
16995 * rtl.c (rtvec_series_p): Define predicate to determine
16996 whether a vector contains a linear series of integers.
16997 * rtl.h (rtvec_series_p): Define.
16998 * rtlanal.c (vec_series_lowpart_p): Define predicate to
16999 determine if a vector selection is equivalent to the low part
17001 * rtlanal.h (vec_series_lowpart_p): Define.
17002 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
17003 Add vec_select -> subreg simplification.
17005 2021-07-13 Paul A. Clarke <pc@us.ibm.com>
17007 * config/rs6000/smmintrin.h (_mm_testz_si128, _mm_testc_si128,
17008 _mm_testnzc_si128, _mm_test_all_ones, _mm_test_all_zeros,
17009 _mm_test_mix_ones_zeros): New.
17011 2021-07-13 Roger Sayle <roger@nextmovesoftware.com>
17012 Richard Biener <rguenther@suse.de>
17014 * gimple.c (gimple_could_trap_p_1): Make S argument a
17015 "const gimple*". Preserve constness in call to
17016 gimple_asm_volatile_p.
17017 (gimple_could_trap_p): Make S argument a "const gimple*".
17018 * gimple.h (gimple_could_trap_p_1, gimple_could_trap_p):
17019 Update function prototypes.
17021 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
17023 * tree-vectorizer.h (vect_reusable_accumulator): New structure.
17024 (_loop_vec_info::main_loop_edge): New field.
17025 (_loop_vec_info::skip_main_loop_edge): Likewise.
17026 (_loop_vec_info::skip_this_loop_edge): Likewise.
17027 (_loop_vec_info::reusable_accumulators): Likewise.
17028 (_stmt_vec_info::reduc_scalar_results): Likewise.
17029 (_stmt_vec_info::reused_accumulator): Likewise.
17030 (vect_get_main_loop_result): Declare.
17031 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize
17032 reduc_scalar_inputs.
17033 (vec_info::free_stmt_vec_info): Free reduc_scalar_inputs.
17034 * tree-vect-loop-manip.c (vect_get_main_loop_result): New function.
17035 (vect_do_peeling): Fill an epilogue loop's main_loop_edge,
17036 skip_main_loop_edge and skip_this_loop_edge fields.
17037 * tree-vect-loop.c (INCLUDE_ALGORITHM): Define.
17038 (vect_emit_reduction_init_stmts): New function.
17039 (get_initial_def_for_reduction): Use it.
17040 (get_initial_defs_for_reduction): Likewise. Change the vinfo
17041 parameter to a loop_vec_info.
17042 (vect_create_epilog_for_reduction): Store the scalar results
17043 in the reduc_info. If an epilogue loop is reusing an accumulator
17044 from the main loop, and if the epilogue loop can also be skipped,
17045 try to place the reduction code in the join block. Record
17046 accumulators that could potentially be reused by epilogue loops.
17047 (vect_transform_cycle_phi): When vectorizing epilogue loops,
17048 try to reuse accumulators from the main loop. Record the initial
17049 value in reduc_info for non-SLP reductions too.
17051 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
17053 * tree-vect-loop.c (get_initial_def_for_reduction): Remove
17054 adjustment handling. Take the neutral value as an argument,
17055 in place of the code argument.
17056 (vect_transform_cycle_phi): Update accordingly. Handle the
17057 initial values of cond reductions separately from code reductions.
17058 Choose the adjustment here rather than in
17059 get_initial_def_for_reduction. Sink the splat of vec_initial_def.
17061 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
17063 * tree-vect-loop.c (neutral_op_for_slp_reduction): Replace with...
17064 (neutral_op_for_reduction): ...this, providing a more general
17066 (vect_create_epilog_for_reduction): Update accordingly.
17067 (vectorizable_reduction): Likewise.
17068 (vect_transform_cycle_phi): Likewise.
17070 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
17072 * tree-vect-loop.c (get_initial_def_for_reduction): Take the
17073 reduc_info instead of the original stmt_vec_info.
17074 (vect_transform_cycle_phi): Update accordingly.
17076 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
17078 * tree-vect-loop.c (get_initial_defs_for_reduction): Take the
17079 reduc_info as an additional parameter.
17080 (vect_transform_cycle_phi): Update accordingly.
17082 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
17084 * tree-vectorizer.h: Include tree-ssa-operands.h.
17085 (vect_phi_initial_value): New function.
17086 * tree-vect-loop.c (neutral_op_for_slp_reduction): Use it.
17087 (get_initial_defs_for_reduction, info_for_reduction): Likewise.
17088 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
17089 (vect_transform_cycle_phi, vectorizable_induction): Likewise.
17091 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
17093 * tree-vect-loop.c (vect_create_epilog_for_reduction): Convert
17094 the phi results to vectype after creating them. Remove later
17095 conversion code that thus becomes redundant.
17097 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
17099 * tree-vect-loop.c (vect_create_epilog_for_reduction): Replace
17100 the new_phis vector with a reduc_inputs vector. Combine handling
17101 of reduction chains and ncopies > 1.
17103 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
17105 * tree-vect-loop.c (vect_create_epilog_for_reduction): Truncate
17106 scalar_results to group_size elements after reducing down from
17107 N*group_size elements. Construct an array_slice of the live-out
17108 stmts and assert that there is one stmt per scalar result.
17110 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
17112 * tree-vect-loop.c (vect_create_epilog_for_reduction): Remove
17113 nested_in_vect_loop and use double_reduc everywhere. Remove dead
17114 assignment to "loop".
17116 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
17118 * internal-fn.c (vectorized_internal_fn_supported_p): Handle
17119 vector types first. For scalar types, consider both the preferred
17120 vector mode and the alternative vector modes.
17121 * optabs-query.c (can_vec_mask_load_store_p): Use the same
17122 structure as above, in particular using related_vector_mode
17123 for modes provided by autovectorize_vector_modes.
17125 2021-07-13 Jakub Jelinek <jakub@redhat.com>
17126 Richard Biener <rguenther@suse.de>
17128 PR tree-optimization/101419
17129 * tree-pass.h (PROP_objsz): Define.
17130 (make_pass_early_object_sizes): Declare.
17131 * passes.def (pass_all_early_optimizations): Rename pass_object_sizes
17132 there to pass_early_object_sizes, drop parameter.
17133 (pass_all_optimizations): Move pass_object_sizes right after pass_ccp,
17134 drop parameter, move pass_post_ipa_warn right after that.
17135 * tree-object-size.c (pass_object_sizes::execute): Rename to...
17136 (object_sizes_execute): ... this. Add insert_min_max_p argument.
17137 (pass_data_object_sizes): Move after object_sizes_execute.
17138 (pass_object_sizes): Likewise. In execute method call
17139 object_sizes_execute, drop set_pass_param method and insert_min_max_p
17140 non-static data member and its initializer in the ctor.
17141 (pass_data_early_object_sizes, pass_early_object_sizes,
17142 make_pass_early_object_sizes): New.
17143 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Use
17144 (cfun->curr_properties & PROP_objsz) instead of cfun->after_inlining.
17146 2021-07-13 Kito Cheng <kito.cheng@sifive.com>
17149 * config/riscv/constraints.md ("S"): Update description and remove
17151 * doc/md.texi (Machine Constraints): Document the 'S' constraints
17154 2021-07-13 Richard Biener <rguenther@suse.de>
17157 2021-07-12 Richard Biener <rguenther@suse.de>
17159 * tree-vect-slp.c (vect_slp_region): Show the number of
17160 SLP graph entries in the optimization message.
17162 2021-07-13 Michael Meissner <meissner@linux.ibm.com>
17164 * config/rs6000/altivec.md (xxspltiw_v4sf): Change local variable
17166 * config/rs6000/rs6000-protos.h (rs6000_const_f32_to_i32): Change
17167 return type to long.
17168 * config/rs6000/rs6000.c (rs6000_const_f32_to_i32): Change return
17171 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
17173 * gimple-range-fold.cc (fold_using_range::range_of_builtin_ubsan_call):
17174 Query relation between the 2 operands and use it.
17176 2021-07-12 Sergei Trofimovich <siarheit@google.com>
17178 * doc/cfg.texi: Fix s/ei_safe_safe/ei_safe_edge/ typo.
17180 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
17183 * config/i386/predicates.md (vec_setm_sse41_operand):
17184 Rename from vec_setm_operand.
17185 (vec_setm_avx2_operand): New predicate.
17186 * config/i386/sse.md (vec_set<V_128:mode>): Use V_128 mode iterator.
17187 Use vec_setm_sse41_operand as operand 2 predicate.
17188 (vec_set<V_256_512:mode): New expander.
17189 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_sse41_operand
17190 as operand 2 predicate.
17192 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
17194 PR tree-optimization/101335
17195 * range-op.cc (operator_cast::lhs_op1_relation): Delete.
17197 2021-07-12 Andrew Pinski <apinski@marvell.com>
17199 * tree-ssa-phiopt.c (match_simplify_replacement): Move
17200 insert of the sequence before the movement of the
17201 statement. Check if to see if the statement is used
17202 outside of the original phi to see if we should move it.
17204 2021-07-12 Richard Biener <rguenther@suse.de>
17206 * dump-context.h (debug_dump_context::debug_dump_context):
17207 Add FILE * parameter defaulted to stderr.
17208 * dumpfile.c (debug_dump_context::debug_dump_context): Adjust.
17209 * tree-vect-slp.c (dot_slp_tree): New functions.
17211 2021-07-12 Richard Biener <rguenther@suse.de>
17213 PR tree-optimization/101373
17214 * tree-ssa-pre.c (prune_clobbered_mems): Also prune trapping
17215 references when the BB may not return.
17216 (compute_avail): Pass in the function we're working on and
17217 replace cfun references with it. Externally throwing
17218 const calls also possibly terminate the function.
17219 (pass_pre::execute): Pass down the function we're working on.
17220 * gcse.c (compute_hash_table_work): Externally throwing
17221 const/pure calls also need record_last_mem_set_info.
17222 * postreload-gcse.c (record_opr_changes): Looping or externally
17223 throwing const/pure calls also need record_last_mem_set_info.
17225 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
17227 * recog.c (memory_address_addr_space_p): Change the type to bool.
17228 Return true/false instead of 1/0.
17229 (offsettable_memref_p): Ditto.
17230 (offsettable_nonstrict_memref_p): Ditto.
17231 (offsettable_address_addr_space_p): Ditto.
17232 Change the type of addressp indirect function to bool.
17233 * recog.h (memory_address_addr_space_p): Change the type to bool.
17234 (strict_memory_address_addr_space_p): Ditto.
17235 (offsettable_memref_p): Ditto.
17236 (offsettable_nonstrict_memref_p): Ditto.
17237 (offsettable_address_addr_space_p): Ditto.
17238 * reload.c (maybe_memory_address_addr_space_p): Ditto.
17239 (strict_memory_address_addr_space_p): Change the type to bool.
17240 Return true/false instead of 1/0.
17241 (maybe_memory_address_addr_space_p): Change the type to bool.
17243 2021-07-12 Richard Biener <rguenther@suse.de>
17245 * tree-vect-slp.c (vect_slp_region): Show the number of
17246 SLP graph entries in the optimization message.
17248 2021-07-12 Richard Biener <rguenther@suse.de>
17250 PR tree-optimization/101394
17251 * tree-ssa-pre.c (do_pre_regular_insertion): Avoid inserting
17252 copies from abnormals for a full redundancy.
17254 2021-07-12 Richard Biener <rguenther@suse.de>
17256 PR middle-end/101423
17257 * gimple.c (gimple_could_trap_p_1): Internal function calls
17259 * tree-eh.c (tree_could_trap_p): Likewise.
17261 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
17264 * config/arm/arm_neon.h (vmul_n_u32): Replace call to builtin with
17266 (vmulq_n_u32): Likewise.
17267 (vmul_n_f32): Gate __a * __b on __FAST_MATH__.
17268 (vmulq_n_f32): Likewise.
17269 (vmul_n_f16): Likewise.
17270 (vmulq_n_f16): Likewise.
17272 2021-07-12 Martin Liska <mliska@suse.cz>
17274 PR sanitizer/101425
17275 * gcc.c (check_offload_target_name): Call
17276 candidates_list_and_hint only if we have a candidate.
17278 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
17281 * config/arm/neon.md (vec_init): Move to ...
17282 * config/arm/vec-common.md (vec_init): ... here.
17283 Change the pattern's mode to VDQX and gate it on VALID_MVE_MODE.
17285 2021-07-12 Roger Sayle <roger@nextmovesoftware.com>
17287 PR tree-optimization/101403
17288 * match.pd ((T)bswap(X)>>C): Correctly handle cases where
17289 signedness of the shift is not the same as the signedness of
17290 the type extension.
17292 2021-07-09 Roger Sayle <roger@nextmovesoftware.com>
17293 Uroš Bizjak <ubizjak@gmail.com>
17295 * config/i386/i386.md (*divmodsi4_const): Optimize SImode
17296 divmod of a constant numerator with new define_insn_and_split.
17298 2021-07-09 Iain Sandoe <iain@sandoe.co.uk>
17301 * config/i386/i386-expand.c (ix86_expand_call): If a call is
17302 to a non-local-binding, or local but to a public symbol, then
17303 assume that it might be indirected via the lazy symbol binder.
17304 Mark R10 and R10 as clobbered in that case.
17306 2021-07-09 Eric Botcazou <ebotcazou@adacore.com>
17309 * gcc.c (ASM_DEBUG_DWARF_OPTION): Set again to --gdwarf2 in
17310 the case where HAVE_AS_WORKING_DWARF_N_FLAG is not defined
17311 and HAVE_LD_BROKEN_PE_DWARF5 is defined.
17313 2021-07-09 Uroš Bizjak <ubizjak@gmail.com>
17315 * config/i386/i386.md (*udivmodsi4_pow2_zext_1): Limit the
17316 log2 range of operands[3] to [1,31].
17317 (*udivmodsi4_pow2_zext_2): Ditto. Correct insn RTX pattern.
17319 2021-07-09 Sergei Trofimovich <siarheit@google.com>
17321 * doc/md.texi: Don't split @smallexample in multiple @groups.
17323 2021-07-09 Sergei Trofimovich <siarheit@google.com>
17325 * doc/md.texi: Add missing 'see' word.
17327 2021-07-09 Andrew Pinski <apinski@marvell.com>
17329 * tree-ssa-phiopt.c (phiopt_early_allow): Change arguments
17330 to take sequence and gimple_match_op. Accept the case where
17331 op is a SSA_NAME and one statement in the sequence.
17332 Also allow constants.
17333 (gimple_simplify_phiopt): Always pass a sequence to resimplify.
17334 Update call to phiopt_early_allow. Discard the sequence if not
17337 2021-07-09 Xi Ruoyao <xry111@mengyan1223.wang>
17342 * config/mips/mips.c (mips_const_insns): Use MSA_SUPPORTED_MODE_P
17343 instead of ISA_HAS_MSA.
17344 (mips_expand_vec_unpack): Likewise.
17345 (mips_expand_vector_init): Likewise.
17347 2021-07-09 Kewen Lin <linkw@linux.ibm.com>
17349 * config/rs6000/vsx.md (mods_<mode>): Rename to...
17350 (mod<mode>3): ... this.
17351 (modu_<mode>): Rename to...
17352 (umod<mode>3): ... this.
17353 * config/rs6000/rs6000-builtin.def (MODS_V2DI, MODS_V4SI, MODU_V2DI,
17354 MODU_V4SI): Adjust.
17356 2021-07-08 Jeff Law <jeffreyalaw@gmail.com>
17358 * config/h8300/shiftrotate.md (variable shifts): Expose condition
17359 code handling for the test before the loop.
17361 2021-07-08 Martin Jambor <mjambor@suse.cz>
17364 * ipa-sra.c (class isra_call_summary): New member
17365 m_before_any_store, initialize it in the constructor.
17366 (isra_call_summary::dump): Dump the new field.
17367 (ipa_sra_call_summaries::duplicate): Copy it.
17368 (process_scan_results): Set it.
17369 (isra_write_edge_summary): Stream it.
17370 (isra_read_edge_summary): Likewise.
17371 (param_splitting_across_edge): Only override
17372 safe_to_import_accesses if m_before_any_store is set.
17374 2021-07-08 Martin Sebor <msebor@redhat.com>
17376 PR bootstrap/101374
17377 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref):
17378 Use Object Size Type 0 instead of 1.
17380 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
17382 * tree-vect-loop.c (vectorizable_reduction): Remove always-true
17385 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
17387 * match.pd: Simplify an extend-operate-truncate sequence involving
17390 2021-07-08 Roger Sayle <roger@nextmovesoftware.com>
17391 Richard Biener <rguenther@suse.de>
17393 PR tree-optimization/40210
17394 * match.pd (bswap optimizations): Simplify (bswap(x)>>C1)&C2 as
17395 (x>>C3)&C2 when possible. Simplify bswap(x)>>C1 as ((T)x)>>C2
17396 when possible. Simplify bswap(x)&C1 as (x>>C2)&C1 when 0<=C1<=255.
17398 2021-07-08 Uroš Bizjak <ubizjak@gmail.com>
17401 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
17403 * config/i386/mmx.md (V_32): New mode iterator.
17404 (mov<V_32:mode>): Use V_32 mode iterator.
17405 (*mov<V_32:mode>_internal): Ditto.
17406 (*push<V_32:mode>2_rex64): Ditto.
17407 (*push<V_32:mode>2): Ditto.
17408 (movmisalign<V_32:mode>): Ditto.
17409 (mmx_<any_shiftrt:insn>v1si3): New insn pattern.
17410 (sse4_1_<any_extend:code>v2qiv2hi2): Ditto.
17411 (vec_unpacks_lo_v4qi): New expander.
17412 (vec_unpacks_hi_v4qi): Ditto.
17413 (vec_unpacku_lo_v4qi): Ditto.
17414 (vec_unpacku_hi_v4qi): Ditto.
17415 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add V1SImode.
17416 (VALID_INT_MODE_P): Ditto.
17418 2021-07-08 Michael Meissner <meissner@linux.ibm.com>
17421 * config/rs6000/rs6000.md (udivti3): New insn.
17422 (divti3): New insn.
17423 (umodti3): New insn.
17424 (modti3): New insn.
17426 2021-07-07 Martin Sebor <msebor@redhat.com>
17428 PR tree-optimization/100137
17429 PR tree-optimization/99121
17430 PR tree-optimization/97027
17431 * builtins.c (access_ref::access_ref): Also set offmax.
17432 (access_ref::offset_in_range): Define new function.
17433 (access_ref::add_offset): Set offmax.
17434 (access_ref::inform_access): Handle access_none.
17435 (handle_mem_ref): Clear ostype.
17436 (compute_objsize_r): Handle ASSERT_EXPR.
17437 * builtins.h (struct access_ref): Add offmax member.
17438 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Use
17439 compute_objsize() and simplify.
17441 2021-07-07 Peter Bergner <bergner@linux.ibm.com>
17443 * config/rs6000/rs6000-call.c (mma_init_builtins): Use VSX_BUILTIN_LXVP
17444 and VSX_BUILTIN_STXVP.
17446 2021-07-07 Martin Sebor <msebor@redhat.com>
17449 * config/aarch64/aarch64.c (aarch64_simd_lane_bounds): Remove
17450 a stray %K from error_at() missed in r12-2088.
17452 2021-07-07 Richard Biener <rguenther@suse.de>
17454 PR tree-optimization/99728
17455 * tree-ssa-loop-im.c (gather_mem_refs_stmt): Record
17457 (mem_refs_may_alias_p): Add assert we handled aggregate
17459 (sm_seq_valid_bb): Give up when running into aggregate copies.
17460 (ref_indep_loop_p): Handle aggregate copies as never
17461 being invariant themselves but allow other refs to be
17462 disambiguated against them.
17463 (can_sm_ref_p): Do not try to apply store-motion to aggregate
17466 2021-07-06 Indu Bhagat <indu.bhagat@oracle.com>
17469 * dwarf2ctf.c (ctf_get_AT_data_member_location): Multiply by 8 to get
17472 2021-07-06 Martin Sebor <msebor@redhat.com>
17474 * gimple-pretty-print.c (percent_G_format): Remove.
17475 * tree-diagnostic.c (default_tree_printer): Remove calls.
17476 * tree-pretty-print.c (percent_K_format): Remove.
17477 * tree-pretty-print.h (percent_K_format): Remove.
17479 2021-07-06 Martin Sebor <msebor@redhat.com>
17481 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_builtin):
17482 Remove %K and use error_at.
17483 (aarch64_expand_fcmla_builtin): Same.
17484 (aarch64_expand_builtin_tme): Same.
17485 (aarch64_expand_builtin_memtag): Same.
17486 * config/arm/arm-builtins.c (arm_expand_acle_builtin): Same.
17487 (arm_expand_builtin): Same.
17488 * config/arm/arm.c (bounds_check): Same.
17490 2021-07-06 Martin Sebor <msebor@redhat.com>
17492 * builtins.c (warn_string_no_nul): Remove %G.
17493 (maybe_warn_for_bound): Same.
17494 (warn_for_access): Same.
17495 (check_access): Same.
17496 (check_strncat_sizes): Same.
17497 (expand_builtin_strncat): Same.
17498 (expand_builtin_strncmp): Same.
17499 (expand_builtin): Same.
17500 (expand_builtin_object_size): Same.
17501 (warn_dealloc_offset): Same.
17502 (maybe_emit_free_warning): Same.
17503 * calls.c (maybe_warn_alloc_args_overflow): Same.
17504 (maybe_warn_nonstring_arg): Same.
17505 (maybe_warn_rdwr_sizes): Same.
17506 * expr.c (expand_expr_real_1): Remove %K.
17507 * gimple-fold.c (gimple_fold_builtin_strncpy): Remove %G.
17508 (gimple_fold_builtin_strncat): Same.
17509 * gimple-ssa-sprintf.c (format_directive): Same.
17510 (handle_printf_call): Same.
17511 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Same.
17512 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
17513 (maybe_diag_access_bounds): Same. Call gimple_location.
17514 (check_bounds_or_overlap): Same.
17515 * trans-mem.c (ipa_tm_scan_irr_block): Remove %K. Simplify.
17516 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Remove %G.
17517 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
17518 (maybe_diag_stxncpy_trunc): Same.
17519 (handle_builtin_stxncpy_strncat): Same.
17520 (maybe_warn_pointless_strcmp): Same.
17521 * tree-ssa-uninit.c (maybe_warn_operand): Same.
17523 2021-07-06 Uroš Bizjak <ubizjak@gmail.com>
17526 * config/i386/predicates.md (vec_setm_operand): Enable
17527 register_operand for TARGET_SSE4_1.
17528 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_operand
17529 as operand 2 predicate. Call ix86_expand_vector_set_var
17530 for non-constant index operand.
17531 (vec_setv4qi): Use vec_setm_mmx_operand as operand 2 predicate.
17532 Call ix86_expand_vector_set_var for non-constant index operand.
17534 2021-07-06 Jeff Law <jeffreyalaw@gmail.com>
17536 * config/h8300/jumpcall.md (*branch): When possible, generate
17537 the comparison in CCZN mode.
17538 * config/h8300/predicates.md (simple_memory_operand): Reject all
17539 auto-increment addressing modes.
17541 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
17543 PR bootstrap/100246
17544 * config/i386/i386.h (struct stringop_algs): Define a CTOR for
17547 2021-07-06 Richard Biener <rguenther@suse.de>
17549 * doc/md.texi (vec_fmaddsub<mode>4): Document.
17550 (vec_fmsubadd<mode>4): Likewise.
17551 * optabs.def (vec_fmaddsub$a4): Add.
17552 (vec_fmsubadd$a4): Likewise.
17553 * internal-fn.def (IFN_VEC_FMADDSUB): Add.
17554 (IFN_VEC_FMSUBADD): Likewise.
17555 * tree-vect-slp-patterns.c (addsub_pattern::recognize):
17556 Refactor to handle IFN_VEC_FMADDSUB and IFN_VEC_FMSUBADD.
17557 (addsub_pattern::build): Likewise.
17558 * tree-vect-slp.c (vect_optimize_slp): CFN_VEC_FMADDSUB
17559 and CFN_VEC_FMSUBADD are not transparent for permutes.
17560 * config/i386/sse.md (vec_fmaddsub<mode>4): New expander.
17561 (vec_fmsubadd<mode>4): Likewise.
17563 2021-07-06 Richard Biener <rguenther@suse.de>
17565 * doc/invoke.texi (fmove-loop-stores): Document.
17566 * common.opt (fmove-loop-stores): New option.
17567 * opts.c (default_options_table): Enable -fmove-loop-stores
17568 at -O1 but not -Og.
17569 * tree-ssa-loop-im.c (pass_lim::execute): Pass
17570 flag_move_loop_stores instead of true to
17571 loop_invariant_motion_in_fun.
17573 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
17575 * doc/install.texi: Document --with-dsymutil.
17577 2021-07-06 Andrew Pinski <apinski@marvell.com>
17579 PR tree-optimization/101256
17580 * dbgcnt.def (phiopt_edge_range): New counter.
17581 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
17582 Check to make sure the new name is defined in the same
17583 bb as the conditional before duplicating range info.
17584 Also add debug counter.
17586 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
17588 PR rtl-optimization/100328
17589 * config/i386/i386-options.c (ix86_option_override_internal):
17590 Set param_ira_consider_dup_in_all_alts to 0.
17592 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
17594 PR rtl-optimization/100328
17595 * doc/invoke.texi (ira-consider-dup-in-all-alts): Document new
17597 * ira.c (ira_get_dup_out_num): Adjust as parameter
17598 param_ira_consider_dup_in_all_alts.
17599 * params.opt (ira-consider-dup-in-all-alts): New.
17600 * ira-conflicts.c (process_regs_for_copy): Add one parameter
17601 single_input_op_has_cstr_p.
17602 (get_freq_for_shuffle_copy): New function.
17603 (add_insn_allocno_copies): Adjust as single_input_op_has_cstr_p.
17604 * ira-int.h (ira_get_dup_out_num): Add one bool parameter.
17606 2021-07-05 Jeff Law <jeffreyalaw@gmail.com>
17608 * config/h8300/shiftrotate.md (shift-by-variable patterns): Update to
17609 generate condition code aware RTL directly.
17611 2021-07-05 Andrew Pinski <apinski@marvell.com>
17613 PR tree-optimization/101039
17614 * match.pd (A CMP 0 ? A : -A): New patterns.
17615 * tree-ssa-phiopt.c (abs_replacement): Delete function.
17616 (tree_ssa_phiopt_worker): Don't call abs_replacement.
17617 Update comment about abs_replacement.
17619 2021-07-05 Andrew Pinski <apinski@marvell.com>
17621 * tree-ssa-phiopt.c (gimple_simplify_phiopt):
17622 If "A ? B : C" fails to simplify, try "(!A) ? C : B".
17624 2021-07-05 Andrew Pinski <apinski@marvell.com>
17626 * tree-ssa-phiopt.c (match_simplify_replacement):
17627 Add early_p argument. Call gimple_simplify_phiopt
17628 instead of gimple_simplify.
17629 (tree_ssa_phiopt_worker): Update call to
17630 match_simplify_replacement and allow unconditionally.
17631 (phiopt_early_allow): New function.
17632 (gimple_simplify_phiopt): New function.
17634 2021-07-05 Andrew Pinski <apinski@marvell.com>
17636 PR middle-end/101237
17637 * fold-const.c (negate_expr_p): Remove call to element_mode
17638 and TREE_MODE/TREE_TYPE when calling HONOR_SIGNED_ZEROS,
17639 HONOR_SIGN_DEPENDENT_ROUNDING, and HONOR_SNANS.
17640 (fold_negate_expr_1): Likewise.
17641 (const_unop): Likewise.
17642 (fold_cond_expr_with_comparison): Likewise.
17643 (fold_binary_loc): Likewise.
17644 (fold_ternary_loc): Likewise.
17645 (tree_call_nonnegative_warnv_p): Likewise.
17646 * match.pd (-(A + B) -> (-B) - A): Likewise.
17648 2021-07-05 Iain Sandoe <iain@sandoe.co.uk>
17650 * configure.ac: Handle --with-dsymutil in the same way as we
17651 do for the assembler and linker. (DEFAULT_DSYMUTIL): New.
17652 Extract the type and version for the dsymutil configured or
17653 found by the default searches.
17654 * config.in: Regenerated.
17655 * configure: Regenerated.
17656 * collect2.c (do_dsymutil): Handle locating dsymutil in the
17657 same way as for the assembler and linker.
17658 * config/darwin.h (DSYMUTIL): Delete.
17659 * gcc.c: Report a configured dsymutil correctly.
17660 * exec-tool.in: Allow for dsymutil.
17662 2021-07-05 Uroš Bizjak <ubizjak@gmail.com>
17664 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
17665 Handle V4QI and V2HI modes.
17666 (expand_vec_perm_blend): Allow 4-byte vector modes with TARGET_SSE4_1.
17667 Handle V4QI mode. Emit mmx_pblendvb32 for 4-byte modes.
17668 (expand_vec_perm_pshufb): Rewrite to use switch statemets.
17669 Handle 4-byte dual operands with TARGET_XOP and single operands
17670 with TARGET_SSSE3. Emit mmx_ppermv32 for TARGET_XOP and
17671 mmx_pshufbv4qi3 for TARGET_SSSE3.
17672 (expand_vec_perm_pblendv): Allow 4-byte vector modes with TARGET_SSE4_1.
17673 (expand_vec_perm_interleave2): Allow 4-byte vector modes.
17674 (expand_vec_perm_pshufb2): Allow 4-byte vector modes with TARGET_SSSE3.
17675 (expand_vec_perm_even_odd_1): Handle V4QI mode.
17676 (expand_vec_perm_broadcast_1): Handle V4QI mode.
17677 (ix86_vectorize_vec_perm_const): Handle V4QI mode.
17678 * config/i386/mmx.md (mmx_ppermv32): New insn pattern.
17679 (mmx_pshufbv4qi3): Ditto.
17680 (*mmx_pblendw32): Ditto.
17681 (*mmx_pblendw64): Rename from *mmx_pblendw.
17682 (mmx_punpckhbw_low): New insn_and_split pattern.
17683 (mmx_punpcklbw_low): Ditto.
17685 2021-07-05 Richard Biener <rguenther@suse.de>
17687 * tree-vect-loop-manip.c (vect_loop_versioning): Do not
17688 set LOOP_C_INFINITE on the vectorized loop.
17690 2021-07-05 Richard Biener <rguenther@suse.de>
17692 PR middle-end/101291
17693 * cfgloopmanip.c (loop_version): Set the loop copy of the
17694 versioned loop to the new loop.
17696 2021-07-04 Iain Sandoe <iain@sandoe.co.uk>
17699 * config.gcc: Ensure that Darwin biarch definitions are
17700 added before i386.h.
17701 * config/i386/darwin.h (TARGET_64BIT): Remove.
17702 (PR80556_WORKAROUND): New.
17703 (REAL_LIBGCC_SPEC): Amend to use PR80556_WORKAROUND.
17704 (DARWIN_SUBARCH_SPEC): New.
17705 * config/i386/darwin32-biarch.h (TARGET_64BIT_DEFAULT,
17706 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
17707 (REAL_LIBGCC_SPEC): Remove.
17708 * config/i386/darwin64-biarch.h (TARGET_64BIT_DEFAULT,
17709 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
17710 (REAL_LIBGCC_SPEC): Remove.
17712 2021-07-03 H.J. Lu <hjl.tools@gmail.com>
17714 PR middle-end/101294
17715 * expr.c (store_constructor): Don't use vec_duplicate on vector.
17717 2021-07-02 Martin Sebor <msebor@redhat.com>
17719 PR middle-end/98871
17720 PR middle-end/98512
17721 * diagnostic.c (get_any_inlining_info): New.
17722 (update_effective_level_from_pragmas): Handle inlining context.
17723 (diagnostic_enabled): Same.
17724 (diagnostic_report_diagnostic): Same.
17725 * diagnostic.h (struct diagnostic_info): Add ctor.
17726 (struct diagnostic_context): Add new member.
17727 * tree-diagnostic.c (set_inlining_locations): New.
17728 (tree_diagnostics_defaults): Set new callback pointer.
17730 2021-07-02 Peter Bergner <bergner@linux.ibm.com>
17732 * config/rs6000/rs6000-builtin.def (BU_MMA_PAIR_LD, BU_MMA_PAIR_ST):
17734 (__builtin_vsx_lxvp, __builtin_vsx_stxvp): New built-ins.
17735 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Expand
17736 lxvp and stxvp built-ins.
17737 (mma_init_builtins): Handle lxvp and stxvp built-ins.
17738 (builtin_function_type): Likewise.
17739 * doc/extend.texi (__builtin_vsx_lxvp, __builtin_mma_stxvp): Document.
17741 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
17743 * config/h8300/h8300-protos.h (compute_a_shift_cc): Accept
17744 additional argument for the code.
17745 * config/h8300/h8300.c (compute_a_shift_cc): Accept additional
17746 argument for the code. Just return if the ZN bits are useful or
17747 not rather than the old style CC_* enums.
17748 * config/h8300/shiftrotate.md (shiftqi_noscratch): Move before
17749 more generic shiftqi patterns.
17750 (shifthi_noscratch, shiftsi_noscratch): Similarly.
17751 (shiftqi_noscratch_set_flags): New pattern.
17752 (shifthi_noscratch_set_flags, shiftsi_noscratch_set_flags): Likewise.
17754 2021-07-02 Andrew MacLeod <amacleod@redhat.com>
17756 PR tree-optimization/101223
17757 * range-op.cc (build_lt): Add -1 for signed values.
17758 (built_gt): Subtract -1 for signed values.
17760 2021-07-02 David Faust <david.faust@oracle.com>
17762 * btfout.c (get_btf_kind): Support BTF_KIND_FLOAT.
17763 (btf_asm_type): Likewise.
17765 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
17767 * config/h8300/h8300-protos.h (output_a_shift): Make first argument
17768 an array of rtx rather than a pointer to rtx. Add code argument.
17769 (compute_a_shift_length): Similarly.
17770 * config/h8300/h8300.c (h8300_shift_costs): Adjust now that the
17771 shift itself isn't an operand. Create dummy operand[0] to carry
17772 a mode and pass a suitable rtx code to compute_a_shift_length.
17773 (get_shift_alg): Adjust operand number of clobber in output templates.
17774 (output_a_shift): Make first argument an array of rtx rather than
17775 a pointer to rtx. Add code argument for the type of shift.
17776 Adjust now that the shift itself is no longer an operand.
17777 (compute_a_shift_length): Similarly.
17778 * config/h8300/shiftrotate.md (shiftqi, shifthi, shiftsi): Use an
17779 iterator rather than nshift_operator.
17780 (shiftqi_noscratch, shifthi_noscratch, shiftsi_noscratch): Likewise.
17781 (shiftqi_clobber_flags): Adjust to API changes in output_a_shift
17782 and compute_a_shift_length.
17783 (shiftqi_noscratch_clobber_flags): Likewise.
17784 (shifthi_noscratch_clobber_flags): Likewise.
17785 (shiftsi_noscratch_clobber_flags): Likewise.
17787 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
17790 * config/darwin.h (DSYMUTIL_SPEC): Do not try to run
17791 dsymutil for BTF/CTF.
17793 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
17796 * config/darwin.h (CTF_INFO_SECTION_NAME): Update the
17797 segment to include BTF.
17798 (BTF_INFO_SECTION_NAME): New.
17800 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
17802 * config/m32r/m32r-protos.h (call_operand): Adjust return type.
17803 (small_data_operand, memreg_operand, small_insn_p): Likewise.
17804 * config/m32r/m32r.c (call_operand): Adjust return type.
17805 (small_data_operand, memreg_operand): Likewise.
17807 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
17809 * config/frv/frv-protos.h (integer_register_operand): Adjust return
17811 (frv_load_operand, gpr_or_fpr_operand, gpr_no_subreg_operand): Likewise.
17812 (fpr_or_int6_operand, gpr_or_int_operand); Likewise.
17813 (gpr_or_int12_operand, gpr_or_int10_operand); Likewise.
17814 (move_source_operand, move_destination_operand): Likewise.
17815 (condexec_source_operand, condexec_dest_operand): Likewise.
17816 (lr_operand, gpr_or_memory_operand, fpr_or_memory_operand): Likewise.
17817 (reg_or_0_operand, fcc_operand, icc_operand, cc_operand): Likewise.
17818 (fcr_operand, icr_operand, cr_operand, call_operand): Likewise.
17819 (fpr_operand, even_reg_operand, odd_reg_operand): Likewise.
17820 (even_gpr_operand, odd_gpr_operand, quad_fpr_operand): Likewise.
17821 (even_fpr_operand, odd_fpr_operand): Likewise.
17822 (dbl_memory_one_insn_operand, dbl_memory_two_insn_operand): Likewise.
17823 (int12_operand, int6_operand, int5_operand, uint5_operand): Likewise.
17824 (uint4_operand, uint1_operand, int_2word_operand): Likewise
17825 (upper_int16_operand, uint16_operand, symbolic_operand): Likewise.
17826 (relational_operator, float_relational_operator): Likewise.
17827 (ccr_eqne_operator, minmax_operator): Likewise.
17828 (condexec_si_binary_operator, condexec_si_media_operator): Likewise.
17829 (condexec_si_divide_operator, condexec_si_unary_operator): Likewise.
17830 (condexec_sf_conv_operator, condexec_sf_add_operator): Likewise.
17831 (intop_compare_operator, acc_operand, even_acc_operand): Likewise.
17832 (quad_acc_operand, accg_operand): Likewise.
17834 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
17836 * config/stormy16/stormy16-protos.h (xstormy16_below_100_symbol): Change
17837 return type to a bool.
17838 (nonimmediate_nonstack_operand): Likewise.
17839 (xstormy16_splittable_below100_operand): Likewise.
17840 * config/stormy16/stormy16.c (xstormy16_below_100_symbol): Fix
17842 (xstormy16_splittable_below100_operand): Likewise.
17844 2021-07-02 Richard Biener <rguenther@suse.de>
17846 PR tree-optimization/101293
17847 * tree-ssa-loop-im.c (mem_ref_hasher::equal): Compare MEM_REF bases
17848 with combined offsets.
17849 (gather_mem_refs_stmt): Hash MEM_REFs as if their offset were
17850 combined with the rest of the offset.
17852 2021-07-02 Eric Botcazou <ebotcazou@adacore.com>
17854 * config/i386/i386.c (asm_preferred_eh_data_format): Always use the
17855 PIC encodings for PE-COFF targets.
17857 2021-07-02 Jakub Jelinek <jakub@redhat.com>
17860 * config/i386/i386-expand.c (ix86_broadcast_from_integer_constant):
17861 Return nullptr for TImode inner mode.
17863 2021-07-02 Richard Biener <rguenther@suse.de>
17865 PR tree-optimization/101280
17866 PR tree-optimization/101173
17867 * gimple-loop-interchange.cc
17868 (tree_loop_interchange::valid_data_dependences): Properly
17869 guard all dependence checks with DDR_REVERSED_P or its
17872 2021-07-02 Hongyu Wang <hongyu.wang@intel.com>
17874 * config/i386/i386-expand.c (ix86_expand_builtin):
17875 Add branch to clear odata when ZF is set for asedecenc_expand
17876 and wideaesdecenc_expand.
17878 2021-07-02 Eugene Rozenfeld <erozen@microsoft.com>
17880 * config/i386/gcc-auto-profile: regenerate
17882 2021-07-02 liuhongt <hongtao.liu@intel.com>
17884 * config/i386/sse.md (trunc<mode><pmov_dst_4>2): Refined to ..
17885 (trunc<mode><pmov_dst_4_lower>2): this.
17887 2021-07-01 David Malcolm <dmalcolm@redhat.com>
17889 * diagnostic.h (diagnostic_context::m_file_cache): New field.
17890 * input.c (class fcache): Rename to...
17891 (class file_cache_slot): ...this, making most members private and
17892 prefixing fields with "m_".
17893 (file_cache_slot::get_file_path): New accessor.
17894 (file_cache_slot::get_use_count): New accessor.
17895 (file_cache_slot::missing_trailing_newline_p): New accessor.
17896 (file_cache_slot::inc_use_count): New.
17897 (fcache_buffer_size): Move to...
17898 (file_cache_slot::buffer_size): ...here.
17899 (fcache_line_record_size): Move to...
17900 (file_cache_slot::line_record_size): ...here.
17901 (fcache_tab): Delete, in favor of global_dc->m_file_cache.
17902 (fcache_tab_size): Move to file_cache::num_file_slots.
17903 (diagnostic_file_cache_init): Update for move of fcache_tab
17904 to global_dc->m_file_cache.
17905 (diagnostic_file_cache_fini): Likewise.
17906 (lookup_file_in_cache_tab): Convert to...
17907 (file_cache::lookup_file): ...this.
17908 (diagnostics_file_cache_forcibly_evict_file): Update for move of
17909 fcache_tab to global_dc->m_file_cache, moving most of
17910 implementation to...
17911 (file_cache::forcibly_evict_file): ...this new function and...
17912 (file_cache_slot::evict): ...this new function.
17913 (evicted_cache_tab_entry): Convert to...
17914 (file_cache::evicted_cache_tab_entry): ...this.
17915 (add_file_to_cache_tab): Convert to...
17916 (file_cache::add_file): ...this, moving bulk of implementation
17918 (file_cache_slot::create): ..this new function.
17919 (file_cache::file_cache): New.
17920 (file_cache::~file_cache): New.
17921 (lookup_or_add_file_to_cache_tab): Convert to...
17922 (file_cache::lookup_or_add_file): ..this new function.
17923 (fcache::fcache): Rename to...
17924 (file_cache_slot::file_cache_slot): ...this, adding "m_" prefixes
17926 (fcache::~fcache): Rename to...
17927 (file_cache_slot::~file_cache_slot): ...this, adding "m_" prefixes
17929 (needs_read): Convert to...
17930 (file_cache_slot::needs_read_p): ...this.
17931 (needs_grow): Convert to...
17932 (file_cache_slot::needs_grow_p): ...this.
17933 (maybe_grow): Convert to...
17934 (file_cache_slot::maybe_grow): ...this.
17935 (read_data): Convert to...
17936 (file_cache_slot::read_data): ...this.
17937 (maybe_read_data): Convert to...
17938 (file_cache_slot::maybe_read_data): ...this.
17939 (get_next_line): Convert to...
17940 (file_cache_slot::get_next_line): ...this.
17941 (goto_next_line): Convert to...
17942 (file_cache_slot::goto_next_line): ...this.
17943 (read_line_num): Convert to...
17944 (file_cache_slot::read_line_num): ...this.
17945 (location_get_source_line): Update for moving of globals to
17946 global_dc->m_file_cache.
17947 (location_missing_trailing_newline): Likewise.
17948 * input.h (class file_cache_slot): New forward decl.
17949 (class file_cache): New.
17951 2021-07-01 Michael Meissner <meissner@linux.ibm.com>
17953 * config/rs6000/rs6000.c (rs6000_maybe_emit_fp_cmove): Add IEEE
17954 128-bit floating point conditional move support.
17955 (have_compare_and_set_mask): Add IEEE 128-bit floating point
17957 * config/rs6000/rs6000.md (mov<mode>cc, IEEE128 iterator): New insn.
17958 (mov<mode>cc_p10, IEEE128 iterator): New insn.
17959 (mov<mode>cc_invert_p10, IEEE128 iterator): New insn.
17960 (fpmask<mode>, IEEE128 iterator): New insn.
17961 (xxsel<mode>, IEEE128 iterator): New insn.
17963 2021-07-01 Iain Sandoe <iain@sandoe.co.uk>
17966 * config/darwin.h (CTF_INFO_SECTION_NAME): New.
17968 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
17970 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
17972 * config/i386/i386-protos.h (ix86_expand_vector_init_duplicate):
17974 * config/i386/sse.md (INT_BROADCAST_MODE): New mode iterator.
17975 (vec_duplicate<mode>): New expander.
17977 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
17980 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
17982 (ix86_byte_broadcast): New function.
17983 (ix86_convert_const_wide_int_to_broadcast): Likewise.
17984 (ix86_expand_move): Convert CONST_WIDE_INT to broadcast if mode
17985 size is 16 bytes or bigger.
17986 (ix86_broadcast_from_integer_constant): New function.
17987 (ix86_expand_vector_move): Convert CONST_WIDE_INT and CONST_VECTOR
17988 to broadcast if mode size is 16 bytes or bigger.
17989 * config/i386/i386-protos.h (ix86_gen_scratch_sse_rtx): New
17991 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): New function.
17993 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
17995 * config/i386/predicates.md (ix86_endbr_immediate_operand):
17996 Return true/false instead of 1/0.
17997 (movq_parallel): Ditto.
17999 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
18001 * recog.c (general_operand): Return true/false instead of 1/0.
18002 (register_operand): Ditto.
18003 (immediate_operand): Ditto.
18004 (const_int_operand): Ditto.
18005 (const_scalar_int_operand): Ditto.
18006 (const_double_operand): Ditto.
18007 (push_operand): Ditto.
18008 (pop_operand): Ditto.
18009 (memory_operand): Ditto.
18010 (indirect_operand): Ditto.
18012 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
18014 * genpreds.c (write_predicate_subfunction):
18015 Change the type of written subfunction to bool.
18016 (write_one_predicate_function):
18017 Change the type of written function to bool.
18018 (write_tm_preds_h): Ditto.
18019 * recog.h (*insn_operand_predicate_fn): Change the type to bool.
18020 * recog.c (general_operand): Change the type to bool.
18021 (address_operand): Ditto.
18022 (register_operand): Ditto.
18023 (pmode_register_operand): Ditto.
18024 (scratch_operand): Ditto.
18025 (immediate_operand): Ditto.
18026 (const_int_operand): Ditto.
18027 (const_scalar_int_operand): Ditto.
18028 (const_double_operand): Ditto.
18029 (nonimmediate_operand): Ditto.
18030 (nonmemory_operand): Ditto.
18031 (push_operand): Ditto.
18032 (pop_operand): Ditto.
18033 (memory_operand): Ditto.
18034 (indirect_operand): Ditto.
18035 (ordered_comparison_operator): Ditto.
18036 (comparison_operator): Ditto.
18037 * config/i386/i386-expand.c (ix86_expand_sse_cmp):
18038 Change the type of indirect predicate function to bool.
18039 * config/rs6000/rs6000.c (easy_vector_constant):
18040 Change the type to bool.
18041 * config/mips/mips-protos.h (m16_based_address_p):
18042 Change the type of operand 3 to bool.
18044 2021-07-01 Richard Biener <rguenther@suse.de>
18046 PR tree-optimization/101280
18047 PR tree-optimization/101173
18048 * gimple-loop-interchange.cc
18049 (tree_loop_interchange::valid_data_dependences): Revert
18050 previous change and instead correctly handle DDR_REVERSED_P
18053 2021-07-01 Richard Biener <rguenther@suse.de>
18055 PR tree-optimization/101278
18056 * tree-ssa-dse.c (dse_classify_store): First check for
18057 uses, then ignore stmt for chaining purposes.
18059 2021-07-01 Richard Biener <rguenther@suse.de>
18061 PR tree-optimization/100778
18062 * tree-vect-slp.c (vect_schedule_slp_node): Do not place trapping
18063 vectorized ops ahead of their scalar BB.
18065 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
18068 * config/i386/i386.md (*nabs<dwi>2_doubleword):
18069 New insn_and_split pattern.
18070 (*nabs<dwi>2_1): Ditto.
18071 * config/i386/i386-features.c
18072 (general_scalar_chain::compute_convert_gain):
18073 Handle (NEG (ABS (...))) RTX. Rewrite src code
18074 scanner as switch statement.
18075 (general_scalar_chain::convert_insn):
18076 Handle (NEG (ABS (...))) RTX.
18077 (general_scalar_to_vector_candidate_p):
18078 Detect (NEG (ABS (...))) RTX. Reorder case statements
18079 for (AND (NOT (...) ...)) fallthrough.
18081 2021-07-01 Richard Biener <rguenther@suse.de>
18083 PR tree-optimization/101178
18084 * tree-vect-slp.c (slpg_vertex::materialize): Remove.
18085 (slpg::perm_in): Add.
18086 (slpg::get_perm_in): Remove.
18087 (slpg::get_perm_materialized): Add.
18088 (vect_optimize_slp): Handle VEC_PERM nodes more optimally
18089 during permute propagation and materialization.
18091 2021-07-01 Jakub Jelinek <jakub@redhat.com>
18094 * dwarf2out.c (loc_list_from_tree_1): Handle COMPOUND_LITERAL_EXPR.
18096 2021-07-01 Jakub Jelinek <jakub@redhat.com>
18098 PR middle-end/94366
18099 * omp-low.c (lower_rec_input_clauses): Rename is_fp_and_or to
18100 is_truth_op, set it for TRUTH_*IF_EXPR regardless of new_var's type,
18101 use boolean_type_node instead of integer_type_node as NE_EXPR type.
18102 (lower_reduction_clauses): Likewise.
18104 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
18106 * config/gcn/gcn.c: Include dwarf2.h.
18107 (gcn_addr_space_debug): New function.
18108 (TARGET_ADDR_SPACE_DEBUG): New hook.
18110 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
18112 * common/config/gcn/gcn-common.c
18113 (gcn_option_optimization_table): Change OPT_fomit_frame_pointer to -O3.
18114 * config/gcn/gcn.c (gcn_expand_prologue): Prefer the frame pointer
18116 (gcn_expand_prologue): Prefer the frame pointer when emitting CFI.
18117 (gcn_frame_pointer_rqd): New function.
18118 (TARGET_FRAME_POINTER_REQUIRED): New hook.
18120 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
18122 * config/gcn/gcn.c (move_callee_saved_registers): Emit CFI notes for
18123 prologue register saves.
18124 (gcn_debug_unwind_info): Use UI_DWARF2.
18125 (gcn_dwarf_register_number): Map DWARF_LINK_REGISTER to DWARF PC.
18126 (gcn_dwarf_register_span): DWARF_LINK_REGISTER doesn't span.
18127 * config/gcn/gcn.h: (DWARF_FRAME_RETURN_COLUMN): New define.
18128 (DWARF_LINK_REGISTER): New define.
18129 (FIRST_PSEUDO_REGISTER): Increment.
18130 (FIXED_REGISTERS): Add entry for DWARF_LINK_REGISTER.
18131 (CALL_USED_REGISTERS): Likewise.
18132 (REGISTER_NAMES): Likewise.
18134 2021-06-30 Richard Biener <rguenther@suse.de>
18136 PR tree-optimization/101267
18137 * tree-vect-stmts.c (vect_check_scalar_mask): Adjust
18138 API and use SLP compatible interface of vect_is_simple_use.
18139 Reject not vectorized SLP defs for callers that do not support
18141 (vect_check_store_rhs): Handle masked stores and pass down
18142 the appropriate operator index.
18143 (vectorizable_call): Adjust.
18144 (vectorizable_store): Likewise.
18145 (vectorizable_load): Likewise. Handle SLP pecularity of
18147 (vect_is_simple_use): Remove special-casing of masked stores.
18149 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
18151 * common.opt (foffload): Remove help as Driver only.
18152 * gcc.c (display_help): Add -foffload.
18154 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
18156 * gcc.c (close_at_file, execute): Replace alloca by XALLOCAVEC.
18157 (check_offload_target_name): Fix splitting OFFLOAD_TARGETS into
18158 a candidate list; better inform no offload target is configured
18159 and fix hint extraction when passed target is not '\0' at [len].
18160 * common.opt (foffload): Add tailing '.'.
18161 (foffload-options): Likewise; fix flag name in the help string.
18163 2021-06-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
18166 * config/arm/arm_neon.h: Move vabs intrinsics before vcage_f32.
18167 (vcage_f32): Gate comparison on __FAST_MATH__.
18168 (vcageq_f32): Likewise.
18169 (vcale_f32): Likewise.
18170 (vcaleq_f32): Likewise.
18171 (vcagt_f32): Likewise.
18172 (vcagtq_f32): Likewise.
18173 (vcalt_f32): Likewise.
18174 (vcaltq_f32): Likewise.
18175 (vcage_f16): Likewise.
18176 (vcageq_f16): Likewise.
18177 (vcale_f16): Likewise.
18178 (vcaleq_f16): Likewise.
18179 (vcagt_f16): Likewise.
18180 (vcagtq_f16): Likewise.
18181 (vcalt_f16): Likewise.
18182 (vcaltq_f16): Likewise.
18184 2021-06-30 Richard Biener <rguenther@suse.de>
18186 PR tree-optimization/101264
18187 * tree-vect-slp.c (vect_optimize_slp): Propagate the
18188 computed perm_in to all "any" permute successors
18189 we cannot de-duplicate immediately.
18191 2021-06-30 liuhongt <hongtao.liu@intel.com>
18194 * config/i386/sse.md
18195 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>):
18197 (avx512f_sfixupimm<mode><maskz_scalar_name><round_saeonly_name>):
18199 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>"): Refined.
18200 * config/i386/subst.md (maskz_scalar): New define_subst.
18201 (maskz_scalar_name): New subst_attr.
18202 (maskz_scalar_op5): Ditto.
18203 (round_saeonly_maskz_scalar_op5): Ditto.
18204 (round_saeonly_maskz_scalar_operand5): Ditto.
18206 2021-06-30 David Edelsohn <dje.gcc@gmail.com>
18208 * config/rs6000/rs6000.c (rs6000_xcoff_section_type_flags):
18209 Increase code CSECT alignment to at least 32 bytes.
18210 * config/rs6000/xcoff.h (TEXT_SECTION_ASM_OP): Add 32 byte
18211 alignment designation.
18213 2021-06-29 Sergei Trofimovich <siarheit@google.com>
18215 * doc/generic.texi: Fix s/net yet/not yet/ typo.
18217 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
18219 PR tree-optimization/101254
18220 * range-op.cc (operator_minus::op1_op2_relation_effect): Check for
18221 wrapping/non-wrapping when setting the result range.
18223 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
18225 * value-query.cc (gimple_range_global): Allow phis.
18227 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
18229 * vr-values.c (vr_values::vrp_stmt_computes_nonzero): Use stmt.
18230 (simplify_using_ranges::op_with_boolean_value_range_p): Add a
18231 statement for location context.
18232 (check_for_binary_op_overflow): Ditto.
18233 (simplify_using_ranges::get_vr_for_comparison): Ditto.
18234 (simplify_using_ranges::compare_name_with_value): Ditto.
18235 (simplify_using_ranges::compare_names): Ditto.
18236 (vrp_evaluate_conditional_warnv_with_ops_using_ranges): Ditto.
18237 (simplify_using_ranges::simplify_truth_ops_using_ranges): Ditto.
18238 (simplify_using_ranges::simplify_min_or_max_using_ranges): Ditto.
18239 (simplify_using_ranges::simplify_internal_call_using_ranges): Ditto.
18240 (simplify_using_ranges::two_valued_val_range_p): Ditto.
18241 (simplify_using_ranges::simplify): Ditto.
18242 * vr-values.h: Adjust prototypes.
18244 2021-06-29 Uroš Bizjak <ubizjak@gmail.com>
18247 * config/i386/mmx.md (vec_addsubv2sf3): New insn pattern.
18249 2021-06-29 Julian Brown <julian@codesourcery.com>
18251 * config/gcn/gcn.c (gcn_init_libfuncs): New function.
18252 (TARGET_INIT_LIBFUNCS): Define target hook using above function.
18253 * config/gcn/gcn.h (UNITS_PER_WORD): Define to 8 for IN_LIBGCC2, 4
18255 (LIBGCC2_UNITS_PER_WORD, BITS_PER_WORD): Remove definitions.
18256 (MAX_FIXED_MODE_SIZE): Change to 128.
18258 2021-06-29 Julian Brown <julian@codesourcery.com>
18260 * config/gcn/gcn.md (UNSPEC_FLBIT_INT): New unspec constant.
18261 (s_mnemonic): Add clrsb.
18262 (gcn_flbit<mode>_int): Add insn pattern for SImode/DImode.
18263 (clrsb<mode>2): Add expander for SImode/DImode.
18265 2021-06-29 Julian Brown <julian@codesourcery.com>
18267 * config/gcn/gcn.md (<su>mulsidi3, <su>mulsidi3_reg, <su>mulsidi3_imm,
18268 muldi3): Add patterns.
18270 2021-06-29 Julian Brown <julian@codesourcery.com>
18272 * config/gcn/gcn.md (<su>mulsi3_highpart): Change to expander.
18273 (<su>mulsi3_highpart_reg, <su>mulsi3_highpart_imm): New patterns.
18275 2021-06-29 Julian Brown <julian@codesourcery.com>
18277 * config/gcn/gcn.md (mulsi3): Make s_mulk_i32 variant clobber SCC.
18279 2021-06-29 Joseph Myers <joseph@codesourcery.com>
18281 * btfout.c, ctfout.c: Include "memmodel.h".
18283 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
18285 * gcc.c (check_offload_target_name): Cast len argument to
18286 %q.*s to 'int'; avoid -Wstringop-truncation warning.
18288 2021-06-29 Richard Biener <rguenther@suse.de>
18290 * tree-vect-slp.c (vect_optimize_slp): Forward propagate
18291 to "any" permute nodes and relax "any" permute proapgation
18292 during iterative backward propagation.
18294 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
18297 * common.opt (-foffload=): Update description.
18298 (-foffload-options=): New.
18299 * doc/invoke.texi (C Language Options): Document
18300 -foffload and -foffload-options.
18301 * gcc.c (check_offload_target_name): New, split off from
18302 handle_foffload_option.
18303 (check_foffload_target_names): New.
18304 (handle_foffload_option): Handle -foffload=default.
18305 (driver_handle_option): Update for -foffload-options.
18306 * lto-opts.c (lto_write_options): Use -foffload-options
18307 instead of -foffload.
18308 * lto-wrapper.c (merge_and_complain, append_offload_options):
18310 * opts.c (common_handle_option): Likewise.
18312 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
18314 * doc/invoke.texi (C Language Options): Sort options
18315 alphabetically in optlist and also the description itself.
18316 Remove leftover -fallow-single-precision from and add missing
18317 -fgnu-tm to the optlist.
18319 2021-06-29 Richard Biener <rguenther@suse.de>
18321 * tree-vect-slp.c (slpg_vertex::visited): Remove.
18322 (vect_slp_perms_eq): Handle -1 permutes.
18323 (vect_optimize_slp): Rewrite permute propagation.
18325 2021-06-29 Jakub Jelinek <jakub@redhat.com>
18328 * match.pd ((intptr_t)x eq/ne CST to x eq/ne (typeof x) CST): Don't
18329 perform the optimization in GENERIC when sanitizing and x has a
18332 2021-06-29 Richard Biener <rguenther@suse.de>
18334 PR tree-optimization/101242
18335 * tree-vect-slp.c (vect_slp_build_vertices): Force-add
18336 PHIs with not represented initial values as leafs.
18338 2021-06-29 Jan-Benedict Glaw <jbglaw@getslash.de>
18340 * config/pdp11/pdp11.h (ASM_OUTPUT_SKIP): Fix signedness warning.
18341 * config/pdp11/pdp11.c (pdp11_asm_print_operand_punct_valid_p): Remove
18342 "register" keyword.
18343 (pdp11_initial_elimination_offset) Remove unused variable.
18344 (pdp11_cmp_length) Ditto.
18345 (pdp11_insn_cost): Ditto, and fix signedness warning.
18347 2021-06-29 David Edelsohn <dje.gcc@gmail.com>
18349 * btfout.c: Include tm_p.h.
18352 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
18354 * config/bpf/bpf.c (bpf_expand_prologue): Do not mark insns as
18356 (bpf_expand_epilogue): Likewise.
18357 * config/bpf/bpf.h (DWARF2_FRAME_INFO): Define to 0.
18358 Do not define DBX_DEBUGGING_INFO.
18360 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
18362 * doc/invoke.texi: Document the CTF and BTF debug info options.
18364 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
18365 David Faust <david.faust@oracle.com>
18366 Jose E. Marchesi <jose.marchesi@oracle.com>
18367 Weimin Pan <weimin.pan@oracle.com>
18369 * Makefile.in: Add ctfc.*, ctfout.c and btfout.c files to
18370 GTFILES. Add new object files.
18371 * common.opt: Add CTF and BTF debug info options.
18372 * btfout.c: New file.
18373 * ctfc.c: Likewise.
18374 * ctfc.h: Likewise.
18375 * ctfout.c: Likewise.
18376 * dwarf2ctf.c: Likewise.
18377 * dwarf2ctf.h: Likewise.
18378 * dwarf2cfi.c (dwarf2out_do_frame): Acknowledge CTF_DEBUG and
18380 * dwarf2out.c (dwarf2out_source_line): Likewise.
18381 (dwarf2out_finish): Skip emitting DWARF if CTF or BTF are to
18383 (debug_format_do_cu): New function.
18384 (dwarf2out_early_finish): Traverse DIEs and emit CTF/BTF for
18386 Include dwarf2ctf.c.
18387 * final.c (dwarf2_debug_info_emitted_p): Acknowledge DWARF-based debug
18389 * flag-types.h (enum debug_info_type): Add CTF_DEBUG and BTF_DEBUG.
18390 (CTF_DEBUG): New bitmask.
18391 (BTF_DEBUG): Likewise.
18392 (enum ctf_debug_info_levels): New enum.
18393 * gengtype.c (open_base_files): Handle ctfc.h.
18394 (main): Handle uint32_t type.
18395 * flags.h (btf_debuginfo_p): New definition.
18396 (dwarf_based_debuginfo_p): Likewise.
18397 * opts.c (debug_type_names): Add entries for CTF and BTF.
18398 (btf_debuginfo_p): New function.
18399 (dwarf_based_debuginfo_p): Likewise.
18400 (common_handle_option): Handle -gctfN and -gbtf options.
18401 (set_debug_level): Set CTF_DEBUG, BTF_DEBUG whenever appropriate.
18402 * toplev.c (process_options): Inform the user and ignore -gctfLEVEL if
18405 2021-06-28 Jose E. Marchesi <jose.marchesi@oracle.com>
18407 * dwarf2out.c (AT_class): Function is no longer static.
18408 (AT_int): Likewise.
18409 (AT_unsigned): Likewise.
18410 (AT_loc): Likewise.
18411 (get_AT): Likewise.
18412 (get_AT_string): Likewise.
18413 (get_AT_flag): Likewise.
18414 (get_AT_unsigned): Likewise.
18415 (get_AT_ref): Likewise.
18416 (new_die_raw): Likewise.
18417 (lookup_decl_die): Likewise.
18418 (base_type_die): Likewise.
18419 (add_name_attribute): Likewise.
18420 (add_AT_int): Likewise.
18421 (add_AT_unsigned): Likewise.
18422 (add_AT_loc): Likewise.
18423 (dw_get_die_tag): New function.
18424 (dw_get_die_child): Likewise.
18425 (dw_get_die_sib): Likewise.
18426 (struct dwarf_file_data): Move from here to dwarf2out.h
18427 (struct dw_attr_struct): Likewise.
18428 * dwarf2out.h: Analogous changes.
18430 2021-06-28 Martin Jambor <mjambor@suse.cz>
18433 * ipa-param-manipulation.h (class ipa_param_body_adjustments): New
18434 members m_dead_stmts and m_dead_ssas.
18435 * ipa-param-manipulation.c
18436 (ipa_param_body_adjustments::mark_dead_statements): New function.
18437 (ipa_param_body_adjustments::common_initialization): Call it on
18438 all removed but not split parameters.
18439 (ipa_param_body_adjustments::ipa_param_body_adjustments): Initialize
18441 (ipa_param_body_adjustments::modify_call_stmt): Remove arguments that
18443 * tree-inline.c (remap_gimple_stmt): Do not copy dead statements, reset
18444 dead debug statements.
18445 (copy_phis_for_bb): Do not copy dead PHI nodes.
18447 2021-06-28 Martin Jambor <mjambor@suse.cz>
18450 * symtab-clones.h (clone_info): Removed member param_adjustments.
18451 * ipa-param-manipulation.h: Adjust initial comment to reflect how we
18452 deal with pass-through splits now.
18453 (ipa_param_performed_split): Removed.
18454 (ipa_param_adjustments::modify_call): Adjusted parameters.
18455 (class ipa_param_body_adjustments): Adjusted parameters of
18456 register_replacement, modify_gimple_stmt and modify_call_stmt.
18457 (ipa_verify_edge_has_no_modifications): Declare.
18458 (ipa_edge_modifications_finalize): Declare.
18459 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Remove
18460 performed_splits processing, pas only edge to padjs->modify_call,
18461 check that call arguments were not modified if they should not have
18463 * cgraphclones.c (cgraph_node::create_clone): Do not copy performed
18465 * ipa-param-manipulation.c (struct pass_through_split_map): New type.
18466 (ipa_edge_modification_info): Likewise.
18467 (ipa_edge_modification_sum): Likewise.
18468 (ipa_edge_modifications): New edge summary.
18469 (ipa_verify_edge_has_no_modifications): New function.
18470 (transitive_split_p): Removed.
18471 (transitive_split_map): Likewise.
18472 (init_transitive_splits): Likewise.
18473 (ipa_param_adjustments::modify_call): Adjusted to use the new edge
18474 summary instead of performed_splits.
18475 (ipa_param_body_adjustments::register_replacement): Drop dummy
18476 parameter, set base_index of the created ipa_param_body_replacement.
18477 (phi_arg_will_live_p): New function.
18478 (ipa_param_body_adjustments::common_initialization): Do not create
18479 IPA_SRA dummy decls.
18480 (simple_tree_swap_info): Removed.
18481 (remap_split_decl_to_dummy): Likewise.
18482 (record_argument_state_1): New function.
18483 (record_argument_state): Likewise.
18484 (ipa_param_body_adjustments::modify_call_stmt): New parameter
18485 orig_stmt. Do not work with dummy decls, save necessary info about
18486 changes to ipa_edge_modifications.
18487 (ipa_param_body_adjustments::modify_gimple_stmt): New parameter
18488 orig_stmt, pass it to modify_call_stmt.
18489 (ipa_param_body_adjustments::modify_cfun_body): Adjust call to
18490 modify_gimple_stmt.
18491 (ipa_edge_modifications_finalize): New function.
18492 * tree-inline.c (remap_gimple_stmt): Pass original statement to
18493 modify_gimple_stmt.
18494 (copy_phis_for_bb): Do not copy dead PHI nodes.
18495 (expand_call_inline): Do not remap performed_splits.
18496 (update_clone_info): Likewise.
18497 * toplev.c: Include ipa-param-manipulation.h.
18498 (toplev::finalize): Call ipa_edge_modifications_finalize.
18500 2021-06-28 Andrew Pinski <apinski@marvell.com>
18502 * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Duplicate range
18503 info if we're the only things setting the target PHI.
18504 (value_replacement): Don't duplicate range here.
18505 (minmax_replacement): Likewise.
18507 2021-06-28 Richard Biener <rguenther@suse.de>
18509 PR tree-optimization/101229
18510 * gimple-walk.c (gimple_walk_op): Handle PHIs.
18512 2021-06-28 Martin Liska <mliska@suse.cz>
18514 * config/v850/v850.c (construct_dispose_instruction): Allocate
18516 (construct_prepare_instruction): Likewise.
18518 2021-06-28 Martin Liska <mliska@suse.cz>
18520 * config/v850/v850.c (v850_option_override): Build default
18522 (v850_can_inline_p): New. Allow MASK_PROLOG_FUNCTION to be
18523 ignored for inlining.
18524 (TARGET_CAN_INLINE_P): New.
18526 2021-06-28 Richard Biener <rguenther@suse.de>
18528 PR tree-optimization/101207
18529 * tree-vect-slp.c (vect_optimize_slp): Do BB reduction
18530 permute eliding for load permutations properly.
18532 2021-06-28 Richard Biener <rguenther@suse.de>
18534 PR tree-optimization/101173
18535 * gimple-loop-interchange.cc
18536 (tree_loop_interchange::valid_data_dependences): Disallow outer
18537 loop dependence distance of zero.
18539 2021-06-28 liuhongt <hongtao.liu@intel.com>
18542 * config/i386/sse.md (*avx_cmp<mode>3_lt): New
18543 define_insn_and_split.
18544 (*avx_cmp<mode>3_ltint): Ditto.
18545 (*avx2_pcmp<mode>3_3): Ditto.
18546 (*avx2_pcmp<mode>3_4): Ditto.
18547 (*avx2_pcmp<mode>3_5): Ditto.
18549 2021-06-28 liuhongt <hongtao.liu@intel.com>
18551 * config/i386/i386-builtin.def (IX86_BUILTIN_BLENDVPD256,
18552 IX86_BUILTIN_BLENDVPS256, IX86_BUILTIN_PBLENDVB256,
18553 IX86_BUILTIN_BLENDVPD, IX86_BUILTIN_BLENDVPS,
18554 IX86_BUILTIN_PBLENDVB128): Replace icode with
18556 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold blendv
18558 * config/i386/sse.md (*<sse4_1_avx2>_pblendvb_lt_subreg_not):
18559 New pre_reload splitter.
18561 2021-06-27 Andrew Pinski <apinski@marvell.com>
18563 PR middle-end/101230
18564 * fold-const.c (fold_ternary_loc): Check
18565 the return value of invert_tree_comparison.
18567 2021-06-27 David Edelsohn <dje.gcc@gmail.com>
18569 * config.gcc: Add SPDX License Identifier.
18570 (powerpc-ibm-aix789): Default to aix73.h.
18571 (powerpc-ibm-aix7.2.*.*): New stanza.
18572 * config/rs6000/aix72.h: Add SPDX License Identifier.
18573 * config/rs6000/aix73.h: New file.
18575 2021-06-26 Jason Merrill <jason@redhat.com>
18577 * except.c: #include "dwarf2.h" instead of "dwarf2out.h".
18579 2021-06-26 Andrew Pinski <apinski@marvell.com>
18581 * genmatch.c (lower_cond): Copy for_subst_vec
18582 for the simplify also.
18583 (lower): Swap the order for lower_for and lower_cond.
18585 2021-06-26 Andrew Pinski <apinski@marvell.com>
18587 * tree-ssa-phiopt.c (match_simplify_replacement): Reset
18588 flow senatitive info on the moved ssa set.
18590 2021-06-26 Andrew Pinski <apinski@marvell.com>
18592 * fold-const.c (fold_cond_expr_with_comparison):
18593 Exand arg0 into comp_code, arg00, and arg01.
18594 (fold_ternary_loc): Use invert_tree_comparison
18595 instead of fold_invert_truthvalue for the case
18596 where we have A CMP B ? C : A.
18598 2021-06-25 Martin Sebor <msebor@redhat.com>
18600 PR middle-end/101216
18601 * calls.c (maybe_warn_rdwr_sizes): Use the no_warning constant.
18603 2021-06-25 Jeff Law <jeffreyalaw@gmail.com>
18605 * config/h8300/h8300.c (select_cc_mode): Handle ASHIFTRT and LSHIFTRT.
18607 2021-06-25 Richard Biener <rguenther@suse.de>
18609 PR tree-optimization/101202
18610 * tree-vect-slp.c (vect_optimize_slp): Explicitely handle
18613 2021-06-25 Richard Biener <rguenther@suse.de>
18615 * tree-vect-slp-patterns.c (addsub_pattern::build): Copy
18616 STMT_VINFO_REDUC_DEF from the original representative.
18618 2021-06-25 Martin Sebor <msebor@redhat.com>
18620 * builtins.c (warn_string_no_nul): Replace uses of TREE_NO_WARNING,
18621 gimple_no_warning_p and gimple_set_no_warning with
18622 warning_suppressed_p, and suppress_warning.
18624 (maybe_warn_for_bound): Same.
18625 (warn_for_access): Same.
18626 (check_access): Same.
18627 (expand_builtin_strncmp): Same.
18628 (fold_builtin_varargs): Same.
18629 * calls.c (maybe_warn_nonstring_arg): Same.
18630 (maybe_warn_rdwr_sizes): Same.
18631 * cfgexpand.c (expand_call_stmt): Same.
18632 * cgraphunit.c (check_global_declaration): Same.
18633 * fold-const.c (fold_undefer_overflow_warnings): Same.
18634 (fold_truth_not_expr): Same.
18635 (fold_unary_loc): Same.
18636 (fold_checksum_tree): Same.
18637 * gimple-array-bounds.cc (array_bounds_checker::check_array_ref): Same.
18638 (array_bounds_checker::check_mem_ref): Same.
18639 (array_bounds_checker::check_addr_expr): Same.
18640 (array_bounds_checker::check_array_bounds): Same.
18641 * gimple-expr.c (copy_var_decl): Same.
18642 * gimple-fold.c (gimple_fold_builtin_strcpy): Same.
18643 (gimple_fold_builtin_strncat): Same.
18644 (gimple_fold_builtin_stxcpy_chk): Same.
18645 (gimple_fold_builtin_stpcpy): Same.
18646 (gimple_fold_builtin_sprintf): Same.
18647 (fold_stmt_1): Same.
18648 * gimple-ssa-isolate-paths.c (diag_returned_locals): Same.
18649 * gimple-ssa-nonnull-compare.c (do_warn_nonnull_compare): Same.
18650 * gimple-ssa-sprintf.c (handle_printf_call): Same.
18651 * gimple-ssa-store-merging.c (imm_store_chain_info::output_merged_store): Same.
18652 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
18653 * gimple-ssa-warn-restrict.h: Adjust declarations.
18654 (maybe_diag_access_bounds): Replace uses of TREE_NO_WARNING,
18655 gimple_no_warning_p and gimple_set_no_warning with
18656 warning_suppressed_p, and suppress_warning.
18657 (check_call): Same.
18658 (check_bounds_or_overlap): Same.
18659 * gimple.c (gimple_build_call_from_tree): Same.
18660 * gimplify.c (gimplify_return_expr): Same.
18661 (gimplify_cond_expr): Same.
18662 (gimplify_modify_expr_complex_part): Same.
18663 (gimplify_modify_expr): Same.
18664 (gimple_push_cleanup): Same.
18665 (gimplify_expr): Same.
18666 * omp-expand.c (expand_omp_for_generic): Same.
18667 (expand_omp_taskloop_for_outer): Same.
18668 * omp-low.c (lower_rec_input_clauses): Same.
18669 (lower_lastprivate_clauses): Same.
18670 (lower_send_clauses): Same.
18671 (lower_omp_target): Same.
18672 * tree-cfg.c (pass_warn_function_return::execute): Same.
18673 * tree-complex.c (create_one_component_var): Same.
18674 * tree-inline.c (remap_gimple_op_r): Same.
18675 (copy_tree_body_r): Same.
18676 (declare_return_variable): Same.
18677 (expand_call_inline): Same.
18678 * tree-nested.c (lookup_field_for_decl): Same.
18679 * tree-sra.c (create_access_replacement): Same.
18680 (generate_subtree_copies): Same.
18681 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Same.
18682 * tree-ssa-forwprop.c (combine_cond_expr_cond): Same.
18683 * tree-ssa-loop-ch.c (ch_base::copy_headers): Same.
18684 * tree-ssa-loop-im.c (execute_sm): Same.
18685 * tree-ssa-phiopt.c (cond_store_replacement): Same.
18686 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
18687 (handle_builtin_strcpy): Same.
18688 (maybe_diag_stxncpy_trunc): Same.
18689 (handle_builtin_stxncpy_strncat): Same.
18690 (handle_builtin_strcat): Same.
18691 * tree-ssa-uninit.c (get_no_uninit_warning): Same.
18692 (set_no_uninit_warning): Same.
18693 (uninit_undefined_value_p): Same.
18694 (warn_uninit): Same.
18695 (maybe_warn_operand): Same.
18696 * tree-vrp.c (compare_values_warnv): Same.
18697 * vr-values.c (vr_values::extract_range_for_var_from_comparison_expr): Same.
18698 (test_for_singularity): Same.
18699 * gimple.h (warning_suppressed_p): New function.
18700 (suppress_warning): Same.
18701 (copy_no_warning): Same.
18702 (gimple_set_block): Call gimple_set_location.
18703 (gimple_set_location): Call copy_warning.
18705 2021-06-25 Martin Sebor <msebor@redhat.com>
18707 * tree.h (warning_suppressed_at, copy_warning,
18708 warning_suppressed_p, suppress_warning): New functions.
18710 2021-06-25 Martin Sebor <msebor@redhat.com>
18712 * Makefile.in (OBJS-libcommon): Add diagnostic-spec.o.
18713 * gengtype.c (open_base_files): Add diagnostic-spec.h.
18714 * diagnostic-spec.c: New file.
18715 * diagnostic-spec.h: New file.
18716 * tree.h (no_warning, all_warnings, suppress_warning_at): New
18718 * warning-control.cc: New file.
18720 2021-06-25 liuhongt <hongtao.liu@intel.com>
18723 * config/i386/i386.c (x86_order_regs_for_local_alloc):
18726 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
18728 PR tree-optimization/101189
18729 * gimple-range-fold.cc (fold_using_range::range_of_range_op): Pass
18730 LHS range of condition to postfold routine.
18731 (fold_using_range::postfold_gcond_edges): Only process the TRUE or
18732 FALSE edge if the LHS range supports it being taken.
18733 * gimple-range-fold.h (postfold_gcond_edges): Add range parameter.
18735 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
18737 * value-relation.cc (equiv_oracle::dump): Do not dump NULL blocks.
18738 (relation_oracle::find_relation_block): Check correct bitmap.
18739 (relation_oracle::dump): Do not dump NULL blocks.
18741 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
18743 * gimple-range-cache.cc (ranger_cache::propagate_cache): Call
18744 range_on_edge instead of manually calculating.
18746 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
18748 * range-op.cc: Fix comment.
18750 2021-06-24 Uroš Bizjak <ubizjak@gmail.com>
18753 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
18754 Handle V8QI and V4HI modes.
18755 * config/i386/mmx.md (sse4_1_<any_extend:code>v4qiv4hi2):
18757 (sse4_1_<any_extend:code>v4qiv4hi2): Ditto.
18758 (mmxpackmode): New mode attribute.
18759 (vec_pack_trunc_<mmxpackmode:mode>): New expander.
18760 (mmxunpackmode): New mode attribute.
18761 (vec_unpacks_lo_<mmxunpackmode:mode>): New expander.
18762 (vec_unpacks_hi_<mmxunpackmode:mode>): Ditto.
18763 (vec_unpacku_lo_<mmxunpackmode:mode>): Ditto.
18764 (vec_unpacku_hi_<mmxunpackmode:mode>): Ditto.
18765 * config/i386/i386.md (extsuffix): Move from ...
18766 * config/i386/sse.md: ... here.
18768 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
18770 * dwarf2out.c (dwarf2out_assembly_start): Emit .file 0 marker here..
18771 (dwarf2out_finish): ...instead of here.
18773 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
18775 * configure.ac (--gdwarf-5 option): Use objdump instead of readelf.
18776 (working --gdwarf-4/--gdwarf-5 for all sources): Likewise.
18777 (--gdwarf-4 not refusing generated .debug_line): Adjust for Windows.
18778 * configure: Regenerate.
18780 2021-06-24 Richard Biener <rguenther@suse.de>
18782 * config/i386/sse.md (vec_addsubv4df3, vec_addsubv2df3,
18783 vec_addsubv8sf3, vec_addsubv4sf3): Merge into ...
18784 (vec_addsub<mode>3): ... using a new addsub_cst mode attribute.
18786 2021-06-24 Richard Biener <rguenther@suse.de>
18788 * config/i386/sse.md (avx_addsubv4df3): Rename to
18790 (avx_addsubv8sf3): Rename to vec_addsubv8sf3.
18791 (sse3_addsubv2df3): Rename to vec_addsubv2df3.
18792 (sse3_addsubv4sf3): Rename to vec_addsubv4sf3.
18793 * config/i386/i386-builtin.def: Adjust.
18794 * internal-fn.def (VEC_ADDSUB): New internal optab fn.
18795 * optabs.def (vec_addsub_optab): New optab.
18796 * tree-vect-slp-patterns.c (class addsub_pattern): New.
18797 (slp_patterns): Add addsub_pattern.
18798 * tree-vect-slp.c (vect_optimize_slp): Disable propagation
18799 across CFN_VEC_ADDSUB.
18800 * tree-vectorizer.h (vect_pattern::vect_pattern): Make
18802 * doc/md.texi (vec_addsub<mode>3): Document.
18804 2021-06-24 Jakub Jelinek <jakub@redhat.com>
18806 PR middle-end/101170
18807 * df-scan.c (df_ref_record): For paradoxical big-endian SUBREGs
18808 where regno + subreg_regno_offset wraps around use 0 as starting
18811 2021-06-24 Jakub Jelinek <jakub@redhat.com>
18813 PR middle-end/101172
18814 * stor-layout.c (finish_bitfield_representative): If nextf has
18815 error_mark_node type, set repr type to error_mark_node too.
18817 2021-06-24 Ilya Leoshkevich <iii@linux.ibm.com>
18819 * config/s390/s390.c (s390_function_profiler): Ignore labelno
18821 * config/s390/s390.h (NO_PROFILE_COUNTERS): Define.
18823 2021-06-24 Richard Biener <rguenther@suse.de>
18825 * tree-vect-slp.c (vect_optimize_slp): Do not propagate
18826 across operations that have different semantics on different
18829 2021-06-24 Jakub Jelinek <jakub@redhat.com>
18831 * tree.h (OMP_CLAUSE_MAP_IN_REDUCTION): Document meaning for OpenMP.
18832 * gimplify.c (gimplify_scan_omp_clauses): For OpenMP map clauses
18833 with OMP_CLAUSE_MAP_IN_REDUCTION flag partially defer gimplification
18834 of non-decl OMP_CLAUSE_DECL. For OMP_CLAUSE_IN_REDUCTION on
18835 OMP_TARGET user outer_ctx instead of ctx for placeholders and
18836 initializer/combiner gimplification.
18837 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_MAP_IN_REDUCTION
18838 on target constructs.
18839 (lower_rec_input_clauses): Likewise.
18840 (lower_omp_target): Likewise.
18841 * omp-expand.c (expand_omp_target): Temporarily ignore nowait clause
18842 on target if in_reduction is present.
18844 2021-06-24 Kewen Lin <linkw@linux.ibm.com>
18846 * tree-predcom.c (class pcom_worker): New class.
18847 (release_chain): Renamed to...
18848 (pcom_worker::release_chain): ...this.
18849 (release_chains): Renamed to...
18850 (pcom_worker::release_chains): ...this.
18851 (aff_combination_dr_offset): Renamed to...
18852 (pcom_worker::aff_combination_dr_offset): ...this.
18853 (determine_offset): Renamed to...
18854 (pcom_worker::determine_offset): ...this.
18855 (class comp_ptrs): New class.
18856 (split_data_refs_to_components): Renamed to...
18857 (pcom_worker::split_data_refs_to_components): ...this,
18858 and update with class comp_ptrs.
18859 (suitable_component_p): Renamed to...
18860 (pcom_worker::suitable_component_p): ...this.
18861 (filter_suitable_components): Renamed to...
18862 (pcom_worker::filter_suitable_components): ...this.
18863 (valid_initializer_p): Renamed to...
18864 (pcom_worker::valid_initializer_p): ...this.
18865 (find_looparound_phi): Renamed to...
18866 (pcom_worker::find_looparound_phi): ...this.
18867 (add_looparound_copies): Renamed to...
18868 (pcom_worker::add_looparound_copies): ...this.
18869 (determine_roots_comp): Renamed to...
18870 (pcom_worker::determine_roots_comp): ...this.
18871 (determine_roots): Renamed to...
18872 (pcom_worker::determine_roots): ...this.
18873 (single_nonlooparound_use): Renamed to...
18874 (pcom_worker::single_nonlooparound_use): ...this.
18875 (remove_stmt): Renamed to...
18876 (pcom_worker::remove_stmt): ...this.
18877 (execute_pred_commoning_chain): Renamed to...
18878 (pcom_worker::execute_pred_commoning_chain): ...this.
18879 (execute_pred_commoning): Renamed to...
18880 (pcom_worker::execute_pred_commoning): ...this.
18881 (struct epcc_data): New member worker.
18882 (execute_pred_commoning_cbck): Call execute_pred_commoning
18883 with pcom_worker pointer.
18884 (find_use_stmt): Renamed to...
18885 (pcom_worker::find_use_stmt): ...this.
18886 (find_associative_operation_root): Renamed to...
18887 (pcom_worker::find_associative_operation_root): ...this.
18888 (find_common_use_stmt): Renamed to...
18889 (pcom_worker::find_common_use_stmt): ...this.
18890 (combinable_refs_p): Renamed to...
18891 (pcom_worker::combinable_refs_p): ...this.
18892 (reassociate_to_the_same_stmt): Renamed to...
18893 (pcom_worker::reassociate_to_the_same_stmt): ...this.
18894 (stmt_combining_refs): Renamed to...
18895 (pcom_worker::stmt_combining_refs): ...this.
18896 (combine_chains): Renamed to...
18897 (pcom_worker::combine_chains): ...this.
18898 (try_combine_chains): Renamed to...
18899 (pcom_worker::try_combine_chains): ...this.
18900 (prepare_initializers_chain): Renamed to...
18901 (pcom_worker::prepare_initializers_chain): ...this.
18902 (prepare_initializers): Renamed to...
18903 (pcom_worker::prepare_initializers): ...this.
18904 (prepare_finalizers_chain): Renamed to...
18905 (pcom_worker::prepare_finalizers_chain): ...this.
18906 (prepare_finalizers): Renamed to...
18907 (pcom_worker::prepare_finalizers): ...this.
18908 (tree_predictive_commoning_loop): Renamed to...
18909 (pcom_worker::tree_predictive_commoning_loop): ...this, adjust
18910 some calls and remove some cleanup code.
18911 (tree_predictive_commoning): Adjusted to use pcom_worker instance.
18912 (static variable looparound_phis): Remove.
18913 (static variable name_expansions): Remove.
18915 2021-06-24 Richard Biener <rguenther@suse.de>
18917 * tree-vect-slp.c (slpg_vertex): New struct.
18918 (vect_slp_build_vertices): Adjust.
18919 (vect_optimize_slp): Likewise. Maintain an outgoing permute
18920 and a materialized one.
18922 2021-06-24 Richard Biener <rguenther@suse.de>
18924 PR tree-optimization/101105
18925 * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
18926 Only ignore steps when they are equal or scalar order is preserved.
18928 2021-06-24 liuhongt <hongtao.liu@intel.com>
18931 * config/i386/i386-expand.c (ix86_expand_vec_interleave):
18932 Adjust comments for ix86_expand_vecop_qihi2.
18933 (ix86_expand_vecmul_qihi): Renamed to ..
18934 (ix86_expand_vecop_qihi2): Adjust function prototype to
18935 support shift operation, add static to definition.
18936 (ix86_expand_vec_shift_qihi_constant): Add static to definition.
18937 (ix86_expand_vecop_qihi): Call ix86_expand_vecop_qihi2 and
18938 ix86_expand_vec_shift_qihi_constant.
18939 * config/i386/i386-protos.h (ix86_expand_vecmul_qihi): Deleted.
18940 (ix86_expand_vec_shift_qihi_constant): Deleted.
18941 * config/i386/sse.md (VI12_256_512_AVX512VL): New mode
18943 (mulv8qi3): Call ix86_expand_vecop_qihi directly, add
18944 condition TARGET_64BIT.
18945 (mul<mode>3): Ditto.
18946 (<insn><mode>3): Ditto.
18947 (vlshr<mode>3): Extend to support avx512 vlshr.
18948 (v<insn><mode>3): New expander for
18950 (v<insn>v8qi3): Ditto.
18951 (vashrv8hi3<mask_name>): Renamed to ..
18952 (vashr<mode>3): And extend to support V16QImode for avx512.
18953 (vashrv16qi3): Deleted.
18954 (vashrv2di3<mask_name>): Extend expander to support avx512
18957 2021-06-23 Dimitar Dimitrov <dimitar@dinux.eu>
18959 * doc/lto.texi (Design Overview): Update that slim objects are
18962 2021-06-23 Aaron Sawdey <acsawdey@linux.ibm.com>
18964 * config/rs6000/rs6000-cpus.def: Take OPTION_MASK_PCREL_OPT out
18965 of OTHER_POWER10_MASKS so it will not be enabled by default.
18967 2021-06-23 Richard Biener <rguenther@suse.de>
18968 Martin Jambor <mjambor@suse.cz>
18970 * tree-inline.c (setup_one_parameter): Set TREE_READONLY of the
18971 param replacement unconditionally. Adjust comment.
18973 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
18975 * Makefile.in (OBJS): Add gimple-range-fold.o
18976 * gimple-range-fold.cc: New.
18977 * gimple-range-fold.h: New.
18978 * gimple-range-gori.cc (gimple_range_calc_op1): Move to here.
18979 (gimple_range_calc_op2): Ditto.
18980 * gimple-range-gori.h: Move prototypes to here.
18981 * gimple-range.cc: Adjust include files.
18982 (fur_source:fur_source): Relocate to gimple-range-fold.cc.
18983 (fur_source::get_operand): Ditto.
18984 (fur_source::get_phi_operand): Ditto.
18985 (fur_source::query_relation): Ditto.
18986 (fur_source::register_relation): Ditto.
18987 (class fur_edge): Ditto.
18988 (fur_edge::fur_edge): Ditto.
18989 (fur_edge::get_operand): Ditto.
18990 (fur_edge::get_phi_operand): Ditto.
18991 (fur_stmt::fur_stmt): Ditto.
18992 (fur_stmt::get_operand): Ditto.
18993 (fur_stmt::get_phi_operand): Ditto.
18994 (fur_stmt::query_relation): Ditto.
18995 (class fur_depend): Relocate to gimple-range-fold.h.
18996 (fur_depend::fur_depend): Relocate to gimple-range-fold.cc.
18997 (fur_depend::register_relation): Ditto.
18998 (fur_depend::register_relation): Ditto.
18999 (class fur_list): Ditto.
19000 (fur_list::fur_list): Ditto.
19001 (fur_list::get_operand): Ditto.
19002 (fur_list::get_phi_operand): Ditto.
19003 (fold_range): Ditto.
19004 (adjust_pointer_diff_expr): Ditto.
19005 (gimple_range_adjustment): Ditto.
19006 (gimple_range_base_of_assignment): Ditto.
19007 (gimple_range_operand1): Ditto.
19008 (gimple_range_operand2): Ditto.
19009 (gimple_range_calc_op1): Relocate to gimple-range-gori.cc.
19010 (gimple_range_calc_op2): Ditto.
19011 (fold_using_range::fold_stmt): Relocate to gimple-range-fold.cc.
19012 (fold_using_range::range_of_range_op): Ditto.
19013 (fold_using_range::range_of_address): Ditto.
19014 (fold_using_range::range_of_phi): Ditto.
19015 (fold_using_range::range_of_call): Ditto.
19016 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
19017 (fold_using_range::range_of_builtin_call): Ditto.
19018 (fold_using_range::range_of_cond_expr): Ditto.
19019 (fold_using_range::range_of_ssa_name_with_loop_info): Ditto.
19020 (fold_using_range::relation_fold_and_or): Ditto.
19021 (fold_using_range::postfold_gcond_edges): Ditto.
19022 * gimple-range.h: Add gimple-range-fold.h to include files. Change
19023 GIMPLE_RANGE_STMT_H to GIMPLE_RANGE_H.
19024 (gimple_range_handler): Relocate to gimple-range-fold.h.
19025 (gimple_range_ssa_p): Ditto.
19026 (range_compatible_p): Ditto.
19027 (class fur_source): Ditto.
19028 (class fur_stmt): Ditto.
19029 (class fold_using_range): Ditto.
19030 (gimple_range_calc_op1): Relocate to gimple-range-gori.h
19031 (gimple_range_calc_op2): Ditto.
19033 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
19035 PR tree-optimization/101148
19036 PR tree-optimization/101014
19037 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
19038 (ranger_cache::~ranger_cache): Adjust.
19039 (ranger_cache::block_range): Check if propagation disallowed.
19040 (ranger_cache::propagate_cache): Disallow propagation if new value
19041 can't be stored properly.
19042 * gimple-range-cache.h (ranger_cache::m_propfail): New member.
19044 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
19046 * gimple-range-cache.cc (class ssa_block_ranges): Adjust prototype.
19047 (sbr_vector::set_bb_range): Return true.
19048 (class sbr_sparse_bitmap): Adjust.
19049 (sbr_sparse_bitmap::set_bb_range): Return value.
19050 (block_range_cache::set_bb_range): Return value.
19051 (ranger_cache::propagate_cache): Use return value to print msg.
19052 * gimple-range-cache.h (class block_range_cache): Adjust.
19054 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
19056 * gimple-range.cc (dump_bb): Use range_on_edge from the cache.
19058 2021-06-23 Jeff Law <jeffreyalaw@gmail.com>
19060 * config/h8300/logical.md (<code><mode>3<ccnz>): Use <cczn>
19061 so this pattern can be used for test/compare removal. Pass
19062 current insn to compute_logical_op_length and output_logical_op.
19063 * config/h8300/h8300.c (compute_logical_op_cc): Remove.
19064 (h8300_and_costs): Add argument to compute_logical_op_length.
19065 (output_logical_op): Add new argument. Use it to determine if the
19066 condition codes are used and adjust the output accordingly.
19067 (compute_logical_op_length): Add new argument and update length
19068 computations when condition codes are used.
19069 * config/h8300/h8300-protos.h (compute_logical_op_length): Update
19071 (output_logical_op): Likewise.
19073 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
19076 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
19077 Handle 64bit modes for TARGET_XOP. Use indirect gen_* functions.
19078 * config/i386/mmx.md (mmx_ppermv64): New insn pattern.
19079 * config/i386/i386.md (unspec): Move UNSPEC_XOP_PERMUTE from ...
19080 * config/i386/sse.md (unspec): ... here.
19082 2021-06-23 Martin Liska <mliska@suse.cz>
19085 * optc-save-gen.awk: Put back arm_fp16_format to
19088 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
19091 * config/i386/i386.md (bsr_rex64): Add zero-flag setting RTX.
19094 (clz<mode>2): Update RTX pattern for additions.
19096 2021-06-23 Jakub Jelinek <jakub@redhat.com>
19098 PR middle-end/101167
19099 * omp-low.c (lower_omp_regimplify_p): Regimplify also PARM_DECLs
19100 and RESULT_DECLs that have DECL_HAS_VALUE_EXPR_P set.
19102 2021-06-22 Sergei Trofimovich <siarheit@google.com>
19104 * doc/rtl.texi: drop unbalanced parenthesis.
19106 2021-06-22 Richard Biener <rguenther@suse.de>
19108 PR middle-end/101156
19109 * gimplify.c (gimplify_expr): Remove premature incorrect
19112 2021-06-22 Jakub Jelinek <jakub@redhat.com>
19114 PR tree-optimization/101159
19115 * tree-vect-patterns.c (vect_recog_popcount_pattern): Fix some
19118 2021-06-22 Jakub Jelinek <jakub@redhat.com>
19120 PR middle-end/101160
19121 * function.c (assign_parms): For decl_result with TYPE_EMPTY_P type
19122 clear crtl->return_rtx instead of keeping it referencing a pseudo.
19124 2021-06-22 Jakub Jelinek <jakub@redhat.com>
19125 Andrew Pinski <apinski@marvell.com>
19127 PR tree-optimization/101162
19128 * fold-const.c (range_check_type): Handle OFFSET_TYPE like pointer
19131 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
19133 * range-op.cc (range_relational_tests): New.
19134 (range_op_tests): Call range_relational_tests.
19136 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
19138 * range-op.cc (operator_cast::lhs_op1_relation): New.
19139 (operator_identity::lhs_op1_relation): Mew.
19141 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
19143 * range-op.cc (operator_minus::op1_op2_relation_effect): New.
19145 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
19147 * range-op.cc (operator_plus::lhs_op1_relation): New.
19148 (operator_plus::lhs_op2_relation): New.
19150 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
19152 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a
19153 relation_oracle if dominators exist.
19154 (ranger_cache::~ranger_cache): Dispose of oracle.
19155 (ranger_cache::dump_bb): Dump oracle.
19156 * gimple-range.cc (fur_source::fur_source): New.
19157 (fur_source::get_operand): Use mmeber query.
19158 (fur_source::get_phi_operand): Use member_query.
19159 (fur_source::query_relation): New.
19160 (fur_source::register_dependency): Delete.
19161 (fur_source::register_relation): New.
19162 (fur_edge::fur_edge): Adjust.
19163 (fur_edge::get_phi_operand): Fix comment.
19164 (fur_edge::query): Delete.
19165 (fur_stmt::fur_stmt): Adjust.
19166 (fur_stmt::query): Delete.
19167 (fur_depend::fur_depend): Adjust.
19168 (fur_depend::register_relation): New.
19169 (fur_depend::register_relation): New.
19170 (fur_list::fur_list): Adjust.
19171 (fur_list::get_operand): Use member query.
19172 (fold_using_range::range_of_range_op): Process and query relations.
19173 (fold_using_range::range_of_address): Adjust dependency call.
19174 (fold_using_range::range_of_phi): Ditto.
19175 (gimple_ranger::gimple_ranger): New. Use ranger_ache oracle.
19176 (fold_using_range::relation_fold_and_or): New.
19177 (fold_using_range::postfold_gcond_edges): New.
19178 * gimple-range.h (class gimple_ranger): Adjust.
19179 (class fur_source): Adjust members.
19180 (class fur_stmt): Ditto.
19181 (class fold_using_range): Ditto.
19183 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
19185 * range-op.cc (range_operator::wi_fold): Apply relation effect.
19186 (range_operator::fold_range): Adjust and apply relation effect.
19187 (*::fold_range): Add relation parameters.
19188 (*::op1_range): Ditto.
19189 (*::op2_range): Ditto.
19190 (range_operator::lhs_op1_relation): New.
19191 (range_operator::lhs_op2_relation): New.
19192 (range_operator::op1_op2_relation): New.
19193 (range_operator::op1_op2_relation_effect): New.
19194 (relop_early_resolve): New.
19195 (operator_equal::op1_op2_relation): New.
19196 (operator_equal::fold_range): Call relop_early_resolve.
19197 (operator_not_equal::op1_op2_relation): New.
19198 (operator_not_equal::fold_range): Call relop_early_resolve.
19199 (operator_lt::op1_op2_relation): New.
19200 (operator_lt::fold_range): Call relop_early_resolve.
19201 (operator_le::op1_op2_relation): New.
19202 (operator_le::fold_range): Call relop_early_resolve.
19203 (operator_gt::op1_op2_relation): New.
19204 (operator_gt::fold_range): Call relop_early_resolve.
19205 (operator_ge::op1_op2_relation): New.
19206 (operator_ge::fold_range): Call relop_early_resolve.
19207 * range-op.h (class range_operator): Adjust parameters and methods.
19209 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
19211 * Makefile.in (OBJS): Add value-relation.o.
19212 * gimple-range.h: Adjust include files.
19213 * tree-data-ref.c: Adjust include file order.
19214 * value-query.cc (range_query::get_value_range): Default to no oracle.
19215 (range_query::query_relation): New.
19216 (range_query::query_relation): New.
19217 * value-query.h (class range_query): Adjust.
19218 * value-relation.cc: New.
19219 * value-relation.h: New.
19221 2021-06-22 Richard Biener <rguenther@suse.de>
19223 PR tree-optimization/101151
19224 * tree-ssa-sink.c (statement_sink_location): Expand irreducible
19227 2021-06-22 Jojo R <rjiejie@linux.alibaba.com>
19229 * config/riscv/riscv.c (thead_c906_tune_info): New.
19230 (riscv_tune_info_table): Use new tune.
19232 2021-06-22 Richard Biener <rguenther@suse.de>
19234 PR tree-optimization/101158
19235 * tree-vect-slp.c (vect_build_slp_tree_1): Move same operand
19236 checking after checking for matching operation.
19238 2021-06-22 Richard Biener <rguenther@suse.de>
19240 PR tree-optimization/101159
19241 * tree-vect-patterns.c (vect_recog_popcount_pattern): Add
19242 missing NULL vectype check.
19244 2021-06-22 Richard Biener <rguenther@suse.de>
19246 PR tree-optimization/101154
19247 * tree-vect-slp.c (vect_build_slp_tree_2): Fix out-of-bound access.
19249 2021-06-22 Jakub Jelinek <jakub@redhat.com>
19252 * config/i386/i386-protos.h (ix86_last_zero_store_uid): Declare.
19253 * config/i386/i386-expand.c (ix86_last_zero_store_uid): New variable.
19254 * config/i386/i386.c (ix86_expand_prologue): Clear it.
19255 * config/i386/i386.md (peephole2s for 1/2/4 stores of const0_rtx):
19256 Remove "" from match_operand. Emit new insns using emit_move_insn and
19257 set ix86_last_zero_store_uid to INSN_UID of the last store.
19258 Add peephole2s for 1/2/4 stores of const0_rtx following previous
19261 2021-06-22 Martin Liska <mliska@suse.cz>
19263 * auto-profile.c (AUTO_PROFILE_VERSION): Bump as string format
19266 2021-06-22 Martin Liska <mliska@suse.cz>
19268 * gcov-io.h: Remove padding entries.
19270 2021-06-22 liuhongt <hongtao.liu@intel.com>
19272 PR tree-optimization/97770
19273 * tree-vect-patterns.c (vect_recog_popcount_pattern):
19275 (vect_recog_func vect_vect_recog_func_ptrs): Add new pattern.
19277 2021-06-22 liuhongt <hongtao.liu@intel.com>
19280 * config/i386/i386-builtin.def (BDESC): Adjust builtin name.
19281 * config/i386/sse.md (<avx512>_expand<mode>_mask): Rename to ..
19282 (expand<mode>_mask): this ..
19283 (*expand<mode>_mask): New pre_reload splitter to transform
19284 v{,p}expand* to vmov* when mask is zero, all ones, or has all
19285 ones in it's lower part, otherwise still generate
19288 2021-06-22 liuhongt <hongtao.liu@intel.com>
19291 * config/i386/i386-expand.c
19292 (ix86_expand_special_args_builtin): Keep constm1_operand only
19293 if it satisfies insn's operand predicate.
19295 2021-06-21 Jason Merrill <jason@redhat.com>
19298 * df-scan.c (df_ref_record): Check that regno < endregno.
19299 * function.c (assign_parms, expand_function_end): Do nothing with a
19300 TYPE_EMPTY_P result.
19302 2021-06-21 Richard Biener <rguenther@suse.de>
19304 PR tree-optimization/101120
19305 * tree-vect-data-refs.c (bump_vector_ptr): Fold the
19307 * tree-vect-slp.c (vect_transform_slp_perm_load): Add
19308 DR chain DCE capability.
19309 * tree-vectorizer.h (vect_transform_slp_perm_load): Adjust.
19310 * tree-vect-stmts.c (vectorizable_load): Remove unused
19311 loads in the DR chain for SLP.
19313 2021-06-21 Jakub Jelinek <jakub@redhat.com>
19315 PR inline-asm/100785
19316 * gimplify.c (gimplify_asm_expr): Don't diagnose errors if
19317 output or input operands were already error_mark_node.
19318 * cfgexpand.c (expand_asm_stmt): If errors are emitted,
19319 remove all inputs, outputs and clobbers from the asm and
19320 set template to "".
19322 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
19324 * config/arm/arm_neon.h (vceq_s8): Replace builtin with __a == __b.
19325 (vceq_s16): Likewise.
19326 (vceq_s32): Likewise.
19327 (vceq_u8): Likewise.
19328 (vceq_u16): Likewise.
19329 (vceq_u32): Likewise.
19330 (vceq_p8): Likewise.
19331 (vceqq_s8): Likewise.
19332 (vceqq_s16): Likewise.
19333 (vceqq_s32): Likewise.
19334 (vceqq_u8): Likewise.
19335 (vceqq_u16): Likewise.
19336 (vceqq_u32): Likewise.
19337 (vceqq_p8): Likewise.
19338 (vceq_f32): Gate __a == __b on __FAST_MATH__.
19339 (vceqq_f32): Likewise.
19340 (vceq_f16): Likewise.
19341 (vceqq_f16): Likewise.
19343 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
19346 * config/arm/iterators.md (NEON_VACMP): Remove.
19347 * config/arm/neon.md (neon_vca<cmp_op><mode>): Use GLTE instead of GTGE
19349 (neon_vca<cmp_op><mode>_insn): Likewise.
19350 (neon_vca<cmp_op_unsp><mode>_insn_unspec): Use NEON_VAGLTE instead of
19353 2021-06-21 Richard Biener <rguenther@suse.de>
19355 PR tree-optimization/101121
19356 * tree-vect-slp.c (vect_build_slp_tree_2): To not fail fatally
19357 when we just lack a stmt with the desired op when doing permutation.
19358 (vect_build_slp_tree): When caching a failed SLP build attempt
19359 assert that at least one lane is marked as not matching.
19361 2021-06-21 liuhongt <hongtao.liu@intel.com>
19364 * config/i386/i386.md: (*anddi_1): Disparage slightly the mask
19365 register alternative.
19366 (*and<mode>_1): Ditto.
19368 (*andn<mode>_1): Ditto.
19369 (*<code><mode>_1): Ditto.
19370 (*<code>qi_1): Ditto.
19371 (*one_cmpl<mode>2_1): Ditto.
19372 (*one_cmplsi2_1_zext): Ditto.
19373 (*one_cmplqi2_1): Ditto.
19374 * config/i386/i386.c (x86_order_regs_for_local_alloc): Change
19375 the order of mask registers to be before general registers.
19377 2021-06-21 Roger Sayle <roger@nextmovesoftware.com>
19380 * config/i386/i386.md: New define_peephole2s to shrink writing
19381 1, 2 or 4 consecutive zeros to memory when optimizing for size.
19383 2021-06-18 Jeff Law <jeffreyalaw@gmail.com>
19385 * config/h8300/h8300.c (h8300_select_cc_mode): Handle SYMBOL_REF.
19386 * config/h8300/logical.md (<code><mode>3 logcial expander): Generate
19387 more efficient code when the source can be trivially simplified.
19389 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
19391 * gimple-range-cache.cc (ranger_cache::range_of_def): Calculate
19392 a range if global is not available.
19393 (ranger_cache::entry_range): Fallback to range_of_def.
19394 * gimple-range-cache.h (range_of_def): Adjust prototype.
19396 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
19398 PR tree-optimization/101014
19399 * gimple-range-cache.cc (ranger_cache::ranger_cache): Remove poor
19401 (ranger_cache::~ranger_cache): Ditto.
19402 (ranger_cache::enable_new_values): Delete.
19403 (ranger_cache::push_poor_value): Delete.
19404 (ranger_cache::range_of_def): Remove poor value processing.
19405 (ranger_cache::entry_range): Ditto.
19406 (ranger_cache::fill_block_cache): Ditto.
19407 * gimple-range-cache.h (class ranger_cache): Remove poor value members.
19408 * gimple-range.cc (gimple_ranger::range_of_expr): Remove call.
19409 * gimple-range.h (class gimple_ranger): Adjust.
19411 2021-06-18 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
19414 * common/config/arm/arm-common.c (arm_canon_arch_option_1): New function
19415 derived from arm_canon_arch.
19416 (arm_canon_arch_option): Call it.
19417 (arm_canon_arch_multilib_option): New function.
19418 * config/arm/arm-cpus.in (IGNORE_FOR_MULTILIB): New fgroup.
19419 * config/arm/arm.h (arm_canon_arch_multilib_option): New prototype.
19420 (CANON_ARCH_MULTILIB_SPEC_FUNCTION): New macro.
19421 (MULTILIB_ARCH_CANONICAL_SPECS): New macro.
19422 (DRIVER_SELF_SPECS): Add MULTILIB_ARCH_CANONICAL_SPECS.
19423 * config/arm/arm.opt (mlibarch): New option.
19424 * config/arm/t-rmprofile (MULTILIB_MATCHES): For armv8*-m, replace use
19425 of march on RHS with mlibarch.
19427 2021-06-18 Marcel Vollweiler <marcel@codesourcery.com>
19429 * config.in: Regenerate.
19430 * config/gcn/gcn.c (print_operand_address): Fix for global_load assembler
19432 * configure: Regenerate.
19433 * configure.ac: Fix for global_load assembler functions.
19435 2021-06-18 Richard Biener <rguenther@suse.de>
19437 PR tree-optimization/101112
19438 * tree-vect-slp.c (vect_slp_linearize_chain): Fix condition
19439 to lookup a pattern stmt def.
19441 2021-06-18 Jakub Jelinek <jakub@redhat.com>
19443 PR middle-end/101062
19444 * stor-layout.c (finish_bitfield_layout): Don't add bitfield
19445 representatives in QUAL_UNION_TYPE.
19447 2021-06-18 Andrew Pinski <apinski@marvell.com>
19449 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
19450 Add counting of how many times it is done.
19451 (factor_out_conditional_conversion): Likewise.
19452 (match_simplify_replacement): Likewise.
19453 (value_replacement): Likewise.
19454 (spaceship_replacement): Likewise.
19455 (cond_store_replacement): Likewise.
19456 (cond_if_else_store_replacement_1): Likewise.
19457 (hoist_adjacent_loads): Likewise.
19459 2021-06-18 Andrew Pinski <apinski@marvell.com>
19461 * tree-cfg.c (verify_gimple_assign_unary): Reject point and offset
19462 types on NEGATE_EXPR, ABS_EXPR, BIT_NOT_EXPR, PAREN_EXPR and CNONJ_EXPR.
19463 (verify_gimple_assign_binary): Reject point and offset types on
19464 MULT_EXPR, MULT_HIGHPART_EXPR, TRUNC_DIV_EXPR, CEIL_DIV_EXPR,
19465 FLOOR_DIV_EXPR, ROUND_DIV_EXPR, TRUNC_MOD_EXPR, CEIL_MOD_EXPR,
19466 FLOOR_MOD_EXPR, ROUND_MOD_EXPR, RDIV_EXPR, and EXACT_DIV_EXPR.
19468 2021-06-18 Michael Meissner <meissner@linux.ibm.com>
19470 * config/rs6000/rs6000.c (rs6000_emit_minmax): Add support for ISA
19471 3.1 IEEE 128-bit floating point xsmaxcqp/xsmincqp instructions.
19472 * config/rs6000/rs6000.md (s<minmax><mode>3, IEEE128 iterator):
19475 2021-06-17 Aaron Sawdey <acsawdey@linux.ibm.com>
19477 * config/rs6000/genfusion.pl (gen_logical_addsubf): Add
19478 earlyclobber to alts 0/1.
19479 (gen_addadd): Add earlyclobber to alts 0/1.
19480 * config/rs6000/fusion.md: Regenerate file.
19482 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
19484 * cfgloopanal.c (get_loop_hot_path): Make path an auto_vec.
19486 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
19488 * gimple-range-cache.cc: Comment cleanups.
19489 * gimple-range-gori.cc: Comment cleanups.
19490 * gimple-range.cc: Comment/spacing cleanups
19491 * value-range.h: Comment cleanups.
19493 2021-06-17 H.J. Lu <hjl.tools@gmail.com>
19496 * calls.c (expand_call): Replace PUSH_ARGS with
19497 targetm.calls.push_argument (0).
19498 (emit_library_call_value_1): Likewise.
19499 * defaults.h (PUSH_ARGS): Removed.
19500 (PUSH_ARGS_REVERSED): Replace PUSH_ARGS with
19501 targetm.calls.push_argument (0).
19502 * expr.c (block_move_libcall_safe_for_call_parm): Likewise.
19503 (emit_push_insn): Pass the number bytes to push to
19504 targetm.calls.push_argument and pass 0 if ARGS_ADDR is 0.
19505 * hooks.c (hook_bool_uint_true): New.
19506 * hooks.h (hook_bool_uint_true): Likewise.
19507 * rtlanal.c (nonzero_bits1): Replace PUSH_ARGS with
19508 targetm.calls.push_argument (0).
19509 * target.def (push_argument): Add a targetm.calls hook.
19510 * targhooks.c (default_push_argument): New.
19511 * targhooks.h (default_push_argument): Likewise.
19512 * config/bpf/bpf.h (PUSH_ARGS): Removed.
19513 * config/cr16/cr16.c (TARGET_PUSH_ARGUMENT): New.
19514 * config/cr16/cr16.h (PUSH_ARGS): Removed.
19515 * config/i386/i386.c (ix86_push_argument): New.
19516 (TARGET_PUSH_ARGUMENT): Likewise.
19517 * config/i386/i386.h (PUSH_ARGS): Removed.
19518 * config/m32c/m32c.c (TARGET_PUSH_ARGUMENT): New.
19519 * config/m32c/m32c.h (PUSH_ARGS): Removed.
19520 * config/nios2/nios2.h (PUSH_ARGS): Likewise.
19521 * config/pru/pru.h (PUSH_ARGS): Likewise.
19522 * doc/tm.texi.in: Remove PUSH_ARGS documentation. Add
19523 TARGET_PUSH_ARGUMENT hook.
19524 * doc/tm.texi: Regenerated.
19526 2021-06-17 Uroš Bizjak <ubizjak@gmail.com>
19529 * config/i386/i386-expand.c (expand_vector_set_var):
19530 Handle V2FS mode remapping. Pass TARGET_MMX_WITH_SSE to
19531 ix86_expand_vector_init_duplicate.
19532 (ix86_expand_vector_init_duplicate): Emit insv_1 for
19533 QImode for !TARGET_PARTIAL_REG_STALL.
19534 * config/i386/predicates.md (vec_setm_mmx_operand): New predicate.
19535 * config/i386/mmx.md (vec_setv2sf): Use vec_setm_mmx_operand
19536 as operand 2 predicate. Call ix86_expand_vector_set_var
19537 for non-constant index operand.
19538 (vec_setv2si): Ditto.
19539 (vec_setv4hi): Ditto.
19540 (vec_setv8qi): ditto.
19542 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
19544 PR tree-optimization/100790
19545 * gimple-range.cc (range_of_builtin_call): Cleanup clz and ctz
19548 2021-06-17 Martin Liska <mliska@suse.cz>
19550 * doc/invoke.texi: Use consistently -O1 instead of -O.
19552 2021-06-17 Martin Liska <mliska@suse.cz>
19554 * gcov-io.h: Update documentation entry about string format.
19556 2021-06-17 Marius Hillenbrand <mhillen@linux.ibm.com>
19559 * config/s390/vecintrin.h (vec_doublee): Fix to use
19560 __builtin_s390_vflls.
19561 (vec_floate): Fix to use __builtin_s390_vflrd.
19563 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
19565 * dominance.c (get_dominated_to_depth): Return auto_vec<basic_block>.
19566 * dominance.h (get_dominated_to_depth): Likewise.
19567 (get_all_dominated_blocks): Likewise.
19568 * cfgcleanup.c (delete_unreachable_blocks): Adjust.
19569 * gcse.c (hoist_code): Likewise.
19570 * tree-cfg.c (remove_edge_and_dominated_blocks): Likewise.
19571 * tree-parloops.c (oacc_entry_exit_ok): Likewise.
19572 * tree-ssa-dce.c (eliminate_unnecessary_stmts): Likewise.
19573 * tree-ssa-phiprop.c (pass_phiprop::execute): Likewise.
19575 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
19577 * dominance.c (get_dominated_by_region): Return auto_vec<basic_block>.
19578 * dominance.h (get_dominated_by_region): Likewise.
19579 * tree-cfg.c (gimple_duplicate_sese_region): Adjust.
19580 (gimple_duplicate_sese_tail): Likewise.
19581 (move_sese_region_to_fn): Likewise.
19583 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
19585 * dominance.c (get_dominated_by): Return auto_vec<basic_block>.
19586 * dominance.h (get_dominated_by): Likewise.
19587 * auto-profile.c (afdo_find_equiv_class): Adjust.
19588 * cfgloopmanip.c (duplicate_loop_to_header_edge): Likewise.
19589 * loop-unroll.c (unroll_loop_runtime_iterations): Likewise.
19590 * tree-cfg.c (test_linear_chain): Likewise.
19591 (test_diamond): Likewise.
19593 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
19595 * cfgloop.h (get_loop_hot_path): Return auto_vec<basic_block>.
19596 * cfgloopanal.c (get_loop_hot_path): Likewise.
19597 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Likewise.
19599 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
19601 * cgraph.c (cgraph_node::collect_callers): Return
19602 auto_vec<cgraph_edge *>.
19603 * cgraph.h (cgraph_node::collect_callers): Likewise.
19604 * ipa-cp.c (create_specialized_node): Adjust.
19605 (decide_about_value): Likewise.
19606 (decide_whether_version_node): Likewise.
19607 * ipa-sra.c (process_isra_node_results): Likewise.
19609 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
19611 * vec.h (vl_ptr>::using_auto_storage): Handle null m_vec.
19612 (auto_vec<T, 0>::auto_vec): Define move constructor, and delete copy
19614 (auto_vec<T, 0>::operator=): Define move assignment and delete copy
19617 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
19619 * gimple-range.cc (debug_seed_ranger): New.
19620 (dump_ranger): New.
19621 (debug_ranger): New.
19623 2021-06-17 Richard Biener <rguenther@suse.de>
19625 PR tree-optimization/54400
19626 * tree-vectorizer.h (enum slp_instance_kind): Add
19627 slp_inst_kind_bb_reduc.
19628 (reduction_fn_for_scalar_code): Declare.
19629 * tree-vect-data-refs.c (vect_slp_analyze_instance_dependence):
19630 Check SLP_INSTANCE_KIND instead of looking at the
19632 (vect_slp_analyze_instance_alignment): Likewise.
19633 * tree-vect-loop.c (reduction_fn_for_scalar_code): Export.
19634 * tree-vect-slp.c (vect_slp_linearize_chain): Split out
19635 chain linearization from vect_build_slp_tree_2 and generalize
19636 for the use of BB reduction vectorization.
19637 (vect_build_slp_tree_2): Adjust accordingly.
19638 (vect_optimize_slp): Elide permutes at the root of BB reduction
19640 (vectorizable_bb_reduc_epilogue): New function.
19641 (vect_slp_prune_covered_roots): Likewise.
19642 (vect_slp_analyze_operations): Use them.
19643 (vect_slp_check_for_constructors): Recognize associatable
19644 chains for BB reduction vectorization.
19645 (vectorize_slp_instance_root_stmt): Generate code for the
19646 BB reduction epilogue.
19648 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
19650 * gimple-range-gori.cc (gori_compute::has_edge_range_p): Check with
19652 (gori_compute::may_recompute_p): New.
19653 (gori_compute::outgoing_edge_range_p): Perform recomputations.
19654 * gimple-range-gori.h (class gori_compute): Add prototype.
19656 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
19658 * gimple-range-cache.cc (ranger_cache::range_on_edge): Always return
19659 true when a range can be calculated.
19660 * gimple-range.cc (gimple_ranger::dump_bb): Check has_edge_range_p.
19662 2021-06-16 Martin Sebor <msebor@redhat.com>
19664 * doc/invoke.texi (-Wmismatched-dealloc, -Wmismatched-new-delete):
19665 Correct documented defaults.
19667 2021-06-16 Andrew MacLeod <amacleod@redhat.com>
19669 * gimple-range-cache.cc (ranger_cache::ranger_cache): Initialize
19670 m_new_value_p directly.
19672 2021-06-16 Uroš Bizjak <ubizjak@gmail.com>
19675 * config/i386/i386-expand.c (expand_vec_perm_2perm_pblendv):
19676 Handle 64bit modes for TARGET_SSE4_1.
19677 (expand_vec_perm_pshufb2): Handle 64bit modes for TARGET_SSSE3.
19678 (expand_vec_perm_even_odd_pack): Handle V4HI mode.
19679 (expand_vec_perm_even_odd_1) <case E_V4HImode>: Expand via
19680 expand_vec_perm_pshufb2 for TARGET_SSSE3 and via
19681 expand_vec_perm_even_odd_pack for TARGET_SSE4_1.
19682 * config/i386/mmx.md (mmx_packusdw): New insn pattern.
19684 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
19686 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn<mode>):
19687 Change to an expander that emits the correct instruction
19688 depending on endianness.
19689 (aarch64_<sur><addsub>hn<mode>_insn_le): Define.
19690 (aarch64_<sur><addsub>hn<mode>_insn_be): Define.
19692 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
19694 * config/aarch64/aarch64-simd-builtins.def: Split generator
19695 for aarch64_<su>qmovn builtins into scalar and vector
19697 * config/aarch64/aarch64-simd.md (aarch64_<su>qmovn<mode>_insn_le):
19699 (aarch64_<su>qmovn<mode>_insn_be): Define.
19700 (aarch64_<su>qmovn<mode>): Split into scalar and vector
19701 variants. Change vector variant to an expander that emits the
19702 correct instruction depending on endianness.
19704 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
19706 * config/aarch64/aarch64-simd-builtins.def: Split generator
19707 for aarch64_sqmovun builtins into scalar and vector variants.
19708 * config/aarch64/aarch64-simd.md (aarch64_sqmovun<mode>):
19709 Split into scalar and vector variants. Change vector variant
19710 to an expander that emits the correct instruction depending
19712 (aarch64_sqmovun<mode>_insn_le): Define.
19713 (aarch64_sqmovun<mode>_insn_be): Define.
19715 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
19717 * config/aarch64/aarch64-simd.md (aarch64_xtn<mode>_insn_le):
19718 Define - modeling zero-high-half semantics.
19719 (aarch64_xtn<mode>): Change to an expander that emits the
19720 appropriate instruction depending on endianness.
19721 (aarch64_xtn<mode>_insn_be): Define - modeling zero-high-half
19723 (aarch64_xtn2<mode>_le): Rename to...
19724 (aarch64_xtn2<mode>_insn_le): This.
19725 (aarch64_xtn2<mode>_be): Rename to...
19726 (aarch64_xtn2<mode>_insn_be): This.
19727 (vec_pack_trunc_<mode>): Emit truncation instruction instead
19729 * config/aarch64/iterators.md (Vnarrowd): Add Vnarrowd mode
19730 attribute iterator.
19732 2021-06-16 Martin Jambor <mjambor@suse.cz>
19734 PR tree-optimization/100453
19735 * tree-sra.c (create_access): Disqualify any const candidates
19736 which are written to.
19737 (sra_modify_expr): Do not store sub-replacements back to a const base.
19738 (handle_unscalarized_data_in_subtree): Likewise.
19739 (sra_modify_assign): Likewise. Earlier, use TREE_READONLy test
19740 instead of constant_decl_p.
19742 2021-06-16 Jakub Jelinek <jakub@redhat.com>
19744 PR middle-end/101062
19745 * stor-layout.c (finish_bitfield_representative): For fields in unions
19746 assume nextf is always NULL.
19747 (finish_bitfield_layout): Compute bit field representatives also in
19748 unions, but handle it as if each bitfield was the only field in the
19751 2021-06-16 Richard Biener <rguenther@suse.de>
19753 PR tree-optimization/101088
19754 * tree-ssa-loop-im.c (sm_seq_valid_bb): Only look for
19755 supported refs on edges. Do not assert same ref but
19756 different kind stores are unsuported but mark them so.
19757 (hoist_memory_references): Only look for supported refs
19760 2021-06-16 Roger Sayle <roger@nextmovesoftware.com>
19762 PR rtl-optimization/46235
19763 * config/i386/i386.md: New define_split for bt followed by cmov.
19764 (*bt<mode>_setcqi): New define_insn_and_split for bt followed by setc.
19765 (*bt<mode>_setncqi): New define_insn_and_split for bt then setnc.
19766 (*bt<mode>_setnc<mode>): New define_insn_and_split for bt followed
19767 by setnc with zero extension.
19769 2021-06-16 Richard Biener <rguenther@suse.de>
19771 PR tree-optimization/101083
19772 * tree-vect-slp.c (vect_slp_build_two_operator_nodes): Get
19773 vectype as argument.
19774 (vect_build_slp_tree_2): Adjust.
19776 2021-06-15 Martin Sebor <msebor@redhat.com>
19778 PR middle-end/100876
19779 * builtins.c: (gimple_call_return_array): Account for size_t
19780 mangling as either unsigned int or unsigned long
19782 2021-06-15 Jeff Law <jeffreyalaw@gmail.com>
19784 * compare-elim.c (try_eliminate_compare): Run DCE to clean things
19785 up before eliminating comparisons.
19787 2021-06-15 Aldy Hernandez <aldyh@redhat.com>
19789 * range-op.cc (operator_bitwise_or::wi_fold): Make sure
19790 nonzero|X is nonzero.
19791 (range_op_bitwise_and_tests): Add tests for above.
19793 2021-06-15 Carl Love <cel@us.ibm.com>
19796 * config/rs6000/rs6000-builtin.def (VCMPEQUT): Fix the ICODE for the
19798 (VRLQ, VSLQ, VSRQ, VSRAQ): Remove unused BU_P10_OVERLOAD_2
19801 2021-06-15 Tobias Burnus <tobias@codesourcery.com>
19804 * gimplify.c (enum gimplify_defaultmap_kind): Add GDMK_SCALAR_TARGET.
19805 (struct gimplify_omp_ctx): Extend defaultmap array by one.
19806 (new_omp_context): Init defaultmap[GDMK_SCALAR_TARGET].
19807 (omp_notice_variable): Update type classification for Fortran.
19808 (gimplify_scan_omp_clauses): Update calls for new argument; handle
19809 GDMK_SCALAR_TARGET; for Fortran, GDMK_POINTER avoid GOVD_MAP_0LEN_ARRAY.
19810 * langhooks-def.h (lhd_omp_scalar_p): Add 'ptr_ok' argument.
19811 * langhooks.c (lhd_omp_scalar_p): Likewise.
19812 (LANG_HOOKS_OMP_ALLOCATABLE_P, LANG_HOOKS_OMP_SCALAR_TARGET_P): New.
19813 (LANG_HOOKS_DECLS): Add them.
19814 * langhooks.h (struct lang_hooks_for_decls): Add new hooks, update
19815 omp_scalar_p pointer type to include the new bool argument.
19817 2021-06-15 David Malcolm <dmalcolm@redhat.com>
19819 * doc/analyzer.texi
19820 (Special Functions for Debugging the Analyzer): Add
19821 __analyzer_dump_capacity.
19823 2021-06-15 Jakub Jelinek <jakub@redhat.com>
19826 * expr.c (expand_expr_real_2) <case VEC_PACK_FIX_TRUNC_EXPR,
19827 case VEC_PACK_TRUNC_EXPR>: Clear subtarget when changing mode.
19829 2021-06-15 Richard Biener <rguenther@suse.de>
19831 * cfgloopanal.c (mark_irreducible_loops): Use a dominance
19832 check to identify loop latches.
19833 * cfgloop.c (verify_loop_structure): Likewise.
19834 * loop-init.c (apply_loop_flags): Allow marked irreducible
19835 regions even with multiple latches.
19836 * predict.c (rebuild_frequencies): Simplify.
19838 2021-06-15 Richard Biener <rguenther@suse.de>
19840 * tree-ssa-threadupdate.c
19841 (jump_thread_path_registry::mark_threaded_blocks): Assert we
19842 have marked irreducible regions.
19844 2021-06-14 Martin Sebor <msebor@redhat.com>
19847 * builtins.c (gimple_call_return_array): Check for attribute fn spec.
19848 Handle calls to placement new.
19849 (ndecl_dealloc_argno): Avoid placement delete.
19851 2021-06-14 Peter Bergner <bergner@linux.ibm.com>
19854 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Use
19855 create_tmp_reg_or_ssa_name().
19857 2021-06-14 Andrew MacLeod <amacleod@redhat.com>
19859 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
19860 (ranger_cache::enable_new_values): Set to specified value and
19861 return the old value.
19862 (ranger_cache::disable_new_values): Delete.
19863 (ranger_cache::fill_block_cache): Disable non 1st order derived
19865 * gimple-range-cache.h (ranger_cache): Adjust prototypes.
19866 * gimple-range.cc (gimple_ranger::range_of_expr): Adjust.
19868 2021-06-14 Uroš Bizjak <ubizjak@gmail.com>
19871 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const):
19872 Return true early when testing with V2HImode.
19873 * config/i386/mmx.md (*punpckwd): Split to sse2_pshuflw_1.
19875 2021-06-14 Christophe Lyon <christophe.lyon@linaro.org>
19877 * config/arm/mve.md (mve_vec_unpack<US>_lo_<mode>): New pattern.
19878 (mve_vec_unpack<US>_hi_<mode>): New pattern.
19879 (@mve_vec_pack_trunc_lo_<mode>): New pattern.
19880 (mve_vmovntq_<supf><mode>): Prefix with '@'.
19881 * config/arm/neon.md (vec_unpack<US>_hi_<mode>): Move to
19883 (vec_unpack<US>_lo_<mode>): Likewise.
19884 (vec_pack_trunc_<mode>): Rename to
19885 neon_quad_vec_pack_trunc_<mode>.
19886 * config/arm/vec-common.md (vec_unpack<US>_hi_<mode>): New
19888 (vec_unpack<US>_lo_<mode>): New.
19889 (vec_pack_trunc_<mode>): New.
19891 2021-06-14 Richard Biener <rguenther@suse.de>
19893 PR tree-optimization/100934
19894 * tree-ssa-dom.c (pass_dominator::execute): Properly
19895 mark irreducible regions.
19897 2021-06-14 Martin Liska <mliska@suse.cz>
19899 * doc/invoke.texi: Put r{...} on the same line as @item.
19901 2021-06-14 Martin Liska <mliska@suse.cz>
19903 * doc/invoke.texi: Add missing newline.
19905 2021-06-14 Martin Liska <mliska@suse.cz>
19907 * doc/invoke.texi: Remove '+' charasters.
19909 2021-06-14 Claudiu Zissulescu <claziss@synopsys.com>
19911 * config.gcc (arc): Add support for with_cpu option.
19912 * config/arc/arc.h (OPTION_DEFAULT_SPECS): Add fpu.
19914 2021-06-14 Richard Biener <rguenther@suse.de>
19916 PR tree-optimization/101031
19917 * tree-ssa-strlen.c (maybe_invalidate): Increment max_size
19918 instead of size when accounting for a possibly string
19921 2021-06-14 Martin Liska <mliska@suse.cz>
19923 * gimple-ssa-evrp.c (pointer_equiv_analyzer::~pointer_equiv_analyzer): Use delete[].
19925 2021-06-14 Aldy Hernandez <aldyh@redhat.com>
19927 * value-query.cc (gimple_range_global): Call get_range_global
19928 if called after inlining.
19930 2021-06-13 Uroš Bizjak <ubizjak@gmail.com>
19933 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
19934 Emit constant permutation insn directly from here.
19936 2021-06-13 Trevor Saunders <tbsaunde@tbsaunde.org>
19938 * attribs.c (find_attribute_namespace): Iterate over vec<> with
19940 * auto-profile.c (afdo_find_equiv_class): Likewise.
19941 * gcc.c (do_specs_vec): Likewise.
19942 (do_spec_1): Likewise.
19943 (driver::set_up_specs): Likewise.
19944 * gimple-loop-jam.c (any_access_function_variant_p): Likewise.
19945 * gimple-ssa-store-merging.c (compatible_load_p): Likewise.
19946 (imm_store_chain_info::try_coalesce_bswap): Likewise.
19947 (imm_store_chain_info::coalesce_immediate_stores): Likewise.
19948 (get_location_for_stmts): Likewise.
19949 * graphite-poly.c (print_iteration_domains): Likewise.
19950 (free_poly_bb): Likewise.
19951 (remove_gbbs_in_scop): Likewise.
19952 (free_scop): Likewise.
19953 (dump_gbb_cases): Likewise.
19954 (dump_gbb_conditions): Likewise.
19955 (print_pdrs): Likewise.
19956 (print_scop): Likewise.
19957 * ifcvt.c (cond_move_process_if_block): Likewise.
19958 * lower-subreg.c (decompose_multiword_subregs): Likewise.
19959 * regcprop.c (pass_cprop_hardreg::execute): Likewise.
19960 * sanopt.c (sanitize_rewrite_addressable_params): Likewise.
19961 * sel-sched-dump.c (dump_insn_vector): Likewise.
19962 * store-motion.c (store_ops_ok): Likewise.
19963 (store_killed_in_insn): Likewise.
19964 * timevar.c (timer::named_items::print): Likewise.
19965 * tree-cfgcleanup.c (cleanup_control_flow_pre): Likewise.
19966 (cleanup_tree_cfg_noloop): Likewise.
19967 * tree-data-ref.c (dump_data_references): Likewise.
19968 (print_dir_vectors): Likewise.
19969 (print_dist_vectors): Likewise.
19970 (dump_data_dependence_relations): Likewise.
19971 (dump_dist_dir_vectors): Likewise.
19972 (dump_ddrs): Likewise.
19973 (create_runtime_alias_checks): Likewise.
19974 (free_subscripts): Likewise.
19975 (save_dist_v): Likewise.
19976 (save_dir_v): Likewise.
19977 (invariant_access_functions): Likewise.
19978 (same_access_functions): Likewise.
19979 (access_functions_are_affine_or_constant_p): Likewise.
19980 (find_data_references_in_stmt): Likewise.
19981 (graphite_find_data_references_in_stmt): Likewise.
19982 (free_dependence_relations): Likewise.
19983 (free_data_refs): Likewise.
19984 * tree-inline.c (copy_debug_stmts): Likewise.
19985 * tree-into-ssa.c (dump_currdefs): Likewise.
19986 (rewrite_update_phi_arguments): Likewise.
19987 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
19988 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
19990 (vect_slp_analyze_node_dependences): Likewise.
19991 (vect_slp_analyze_instance_dependence): Likewise.
19992 (vect_record_base_alignments): Likewise.
19993 (vect_get_peeling_costs_all_drs): Likewise.
19994 (vect_peeling_supportable): Likewise.
19995 * tree-vectorizer.c (vec_info::~vec_info): Likewise.
19996 (vec_info::free_stmt_vec_infos): Likewise.
19998 2021-06-13 Jeff Law <jeffreyalaw@gmail.com>
20000 * config/h8300/logical.md (<code>qi3_1<cczn>): New pattern.
20001 (andqi3_1<cczn>): Removed.
20002 (<ors>qi3_1): Do not split for IOR/XOR a single bit.
20003 (H8/SX bit logicals): Split out from other patterns.
20004 * config/h8300/multiply.md (mulqihi3_const<cczn>): Renamed from
20005 mulqihi3_const_clobber_flags.
20006 (mulqihi3<cczn>, mulhisi3_const<cczn>, mulhisi3<cczn>): Similarly
20008 2021-06-13 H.J. Lu <hjl.tools@gmail.com>
20011 * config/i386/i386.c (ix86_expand_prologue): Set red_zone_used
20012 to true if red zone is used.
20013 (ix86_output_indirect_jmp): Replace ix86_red_zone_size with
20014 ix86_red_zone_used.
20015 * config/i386/i386.h (machine_function): Add red_zone_used.
20016 (ix86_red_zone_size): Removed.
20017 (ix86_red_zone_used): New.
20018 * config/i386/i386.md (peephole2 patterns): Replace
20019 ix86_red_zone_size with ix86_red_zone_used.
20021 2021-06-12 Jason Merrill <jason@redhat.com>
20023 * doc/extend.texi (unused variable attribute): Applies to
20024 structure fields as well.
20026 2021-06-12 Eugene Rozenfeld <erozen@microsoft.com>
20028 * auto-profile.c (read_profile): fix a typo in an error string
20030 2021-06-11 Thomas Schwinge <thomas@codesourcery.com>
20032 * tree-pretty-print.h (dump_omp_clauses): Add 'bool = true'
20034 * tree-pretty-print.c (dump_omp_clauses): Update.
20035 (dump_generic_node) <OMP_CLAUSE>: Use it.
20037 2021-06-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
20040 * config/arm/arm_mve.h (__arm_vld1q): Change __ARM_mve_coerce(p0,
20041 int8_t const *) to __ARM_mve_coerce1(p0, int8_t *) in the argument for
20042 the polymorphic variants matching code.
20043 (__arm_vld1q_z): Likewise.
20044 (__arm_vld2q): Likewise.
20045 (__arm_vld4q): Likewise.
20046 (__arm_vldrbq_gather_offset): Likewise.
20047 (__arm_vldrbq_gather_offset_z): Likewise.
20049 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
20051 PR tree-optimization/96392
20052 * fold-const.h (tree_expr_maybe_real_minus_zero_p): Fix prototype.
20054 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
20056 PR tree-optimization/96392
20057 * fold-const.c (fold_real_zero_addition_p): Take both arguments
20058 of the addition or subtraction, not just the zero. Use this
20059 other argument in tests for signaling NaNs and signed zeros.
20060 (tree_expr_maybe_real_minus_zero_p): New predicate.
20061 * fold-const.h (fold_real_zero_addition_p): Update prototype.
20062 (tree_expr_maybe_real_minus_zero_p): New function prototype.
20063 * match.pd: Update calls to fold_real_zero_addition_p.
20064 Replace HONOR_NANS with tree_expr_maybe_nan_p.
20065 Replace HONOR_SIGNED_ZEROS with tree_expr_maybe_real_minus_zero_p.
20066 Replace HONOR_SNANS with tree_expr_maybe_signaling_nan_p.
20067 * tree-ssa-reassoc.c (eliminate_using_constants): Update
20068 call to fold_real_zero_addition_p.
20070 2021-06-11 Richard Biener <rguenther@suse.de>
20072 PR tree-optimization/101025
20073 * tree-ssa-loop-im.c (sm_seq_valid_bb): Make sure to process
20074 all refs that require dependence checking.
20076 2021-06-11 Richard Biener <rguenther@suse.de>
20078 PR tree-optimization/101028
20079 * tree-vect-slp.c (vect_build_slp_tree_2): When SLP
20080 reassoc discovery fails fatally, mark appropriate lanes
20083 2021-06-11 Richard Biener <rguenther@suse.de>
20085 PR tree-optimization/101026
20086 * tree-vect-slp.c (vect_build_slp_tree_2): Make sure we
20087 have a representative for the associated chain nodes.
20089 2021-06-11 Jakub Jelinek <jakub@redhat.com>
20091 PR rtl-optimization/101008
20092 * simplify-rtx.c (relational_result): New function.
20093 (simplify_logical_relational_operation,
20094 simplify_relational_operation): Use it.
20096 2021-06-11 Jakub Jelinek <jakub@redhat.com>
20099 * config/i386/sse.md (*vec_concat<mode>_0_1): Require TARGET_SSE2.
20101 2021-06-11 Uroš Bizjak <ubizjak@gmail.com>
20104 * config/i386/i386-expand.c (expand_vec_perm_pshufb): Return
20105 false if the permutation can be implemented with constant
20106 permutation instruction in wider mode.
20107 (canonicalize_vector_int_perm): Move above expand_vec_perm_pshufb.
20108 Handle V8QImode and V4HImode.
20110 2021-06-11 Martin Liska <mliska@suse.cz>
20112 PR gcov-profile/100788
20113 * common.opt: Add new option.
20114 * coverage.c (coverage_begin_function): Emit warning instead on
20115 the internal compiler error.
20116 * doc/invoke.texi: Document the option.
20117 * toplev.c (process_options): Enable it by default.
20119 2021-06-11 Richard Biener <rguenther@suse.de>
20121 PR middle-end/101009
20122 * tree-data-ref.c (build_classic_dist_vector_1): Make sure
20123 to set *init_b to true when we encounter a constant equal
20125 (compute_affine_dependence): Also dump the actual DR_REF.
20127 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
20129 PR tree-optimization/100984
20130 * gimple-ssa-evrp.c (ssa_equiv_stack): Use auto_vec for
20131 replacements table.
20132 (ssa_equiv_stack::~ssa_equiv_stack): Remove.
20134 2021-06-11 Kewen Lin <linkw@linux.ibm.com>
20136 * config/rs6000/rs6000.md
20137 (floatsi<SFDF:mode>2_lfiwax_<QHI:mode>_mem_zext): New
20138 define_insn_and_split.
20140 2021-06-11 Richard Biener <rguenther@suse.de>
20142 * tree-vect-slp.c (vect_build_slp_tree_2): Use stablesort
20143 to sort operands of the associative chain.
20145 2021-06-11 Richard Biener <rguenther@suse.de>
20147 * system.h (gcc_stablesort_r): Declare.
20148 * sort.cc (gcc_sort_r): Support stable sort.
20149 (gcc_stablesort_r): Define.
20150 * vec.h (vec<>::stablesort): Add.
20152 2021-06-10 Uroš Bizjak <ubizjak@gmail.com>
20155 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
20156 Handle V2SF mode. Emit SHUFPS to fixup unpack-high for V2SF mode.
20157 (expand_vec_perm_blend): Handle 64bit modes for TARGET_SSE4_1.
20158 (expand_vec_perm_pshufb): Handle 64bit modes for TARGET_SSSE3.
20159 (expand_vec_perm_pblendv): Handle 64bit modes for TARGET_SSE4_1.
20160 (expand_vec_perm_interleave2): Handle 64bit modes.
20161 (expand_vec_perm_even_odd_pack): Handle V8QI mode.
20162 (expand_vec_perm_even_odd_1): Ditto.
20163 (ix86_vectorize_vec_perm_const): Ditto.
20164 * config/i386/i386.md (UNSPEC_PSHUFB): Move from ...
20165 * config/i386/sse.md: ... here.
20166 * config/i386/mmx.md (*vec_interleave_lowv2sf):
20167 New insn_and_split pattern.
20168 (*vec_interleave_highv2sf): Ditto.
20169 (mmx_pshufbv8qi3): New insn pattern.
20170 (*mmx_pblendw): Ditto.
20172 2021-06-10 Peter Bergner <bergner@linux.ibm.com>
20174 * config/rs6000/rs6000-builtin.def (build_pair): New built-in.
20175 (build_acc): Likewise.
20176 * config/rs6000/rs6000-call.c (mma_expand_builtin): Swap assemble
20177 source operands in little-endian mode.
20178 (rs6000_gimple_fold_mma_builtin): Handle VSX_BUILTIN_BUILD_PAIR.
20179 (mma_init_builtins): Likewise.
20180 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Handle endianness
20181 ordering for the MMA assemble and build source operands.
20182 * doc/extend.texi (__builtin_vsx_build_acc, __builtin_mma_build_pair):
20184 (__builtin_mma_assemble_acc, __builtin_mma_assemble_pair): Remove
20187 2021-06-10 Jeff Law <jeffreyalaw@gmail.com>
20189 * config/h8300/h8300.c (select_cc_mode): Handle MEM. Use
20191 * config/h8300/extensions.md: Replace _clobber_flags patterns
20194 2021-06-10 Robin Dapp <rdapp@linux.ibm.com>
20196 * config/s390/vector.md (vcond_mask_<mode><mode>): Change to
20197 (vcond_mask_<mode><tointvec>): this.
20199 2021-06-10 Andrew Stubbs <ams@codesourcery.com>
20200 Thomas Schwinge <thomas@codesourcery.com>
20202 * omp-builtins.def (BUILT_IN_GOACC_ENTER_EXIT_DATA): Split into...
20203 (BUILT_IN_GOACC_ENTER_DATA, BUILT_IN_GOACC_EXIT_DATA): ... these.
20204 * gimple.h (enum gf_mask): Split
20205 'GF_OMP_TARGET_KIND_OACC_ENTER_EXIT_DATA' into
20206 'GF_OMP_TARGET_KIND_OACC_ENTER_DATA' and
20207 'GF_OMP_TARGET_KIND_OACC_EXIT_DATA'.
20208 (is_gimple_omp_oacc): Update.
20209 * gimple-pretty-print.c (dump_gimple_omp_target): Likewise.
20210 * gimplify.c (gimplify_omp_target_update): Likewise.
20211 * omp-expand.c (expand_omp_target, build_omp_regions_1)
20212 (omp_make_gimple_edges): Likewise.
20213 * omp-low.c (check_omp_nesting_restrictions, lower_omp_target):
20216 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
20218 * value-query.cc (value_query::value_on_edge): Rename name to
20220 (range_query::range_on_edge): Same.
20221 (range_query::value_of_expr): Same.
20222 (range_query::value_on_edge): Same.
20223 * value-query.h (class value_query): Same.
20224 (class range_query): Same.
20226 2021-06-10 Richard Biener <rguenther@suse.de>
20228 PR tree-optimization/101003
20229 * tree-vect-slp.c (vect_build_slp_tree_2): Appropriately
20230 use the pattern stmt defs when linearizing a chain.
20232 2021-06-10 Jakub Jelinek <jakub@redhat.com>
20235 * ifcvt.c (noce_get_alt_condition, noce_try_abs): Use
20236 prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
20238 2021-06-10 Clement Chigot <clement.chigot@atos.net>
20240 * config/rs6000/aix71.h (ASM_CPU_SPEC): Add Power10 directive.
20241 * config/rs6000/aix72.h (ASM_CPU_SPEC): Likewise.
20243 2021-06-09 Andrew Pinski <apinski@marvell.com>
20245 PR tree-optimization/100925
20246 * match.pd (a ? CST1 : CST2): Limit transformations
20247 that would produce a negative to integeral types only.
20248 Change !POINTER_TYPE_P to INTEGRAL_TYPE_P also.
20250 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
20253 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
20255 * doc/tm.texi: Correctly update.
20257 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
20259 * doc/tm.texi: Correctly update.
20261 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
20264 * doc/tm.texi.in (Trampolines): Add a missing blank line.
20266 2021-06-09 Paul Eggert <eggert@cs.ucla.edu>
20269 * doc/invoke.texi (Code Gen Options); Document that -fno-trampolines
20270 and -ftrampolines work only with Ada.
20271 * doc/tm.texi.in (Trampolines): Likewise.
20272 * doc/tm.texi: Regenerated.
20274 2021-06-09 Carl Love <cel@us.ibm.com>
20276 * config/rs6000/altivec.h (vec_signextll, vec_signexti, vec_signextq):
20277 Add define for new builtins.
20278 * config/rs6000/altivec.md(altivec_vreveti2): Add define_expand.
20279 * config/rs6000/rs6000-builtin.def (VSIGNEXTI, VSIGNEXTLL): Add
20280 overloaded builtin definitions.
20281 (VSIGNEXTSB2W, VSIGNEXTSH2W, VSIGNEXTSB2D, VSIGNEXTSH2D,VSIGNEXTSW2D,
20282 VSIGNEXTSD2Q): Add builtin expansions.
20283 (SIGNEXT): Add P10 overload definition.
20284 * config/rs6000/rs6000-call.c (P9V_BUILTIN_VEC_VSIGNEXTI, P9V_BUILTIN_VEC_VSIGNEXTLL,
20285 P10_BUILTIN_VEC_SIGNEXT): Add overloaded argument definitions.
20286 * config/rs6000/vsx.md (vsx_sign_extend_v2di_v1ti): Add define_insn.
20287 (vsignextend_v2di_v1ti, vsignextend_qi_<mode>, vsignextend_hi_<mode>,
20288 vsignextend_si_v2di)[VIlong]: Add define_expand.
20289 Make define_insn vsx_sign_extend_si_v2di visible.
20290 * doc/extend.texi: Add documentation for the vec_signexti,
20291 vec_signextll builtins and vec_signextq.
20293 2021-06-09 Carl Love <cel@us.ibm.com>
20295 * config/rs6000/rs6000.c (__fixkfti, __fixunskfti, __floattikf,
20296 __floatuntikf): Names changed to __fixkfti_sw, __fixunskfti_sw,
20297 __floattikf_sw, __floatuntikf_sw respectively.
20298 * config/rs6000/rs6000.md (floatti<mode>2, floatunsti<mode>2,
20299 fix_trunc<mode>ti2, fixuns_trunc<mode>ti2): Add
20300 define_insn for mode IEEE 128.
20302 2021-06-09 Carl Love <cel@us.ibm.com>
20304 * config/rs6000/altivec.md (altivec_vslq, altivec_vsrq):
20305 Rename to altivec_vslq_<mode>, altivec_vsrq_<mode>, mode VEC_TI.
20306 * config/rs6000/vector.md (VEC_TI): Was named VSX_TI in vsx.md.
20307 (vashlv1ti3): Change to vashl<mode>3, mode VEC_TI.
20308 (vlshrv1ti3): Change to vlshr<mode>3, mode VEC_TI.
20309 * config/rs6000/vsx.md (VSX_TI): Remove define_mode_iterator. Update
20310 uses of VSX_TI to VEC_TI.
20312 2021-06-09 Carl Love <cel@us.ibm.com>
20314 * config/rs6000/dfp.md (floattitd2, fixtdti2): New define_insns.
20316 2021-06-09 Carl Love <cel@us.ibm.com>
20318 * config/rs6000/altivec.h (vec_dive, vec_mod): Add define for new
20320 * config/rs6000/altivec.md (UNSPEC_VMULEUD, UNSPEC_VMULESD,
20321 UNSPEC_VMULOUD, UNSPEC_VMULOSD): New unspecs.
20322 (altivec_eqv1ti, altivec_gtv1ti, altivec_gtuv1ti, altivec_vmuleud,
20323 altivec_vmuloud, altivec_vmulesd, altivec_vmulosd, altivec_vrlq,
20324 altivec_vrlqmi, altivec_vrlqmi_inst, altivec_vrlqnm,
20325 altivec_vrlqnm_inst, altivec_vslq, altivec_vsrq, altivec_vsraq,
20326 altivec_vcmpequt_p, altivec_vcmpgtst_p, altivec_vcmpgtut_p): New
20328 (vec_widen_umult_even_v2di, vec_widen_smult_even_v2di,
20329 vec_widen_umult_odd_v2di, vec_widen_smult_odd_v2di, altivec_vrlqmi,
20330 altivec_vrlqnm): New define_expands.
20331 * config/rs6000/rs6000-builtin.def (VCMPEQUT_P, VCMPGTST_P,
20332 VCMPGTUT_P): Add macro expansions.
20333 (BU_P10V_AV_P): Add builtin predicate definition.
20334 (VCMPGTUT, VCMPGTST, VCMPEQUT, CMPNET, CMPGE_1TI,
20335 CMPGE_U1TI, CMPLE_1TI, CMPLE_U1TI, VNOR_V1TI_UNS, VNOR_V1TI, VCMPNET_P,
20336 VCMPAET_P, VMULEUD, VMULESD, VMULOUD, VMULOSD, VRLQ,
20337 VSLQ, VSRQ, VSRAQ, VRLQNM, DIV_V1TI, UDIV_V1TI, DIVES_V1TI, DIVEU_V1TI,
20338 MODS_V1TI, MODU_V1TI, VRLQMI): New macro expansions.
20339 (VRLQ, VSLQ, VSRQ, VSRAQ, DIVE, MOD): New overload expansions.
20340 * config/rs6000/rs6000-call.c (P10_BUILTIN_VCMPEQUT,
20341 P10V_BUILTIN_CMPGE_1TI, P10V_BUILTIN_CMPGE_U1TI,
20342 P10V_BUILTIN_VCMPGTUT, P10V_BUILTIN_VCMPGTST,
20343 P10V_BUILTIN_CMPLE_1TI, P10V_BUILTIN_VCMPLE_U1TI,
20344 P10V_BUILTIN_DIV_V1TI, P10V_BUILTIN_UDIV_V1TI,
20345 P10V_BUILTIN_VMULESD, P10V_BUILTIN_VMULEUD,
20346 P10V_BUILTIN_VMULOSD, P10V_BUILTIN_VMULOUD,
20347 P10V_BUILTIN_VNOR_V1TI, P10V_BUILTIN_VNOR_V1TI_UNS,
20348 P10V_BUILTIN_VRLQ, P10V_BUILTIN_VRLQMI,
20349 P10V_BUILTIN_VRLQNM, P10V_BUILTIN_VSLQ,
20350 P10V_BUILTIN_VSRQ, P10V_BUILTIN_VSRAQ,
20351 P10V_BUILTIN_VCMPGTUT_P, P10V_BUILTIN_VCMPGTST_P,
20352 P10V_BUILTIN_VCMPEQUT_P, P10V_BUILTIN_VCMPGTUT_P,
20353 P10V_BUILTIN_VCMPGTST_P, P10V_BUILTIN_CMPNET,
20354 P10V_BUILTIN_VCMPNET_P, P10V_BUILTIN_VCMPAET_P,
20355 P10V_BUILTIN_DIVES_V1TI, P10V_BUILTIN_MODS_V1TI,
20356 P10V_BUILTIN_MODU_V1TI):
20357 New overloaded definitions.
20358 (rs6000_gimple_fold_builtin) [P10V_BUILTIN_VCMPEQUT,
20359 P10V_BUILTIN_CMPNET, P10V_BUILTIN_CMPGE_1TI,
20360 P10V_BUILTIN_CMPGE_U1TI, P10V_BUILTIN_VCMPGTUT,
20361 P10V_BUILTIN_VCMPGTST, P10V_BUILTIN_CMPLE_1TI,
20362 P10V_BUILTIN_CMPLE_U1TI]: New case statements.
20363 (rs6000_init_builtins) [bool_V1TI_type_node, int_ftype_int_v1ti_v1ti]:
20365 (altivec_init_builtins): New E_V1TImode case statement.
20366 (builtin_function_type)[P10_BUILTIN_128BIT_VMULEUD,
20367 P10_BUILTIN_128BIT_VMULOUD, P10_BUILTIN_128BIT_DIVEU_V1TI,
20368 P10_BUILTIN_128BIT_MODU_V1TI, P10_BUILTIN_CMPGE_U1TI,
20369 P10_BUILTIN_VCMPGTUT, P10_BUILTIN_VCMPEQUT]: New case statements.
20370 * config/rs6000/rs6000.c (rs6000_handle_altivec_attribute) [E_TImode,
20371 E_V1TImode]: New case statements.
20372 * config/rs6000/rs6000.h (rs6000_builtin_type_index): New enum
20373 value RS6000_BTI_bool_V1TI.
20374 * config/rs6000/vector.md (vector_gtv1ti,vector_nltv1ti,
20375 vector_gtuv1ti, vector_nltuv1ti, vector_ngtv1ti, vector_ngtuv1ti,
20376 vector_eq_v1ti_p, vector_ne_v1ti_p, vector_ae_v1ti_p,
20377 vector_gt_v1ti_p, vector_gtu_v1ti_p, vrotlv1ti3, vashlv1ti3,
20378 vlshrv1ti3, vashrv1ti3): New define_expands.
20379 * config/rs6000/vsx.md (UNSPEC_VSX_DIVSQ, UNSPEC_VSX_DIVUQ,
20380 UNSPEC_VSX_DIVESQ, UNSPEC_VSX_DIVEUQ, UNSPEC_VSX_MODSQ,
20381 UNSPEC_VSX_MODUQ): New unspecs.
20382 (mulv2di3, vsx_div_v1ti, vsx_udiv_v1ti, vsx_dives_v1ti,
20383 vsx_diveu_v1ti, vsx_mods_v1ti, vsx_modu_v1ti, xxswapd_v1ti): New
20385 (vcmpnet): New define_expand.
20386 * doc/extend.texi: Add documentation for the new builtins vec_rl,
20387 vec_rlmi, vec_rlnm, vec_sl, vec_sr, vec_sra, vec_mule, vec_mulo,
20388 vec_div, vec_dive, vec_mod, vec_cmpeq, vec_cmpne, vec_cmpgt, vec_cmplt,
20389 vec_cmpge, vec_cmple, vec_all_eq, vec_all_ne, vec_all_gt, vec_all_lt,
20390 vec_all_ge, vec_all_le, vec_any_eq, vec_any_ne, vec_any_gt, vec_any_lt,
20391 vec_any_ge, vec_any_le.
20393 2021-06-09 Carl Love <cel@us.ibm.com>
20395 * config/rs6000/altivec.md (altivec_vrl<VI_char>mi): Fix
20396 bug in argument generation.
20398 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
20400 * config/arm/iterators.md (<supf>): Remove VCLZQ_U, VCLZQ_S.
20402 * config/arm/mve.md (mve_vclzq_<supf><mode>): Add '@' prefix,
20403 remove <supf> iterator.
20404 (mve_vclzq_u<mode>): New.
20405 * config/arm/neon.md (clz<mode>2): Rename to neon_vclz<mode>.
20406 (neon_vclz<mode): Move to ...
20407 * config/arm/unspecs.md (VCLZQ_U, VCLZQ_S): Remove.
20408 * config/arm/vec-common.md: ... here. Add support for MVE.
20410 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
20412 * config/arm/mve.md (mve_vhaddq_<supf><mode>): Prefix with '@'.
20413 (@mve_vrhaddq_<supf><mode): Likewise.
20414 * config/arm/neon.md (neon_v<r>hadd<sup><mode>): Likewise.
20415 * config/arm/vec-common.md (avg<mode>3_floor, uavg<mode>3_floor)
20416 (avg<mode>3_ceil", uavg<mode>3_ceil): New patterns.
20418 2021-06-09 imba-tjd <109224573@qq.com>
20420 * doc/invoke.texi: Fix typo.
20422 2021-06-09 Roger Sayle <roger@nextmovesoftware.com>
20424 PR middle-end/53267
20425 * fold-const-call.c (fold_const_call_sss) [CASE_CFN_FMOD]:
20426 Support evaluation of fmod/fmodf/fmodl at compile-time.
20428 2021-06-09 Richard Biener <rguenther@suse.de>
20430 PR tree-optimization/100981
20431 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
20432 gimple_get_lhs to also handle calls.
20433 * tree-vect-slp-patterns.c (complex_pattern::build): Transfer
20436 2021-06-09 Richard Biener <rguenther@suse.de>
20438 PR tree-optimization/97832
20439 * tree-vectorizer.h (_slp_tree::failed): New.
20440 * tree-vect-slp.c (_slp_tree::_slp_tree): Initialize
20442 (_slp_tree::~_slp_tree): Free failed.
20443 (vect_build_slp_tree): Retain failed nodes and record
20444 matches in them, copying that back out when running
20445 into a cached fail. Dump start and end of discovery.
20446 (dt_sort_cmp): New.
20447 (vect_build_slp_tree_2): Handle associatable chains
20448 together doing more aggressive operand swapping.
20450 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
20453 * config.gcc (gcc_cv_initfini_array): Set to yes for Linux and
20455 * doc/install.texi: Require glibc 2.1 and binutils 2.12 for
20456 Linux and GNU targets.
20458 2021-06-09 Richard Biener <rguenther@suse.de>
20460 * tree-vect-stmts.c (vect_is_simple_use): Always get dt
20463 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
20465 * config/arc/arc.md (loop_end): Change it to
20466 define_insn_and_split.
20468 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
20470 * config/arc/arc.md (maddhisi4): Use VMAC2H instruction.
20471 (machi): New pattern.
20472 (umaddhisi4): Use VMAC2HU instruction.
20473 (umachi): New pattern.
20475 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
20477 * config/arc/arc-protos.h (arc_split_move_p): New prototype.
20478 * config/arc/arc.c (arc_split_move_p): New function.
20479 (arc_split_move): Clean up.
20480 * config/arc/arc.md (movdi_insn): Clean up, use arc_split_move_p.
20481 (movdf_insn): Likewise.
20482 * config/arc/simdext.md (mov<VWH>_insn): Likewise.
20484 2021-06-09 Uroš Bizjak <ubizjak@gmail.com>
20487 * config/i386/i386.c (print_operand_address_as): Rename "no_rip"
20488 argument to "raw". Do not emit segment overrides when "raw" is true.
20490 2021-06-09 Martin Liska <mliska@suse.cz>
20492 * doc/gcov.texi: Create a proper JSON files.
20493 * doc/invoke.texi: Remove dots in order to make it a valid
20496 2021-06-09 Xionghu Luo <luoxhu@linux.ibm.com>
20498 * config/rs6000/rs6000-p8swap.c (pattern_is_rotate64): New.
20499 (insn_is_load_p): Use pattern_is_rotate64.
20500 (insn_is_swap_p): Likewise.
20501 (quad_aligned_load_p): Likewise.
20502 (const_load_sequence_p): Likewise.
20503 (replace_swapped_aligned_load): Likewise.
20504 (recombine_lvx_pattern): Likewise.
20505 (recombine_stvx_pattern): Likewise.
20507 2021-06-09 Andrew MacLeod <amacleod@redhat.com>
20509 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Use a
20510 fur_stmt source record.
20511 * gimple-range.cc (fur_source::get_operand): Generic range query.
20512 (fur_source::get_phi_operand): New.
20513 (fur_source::register_dependency): New.
20514 (fur_source::query): New.
20515 (class fur_edge): New. Edge source for operands.
20516 (fur_edge::fur_edge): New.
20517 (fur_edge::get_operand): New.
20518 (fur_edge::get_phi_operand): New.
20519 (fur_edge::query): New.
20520 (fur_stmt::fur_stmt): New.
20521 (fur_stmt::get_operand): New.
20522 (fur_stmt::get_phi_operand): New.
20523 (fur_stmt::query): New.
20524 (class fur_depend): New. Statement source and process dependencies.
20525 (fur_depend::fur_depend): New.
20526 (fur_depend::register_dependency): New.
20527 (class fur_list): New. List source for operands.
20528 (fur_list::fur_list): New.
20529 (fur_list::get_operand): New.
20530 (fur_list::get_phi_operand): New.
20531 (fold_range): New. Instantiate appropriate fur_source class and fold.
20532 (fold_using_range::range_of_range_op): Use new API.
20533 (fold_using_range::range_of_address): Ditto.
20534 (fold_using_range::range_of_phi): Ditto.
20535 (imple_ranger::fold_range_internal): Use fur_depend class.
20536 (fold_using_range::range_of_ssa_name_with_loop_info): Use new API.
20537 * gimple-range.h (class fur_source): Now a base class.
20538 (class fur_stmt): New.
20539 (fold_range): New prototypes.
20540 (fur_source::fur_source): Delete.
20542 2021-06-08 Andrew Pinski <apinski@marvell.com>
20544 PR tree-optimization/25290
20545 * tree-ssa-phiopt.c (xor_replacement): Delete.
20546 (tree_ssa_phiopt_worker): Delete use of xor_replacement.
20547 (match_simplify_replacement): Allow one cheap preparation
20548 statement that can be moved to before the if.
20550 2021-06-08 Pat Haugen <pthaugen@linux.ibm.com>
20552 * config/rs6000/power10.md (power10-fused-load, power10-fused-store,
20553 power10-fused_alu, power10-fused-vec, power10-fused-branch): New.
20555 2021-06-08 Jeff Law <jeffreyalaw@gmail.com>
20557 * config/h8300/logical.md (andqi3_1): Move BCLR case into define_insn_and_split.
20558 Create length attribute on define_insn_and_split. Only split for cases which we
20560 (andqi3_1<cczn>): Renamed from andqi3_1_clobber_flags. Only handle AND here and
20561 fix length computation.
20562 (b<code><mode>msx): Combine QImode and HImode H8/SX patterns using iterator.
20564 2021-06-08 Richard Biener <rguenther@suse.de>
20566 PR tree-optimization/100923
20567 * tree-ssa-sccvn.c (valueize_refs_1): Take a pointer to
20568 the operand vector to be valueized.
20569 (valueize_refs): Likewise.
20570 (valueize_shared_reference_ops_from_ref): Adjust.
20571 (valueize_shared_reference_ops_from_call): Likewise.
20572 (vn_reference_lookup_3): Likewise.
20573 (vn_reference_lookup_pieces): Likewise. Re-valueize
20574 with honoring availability when we are about to create
20575 the ao_ref and valueized before.
20576 (vn_reference_lookup): Likewise.
20577 (vn_reference_insert_pieces): Adjust.
20579 2021-06-08 Richard Biener <rguenther@suse.de>
20581 * tree-vectorizer.h (_slp_instance::root_stmt): Change to...
20582 (_slp_instance::root_stmts): ... a vector.
20583 (SLP_INSTANCE_ROOT_STMT): Rename to ...
20584 (SLP_INSTANCE_ROOT_STMTS): ... this.
20585 (slp_root::root): Change to...
20586 (slp_root::roots): ... a vector.
20587 (slp_root::slp_root): Adjust.
20588 * tree-vect-slp.c (_slp_instance::location): Adjust.
20589 (vect_free_slp_instance): Release the root stmt vector.
20590 (vect_build_slp_instance): Adjust.
20591 (vect_analyze_slp): Likewise.
20592 (_bb_vec_info::~_bb_vec_info): Likewise.
20593 (vect_slp_analyze_operations): Likewise.
20594 (vect_bb_vectorization_profitable_p): Likewise. Adjust
20595 costs for the root stmt.
20596 (vect_slp_check_for_constructors): Gather all BIT_INSERT_EXPRs
20598 (vect_slp_analyze_bb_1): Simplify by marking all root stmts
20600 (vectorize_slp_instance_root_stmt): Adjust.
20601 (vect_schedule_slp): Likewise.
20603 2021-06-08 Aldy Hernandez <aldyh@redhat.com>
20605 * gimple-ssa-evrp.c (class ssa_equiv_stack): New.
20606 (ssa_equiv_stack::ssa_equiv_stack): New.
20607 (ssa_equiv_stack::~ssa_equiv_stack): New.
20608 (ssa_equiv_stack::enter): New.
20609 (ssa_equiv_stack::leave): New.
20610 (ssa_equiv_stack::push_replacement): New.
20611 (ssa_equiv_stack::get_replacement): New.
20612 (is_pointer_ssa): New.
20613 (class pointer_equiv_analyzer): New.
20614 (pointer_equiv_analyzer::pointer_equiv_analyzer): New.
20615 (pointer_equiv_analyzer::~pointer_equiv_analyzer): New.
20616 (pointer_equiv_analyzer::set_global_equiv): New.
20617 (pointer_equiv_analyzer::set_cond_equiv): New.
20618 (pointer_equiv_analyzer::get_equiv): New.
20619 (pointer_equiv_analyzer::enter): New.
20620 (pointer_equiv_analyzer::leave): New.
20621 (pointer_equiv_analyzer::get_equiv_expr): New.
20622 (pta_valueize): New.
20623 (pointer_equiv_analyzer::visit_stmt): New.
20624 (pointer_equiv_analyzer::visit_edge): New.
20625 (hybrid_folder::value_of_expr): Call PTA.
20626 (hybrid_folder::value_on_edge): Same.
20627 (hybrid_folder::pre_fold_bb): New.
20628 (hybrid_folder::post_fold_bb): New.
20629 (hybrid_folder::pre_fold_stmt): New.
20630 (rvrp_folder::pre_fold_bb): New.
20631 (rvrp_folder::post_fold_bb): New.
20632 (rvrp_folder::pre_fold_stmt): New.
20633 (rvrp_folder::value_of_expr): Call PTA.
20634 (rvrp_folder::value_on_edge): Same.
20636 2021-06-08 Jakub Jelinek <jakub@redhat.com>
20639 * tree-inline.c (copy_tree_body_r): For OMP_CLAUSE_DEPEND don't
20640 check TREE_CODE if OMP_CLAUSE_DECL is NULL.
20642 2021-06-08 Richard Biener <rguenther@suse.de>
20644 PR middle-end/100951
20645 * tree-vect-generic.c (expand_vector_piecewise): Build a
20646 VECTOR_CST if all elements are constant.
20647 (expand_vector_condition): Likewise.
20648 (lower_vec_perm): Likewise.
20649 (expand_vector_conversion): Likewise.
20651 2021-06-08 Martin Liska <mliska@suse.cz>
20653 * doc/invoke.texi: Document new param evrp-sparse-threshold.
20655 2021-06-08 Martin Liska <mliska@suse.cz>
20657 * genautomata.c (create_automata): Fix typo.
20659 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
20661 PR tree-optimization/100794
20662 * tree-predcom.c (tree_predictive_commoning_loop): Add parameter
20663 allow_unroll_p and only allow unrolling when it's true.
20664 (tree_predictive_commoning): Add parameter allow_unroll_p and
20666 (run_tree_predictive_commoning): Likewise.
20667 (pass_predcom::gate): Check flag_tree_loop_vectorize and
20668 global_options_set.x_flag_predictive_commoning.
20669 (pass_predcom::execute): Adjust for allow_unroll_p.
20671 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
20673 * tree-predcom.c (execute_pred_commoning): Remove update_ssa call.
20674 (tree_predictive_commoning_loop): Factor some cleanup stuffs into
20675 lambda function cleanup, remove scev_reset call, and adjust return
20677 (tree_predictive_commoning): Adjust for different changed values,
20678 only set flag TODO_update_ssa_only_virtuals if changed.
20679 (pass_data pass_data_predcom): Remove TODO_update_ssa_only_virtuals
20680 from todo_flags_finish.
20682 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
20684 * gimple-range-cache.cc (class sbr_sparse_bitmap): New.
20685 (sbr_sparse_bitmap::sbr_sparse_bitmap): New.
20686 (sbr_sparse_bitmap::bitmap_set_quad): New.
20687 (sbr_sparse_bitmap::bitmap_get_quad): New.
20688 (sbr_sparse_bitmap::set_bb_range): New.
20689 (sbr_sparse_bitmap::get_bb_range): New.
20690 (sbr_sparse_bitmap::bb_range_p): New.
20691 (block_range_cache::block_range_cache): initialize bitmap obstack.
20692 (block_range_cache::~block_range_cache): Destruct obstack.
20693 (block_range_cache::set_bb_range): Decide when to utilze the
20694 sparse on entry cache.
20695 * gimple-range-cache.h (block_range_cache): Add bitmap obstack.
20696 * params.opt (-param=evrp-sparse-threshold): New.
20698 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
20700 * bitmap.c (bitmap_set_aligned_chunk): New.
20701 (bitmap_get_aligned_chunk): New.
20702 (test_aligned_chunk): New.
20703 (bitmap_c_tests): Call test_aligned_chunk.
20704 * bitmap.h (bitmap_set_aligned_chunk, bitmap_get_aligned_chunk): New.
20706 2021-06-07 Uroš Bizjak <ubizjak@gmail.com>
20709 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
20711 (ix86_expand_vector_init_one_nonzero): Ditto.
20712 (ix86_expand_vector_init_one_var): Ditto.
20713 (ix86_expand_vector_init_general): Ditto.
20714 * config/i386/mmx.md (vec_initv4qiqi): New expander.
20716 2021-06-07 Jeff Law <jeffreyalaw@gmail.com>
20718 * config/h8300/movepush.md: Change most _clobber_flags
20719 patterns to instead use <cczn> subst.
20720 (movsi_cczn): New pattern with usable CC cases split out.
20721 (movsi_h8sx_cczn): Likewise.
20723 2021-06-07 Martin Liska <mliska@suse.cz>
20725 * common/common-target.def: Split long lines and replace them
20727 * target.def: Likewise.
20728 * doc/tm.texi: Re-generated.
20730 2021-06-07 Jakub Jelinek <jakub@redhat.com>
20733 * fold-const.c (fold_read_from_vector): Return NULL if trying to
20734 read from a CONSTRUCTOR with vector type elements.
20736 2021-06-07 Jakub Jelinek <jakub@redhat.com>
20738 PR middle-end/100898
20739 * tree-inline.c (copy_bb): Only use gimple_call_arg_ptr if memcpy
20740 should copy any arguments. Don't call gimple_call_num_args
20741 on id->call_stmt or call_stmt more than once.
20743 2021-06-07 liuhongt <hongtao.liu@intel.com>
20746 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): Refine
20748 (<insn>v4siv4di2): Delete constraints for define_expand.
20750 2021-06-07 liuhongt <hongtao.liu@intel.com>
20753 * config/i386/i386-expand.c (ix86_expand_builtin): Remove
20754 assignment of cfun->machine->has_explicit_vzeroupper.
20755 * config/i386/i386-features.c
20756 (ix86_add_reg_usage_to_vzerouppers): Delete.
20757 (ix86_add_reg_usage_to_vzeroupper): Ditto.
20758 (rest_of_handle_insert_vzeroupper): Remove
20759 ix86_add_reg_usage_to_vzerouppers, add df_analyze at the end
20761 (gate): Remove cfun->machine->has_explicit_vzeroupper.
20762 * config/i386/i386-protos.h (ix86_expand_avx_vzeroupper):
20764 * config/i386/i386.c (ix86_insn_callee_abi): New function.
20765 (ix86_initialize_callee_abi): Ditto.
20766 (ix86_expand_avx_vzeroupper): Ditto.
20767 (ix86_hard_regno_call_part_clobbered): Adjust for vzeroupper
20769 (TARGET_INSN_CALLEE_ABI): Define as ix86_insn_callee_abi.
20770 (ix86_emit_mode_set): Call ix86_expand_avx_vzeroupper
20772 * config/i386/i386.h (struct GTY(()) machine_function): Delete
20773 has_explicit_vzeroupper.
20774 * config/i386/i386.md (enum unspec): New member
20776 (ABI_DEFAULT,ABI_VZEROUPPER,ABI_UNKNOWN): New
20777 define_constants for insn callee abi index.
20778 * config/i386/predicates.md (vzeroupper_pattern): Adjust.
20779 * config/i386/sse.md (UNSPECV_VZEROUPPER): Deleted.
20780 (avx_vzeroupper): Call ix86_expand_avx_vzeroupper.
20781 (*avx_vzeroupper): Rename to ..
20782 (avx_vzeroupper_callee_abi): .. this, and adjust pattern as
20783 call_insn which has a special vzeroupper ABI.
20784 (*avx_vzeroupper_1): Deleted.
20786 2021-06-07 liuhongt <hongtao.liu@intel.com>
20789 * df-scan.c (df_get_call_refs): When call_insn is a fake call,
20790 it won't use stack pointer reg.
20791 * final.c (leaf_function_p): When call_insn is a fake call, it
20792 won't affect caller as a leaf function.
20793 * reg-stack.c (callee_clobbers_any_stack_reg): New.
20794 (subst_stack_regs): When call_insn doesn't clobber any stack
20795 reg, don't clear the arguments.
20796 * rtl.c (shallow_copy_rtx): Don't clear flag used when orig is
20798 * shrink-wrap.c (requires_stack_frame_p): No need for stack
20799 frame for a fake call.
20800 * rtl.h (FAKE_CALL_P): New macro.
20802 2021-06-06 Eric Botcazou <ebotcazou@adacore.com>
20804 * config/sparc/sparc-protos.h (order_regs_for_local_alloc): Rename
20806 (sparc_order_regs_for_local_alloc): ...this.
20807 (sparc_leaf_reg_remap): Declare.
20808 * config/sparc/sparc.h (ADJUST_REG_ALLOC_ORDER): Adjust.
20809 (LEAF_REG_REMAP): Reimplement as call to sparc_leaf_reg_remap.
20810 * config/sparc/sparc.c (leaf_reg_remap): Delete.
20811 (order_regs_for_local_alloc): Rename to...
20812 (sparc_order_regs_for_local_alloc): ...this.
20813 (sparc_leaf_reg_remap): New function.
20814 (sparc_conditional_register_usage): Do not modify leaf_reg_remap.
20816 2021-06-06 David Edelsohn <dje.gcc@gmail.com>
20818 * config/rs6000/rs6000.c (rs6000_xcoff_asm_output_aligned_decl_common):
20819 Use assemble_name to output BSS section name.
20821 2021-06-06 Uroš Bizjak <ubizjak@gmail.com>
20823 * config/i386/constraints.md (Bs):
20824 Remove boolean operators from match_test RTX.
20827 (M): Use "mode" variable instead of GET_MODE (op) in match_test RTX.
20830 2021-06-06 Martin Liska <mliska@suse.cz>
20832 * doc/extend.texi: Add missing @headitem.
20833 * doc/invoke.texi: Likewise.
20834 * doc/objc.texi: Likewise.
20836 2021-06-06 Martin Liska <mliska@suse.cz>
20838 * genhooks.c (emit_findices): Remove unused function.
20839 (emit_documentation): Do not call emit_findices
20840 and do not search for @Fcode directives.
20842 2021-06-06 Martin Liska <mliska@suse.cz>
20844 * doc/invoke.texi: Remove extra character.
20846 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
20848 * config/sh/sh.md (doloop_end_split): Fix empty split condition.
20850 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
20852 * config/sparc/sparc.md (*snedi<W:mode>_zero_vis3,
20853 *neg_snedi<W:mode>_zero_subxc, *plus_snedi<W:mode>_zero,
20854 *plus_plus_snedi<W:mode>_zero, *minus_snedi<W:mode>_zero,
20855 *minus_minus_snedi<W:mode>_zero): Fix empty split condition.
20857 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
20859 * config/or1k/or1k.md (*movdi): Fix empty split condition.
20861 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
20863 * config/mips/mips.md (<anonymous>, bswapsi2, bswapdi2): Fix empty
20866 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
20868 * config/m68k/m68k.md (*zero_extend_inc, *zero_extend_dec,
20869 *zero_extendsidi2): Fix empty split condition.
20871 2021-06-05 Jeff Law <jeffreyalaw@gmail.com>
20873 * config/h8300/addsub.md: Fix split condition in define_insn_and_split
20875 * config/h8300/bitfield.md: Likewise.
20876 * config/h8300/combiner.md: Likewise.
20877 * config/h8300/divmod.md: Likewise.
20878 * config/h8300/extensions.md: Likewise.
20879 * config/h8300/jumpcall.md: Likewise.
20880 * config/h8300/movepush.md: Likewise.
20881 * config/h8300/multiply.md: Likewise.
20882 * config/h8300/other.md: Likewise.
20883 * config/h8300/shiftrotate.md: Likewise.
20884 * config/h8300/logical.md: Likewise. Fix split pattern to use
20885 code iterator that somehow slipped through.
20887 2021-06-04 Tobias Burnus <tobias@codesourcery.com>
20889 PR middle-end/100905
20890 * tree-nested.c (convert_nonlocal_omp_clauses,
20891 convert_local_omp_clauses): Handle OMP_CLAUSE_BIND.
20893 2021-06-04 Martin Sebor <msebor@redhat.com>
20895 PR middle-end/100732
20896 * gimple-fold.c (gimple_fold_builtin_sprintf): Avoid folding calls
20897 with either source or destination argument of invalid type.
20898 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Avoid checking
20899 calls with arguments of invalid type.
20901 2021-06-04 Martin Sebor <msebor@redhat.com>
20903 * attribs.c (init_attr_rdwr_indices): Use VLA bounds in the expected
20905 (attr_access::vla_bounds): Also handle VLA bounds.
20907 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
20909 * config/i386/predicates.md (GOT_memory_operand):
20910 Implement using match_code RTXes.
20911 (GOT32_symbol_operand): Ditto.
20913 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
20916 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
20918 (ix86_expand_vector_init_general): Ditto.
20919 Use SImode instead of word_mode for logic operations
20920 when GET_MODE_SIZE (mode) < UNITS_PER_WORD.
20921 (expand_vec_perm_even_odd_1): Assert that V2HI mode should be
20922 implemented by expand_vec_perm_1.
20923 (expand_vec_perm_broadcast_1): Assert that V2HI and V4HI modes
20924 should be implemented using standard shuffle patterns.
20925 (ix86_vectorize_vec_perm_const): Handle V2HImode. Add V4HI and
20926 V2HI modes to modes, implementable with shuffle for one operand.
20927 * config/i386/mmx.md (*punpckwd): New insn_and_split pattern.
20928 (*pshufw_1): New insn pattern.
20929 (*vec_dupv2hi): Ditto.
20930 (vec_initv2hihi): New expander.
20932 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
20934 * config/arm/vfp.md (no_literal_pool_df_immediate,
20935 no_literal_pool_sf_immediate): Fix empty split condition.
20937 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
20939 * config/i386/i386.md (*load_tp_x32_zext, *add_tp_x32_zext,
20940 *tls_dynamic_gnu2_combine_32): Fix empty split condition.
20941 * config/i386/sse.md (*<sse2_avx2>_pmovmskb_lt,
20942 *<sse2_avx2>_pmovmskb_zext_lt, *sse2_pmovmskb_ext_lt,
20943 *<sse4_1_avx2>_pblendvb_lt): Likewise.
20945 2021-06-04 Jakub Jelinek <jakub@redhat.com>
20948 * config/i386/i386-expand.c (ix86_expand_vector_init): Handle
20949 concatenation from half-sized modes with TImode elements.
20951 2021-06-04 Claudiu Zissulescu <claziss@synopsys.com>
20953 * config/arc/arc.c (arc_override_options): Disable millicode
20954 thunks when RF16 is on.
20956 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
20958 * config/rs6000/rs6000.h (PROMOTE_MODE): Remove.
20960 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
20962 * config/rs6000/rs6000-call.c (rs6000_promote_function_mode):
20963 Replace PROMOTE_MODE marco with its content.
20965 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
20967 * config/cris/cris.md (*addi_reload): Fix empty split condition.
20969 2021-06-03 Jim Wilson <jimw@sifive.com>
20971 * config.gcc (riscv*-*-*): If --with-riscv-attribute not used,
20972 turn it on for all riscv targets.
20974 2021-06-03 Uroš Bizjak <ubizjak@gmail.com>
20977 * config/i386/i386-expand.c (ix86_expand_vector_set):
20978 Handle V2HI and V4QI modes.
20979 (ix86_expand_vector_extract): Ditto.
20980 * config/i386/mmx.md (*pinsrw): New insn pattern.
20983 (*pextrw_zext): Ditto.
20985 (*pextrb_zext): Ditto.
20986 (vec_setv2hi): New expander.
20987 (vec_extractv2hihi): Ditto.
20988 (vec_setv4qi): Ditto.
20989 (vec_extractv4qiqi): Ditto.
20990 (vec_setv8qi): Enable only for TARGET_SSE4_1.
20991 (vec_extractv8qiqi): Ditto.
20993 2021-06-03 Aaron Sawdey <acsawdey@linux.ibm.com>
20995 * config/rs6000/genfusion.pl (gen_logical_addsubf): Fix input
20996 order to subf instruction.
20997 * config/rs6000/fusion.md: Regenerate.
20999 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
21001 * calls.c (get_size_range): Use range_of_expr instead of
21002 determine_value_range.
21003 * tree-affine.c (expr_to_aff_combination): Same.
21004 * tree-data-ref.c (split_constant_offset): Same.
21005 * tree-vrp.c (determine_value_range_1): Remove.
21006 (determine_value_range): Remove.
21007 * tree-vrp.h (determine_value_range): Remove.
21009 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
21011 * function-tests.c (test_ranges): Call gimple_range_tests.
21012 * gimple-range-cache.cc (ranger_cache::range_of_expr): Pass stmt
21014 * gimple-range.cc (fur_source::get_operand): Do not call
21015 get_tree_range or gimple_range_global.
21017 (get_tree_range): Move to value-query.cc.
21018 Call get_arith_expr_range.
21019 (gimple_ranger::range_of_expr): Add argument to get_tree_range.
21020 Include gimple-range-tests.cc.
21021 * gimple-range.h (fold_range): Add argument.
21022 (get_tree_range): Remove.
21023 * selftest.h (gimple_range_tests): New.
21024 * value-query.cc (global_range_query::range_of_expr): Add
21026 (range_query::get_tree_range): Move from gimple-range.cc.
21027 * value-query.h (class range_query): Add get_tree_range and
21028 get_arith_expr_range. Make fur_source a friend.
21029 * vr-values.c (vr_values::range_of_expr): Pass stmt to
21031 * gimple-range-tests.cc: New file.
21033 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
21035 * gimple-range.cc (gimple_ranger::export_global_ranges): Call
21036 update_global_range.
21037 * value-query.cc (update_global_range): New.
21038 * value-query.h (update_global_range): New.
21040 2021-06-03 David Malcolm <dmalcolm@redhat.com>
21042 * diagnostic-show-locus.c (diagnostic_show_locus): Don't reject
21043 printing the same location twice if there are fix-it hints,
21044 multiple locations, or a label.
21046 2021-06-03 Andre Vieira <andre.simoesdiasvieira@arm.com>
21048 * tree-vect-loop.c (vect_transform_loop): Use main loop's various'
21049 thresholds to narrow the upper bound on epilogue iterations.
21051 2021-06-03 Christophe Lyon <christophe.lyon@linaro.org>
21053 * config/arm/mve.md (mve_vabsq_f<mode>): Use 'abs' instead of unspec.
21054 (mve_vabsq_s<mode>): Likewise.
21055 * config/arm/neon.md (abs<mode>2): Rename to neon_abs<mode>2.
21056 * config/arm/unspecs.md (VABSQ_F, VABSQ_S): Delete.
21057 * config/arm/vec-common.md (neg<mode>2): Rename to
21058 <absneg_str><mode>2.
21060 2021-06-03 Claudiu Zissulescu <claziss@synopsys.com>
21062 * common/config/arc/arc-common.c (arc_option_optimization_table):
21063 Remove malign-call.
21064 * config/arc/arc.c (arc_unalign_branch_p): Remove unused function.
21065 * config/arc/arc.h (TARGET_MIXED_CODE): Remove macro.
21066 (INDEX_REG_CLASS): Only refer to GENERAL_REGS.
21067 * config/arc/arc.md (abssi2_mixed): Remove pattern.
21068 * config/arc/arc.opt (munalign-prob-threshold): Mark it obsolete.
21069 (malign-call): Likewise.
21070 (mmixed-code): Likewise.
21071 * doc/invoke.texi (ARC): Update doc.
21073 2021-06-03 Martin Liska <mliska@suse.cz>
21075 * common.opt: Use proper Enum values.
21076 * opts.c (COVERAGE_SANITIZER_OPT): Remove.
21077 (parse_sanitizer_options): Handle only sanitizer_opts.
21078 (common_handle_option): Just assign value.
21080 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
21083 * tree-inline.c (inline_forbidden_p): Remove test on return type.
21085 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
21087 * dwarf2out.c (loc_list_from_tree_1) <FUNCTION_DECL>: Also generate
21088 DW_OP_GNU_variable_value referencing an existing DIE at file scope.
21089 (type_byte_size): Inline into...
21090 (add_byte_size_attribute): ...this and call add_scalar_info.
21092 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
21094 * dwarf2out.c (mem_loc_descriptor) <UDIV>: Fix typo.
21095 (typed_binop_from_tree): New function.
21096 (loc_list_from_tree_1) <EXACT_DIV_EXPR>: For an unsigned type,
21097 turn a divide by a power of 2 into a shift.
21098 <CEIL_DIV_EXPR>: For an unsigned type, use a signed divide if the
21099 size of the mode is lower than DWARF2_ADDR_SIZE; otherwise, do a
21100 typed divide by calling typed_binop_from_tree.
21102 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
21104 * dwarf2out.c (scompare_loc_descriptor): Fix head comment.
21105 (is_handled_procedure_type): Likewise.
21106 (struct loc_descr_context): Add strict_signedness field.
21107 (resolve_args_picking_1): Deal with DW_OP_[GNU_]deref_type,
21108 DW_OP_[GNU_]convert and DW_OP_[GNU_]reinterpret.
21109 (resolve_args_picking): Minor tweak.
21110 (function_to_dwarf_procedure): Initialize strict_signedness field.
21111 (type_byte_size): Likewise.
21112 (field_byte_offset): Likewise.
21113 (gen_descr_array_type_die): Likewise.
21114 (gen_variant_part): Likewise.
21115 (loc_list_from_tree_1) <CALL_EXPR>: Tidy up and set strict_signedness
21116 to true when a context is present before evaluating the arguments.
21117 <COND_EXPR>: Do not generate a useless comparison with zero.
21118 When dereferencing an address, if strict_signedness is true and the
21119 type is small and signed, use DW_OP_deref_type to do the dereference
21120 and then DW_OP_convert to convert back to the generic type.
21122 2021-06-03 Jakub Jelinek <jakub@redhat.com>
21125 * tree-inline.c (copy_tree_body_r): Handle iterators on
21126 OMP_CLAUSE_AFFINITY or OMP_CLAUSE_DEPEND.
21128 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
21130 * config/arc/arc.md (*bbit_di): Remove.
21132 2021-06-02 Christoph Muellner <cmuellner@gcc.gnu.org>
21134 PR rtl-optimization/100264
21135 * ree.c (get_sub_rtx): Ignore SET expressions without register
21136 destinations and remove assertion, as it is not valid anymore
21137 with this new behaviour.
21138 (merge_def_and_ext): Eliminate destination check for register
21139 as such SET expressions can't occur anymore.
21140 (combine_reaching_defs): Likewise.
21142 2021-06-02 Jakub Jelinek <jakub@redhat.com>
21145 * config/xtensa/xtensa.h (LEAF_REG_REMAP): Cast REGNO to int to avoid
21146 -Wtype-limits warnings.
21147 (DWARF_FRAME_REGISTER): Rewrite into ternary operator with addition
21148 in operands to avoid -Wsign-compare warnings.
21150 2021-06-02 Pat Haugen <pthaugen@linux.ibm.com>
21152 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): Use
21155 2021-06-02 Vineet Gupta <vgupta@synopsys.com>
21157 * config/arc/arc.h (TARGET_CPU_DEFAULT): Change to hs38_linux.
21159 2021-06-02 Ilya Leoshkevich <iii@linux.ibm.com>
21161 * config/s390/s390.md(*ashrdi3_31<setcc><cconly>): Use a single
21163 * config/s390/subst.md(cconly_subst): Use a single constraint
21164 in (match_scratch).
21166 2021-06-02 Martin Liska <mliska@suse.cz>
21168 * ipa-icf.h: Use auto_vec for memory_access_types.
21170 2021-06-02 Jeff Law <jeffreyalaw@gmail.com>
21172 * config/h8300/h8300-protos.h (compute_a_shift_length): Drop unused
21173 argument from prototype.
21174 (output_logical_op): Add rtx_code argument.
21175 (compute_logical_op_length): Likewise.
21176 * config/h8300/h8300.c (h8300_and_costs): Pass additional argument
21177 to compute_a_shift_length.
21178 (output_logical_op); New argument with the rtx code rather than
21179 extracting it from an operand. Handle QImode too.
21180 (compute_logical_op_length): Similary.
21181 (compute_a_shift_length): Drop unused argument.
21182 * config/h8300/h8300.md (logicals): New code iterator.
21183 * config/h8300/logical.md (<code><mode>3 expander): Combine
21184 the "and" expander with the "ior"/"xor" expander.
21185 (bclr<mode>msx): Combine the QI/HI mode patterns.
21186 (<logical><mode>3 insns): Use code iterator rather than match_operator.
21187 Handle QImode as well. Update call to output_logical_op and
21188 compute_logical_op_length to pass in rtx_code
21189 Fix split condition on all define_insn_and_split patterns.
21190 (one_cmpl<mode>2<cczn>): Use <cczn> to support both clobbering
21191 the flags and setting ZN via existing define_subst.
21192 * config/h8300/shiftrotate.md: Drop unused argument from
21193 calls to compute_a_shift_length.
21194 Signed-off-by: Jeff Law <jeffreyalaw@gmail.com>
21196 2021-06-01 Andrew Pinski <apinski@marvell.com>
21198 PR tree-optimization/25290
21199 * tree-ssa-phiopt.c (match_simplify_replacement):
21201 (tree_ssa_phiopt_worker): Use match_simplify_replacement.
21202 (two_value_replacement): Change the comment about
21203 conditional_replacement.
21204 (conditional_replacement): Delete.
21206 2021-06-01 Andrew Pinski <apinski@marvell.com>
21208 PR tree-optimization/95481
21209 * tree-tailcall.c (find_tail_calls): Handle empty typed
21212 2021-06-01 Andrew Pinski <apinski@marvell.com>
21214 * gimplify.c (zero_sized_field_decl): Delete
21215 (zero_sized_type): Delete
21216 (gimplify_init_ctor_eval): Use is_empty_type instead
21217 of zero_sized_field_decl.
21218 (gimplify_modify_expr): Use is_empty_type instead of
21221 2021-06-01 Jason Merrill <jason@redhat.com>
21224 * tree.h (CALL_FROM_NEW_OR_DELETE_P): Adjust comment.
21226 2021-06-01 Jason Merrill <jason@redhat.com>
21229 * diagnostic.h (warning_enabled_at): Declare.
21230 * diagnostic.c (diagnostic_enabled): Factor out from...
21231 (diagnostic_report_diagnostic): ...here.
21232 (warning_enabled_at): New.
21234 2021-06-01 Aldy Hernandez <aldyh@redhat.com>
21236 * gimple-ssa-evrp.c: Enable exporting of global ranges.
21238 2021-06-01 Martin Liska <mliska@suse.cz>
21241 * doc/invoke.texi: Mention that -fgcse-after-reload
21242 is enabled with -O3.
21244 2021-06-01 liuhongt <hongtao.liu@intel.com>
21246 PR tree-optimization/98365
21247 * tree-if-conv.c (strip_nop_cond_scalar_reduction): New function.
21248 (is_cond_scalar_reduction): Handle nop_expr in cond scalar reduction.
21249 (convert_scalar_cond_reduction): Ditto.
21250 (predicate_scalar_phi): Ditto.
21252 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
21254 PR tree-optimization/100781
21255 * gimple-range-cache.cc (ranger_cache::ranger_cache): Enable new
21256 value calculation by default.
21257 (ranger_cache::enable_new_values): New.
21258 (ranger_cache::disable_new_values): New.
21259 (ranger_cache::push_poor_value): Check if new values are allowed.
21260 * gimple-range-cache.h (class ranger_cache): New member/methods.
21261 * gimple-range.cc (gimple_ranger::range_of_expr): Check for debug
21262 statement, and disable/renable new value calculation.
21264 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
21266 * gimple-range-cache.cc (ranger_cache::ssa_range_in_bb): Delete.
21267 (ranger_cache::range_of_def): New.
21268 (ranger_cache::entry_range): New.
21269 (ranger_cache::exit_range): New.
21270 (ranger_cache::range_of_expr): Adjust.
21271 (ranger_cache::range_on_edge): Adjust.
21272 (ranger_cache::propagate_cache): Call exit_range directly.
21273 * gimple-range-cache.h (class ranger_cache): Adjust.
21275 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
21277 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust for
21278 gori_compute being a member rather than base class.
21279 dervied call to member call.
21280 (ranger_cache::dump): No longer dump gori_map.
21281 (ranger_cache::dump_bb): New.
21282 (ranger_cache::get_non_stale_global_range): Adjust for gori_compute
21283 being a member rather than base class.
21284 (ranger_cache::set_global_range): Ditto.
21285 (ranger_cache::ssa_range_in_bb): Ditto.
21286 (ranger_cache::range_of_expr): New.
21287 (ranger_cache::range_on_edge): New.
21288 (ranger_cache::block_range): Adjust for gori_computes. Debug changes.
21289 (ranger_cache::propagate_cache): Adjust debugging output.
21290 (ranger_cache::fill_block_cache): Adjust for gori_computes. Debug
21292 * gimple-range-cache.h (class ranger_cache): Make gori_compute a
21293 member, and inherit from range_query instead.
21294 (ranger_cache::dump_bb): New. split from dump.
21295 * gimple-range-gori.cc (gori_compute::ssa_range_in_bb): Delete.
21296 (gori_compute::expr_range_at_stmt): Delete.
21297 (gori_compute::compute_name_range_op): Delete.
21298 (gori_compute::compute_operand_range_switch): Add fur_source.
21299 (gori_compute::compute_operand_range): Add fur_source param, inline
21300 old compute_name_range_op and optimize_logical_operands.
21301 (struct tf_range): Delete.
21302 (gori_compute::logical_combine): Adjust
21303 (gori_compute::optimize_logical_operands): Delete.
21304 (gori_compute::compute_logical_operands_in_chain): Delete.
21305 (gori_compute::compute_logical_operands): Adjust.
21306 (gori_compute::compute_operand1_range): Adjust to fur_source.
21307 (gori_compute::compute_operand2_range): Ditto.
21308 (gori_compute::compute_operand1_and_operand2_range): Ditto.
21309 (gori_compute::outgoing_edge_range_p): Add range_query parameter,
21310 and adjust to fur_source.
21311 * gimple-range-gori.h (class gori_compute): Simplify and adjust to
21312 range_query and fur_source.
21313 * gimple-range.cc (gimple_ranger::range_on_edge): Query range_on_edge
21314 from the ranger_cache..
21315 (gimple_ranger::fold_range_internal): Adjust to base class change of
21317 (gimple_ranger::dump_bb): Adjust dump.
21318 * gimple-range.h (gimple_ranger):export gori computes object.
21320 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
21322 PR tree-optimization/100774
21323 * gimple-range-cache.cc (ranger_cache::get_non_stale_global_range):
21324 Constant values are also not stale.
21325 (ranger_cache::set_global_range): Range invariant values should also
21326 have the correct timestamp.
21328 2021-05-31 Martin Liska <mliska@suse.cz>
21330 * tree-streamer-in.c (unpack_ts_function_decl_value_fields):
21331 Unpack FUNCTION_DECL_DECL_TYPE.
21332 * tree-streamer-out.c (pack_ts_function_decl_value_fields):
21333 Stream FUNCTION_DECL_DECL_TYPE instead of
21334 DECL_IS_OPERATOR_NEW_P.
21335 * tree.h (set_function_decl_type): Use FUNCTION_DECL_DECL_TYPE
21337 (DECL_IS_OPERATOR_NEW_P): Likewise.
21338 (DECL_IS_OPERATOR_DELETE_P): Likewise.
21339 (DECL_LAMBDA_FUNCTION_P): Likewise.
21341 2021-05-31 Richard Biener <rguenther@suse.de>
21344 * internal-fn.c (expand_SHUFFLEVECTOR): Define.
21345 * internal-fn.def (SHUFFLEVECTOR): New.
21346 * internal-fn.h (expand_SHUFFLEVECTOR): Declare.
21347 * doc/extend.texi: Document __builtin_shufflevector.
21349 2021-05-31 Peter Bergner <bergner@linux.ibm.com>
21352 * config/rs6000/predicates.md(mma_assemble_input_operand): Allow
21353 indexed form addresses.
21355 2021-05-29 Jeff Law <jlaw@tachyum.com>
21357 * config/h8300/h8300.c (h8300_emit_stack_adjustment): Drop unused
21358 parameter. Call callers fixed.
21360 (output_plussi): Add FALLTHRU markers.
21361 (h8300_shift_needs_scratch_p): Add gcc_unreachable marker.
21363 2021-05-29 Jakub Jelinek <jakub@redhat.com>
21365 PR middle-end/99928
21366 * gimplify.c (gimplify_scan_omp_clauses): For taskloop simd
21367 combined with parallel, make sure to add shared clause to
21368 parallel for explicit linear clause.
21370 2021-05-29 Aldy Hernandez <aldyh@redhat.com>
21372 PR tree-optimization/100787
21373 * gimple-ssa-evrp.c: Disable exporting of global ranges.
21375 2021-05-28 Jason Merrill <jason@redhat.com>
21377 * tree-iterator.h (struct tree_stmt_iterator): Add operator++,
21378 operator--, operator*, operator==, and operator!=.
21379 (class tsi_range): New.
21381 2021-05-28 Richard Biener <rguenther@suse.de>
21383 PR tree-optimization/100778
21384 * tree-vect-slp.c (vect_build_slp_tree_1): Prevent possibly
21385 trapping ops in different BBs.
21387 2021-05-28 Richard Biener <rguenther@suse.de>
21390 * tree-inline.c (copy_bb): When processing __builtin_va_arg_pack
21391 copy fntype from original call.
21393 2021-05-28 Martin Liska <mliska@suse.cz>
21395 PR gcov-profile/100751
21396 * doc/gcov.texi: Revert partially a hunk that was wrong.
21398 2021-05-28 Cooper Qu <cooper.qu@linux.alibaba.com>
21400 * config/csky/csky-linux-elf.h (HAVE_sync_compare_and_swapqi):
21402 (HAVE_sync_compare_and_swaphi): Likewise.
21403 (HAVE_sync_compare_and_swapsi): Likewise.
21405 2021-05-28 Jakub Jelinek <jakub@redhat.com>
21407 PR middle-end/99928
21408 * tree.h (OMP_CLAUSE_MAP_IMPLICIT): Define.
21410 2021-05-28 Tobias Burnus <tobias@codesourcery.com>
21412 * gimplify.c (gimplify_omp_affinity): New.
21413 (gimplify_scan_omp_clauses): Call it; remove affinity clause afterwards.
21414 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_AFFINITY.
21415 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_AFFINITY.
21416 * tree.c (omp_clause_num_ops, omp_clause_code_name): Add clause.
21417 (walk_tree_1): Handle OMP_CLAUSE_AFFINITY.
21419 2021-05-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
21420 Richard Biener <rguenther@suse.de>
21422 * match.pd <popcount & / + pattern matching>:
21423 When generating popcount directly fails, try doing it in two halves.
21425 2021-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
21427 * Makefile.in (generated_files): Add gimple-match.c and
21430 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
21432 * gensupport.c (alter_predicate_for_insn): Handle MATCH_DUP.
21434 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
21436 * gensupport.c (alter_constraints): Add MATCH_SCRATCH case.
21438 2021-05-28 Kewen Lin <linkw@linux.ibm.com>
21440 PR tree-optimization/99398
21441 * tree-ssa-forwprop.c (simplify_permutation): Optimize some cases
21442 where the fed operands are CTOR/CST and propagated through
21443 VIEW_CONVERT_EXPR. Call vec_perm_indices::new_shrunk_vector.
21444 * vec-perm-indices.c (vec_perm_indices::new_shrunk_vector): New
21446 * vec-perm-indices.h (vec_perm_indices::new_shrunk_vector): New
21449 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
21451 * config/i386/mmx.md (addv2sf3): Do not call
21452 ix86_fixup_binary_operands_no_copy.
21455 (<smaxmin:code>v2sf3): Ditto.
21456 (<plusminus:insn><MMXMODEI:mode>3): Ditto.
21457 (<plusminus:insn><VI_32:mode>3): Remove expander.
21458 (<plusminus:insn><VI_32:mode>3): Rename from
21459 "*<plusminus:insn><VI_32:mode>3".
21460 (mulv4hi): Do not call ix86_fixup_binary_operands_no_copy.
21461 (mulv2hi3): Remove expander.
21462 (mulv2hi3): Rename from *mulv2hi3.
21463 (<s>mulv2hi3_highpart): Remove expander.
21464 (<s>mulv2hi3_highpart): Rename from *<s>mulv2hi3_highpart.
21465 (<smaxmin:code><MMXMODE14:mode>3): Rename from
21466 "*mmx_<smaxmin:code><MMXMODE14:mode>3".
21467 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Remove expander.
21468 (SMAXMIN_MMXMODEI): Remove mode iterator.
21469 (<smaxmin:code>v4hi3): New expander.
21470 (<smaxmin:code>v4qi3): Rename from *<smaxmin:code>v4qi3.
21471 (<smaxmin:code>v2hi3): Rename from *<smaxmin:code>v2hi3.
21472 (<smaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
21473 (SMAXMIN_VI_32): Remove mode iterator.
21474 (<umaxmin:code><MMXMODE24:mode>3): Rename from
21475 "*mmx_<umaxmin:code><MMXMODE24:mode>3".
21476 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Remove expander.
21477 (UMAXMIN_MMXMODEI): Remove mode iterator.
21478 (<umaxmin:code>v8qi3): New expander.
21479 (<umaxmin:code>v4qi3): Rename from *<umaxmin:code>v4qi3.
21480 (<umaxmin:code>v2hi3): Rename from *<umaxmin:code>v2hi3.
21481 (<umaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
21482 (UMAXMIN_VI_32): Remove mode iterator.
21483 (<any_shift:insn>v2hi3): Remove expander.
21484 (<any_shift:insn>v2hi3): Rename from *<any_shift:insn>v2hi3.
21485 (<any_logic:code><MMXMODEI:mode>3): Do not call
21486 ix86_fixup_binary_operands_no_copy.
21487 (<any_logic:code><VI_32:mode>3): Remove expander.
21488 (<any_logic:code><VI_32:mode>3): Rename from
21489 "*<any_logic:code><VI_32:mode>3".
21490 (uavg<mode>3_ceil): Do not call ix86_fixup_binary_operands_no_copy.
21491 * config/i386/sse.md (div<VF2:mode>3): Do not call
21492 ix86_fixup_binary_operands_no_copy.
21493 (div<VF1:mode>3): Ditto.
21494 (<maxmin:code><VI8_AVX2_AVX512F:mode>3): Ditto.
21495 (smulhrsv4hi3): Ditto.
21496 (smulhrsv2hi3): Ditto.
21498 2021-05-27 Martin Sebor <msebor@redhat.com>
21500 * ggc.h (gt_ggc_mx): Add overloads for all integers.
21502 * hash-map.h (class hash_map): Add pch_nx_helper overloads for all
21504 (hash_map::operator==): New function.
21506 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
21509 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
21510 For TARGET_XOP bypass SSE comparisons for all supported vector modes.
21511 * config/i386/mmx.md (*xop_maskcmp<MMXMODEI:mode>3): New insn pattern.
21512 (*xop_maskcmp<VI_32:mode>3): Ditto.
21513 (*xop_maskcmp_uns<MMXMODEI:mode>3): Ditto.
21514 (*xop_maskcmp_uns<VI_32:mode>3): Ditto.
21516 2021-05-27 Richard Earnshaw <rearnsha@arm.com>
21519 * config/arm/arm.c (arm_configure_build_target): Remove parameter
21520 opts_set, directly check opts parameters for being non-null.
21521 (arm_option_restore): Update call to arm_configure_build_target.
21522 (arm_option_override): Likewise.
21523 (arm_can_inline_p): Likewise.
21524 (arm_valid_target_attribute_tree): Likewise.
21525 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
21526 * config/arm/arm-protos.h (arm_configure_build_target): Adjust
21529 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
21531 * vr-values.c (simplify_conversion_using_ranges): Use
21532 get_range_query instead of get_global_range_query.
21534 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
21536 * gimple-range.cc (get_range_global): Move to value-query.cc.
21537 (gimple_range_global): Same.
21538 (get_global_range_query): Same.
21539 (global_range_query::range_of_expr): Same.
21540 * gimple-range.h (class global_range_query): Move to
21542 (gimple_range_global): Same.
21543 * tree-ssanames.c (get_range_info): Move to value-query.cc.
21544 (get_ptr_nonnull): Same.
21545 * tree-ssanames.h (get_range_info): Remove.
21546 (get_ptr_nonnull): Remove.
21547 * value-query.cc (get_ssa_name_range_info): Move from
21549 (get_ssa_name_ptr_info_nonnull): Same.
21550 (get_range_global): Move from gimple-range.cc.
21551 (gimple_range_global): Same.
21552 (get_global_range_query): Same.
21553 (global_range_query::range_of_expr): Same.
21554 * value-query.h (class global_range_query): Move from
21556 (gimple_range_global): Same.
21558 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
21561 * config/i386/mmx.md (uavgv4qi3_ceil): New insn pattern.
21562 (uavgv2hi3_ceil): Ditto.
21564 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
21567 * doc/extend.texi (scalar_storage_order): Rephrase slightly.
21569 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
21571 * tree-ssanames.c (get_range_info): Merge both copies of
21572 get_range_info into one that works with irange.
21573 * tree-ssanames.h (get_range_info): Remove version that works on
21576 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
21578 * builtins.c (check_nul_terminated_array): Convert to get_range_query.
21579 (expand_builtin_strnlen): Same.
21580 (determine_block_size): Same.
21581 * fold-const.c (expr_not_equal_to): Same.
21582 * gimple-fold.c (size_must_be_zero_p): Same.
21583 * gimple-match-head.c: Include gimple-range.h.
21584 * gimple-pretty-print.c (dump_ssaname_info): Convert to get_range_query.
21585 * gimple-ssa-warn-restrict.c
21586 (builtin_memref::extend_offset_range): Same.
21587 * graphite-sese-to-poly.c (add_param_constraints): Same.
21588 * internal-fn.c (get_min_precision): Same.
21589 * ipa-fnsummary.c (set_switch_stmt_execution_predicate): Same.
21590 * ipa-prop.c (ipa_compute_jump_functions_for_edge): Same.
21592 * tree-data-ref.c (split_constant_offset): Same.
21593 (dr_step_indicator): Same.
21594 * tree-dfa.c (get_ref_base_and_extent): Same.
21595 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
21596 * tree-ssa-loop-niter.c (refine_value_range_using_guard): Same.
21597 (determine_value_range): Same.
21598 (record_nonwrapping_iv): Same.
21599 (infer_loop_bounds_from_signedness): Same.
21600 (scev_var_range_cant_overflow): Same.
21601 * tree-ssa-phiopt.c (two_value_replacement): Same.
21602 * tree-ssa-pre.c (insert_into_preds_of_block): Same.
21603 * tree-ssa-reassoc.c (optimize_range_tests_to_bit_test): Same.
21604 * tree-ssa-strlen.c (handle_builtin_stxncpy_strncat): Same.
21606 (dump_strlen_info): Same.
21607 (set_strlen_range): Same.
21608 (maybe_diag_stxncpy_trunc): Same.
21609 (get_len_or_size): Same.
21610 (handle_integral_assign): Same.
21611 * tree-ssa-structalias.c (find_what_p_points_to): Same.
21612 * tree-ssa-uninit.c (find_var_cmp_const): Same.
21613 * tree-switch-conversion.c (bit_test_cluster::emit): Same.
21614 * tree-vect-patterns.c (vect_get_range_info): Same.
21615 (vect_recog_divmod_pattern): Same.
21616 * tree-vrp.c (intersect_range_with_nonzero_bits): Same.
21617 (register_edge_assert_for_2): Same.
21618 (determine_value_range_1): Same.
21619 * tree.c (get_range_pos_neg): Same.
21620 * vr-values.c (vr_values::get_lattice_entry): Same.
21621 (vr_values::update_value_range): Same.
21622 (simplify_conversion_using_ranges): Same.
21624 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
21626 * gimple-ssa-warn-alloca.c (alloca_call_type): Use
21627 get_range_query instead of query argument.
21628 (pass_walloca::execute): Enable and disable global ranger.
21630 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
21632 * gimple-ssa-evrp.c (rvrp_folder::rvrp_folder): Call
21634 (rvrp_folder::~rvrp_folder): Call disable_ranger.
21635 (hybrid_folder::hybrid_folder): Call enable_ranger.
21636 (hybrid_folder::~hybrid_folder): Call disable_ranger.
21638 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
21640 * function.c (allocate_struct_function): Set cfun->x_range_query.
21641 * function.h (struct function): Declare x_range_query.
21642 (get_range_query): New.
21643 (get_global_range_query): New.
21644 * gimple-range-cache.cc (ssa_global_cache::ssa_global_cache):
21645 Remove call to safe_grow_cleared.
21646 * gimple-range.cc (get_range_global): New.
21647 (gimple_range_global): Move from gimple-range.h.
21648 (get_global_range_query): New.
21649 (global_range_query::range_of_expr): New.
21650 (enable_ranger): New.
21651 (disable_ranger): New.
21652 * gimple-range.h (gimple_range_global): Move to gimple-range.cc.
21653 (class global_range_query): New.
21654 (enable_ranger): New.
21655 (disable_ranger): New.
21656 * gimple-ssa-evrp.c (evrp_folder::~evrp_folder): Rename
21657 dump_all_value_ranges to dump.
21658 * tree-vrp.c (vrp_prop::finalize): Same.
21659 * value-query.cc (range_query::dump): New.
21660 * value-query.h (range_query::dump): New.
21661 * vr-values.c (vr_values::dump_all_value_ranges): Rename to...
21662 (vr_values::dump): ...this.
21663 * vr-values.h (class vr_values): Rename dump_all_value_ranges to
21664 dump and make virtual.
21666 2021-05-26 Uroš Bizjak <ubizjak@gmail.com>
21668 * config/i386/i386.c (ix86_autovectorize_vector_modes):
21669 Add V4QImode and V16QImode for TARGET_SSE2.
21670 * doc/sourcebuild.texi (Vector-specific attributes):
21671 Add vect64 and vect32 description.
21673 2021-05-26 Bernd Edlinger <bernd.edlinger@hotmail.de>
21675 * gimple-range-gori.cc (range_def_chain::register_dependency):
21676 Resize m_def_chain when needed.
21678 2021-05-26 Christophe Lyon <christophe.lyon@linaro.org>
21680 * config/arm/mve.md (mve_vaddvq_<supf><mode>): Prefix with '@'.
21681 * config/arm/neon.md (reduc_plus_scal_<mode>): Move to ..
21682 * config/arm/vec-common.md: .. here. Add support for MVE.
21684 2021-05-26 Jakub Jelinek <jakub@redhat.com>
21686 * config/epiphany/epiphany.c (epiphany_print_operand_address): Remove
21688 * config/microblaze/microblaze.c (microblaze_legitimize_address,
21690 microblaze_option_override, print_operand): Likewise.
21691 * config/microblaze/microblaze.md (call_internal_plt,
21692 call_value_intern_plt, call_value_intern): Likewise.
21693 * config/arm/aout.h (ASM_OUTPUT_ALIGN): Likewise.
21694 * config/iq2000/iq2000.md (call_internal1, call_value_internal1,
21695 call_value_multiple_internal1): Likewise.
21696 * config/bfin/bfin.c (symbolic_reference_mentioned_p): Likewise.
21698 2021-05-26 Jan-Benedict Glaw <jbglaw@lug-owl.de>
21700 * config/arc/arc.c (arc_address_cost, arc_print_operand_address,
21701 arc_ccfsm_advance, symbolic_reference_mentioned_p,
21702 arc_raw_symbolic_reference_mentioned_p): Remove register
21705 2021-05-26 Jakub Jelinek <jakub@redhat.com>
21708 * omp-low.c: Include omp-offload.h.
21709 (create_omp_child_function): If current_function_decl has
21710 "omp declare target" attribute and is_gimple_omp_offloaded,
21711 remove that attribute from the copy of attribute list and
21712 add "omp target entrypoint" attribute instead.
21713 (lower_omp_target): Mark .omp_data_sizes.* and .omp_data_kinds.*
21714 variables for offloading if in omp_maybe_offloaded_ctx.
21715 * omp-offload.c (pass_omp_target_link::execute): Nullify second
21716 argument to GOMP_target_data_ext in offloaded code.
21718 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
21720 * config/csky/csky.c (csky_can_change_mode_class): Delete.
21721 For csky, HF/SF mode use the low bits of VREGS.
21723 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
21725 * gimplify.c (gimplify_decl_expr): Do not clear TREE_READONLY on a
21726 DECL which is a reference for OMP.
21728 2021-05-26 Martin Liska <mliska@suse.cz>
21730 PR gcov-profile/100751
21731 * doc/gcov.texi: Document that __gcov_dump can be called just
21732 once and that __gcov_reset resets run-time counters.
21734 2021-05-26 Martin Liska <mliska@suse.cz>
21736 * doc/install.texi: Port relevant part from install-old.texi
21737 and re-generate list of CPUs and systems.
21739 2021-05-26 Martin Liska <mliska@suse.cz>
21741 * Makefile.in: Remove it.
21742 * doc/include/fdl.texi: Update next/previous chapters.
21743 * doc/install.texi: Likewise.
21744 * doc/install-old.texi: Removed.
21746 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
21748 * config/csky/csky.c (ck810_legitimate_index_p): Support
21749 "base + index" with DF mode.
21750 * config/csky/constraints.md ("Y"): New constraint for memory operands
21751 without index register.
21752 * config/csky/csky_insn_fpuv2.md (fpuv3_movdf): Use "Y" instead of "m"
21753 when mov between memory and general registers, and lower their priority.
21754 * config/csky/csky_insn_fpuv3.md (fpuv2_movdf): Likewise.
21756 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
21758 * config/csky/csky.c (TARGET_PROMOTE_PROTOTYPES): Delete.
21760 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
21762 * config/csky/csky.md (untyped_call): Emit clobber for return
21763 registers to mark them used.
21765 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
21767 * config/csky/csky.md (cskyv2_sextend_ldbs): New.
21769 2021-05-26 Andrew Pinski <apinski@marvell.com>
21771 * match.pd (x < 0 ? ~y : y): New patterns.
21773 2021-05-26 Andrew Pinski <apinski@marvell.com>
21775 * match.pd (A?CST1:CST2): Add simplifcations for A?0:+-1, A?+-1:0,
21776 A?POW2:0 and A?0:POW2.
21778 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
21780 * gimple-range-gori.cc (class logical_stmt_cache): Delete
21781 (logical_stmt_cache::logical_stmt_cache ): Delete.
21782 (logical_stmt_cache::~logical_stmt_cache): Delete.
21783 (logical_stmt_cache::cache_entry::dump): Delete.
21784 (logical_stmt_cache::get_range): Delete.
21785 (logical_stmt_cache::cached_name ): Delete.
21786 (logical_stmt_cache::same_cached_name): Delete.
21787 (logical_stmt_cache::cacheable_p): Delete.
21788 (logical_stmt_cache::slot_diagnostics ): Delete.
21789 (logical_stmt_cache::dump): Delete.
21790 (gori_compute_cache::gori_compute_cache): Delete.
21791 (gori_compute_cache::~gori_compute_cache): Delete.
21792 (gori_compute_cache::compute_operand_range): Delete.
21793 (gori_compute_cache::cache_stmt): Delete.
21794 * gimple-range-gori.h (gori_compute::compute_operand_range): Remove
21796 (class gori_compute_cache): Delete.
21798 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
21800 * gimple-range.cc (fold_using_range::range_of_range_op): Use m_gori
21802 (fold_using_range::range_of_address): Adjust.
21803 (fold_using_range::range_of_phi): Adjust.
21804 * gimple-range.h (class fur_source): Adjust.
21805 (fur_source::fur_source): Adjust.
21807 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
21809 * gimple-range-gori.cc (gori_compute::expr_range_at_stmt): Rename
21810 from expr_range_in_bb and adjust.
21811 (gori_compute::compute_name_range_op): Adjust.
21812 (gori_compute::optimize_logical_operands): Adjust.
21813 (gori_compute::compute_logical_operands_in_chain): Adjust.
21814 (gori_compute::compute_operand1_range): Adjust.
21815 (gori_compute::compute_operand2_range): Adjust.
21816 (ori_compute_cache::cache_stmt): Adjust.
21817 * gimple-range-gori.h (gori_compute): Rename prototype.
21819 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
21821 * gimple-range.cc (gimple_ranger::range_of_expr): Non-null should be
21822 checked only after range_of_stmt, not range_on_entry.
21823 (gimple_ranger::range_on_entry): Check for non-null in any
21824 predecessor block, if it is not already non-null.
21825 (gimple_ranger::range_on_exit): DOnt check for non-null after
21826 range on entry call.
21827 (gimple_ranger::dump_bb): New. Split from dump.
21828 (gimple_ranger::dump): Adjust.
21829 * gimple-range.h (class gimple_ranger): Adjust.
21831 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
21833 * gimple-range-cache.cc (struct range_timestamp): Delete.
21834 (class temporal_cache): Adjust.
21835 (temporal_cache::get_timestamp): Delete.
21836 (temporal_cache::set_dependency): Delete.
21837 (temporal_cache::temporal_value): Adjust.
21838 (temporal_cache::current_p): Take dependencies as params.
21839 (temporal_cache::set_timestamp): Adjust.
21840 (temporal_cache::set_always_current): Adjust.
21841 (ranger_cache::get_non_stale_global_range): Adjust.
21842 (ranger_cache::register_dependency): Delete.
21843 * gimple-range-cache.h (class range_cache): Adjust.
21845 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
21847 * gimple-range-gori.cc (range_def_chain::range_def_chain): init
21849 (range_def_chain::~range_def_chain): Dispose of obstack rather than
21850 each individual bitmap.
21851 (range_def_chain::set_import): New.
21852 (range_def_chain::get_imports): New.
21853 (range_def_chain::chain_import_p): New.
21854 (range_def_chain::register_dependency): Rename from build_def_chain
21856 (range_def_chain::def_chain_in_bitmap_p): New.
21857 (range_def_chain::add_def_chain_to_bitmap): New.
21858 (range_def_chain::has_def_chain): Just check first depenedence.
21859 (range_def_chain::get_def_chain): Process imports, use generic
21860 register_dependency routine.
21861 (range_def_chain::dump): New.
21862 (gori_map::gori_map): Allocate import list.
21863 (gori_map::~gori_map): Release imports.
21864 (gori_map::exports): Check for past allocated block size.
21865 (gori_map::imports): New.
21866 (gori_map::def_chain_in_export_p): Delete.
21867 (gori_map::is_import_p): New.
21868 (gori_map::maybe_add_gori): Handle imports.
21869 (gori_map::dump): Adjust output, add imports.
21870 (gori_compute::has_edge_range_p): Remove def_chain_in_export call.
21871 (gori_export_iterator::gori_export_iterator): New.
21872 (gori_export_iterator::next): New.
21873 (gori_export_iterator::get_name): New.
21874 * gimple-range-gori.h (range_def_chain): Add imports and direct
21875 dependecies via struct rdc.
21876 (range_def_chain::depend1): New.
21877 (range_def_chain::depend2): New.
21878 (class gori_map): Adjust.
21879 (FOR_EACH_GORI_IMPORT_NAME): New.
21880 (FOR_EACH_GORI_EXPORT_NAME): New.
21881 (class gori_export_iterator): New.
21883 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
21885 * gimple-range-cache.cc (ranger_cache::ranger_cache): Move initial
21886 export cache filling to here.
21887 * gimple-range-gori.cc (gori_compute::gori_compute) : From Here.
21889 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
21891 * gimple-range-gori.cc (range_def_chain): Move to gimple-range-gori.h.
21892 (gori_map): Move to gimple-range-gori.h.
21893 (gori_compute::gori_compute): Adjust.
21894 (gori_compute::~gori_compute): Delete.
21895 (gori_compute::compute_operand_range_switch): Adjust.
21896 (gori_compute::compute_operand_range): Adjust.
21897 (gori_compute::compute_logical_operands): Adjust.
21898 (gori_compute::has_edge_range_p ): Adjust.
21899 (gori_compute::set_range_invariant): Delete.
21900 (gori_compute::dump): Adjust.
21901 (gori_compute::outgoing_edge_range_p): Adjust.
21902 * gimple-range-gori.h (class range_def_chain): Relocate here.
21903 (class gori_map): Relocate here.
21904 (class gori_compute): Inherit from gori_map, and adjust.
21906 2021-05-25 Aldy Hernandez <aldyh@redhat.com>
21908 * value-range.cc (range_tests_legacy): Use
21909 build_nonstandard_integer_type instead of int and short.
21911 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
21913 * gimplify.c (gimplify_decl_expr): Clear TREE_READONLY on the DECL
21914 when really creating an initialization statement for it.
21916 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
21918 * tree-inline.c (setup_one_parameter): Fix thinko in new condition.
21920 2021-05-25 Kito Cheng <kito.cheng@sifive.com>
21922 * config/riscv/riscv.h (ASM_SPEC): Pass -mno-relax.
21924 2021-05-25 Martin Liska <mliska@suse.cz>
21926 PR tree-optimization/92860
21928 * optc-save-gen.awk: Remove exceptions.
21930 2021-05-25 Martin Liska <mliska@suse.cz>
21932 * asan.h (sanitize_coverage_p): New function.
21933 * doc/extend.texi: Document it.
21934 * fold-const.c (fold_range_test): Use sanitize_flags_p
21935 instead of flag_sanitize_coverage.
21936 (fold_truth_andor): Likewise.
21937 * sancov.c: Likewise.
21938 * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
21939 * ipa-inline.c (sanitize_attrs_match_for_inline_p): Handle
21940 -fsanitize-coverage when inlining.
21942 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
21944 * config/csky/csky-modes.def : Fix copyright.
21946 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
21948 * config/csky/csky-modes.def : Amend copyright.
21949 * config/csky/csky_insn_fpuv2.md : Likewise.
21950 * config/csky/csky_insn_fpuv3.md : Likewise.
21952 2021-05-25 Richard Biener <rguenther@suse.de>
21954 PR middle-end/100727
21955 * calls.c (initialize_argument_information): Explicitely test
21956 for WITH_SIZE_EXPR.
21957 * gimple-expr.c (mark_addressable): Skip outer WITH_SIZE_EXPR.
21959 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
21961 * config/csky/csky.h (FRAME_POINTER_REGNUM): Use
21962 HARD_FRAME_POINTER_REGNUM and FRAME_POINTER_REGNUM instead of
21963 the signle definition. The signle definition may not work well
21964 at simplify_subreg_regno().
21965 (HARD_FRAME_POINTER_REGNUM): New.
21966 (ELIMINABLE_REGS): Add for HARD_FRAME_POINTER_REGNUM.
21967 * config/csky/csky.c (get_csky_live_regs, csky_can_eliminate,
21968 csky_initial_elimination_offset, csky_expand_prologue,
21969 csky_expand_epilogue): Add for HARD_FRAME_POINTER_REGNUM.
21971 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
21973 * config/csky/csky.c (csky_option_override):
21974 Init csky_arch_isa_features[] in advance, so TARGET_DSP
21975 and TARGET_DIV can be set well.
21977 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
21979 * config/csky/constraints.md ("l", "h"): Delete.
21980 * config/csky/csky.h (reg_class, REG_CLASS_NAMES,
21981 REG_CLASS_CONTENTS): Delete LO_REGS and HI_REGS.
21982 * config/csky/csky.c (regno_reg_classm,
21983 csky_secondary_reload, csky_register_move_cost):
21984 Use HILO_REGS instead of LO_REGS and HI_REGS.
21986 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
21988 * config/csky/constraints.md ("W"): New constriant for mem operand
21989 with base reg, index register.
21990 ("Q"): Renamed and modified "csky_valid_fpuv2_mem_operand" to
21991 "csky_valid_mem_constraint_operand" to deal with both "Q" and "W"
21993 ("Dv"): New constraint for const double value that can be used at
21995 * config/csky/csky-modes.def (HFmode): New mode.
21996 * config/csky/csky-protos.h (csky_valid_fpuv2_mem_operand): Rename
21997 to "csky_valid_mem_constraint_operand" and support new constraint
21999 (csky_get_movedouble_length): New.
22000 (fpuv3_output_move): New.
22001 (fpuv3_const_double): New.
22002 * config/csky/csky.c (csky_option_override): New arch CK860 with fpv3.
22003 (decompose_csky_address): Refine.
22004 (csky_print_operand): New "CONST_DOUBLE" operand.
22005 (csky_output_move): Support fpv3 instructions.
22006 (csky_get_movedouble_length): New.
22007 (fpuv3_output_move): New.
22008 (fpuv3_const_double): New.
22009 (csky_emit_compare): Cover float comparsion.
22010 (csky_emit_compare_float): Refine.
22011 (csky_vaild_fpuv2_mem_operand): Rename to
22012 "csky_valid_mem_constraint_operand" and support new constraint "W".
22013 (ck860_rtx_costs): New.
22014 (csky_rtx_costs): Add the cost calculation of CK860.
22015 (regno_reg_class): New vregs for fpuv3.
22016 (csky_dbx_regno): Likewise.
22017 (csky_cpu_cpp_builtins): New builtin macro for fpuv3.
22018 (csky_conditional_register_usage): Suporrot fpuv3.
22019 (csky_dwarf_register_span): Suporrot fpuv3.
22020 (csky_init_builtins, csky_mangle_type): Support "__fp16" type.
22021 (ck810_legitimate_index_p): Support fp16.
22022 * config/csky/csky.h (TARGET_TLS): ADD CK860.
22023 (CSKY_VREG_P, CSKY_VREG_LO_P, CSKY_VREG_HI_P): Support fpuv3.
22024 (TARGET_SINGLE_FPU): Support fpuv3.
22025 (TARGET_SUPPORT_FPV3): New.
22026 (FIRST_PSEUDO_REGISTER): Change to 202 to hold the new fpuv3 registers.
22027 (FIXED_REGISTERS, CALL_REALLY_USED_REGISTERS, REGISTER_NAMES,
22028 REG_CLASS_CONTENTS): Support fpuv3.
22029 * config/csky/csky.md (movsf): Move to cksy_insn_fpu.md and refine.
22030 (csky_movsf_fpv2): Likewise.
22031 (ck801_movsf): Likewise.
22032 (csky_movsf): Likewise.
22034 (csky_movdf_fpv2): Likewise.
22035 (ck801_movdf): Likewise.
22036 (csky_movdf): Likewise.
22037 (movsicc): Refine. Use "comparison_operatior" instead of
22038 "ordered_comparison_operatior".
22039 (addsicc): Likewise.
22040 (CSKY_FIRST_VFP3_REGNUM, CSKY_LAST_VFP3_REGNUM): New constant.
22041 (call_value_internal_vh): New.
22042 * config/csky/csky_cores.def (CK860): New arch and cpu.
22047 * config/csky/csky_insn_fpu.md: Refactor. Separate all float patterns
22048 into emit-patterns and match-patterns, remain the emit-patterns here,
22049 and move the match-patterns to csky_insn_fpuv2.md or
22050 csky_insn_fpuv3.md.
22051 * config/csky/csky_insn_fpuv2.md: New file for fpuv2 instructions.
22052 * config/csky/csky_insn_fpuv3.md: New file and new patterns for fpuv3
22054 * config/csky/csky_isa.def (fcr): New.
22059 (CK860): New definition for ck860.
22060 * config/csky/csky_tables.opt (ck860): New processors ck860,
22061 ck860f. And new arch ck860.
22066 * config/csky/predicates.md (csky_float_comparsion_operator): Delete
22067 "geu", "gtu", "leu", "ltu", which will never appear at float comparison.
22068 * config/csky/t-csky-elf: Support 860.
22069 * config/csky/t-csky-linux: Likewise.
22070 * doc/md.texi: Add "Q" and "W" constraints for C-SKY.
22072 2021-05-24 Aaron Sawdey <acsawdey@linux.ibm.com>
22074 * config/rs6000/genfusion.pl (gen_logical_addsubf): Refactor to
22075 add generation of logical-add and add-logical fusion pairs.
22076 * config/rs6000/rs6000-cpus.def: Add new fusion to ISA 3.1 mask
22078 * config/rs6000/rs6000.c (rs6000_option_override_internal): Turn on
22079 logical-add and add-logical fusion by default.
22080 * config/rs6000/rs6000.opt: Add -mpower10-fusion-logical-add and
22081 -mpower10-fusion-add-logical options.
22082 * config/rs6000/fusion.md: Regenerate file.
22084 2021-05-24 Aldy Hernandez <aldyh@redhat.com>
22086 * value-range.cc (irange::legacy_equal_p): Check type when
22087 comparing VR_VARYING types.
22088 (range_tests_legacy): Test comparing VARYING ranges of different
22091 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
22093 * config/aarch64/aarch64.c (neoversen1_tunings):
22094 Enable AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
22096 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
22098 * config/aarch64/aarch64.c (aarch64_classify_symbol): Use GOT for
22099 extern weak symbols. Limit symbol offsets for non-GOT symbols with
22102 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
22104 * config/arm/neon.md (vec_load_lanesxi<mode>)
22105 (vec_store_lanexoi<mode>): Move ...
22106 * config/arm/vec-common.md: here.
22108 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
22110 * config/arm/neon.md (vec_load_lanesoi<mode>)
22111 (vec_store_lanesoi<mode>): Move ...
22112 * config/arm/vec-common.md: here.
22114 2021-05-24 liuhongt <hongtao.liu@intel.com>
22117 * config/i386/i386.c (ix86_gimple_fold_builtin): Replacing
22118 stmt with GIMPLE_NOP when lhs doesn't exist.
22120 2021-05-23 Uroš Bizjak <ubizjak@gmail.com>
22123 * config/i386/mmx.md (*push<VI_32:mode>2_rex64):
22124 New instruction pattern.
22125 (*push<VI_32:mode>2): Ditto.
22126 (push splitter for SSE registers): New splitter.
22128 2021-05-23 Andrew Pinski <apinski@marvell.com>
22130 * match.pd ((A & C) != 0 ? D : 0): Limit to non pointer types.
22132 2021-05-22 Aaron Sawdey <acsawdey@linux.ibm.com>
22134 * config/rs6000/genfusion.pl (gen_addadd): Fix incorrect attr types.
22135 * config/rs6000/fusion.md: Regenerate file.
22137 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
22139 * config/rs6000/genfusion.pl (gen_addadd): New function.
22140 * config/rs6000/fusion.md: Regenerate file.
22141 * config/rs6000/rs6000-cpus.def: Add
22142 OPTION_MASK_P10_FUSION_2ADD to masks.
22143 * config/rs6000/rs6000.c (rs6000_option_override_internal):
22144 Handle default value of OPTION_MASK_P10_FUSION_2ADD.
22145 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2add.
22147 2021-05-21 Jakub Jelinek <jakub@redhat.com>
22149 PR middle-end/99928
22150 * tree.h (OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET): Define.
22151 * gimplify.c (enum gimplify_omp_var_data): Fix up
22152 GOVD_MAP_HAS_ATTACHMENTS value, add GOVD_FIRSTPRIVATE_IMPLICIT.
22153 (omp_lastprivate_for_combined_outer_constructs): If combined target
22154 has GOVD_FIRSTPRIVATE_IMPLICIT set for the decl, change it to
22155 GOVD_MAP | GOVD_SEEN.
22156 (gimplify_scan_omp_clauses): Set GOVD_FIRSTPRIVATE_IMPLICIT for
22157 firstprivate clauses with OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT.
22158 (gimplify_adjust_omp_clauses): For firstprivate clauses with
22159 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT either clear that bit and
22160 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET too, or remove it and
22161 let it be replaced by implicit map clause.
22163 2021-05-21 Jakub Jelinek <jakub@redhat.com>
22165 PR middle-end/99928
22166 * gimplify.c (omp_lastprivate_for_combined_outer_constructs): New
22168 (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LASTPRIVATE>: Use it.
22169 (gimplify_omp_for): Likewise.
22171 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
22173 PR middle-end/90115
22174 * omp-low.c (oacc_privatization_candidate_p): Reject 'static',
22175 'external' in blocks.
22177 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
22179 PR middle-end/90115
22180 * flag-types.h (enum openacc_privatization): New.
22181 * params.opt (-param=openacc-privatization): New.
22182 * doc/invoke.texi (openacc-privatization): Document it.
22183 * omp-general.h (get_openacc_privatization_dump_flags): New
22185 * omp-low.c (oacc_privatization_candidate_p): Add diagnostics.
22186 * omp-offload.c (execute_oacc_device_lower)
22187 <IFN_UNIQUE_OACC_PRIVATE>: Re-work diagnostics.
22188 * target.def (goacc.adjust_private_decl): Add 'location_t'
22190 * doc/tm.texi: Regenerate.
22191 * config/gcn/gcn-protos.h (gcn_goacc_adjust_private_decl): Adjust.
22192 * config/gcn/gcn-tree.c (gcn_goacc_adjust_private_decl): Likewise.
22193 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl):
22194 Likewise. Preserve it for...
22195 (nvptx_goacc_expand_var_decl): ... use here.
22197 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
22199 * doc/sourcebuild.texi (Other attributes): Document '__OPTIMIZE__'
22202 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
22204 PR middle-end/90115
22205 * omp-low.c (oacc_privatization_candidate_p): New function.
22206 (oacc_privatization_scan_clause_chain)
22207 (oacc_privatization_scan_decl_chain): Use it. Also
22208 'gcc_checking_assert' that we're not seeing duplicates.
22210 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
22212 PR middle-end/90115
22213 * omp-offload.c (execute_oacc_device_lower): Skip processing if no
22216 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
22218 PR middle-end/90115
22219 * omp-offload.c (execute_oacc_device_lower): Explain.
22221 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
22223 PR middle-end/90115
22224 * omp-offload.c (execute_oacc_device_lower)
22225 <IFN_UNIQUE_OACC_PRIVATE>: Diagnose and handle for 'level == -1'
22227 * internal-fn.c (expand_UNIQUE): Don't expect
22228 'IFN_UNIQUE_OACC_PRIVATE'.
22230 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
22232 PR middle-end/90115
22233 * omp-low.c (lower_omp_for): Don't evaluate OpenMP 'for' clauses.
22235 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
22237 PR middle-end/90115
22238 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl)
22239 (nvptx_goacc_expand_var_decl): Tighten.
22241 2021-05-21 Julian Brown <julian@codesourcery.com>
22242 Chung-Lin Tang <cltang@codesourcery.com>
22243 Thomas Schwinge <thomas@codesourcery.com>
22245 PR middle-end/90115
22246 * doc/tm.texi.in (TARGET_GOACC_EXPAND_VAR_DECL)
22247 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Add documentation hooks.
22248 * doc/tm.texi: Regenerate.
22249 * expr.c (expand_expr_real_1): Expand decls using the
22250 expand_var_decl OpenACC hook if defined.
22251 * internal-fn.c (expand_UNIQUE): Handle IFN_UNIQUE_OACC_PRIVATE.
22252 * internal-fn.h (IFN_UNIQUE_CODES): Add OACC_PRIVATE.
22253 * omp-low.c (omp_context): Add oacc_privatization_candidates
22255 (lower_oacc_reductions): Add PRIVATE_MARKER parameter. Insert
22257 (lower_oacc_head_tail): Add PRIVATE_MARKER parameter. Modify
22258 private marker's gimple call arguments, and pass it to
22259 lower_oacc_reductions.
22260 (oacc_privatization_scan_clause_chain)
22261 (oacc_privatization_scan_decl_chain, lower_oacc_private_marker):
22263 (lower_omp_for, lower_omp_target, lower_omp_1): Use these.
22264 * omp-offload.c (convert.h): Include.
22265 (oacc_loop_xform_head_tail): Treat private-variable markers like
22266 fork/join when transforming head/tail sequences.
22267 (struct var_decl_rewrite_info): Add struct.
22268 (oacc_rewrite_var_decl, is_sync_builtin_call): New functions.
22269 (execute_oacc_device_lower): Support rewriting gang-private
22270 variables using target hook, and fix up addr_expr and var_decl
22272 * target.def (adjust_private_decl, expand_var_decl): New hooks.
22273 * config/gcn/gcn-protos.h (gcn_goacc_adjust_gangprivate_decl):
22275 (gcn_goacc_adjust_private_decl): ...this.
22276 * config/gcn/gcn-tree.c (gcn_goacc_adjust_gangprivate_decl):
22278 (gcn_goacc_adjust_private_decl): ...this. Add LEVEL parameter.
22279 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_GANGPRIVATE_DECL): Rename
22280 definition using gcn_goacc_adjust_gangprivate_decl...
22281 (TARGET_GOACC_ADJUST_PRIVATE_DECL): ...to this, using
22282 gcn_goacc_adjust_private_decl.
22283 * config/nvptx/nvptx.c (tree-pretty-print.h): Include.
22284 (gang_private_shared_size): New global variable.
22285 (gang_private_shared_align): Likewise.
22286 (gang_private_shared_sym): Likewise.
22287 (gang_private_shared_hmap): Likewise.
22288 (nvptx_option_override): Initialize these.
22289 (nvptx_file_end): Output gang_private_shared_sym.
22290 (nvptx_goacc_adjust_private_decl, nvptx_goacc_expand_var_decl):
22292 (nvptx_set_current_function): Clear gang_private_shared_hmap.
22293 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Define hook.
22294 (TARGET_GOACC_EXPAND_VAR_DECL): Likewise.
22296 2021-05-21 H.J. Lu <hjl.tools@gmail.com>
22298 * config/i386/i386-modes.def (MAX_BITSIZE_MODE_ANY_INT): Removed.
22300 2021-05-21 Richard Biener <rguenther@suse.de>
22301 H.J. Lu <hjl.tools@gmail.com>
22303 PR middle-end/90773
22304 * expr.c (expand_constructor): Elide expand_constructor if
22305 move by pieces is preferred.
22307 2021-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22309 * config/aarch64/aarch64-builtins.c (aarch64_call_properties):
22310 Take a flag and mode value as arguments.
22311 (aarch64_modifies_global_state_p): Likewise.
22312 (aarch64_reads_global_state_p): Likewise.
22313 (aarch64_could_trap_p): Likewise.
22314 (aarch64_get_attributes): Likewise.
22315 (aarch64_init_simd_builtins): Adjust callsite of above.
22316 (aarch64_init_fcmla_laneq_builtins): Use aarch64_get_attributes to get
22317 function attributes to apply to builtins.
22318 (aarch64_init_crc32_builtins): Likewise.
22319 (aarch64_init_builtin_rsqrt): Likewise.
22321 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
22323 * config/rs6000/rs6000.md (define_attr "type"): Add types for fusion.
22324 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10): Use new fusion types.
22325 (gen_2logical): Use new fusion types.
22326 * config/rs6000/fusion.md: Regenerate.
22328 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
22331 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
22332 Handle V4QI and V2HI modes.
22333 (ix86_expand_sse_movcc): Ditto.
22334 * config/i386/mmx.md (*<sat_plusminus:insn><VI_32:mode>3):
22335 New instruction pattern.
22336 (*eq<VI_32:mode>3): Ditto.
22337 (*gt<VI_32:mode>3): Ditto.
22338 (*xop_pcmov_<VI_32:mode>): Ditto.
22339 (mmx_pblendvb32): Ditto.
22340 (mmx_pblendvb64): Rename from mmx_pblendvb.
22341 (vec_cmp<VI_32:mode><VI_32:mode>): New expander.
22342 (vec_cmpu<VI_32:mode><VI_32:mode>): Ditto.
22343 (vcond<VI_32:mode><VI_32:mode>): Ditto.
22344 (vcondu<VI_32:mode><VI_32:mode>): Ditto.
22345 (vcond_mask_<VI_32:mode><VI_32:mode>): Ditto.
22347 2021-05-21 Jakub Jelinek <jakub@redhat.com>
22349 PR tree-optimization/94589
22350 * tree-ssa-phiopt.c (spaceship_replacement): For integral rhs1 and
22351 rhs2, treat x <= 4 equivalently to x < 5 etc. In cmp1 and cmp2 (if
22352 not the same as cmp3) treat <= the same as < and >= the same as >.
22353 Don't require that cond2_phi_edge is true edge, instead take
22354 false/true edges into account based on cmp1/cmp2 comparison kinds.
22356 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
22359 * config/i386/mmx.md (SMAXMIN_MMXMODEI): New mode iterator.
22360 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Macroize expander
22361 from <smaxmin:code>v4hi3> and <smaxmin:code><MMXMODE14:mode>3
22362 using SMAXMIN_MMXMODEI mode iterator.
22363 (*<smaxmin:code>v4qi3): New insn pattern.
22364 (*<smaxmin:code>v2hi3): Ditto.
22365 (SMAXMIN_VI_32): New mode iterator.
22366 (<smaxmin:code><SMAXMIN_VI_32>mode3): New expander.
22367 (UMAXMIN_MMXMODEI): New mode iterator.
22368 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Macroize expander
22369 from <umaxmin:code>v8qi3> and <umaxmin:code><MMXMODE24:mode>3
22370 using UMAXMIN_MMXMODEI mode iterator.
22371 (*<umaxmin:code>v4qi3): New insn pattern.
22372 (*<umaxmin:code>v2hi3): Ditto.
22373 (UMAXMIN_VI_32): New mode iterator.
22374 (<umaxmin:code><UMAXMIN_VI_32>mode3): New expander.
22375 (abs<VI_32:mode>2): New insn pattern.
22376 (ssse3_abs<MMXMODEI:mode>2, abs<MMXMODEI:mode>2): Move from ...
22377 * config/i386/sse.md: ... here.
22379 2021-05-20 Clement Chigot <clement.chigot@atos.net>
22380 David Edelsohn <dje.gcc@gmail.com>
22382 * collect2.c (scan_prog_file): Issue non-fatal warning for
22385 2021-05-20 Jonathan Wakely <jwakely@redhat.com>
22387 * doc/invoke.texi (-Wno-c++11-extensions)
22388 (-Wno-c++14-extensions, -Wno-c++17-extensions)
22389 (-Wno-c++20-extensions, -Wno-c++23-extensions): Document
22392 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
22394 * config/c6x/c6x.c (c6x_output_file_unwind): Use dwarf_debuginfo_p.
22395 * config/darwin.c (darwin_override_options): Likewise.
22396 * config/i386/cygming.h (DBX_REGISTER_NUMBER): Likewise.
22397 * config/i386/darwin.h (DBX_REGISTER_NUMBER): Likewise.
22398 (DWARF2_FRAME_REG_OUT): Likewise.
22399 * config/mips/mips.c (mips_output_filename): Likewise.
22400 * config/rs6000/rs6000.c (rs6000_xcoff_declare_function_name):
22402 (rs6000_dbx_register_number): Likewise.
22403 * dbxout.c: Include flags.h.
22404 * dwarf2cfi.c (cfi_label_required_p): Likewise.
22405 (dwarf2out_do_frame): Likewise.
22406 * except.c: Include flags.h.
22407 * final.c (dwarf2_debug_info_emitted_p): Likewise.
22408 (final_scan_insn_1): Likewise.
22409 * flags.h (dwarf_debuginfo_p): New function declaration.
22410 * opts.c (dwarf_debuginfo_p): New function definition.
22411 * targhooks.c (default_debug_unwind_info): Use dwarf_debuginfo_p.
22412 * toplev.c (process_options): Likewise.
22414 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
22416 * common.opt: Change type to support bitmasks.
22417 * flag-types.h (enum debug_info_type): Rename enumerator constants.
22418 (NO_DEBUG): New bitmask.
22419 (DBX_DEBUG): Likewise.
22420 (DWARF2_DEBUG): Likewise.
22421 (XCOFF_DEBUG): Likewise.
22422 (VMS_DEBUG): Likewise.
22423 (VMS_AND_DWARF2_DEBUG): Likewise.
22424 * flags.h (debug_set_to_format): New function declaration.
22425 (debug_set_count): Likewise.
22426 (debug_set_names): Likewise.
22427 * opts.c (debug_type_masks): Array of bitmasks for debug formats.
22428 (debug_set_to_format): New function definition.
22429 (debug_set_count): Likewise.
22430 (debug_set_names): Likewise.
22431 (set_debug_level): Update access to debug_type_names.
22432 * toplev.c: Likewise.
22434 2021-05-20 Martin Sebor <msebor@redhat.com>
22436 PR middle-end/100684
22437 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Handle C++ lambda.
22439 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
22442 * config/i386/i386.md (isa): Remove x64_bmi.
22443 (enabled): Remove x64_bmi.
22444 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
22445 Remove general register alternative.
22446 (*andnot<VI_32:mode>3): Ditto.
22447 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
22448 (*<any_logic:code><VI_32:mode>3): Ditto.
22450 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
22452 * config/arm/arm.c: Include head files tree-vectorizer.h and
22455 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
22458 * config/i386/mmx.md (Yv_Yw): Revert adding V4QI and V2HI modes.
22459 (*<plusminus:insn><VI_32:mode>3): Use Yw instad of <Yv_Yw> constrint.
22460 (<s>mulv4hi3_highpart): New expander.
22461 (*<s>mulv2hi3_highpart): New insn pattern.
22462 (<s>mulv2hi3_higpart): New expander.
22463 (*<any_shift:insn>v2hi3): New insn pattern.
22464 (<any_shift:insn>v2hi3): New expander.
22465 * config/i386/sse.md (smulhrsv2hi3): New expander.
22466 (*smulhrsv2hi3): New insn pattern.
22468 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
22470 * doc/invoke.texi (vect-inner-loop-cost-factor): Document new
22472 * params.opt (vect-inner-loop-cost-factor): New.
22473 * targhooks.c (default_add_stmt_cost): Replace hardcoded factor
22474 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR, include head file
22475 tree-vectorizer.h and its required ones.
22476 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Replace
22477 hardcoded factor 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR.
22478 * config/arm/arm.c (arm_add_stmt_cost): Likewise.
22479 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
22480 * config/rs6000/rs6000.c (rs6000_add_stmt_cost): Likewise.
22481 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
22483 (_loop_vec_info::_loop_vec_info): Init inner_loop_cost_factor.
22484 * tree-vectorizer.h (_loop_vec_info): Add inner_loop_cost_factor.
22485 (LOOP_VINFO_INNER_LOOP_COST_FACTOR): New macro.
22487 2021-05-20 Christophe Lyon <christophe.lyon@linaro.org>
22488 Torbjörn Svensson <torbjorn.svensson@st.com>
22491 * doc/cpp.texi (Common Predefined Macros): Document __FILE_NAME__.
22493 2021-05-20 Jakub Jelinek <jakub@redhat.com>
22495 PR middle-end/99928
22496 * gimplify.c (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LINEAR>: For
22497 explicit linear clause when combined with target, make it map(tofrom:)
22498 instead of no clause or firstprivate.
22500 2021-05-20 Jakub Jelinek <jakub@redhat.com>
22502 PR tree-optimization/94589
22503 * match.pd ((X & Y) == X -> (X & ~Y) == 0): Simplify even in presence
22504 of integral conversions.
22506 2021-05-19 Andrew MacLeod <amacleod@redhat.com>
22508 * gimple-range.cc (fur_source::get_operand): New.
22509 (gimple_range_fold): Delete.
22510 (fold_using_range::fold_stmt): Move from gimple_ranger::calc_stmt.
22511 (fold_using_range::range_of_range_op): Move from gimple_ranger.
22512 (fold_using_range::range_of_address): Ditto.
22513 (fold_using_range::range_of_phi): Ditto.
22514 (fold_using_range::range_of_call): Ditto.
22515 (fold_using_range::range_of_builtin_ubsan_call): Move from
22516 range_of_builtin_ubsan_call.
22517 (fold_using_range::range_of_builtin_call): Move from
22518 range_of_builtin_call.
22519 (gimple_ranger::range_of_builtin_call): Delete.
22520 (fold_using_range::range_of_cond_expr): Move from gimple_ranger.
22521 (gimple_ranger::fold_range_internal): New.
22522 (gimple_ranger::range_of_stmt): Use new fold_using_range API.
22523 (fold_using_range::range_of_ssa_name_with_loop_info): Move from
22524 gimple_ranger. Improve ranges of SSA_NAMES when possible.
22525 * gimple-range.h (gimple_ranger): Remove various range_of routines.
22526 (class fur_source): New.
22527 (class fold_using_range): New.
22528 (fur_source::fur_source): New.
22530 * vr-values.c (vr_values::extract_range_basic): Use fold_using_range
22531 instead of range_of_builtin_call.
22533 2021-05-19 Jonathan Wakely <jwakely@redhat.com>
22535 * doc/cpp.texi (Common Predefined Macros): Update documentation
22536 for the __GXX_EXPERIMENTAL_CXX0X__ macro.
22538 2021-05-19 Alex Coplan <alex.coplan@arm.com>
22541 * config/arm/arm.md (nonsecure_call_internal): Always ensure
22542 callee's address is in a register.
22544 2021-05-19 Geng Qi <gengqi@linux.alibaba.com>
22546 * common/config/riscv/riscv-common.c
22547 (riscv_subset_list::parsing_subset_version): Properly parse the letter
22549 (riscv_subset_list::parse_std_ext,
22550 riscv_subset_list::parse_multiletter_ext): To handle errors generated
22551 in riscv_subset_list::parsing_subset_version.
22553 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
22555 * config/aarch64/aarch64-simd.md: Use "neon_move_narrow_q"
22556 type attribute in patterns generating XTN(2).
22558 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
22560 * config/aarch64/aarch64-simd.md (aarch64_simd_vec_pack_trunc_<mode>):
22561 Remove as duplicate of...
22562 (aarch64_xtn<mode>): This.
22563 (aarch64_xtn2<mode>_le): Move position in file.
22564 (aarch64_xtn2<mode>_be): Move position in file.
22565 (aarch64_xtn2<mode>): Move position in file.
22566 (vec_pack_trunc_<mode>): Define as an expander.
22568 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
22570 * config/aarch64/aarch64-simd-builtins.def: Split builtin
22571 generation for aarch64_<sur>q<r>shr<u>n_n<mode> pattern into
22572 separate scalar and vector generators.
22573 * config/aarch64/aarch64-simd.md
22574 (aarch64_<sur>q<r>shr<u>n_n<mode>): Define as an expander and
22576 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_le): This and...
22577 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_be): This.
22578 * config/aarch64/iterators.md: Define SD_HSDI iterator.
22580 2021-05-19 Jonathn Wright <jonathan.wright@arm.com>
22582 * config/aarch64/aarch64-simd.md: Use UNSPEC_SQXTUN instead
22584 * config/aarch64/iterators.md: Remove UNSPEC_SQXTUN2.
22586 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
22588 * config/aarch64/aarch64-simd.md (aarch64_<sur>q<r>shr<u>n2_n<mode>):
22589 Implement as an expand emitting a big/little endian
22590 instruction pattern.
22591 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_le): Define.
22592 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_be): Define.
22594 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
22596 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn2<mode>):
22597 Implement as an expand emitting a big/little endian
22598 instruction pattern.
22599 (aarch64_<sur><addsub>hn2<mode>_insn_le): Define.
22600 (aarch64_<sur><addsub>hn2<mode>_insn_be): Define.
22601 * config/aarch64/iterators.md: Remove UNSPEC_[R]ADDHN2 and
22602 UNSPEC_[R]SUBHN2 unspecs and ADDSUBHN2 iterator.
22604 2021-05-19 Richard Biener <rguenther@suse.de>
22606 PR middle-end/100672
22607 * fold-const.c (fold_negate_expr_1): Use element_precision.
22608 (negate_expr_p): Likewise.
22610 2021-05-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
22612 * config/aarch64/iterators.md (SVE_PRED_LOAD): New iterator.
22613 (pred_load): New int attribute.
22614 * config/aarch64/aarch64-sve.md
22615 (aarch64_load_<ANY_EXTEND:optab><SVE_HSDI:mode><SVE_PARTIAL_I:mode>): Use
22616 SVE_PRED_LOAD enum iterator and corresponding pred_load attribute.
22617 * config/aarch64/aarch64-sve-builtins-base.cc (expand): Update call to
22618 code_for_aarch64_load.
22620 2021-05-19 Richard Biener <rguenther@suse.de>
22622 * cfgexpand.c (discover_nonconstant_array_refs_r): Make
22623 sure TARGET_MEM_REF bases are expanded as memory.
22624 * tree-ssa-operands.c (operands_scanner::get_tmr_operands):
22625 Do not mark TARGET_MEM_REF bases addressable.
22626 * tree-ssa.c (non_rewritable_mem_ref_base): Handle
22627 TARGET_MEM_REF bases as never rewritable.
22628 * gimple-walk.c (walk_stmt_load_store_addr_ops): Do not
22629 walk TARGET_MEM_REF bases as address-takens.
22630 * tree-ssa-dce.c (ref_may_be_aliased): Handle TARGET_MEM_REF.
22632 2021-05-19 Richard Biener <rguenther@suse.de>
22634 * builtins.c (get_object_alignment_1): Strip outer
22636 * tree-dfa.c (get_ref_base_and_extent): Handle outer
22637 WITH_SIZE_EXPR for size processing and process the
22639 * tree-ssa-alias.c (ao_ref_base_alias_set): Strip
22640 outer WITH_SIZE_EXPR.
22641 (ao_ref_base_alias_ptr_type): Likewise.
22642 (refs_may_alias_p_2): Allow WITH_SIZE_EXPR in ref->ref
22643 and handle that accordingly, stripping it for the
22644 core alias workers.
22645 * tree.c (get_base_address): Handle WITH_SIZE_EXPR by
22646 looking through it instead of returning NULL.
22648 2021-05-19 Jakub Jelinek <jakub@redhat.com>
22650 PR middle-end/100576
22651 * builtins.c (check_read_access): Convert bound to size_type_node if
22654 2021-05-19 Richard Biener <rguenther@suse.de>
22656 * tree-cfg.c (verify_types_in_gimple_min_lval): Inline...
22657 (verify_types_in_gimple_reference): ... here. Sanitize.
22658 (verify_gimple_call): Verify references in LHS and arguments.
22659 (verify_gimple_assign_single): Reject WITH_SIZE_EXPR.
22661 2021-05-19 Uroš Bizjak <ubizjak@gmail.com>
22663 * config/i386/i386.h (VALID_INT_MODE_P):
22664 Add V8QI, V4HI and V2SI modes for TARGET_64BIT.
22665 * config/i386/i386.md (isa): Add x64_bmi.
22666 (enabled): Handle x64_bmi.
22667 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
22668 Add alternative using 64bit general registers.
22669 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
22671 2021-05-19 Jakub Jelinek <jakub@redhat.com>
22673 PR middle-end/99928
22674 * tree.h (OMP_MASTER_COMBINED): Define.
22675 * gimplify.c (gimplify_scan_omp_clauses): Rewrite lastprivate
22676 handling for outer combined/composite constructs to a loop.
22677 Handle lastprivate on combined target.
22678 (gimplify_expr): Formatting fix.
22680 2021-05-19 Xionghu Luo <luoxhu@linux.ibm.com>
22682 * passes.def: Add sink_code pass before store_merging.
22683 * tree-ssa-sink.c (pass_sink_code:clone): New.
22685 2021-05-18 Bill Schmidt <wschmidt@linux.ibm.com>
22687 * config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Remove call to
22688 rs6000_special_adjust_field_align_p.
22689 * config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
22690 * config/rs6000/rs6000-call.c (rs6000_function_arg_boundary):
22691 Remove ABI warning.
22692 (rs6000_function_arg): Likewise.
22693 * config/rs6000/rs6000-protos.h
22694 (rs6000_special_adjust_field_align_p): Remove prototype.
22695 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
22697 * config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Remove call to
22698 rs6000_special_adjust_field_align_p.
22700 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
22703 * config/i386/i386.h (VALID_SSE2_REG_MODE):
22704 Add V4QI and V2HI modes.
22705 (VALID_INT_MODE_P): Ditto.
22706 * config/i386/mmx.md (VI_32): New mode iterator.
22707 (mmxvecsize): Handle V4QI and V2HI.
22709 (mov<VI_32:mode>): New expander.
22710 (*mov<mode>_internal): New insn pattern.
22711 (movmisalign<VI_32:mode>): New expander.
22712 (neg<VI_32:mode>): New expander.
22713 (<plusminus:insn><VI_32:mode>3): New expander.
22714 (*<plusminus:insn><VI_32:mode>3): New insn pattern.
22715 (mulv2hi3): New expander.
22716 (*mulv2hi3): New insn pattern.
22717 (one_cmpl<VI_32:mode>2): New expander.
22718 (*andnot<VI_32:mode>3): New insn pattern.
22719 (<any_logic:code><VI_32:mode>3): New expander.
22720 (*<any_logic:code><VI_32:mode>3): New insn pattern.
22722 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
22724 * config/i386/sse.md (<any_extend:insn>v4qiv4di2):
22725 Fix a mode mismatch with operand 1.
22727 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
22730 * config/i386/i386-expand.c (split_double_mode): Return
22731 temporary register when simplify_gen_subreg fails with
22732 the high half od the paradoxical subreg.
22734 2021-05-18 Richard Biener <rguenther@suse.de>
22736 * cfgexpand.c (expand_one_var): Pass in forced_stack_var
22737 and honor it when expanding.
22738 (expand_used_vars_for_block): Pass through forced_stack_var.
22739 (expand_used_vars): Likewise.
22740 (discover_nonconstant_array_refs_r): Set bits in
22741 forced_stack_vars instead of marking vars TREE_ADDRESSABLE.
22742 (avoid_type_punning_on_regs): Likewise.
22743 (discover_nonconstant_array_refs): Likewise.
22744 (pass_expand::execute): Create and pass down forced_stack_var
22745 bitmap. For parameters and returns temporarily set
22746 TREE_ADDRESSABLE when expand_function_start.
22748 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
22750 * doc/sourcebuild.texi: Document 'dg-note'.
22752 2021-05-18 Tobias Burnus <tobias@codesourcery.com>
22755 * configure: Regenerate.
22756 * configure.ac (BUILD_CFLAG, BUILD_CXXFLAGS): Add $(CFLAGS-$@).
22758 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
22760 * gimple.h (is_gimple_omp_oacc): Tighten.
22761 * omp-low.c (check_omp_nesting_restrictions): Adjust.
22763 2021-05-18 Richard Biener <rguenther@suse.de>
22765 * tree-ssa-operands.c (mark_address_taken): Simplify.
22767 2021-05-18 Martin Liska <mliska@suse.cz>
22769 * config/gcn/mkoffload.c (STR): Redefine.
22770 * config/i386/intelmic-mkoffload.c (STR): Likewise.
22771 * config/nvptx/mkoffload.c (STR): Likewise.
22773 2021-05-18 Martin Liska <mliska@suse.cz>
22775 * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
22776 Use startswith function instead of strncmp.
22777 * common/config/bfin/bfin-common.c (bfin_handle_option): Likewise.
22778 * common/config/riscv/riscv-common.c (riscv_subset_list::parse): Likewise.
22779 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_type): Likewise.
22780 * config/aarch64/aarch64.c (aarch64_process_one_target_attr): Likewise.
22781 * config/alpha/alpha.c (alpha_elf_section_type_flags): Likewise.
22782 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
22783 * config/arm/arm.c (arm_file_start): Likewise.
22784 (arm_valid_target_attribute_rec): Likewise.
22785 (thumb1_md_asm_adjust): Likewise.
22786 * config/arm/driver-arm.c (host_detect_local_cpu): Likewise.
22787 * config/avr/avr.c (STR_PREFIX_P): Likewise.
22788 (avr_set_current_function): Likewise.
22789 (avr_handle_addr_attribute): Likewise.
22790 (avr_asm_output_aligned_decl_common): Likewise.
22791 (avr_asm_named_section): Likewise.
22792 (avr_section_type_flags): Likewise.
22793 (avr_asm_select_section): Likewise.
22794 * config/c6x/c6x.c (c6x_in_small_data_p): Likewise.
22795 (c6x_section_type_flags): Likewise.
22796 * config/darwin-c.c (darwin_cfstring_ref_p): Likewise.
22797 (darwin_objc_declare_unresolved_class_reference): Likewise.
22798 (darwin_objc_declare_class_definition): Likewise.
22799 * config/darwin.c (indirect_data): Likewise.
22800 (darwin_encode_section_info): Likewise.
22801 (darwin_objc2_section): Likewise.
22802 (darwin_objc1_section): Likewise.
22803 (machopic_select_section): Likewise.
22804 (darwin_globalize_label): Likewise.
22805 (darwin_label_is_anonymous_local_objc_name): Likewise.
22806 (darwin_asm_named_section): Likewise.
22807 (darwin_asm_output_dwarf_offset): Likewise.
22808 * config/frv/frv.c (frv_string_begins_with): Likewise.
22809 (frv_in_small_data_p): Likewise.
22810 * config/gcn/mkoffload.c (STR): Likewise.
22812 * config/i386/i386-builtins.c (get_builtin_code_for_version): Likewise.
22813 * config/i386/i386-options.c (ix86_option_override_internal): Likewise.
22814 * config/i386/i386.c (x86_64_elf_section_type_flags): Likewise.
22815 (ix86_md_asm_adjust): Likewise.
22816 * config/i386/intelmic-mkoffload.c (STR): Likewise.
22817 * config/i386/winnt.c (i386_pe_asm_named_section): Likewise.
22818 (i386_pe_file_end): Likewise.
22819 * config/ia64/ia64.c (ia64_in_small_data_p): Likewise.
22820 (ia64_section_type_flags): Likewise.
22821 * config/mips/driver-native.c (host_detect_local_cpu): Likewise.
22822 * config/mips/mips.c (mips_handle_interrupt_attr): Likewise.
22823 (mips16_stub_function_p): Likewise.
22824 (mips_function_rodata_section): Likewise.
22825 * config/msp430/msp430.c (msp430_mcu_name): Likewise.
22826 (msp430_function_section): Likewise.
22827 (msp430_section_type_flags): Likewise.
22828 (msp430_expand_helper): Likewise.
22829 * config/nios2/nios2.c (nios2_small_section_name_p): Likewise.
22830 (nios2_valid_target_attribute_rec): Likewise.
22831 * config/nvptx/mkoffload.c (process): Likewise.
22833 * config/pa/som.h: Likewise.
22834 * config/pdp11/pdp11.c (pdp11_output_ident): Likewise.
22835 * config/riscv/riscv.c (riscv_elf_select_rtx_section): Likewise.
22836 * config/rs6000/rs6000.c (VTABLE_NAME_P): Likewise.
22837 (rs6000_inner_target_options): Likewise.
22838 * config/s390/driver-native.c (s390_host_detect_local_cpu): Likewise.
22839 * config/sparc/driver-sparc.c (host_detect_local_cpu): Likewise.
22840 * config/vax/vax.c (vax_output_int_move): Likewise.
22841 * config/vms/vms-ld.c (startswith): Likewise.
22842 (process_args): Likewise.
22844 * config/vms/vms.c: Likewise.
22846 2021-05-18 Jakub Jelinek <jakub@redhat.com>
22848 PR rtl-optimization/100590
22849 * regcprop.c (copyprop_hardreg_forward_1): Only DCE dead sets if
22850 they are NONJUMP_INSN_P.
22852 2021-05-18 Jakub Jelinek <jakub@redhat.com>
22855 * function.c (push_dummy_function): Set DECL_ARTIFICIAL and
22856 DECL_ASSEMBLER_NAME on the fn_decl.
22858 2021-05-18 Jakub Jelinek <jakub@redhat.com>
22860 PR tree-optimization/94589
22861 * tree-ssa-phiopt.c (spaceship_replacement): Pattern match
22862 phi result used in (res & ~1) == 0 comparison as res >= 0 as
22863 res == 2 would be UB with -ffinite-math-only.
22865 2021-05-18 Martin Liska <mliska@suse.cz>
22867 * Makefile.in: genversion.o should depend on DATESTAMP.
22869 2021-05-18 Claudiu Zissulescu <claziss@synopsys.com>
22871 * config/arc/simdext.md (negv2si2): Remove round bracket.
22873 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
22875 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Define
22876 _Bool as macro expanding to _Bool.
22878 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
22881 * tree.c (build_reference_type_for_mode)
22882 (build_pointer_type_for_mode): Pick pointer mode if MODE argument
22884 (build_reference_type, build_pointer_type): Invoke
22885 build_*_type_for_mode with VOIDmode.
22887 2021-05-17 Andrew MacLeod <amacleod@redhat.com>
22889 PR tree-optimization/100512
22890 * gimple-range-cache.cc (ranger_cache::set_global_range): Mark const
22891 and non-zero pointer ranges as invariant.
22892 * gimple-range.cc (gimple_ranger::range_of_stmt): Remove pointer
22893 processing from here.
22895 2021-05-17 Tom de Vries <tdevries@suse.de>
22898 * config/nvptx/nvptx-protos.h (nvptx_output_atomic_insn): Declare
22899 * config/nvptx/nvptx.c (nvptx_output_barrier)
22900 (nvptx_output_atomic_insn): New function.
22901 (nvptx_print_operand): Add support for 'B'.
22902 * config/nvptx/nvptx.md: Use nvptx_output_atomic_insn for atomic
22905 2021-05-17 Aldy Hernandez <aldyh@redhat.com>
22907 PR tree-optimization/100349
22908 * vr-values.c (bounds_of_var_in_loop): Bail if scev returns
22911 2021-05-17 Tamar Christina <tamar.christina@arm.com>
22913 * config/aarch64/driver-aarch64.c (DEFAULT_ARCH): New.
22914 (host_detect_local_cpu): Use it.
22916 2021-05-17 Martin Liska <mliska@suse.cz>
22918 * doc/invoke.texi: Add 2 missing dots.
22920 2021-05-17 Marius Hillenbrand <mhillen@linux.ibm.com>
22922 PR bootstrap/100552
22923 * configure.ac: Replace pattern substitution with call to sed.
22924 * configure: Regenerate.
22926 2021-05-17 Richard Biener <rguenther@suse.de>
22928 PR middle-end/100582
22929 * tree.c (array_at_struct_end_p): Get to the base of the
22930 reference before looking for the underlying decl.
22932 2021-05-17 Joern Rennecke <joern.rennecke@embecosm.com>
22934 * genoutput.c (validate_insn_alternatives) Make "wrong number of
22935 alternatives" message more specific, and remove assumption on where
22938 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
22940 * config/arm/iterators.md (V16): New iterator.
22941 (VH_cvtto): New iterator.
22942 (v_cmp_result): Added V4HF and V8HF support.
22943 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result>): Use VDQWH.
22944 (vcond<mode><mode>): Likewise.
22945 (vcond_mask_<mode><v_cmp_result>): Likewise.
22946 (vcond<VH_cvtto><mode>): New expander.
22948 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
22950 * config/arm/arm-protos.h (arm_expand_vector_compare): Update
22952 * config/arm/arm.c (arm_expand_vector_compare): Add support for
22954 (arm_expand_vcond): Likewise.
22955 * config/arm/iterators.md (supf): Remove VCMPNEQ_S, VCMPEQQ_S,
22956 VCMPEQQ_N_S, VCMPNEQ_N_S.
22957 (VCMPNEQ, VCMPEQQ, VCMPEQQ_N, VCMPNEQ_N): Remove.
22958 * config/arm/mve.md (@mve_vcmp<mve_cmp_op>q_<mode>): Add '@' prefix.
22959 (@mve_vcmp<mve_cmp_op>q_f<mode>): Likewise.
22960 (@mve_vcmp<mve_cmp_op>q_n_f<mode>): Likewise.
22961 (@mve_vpselq_<supf><mode>): Likewise.
22962 (@mve_vpselq_f<mode>"): Likewise.
22963 * config/arm/neon.md (vec_cmp<mode><v_cmp_result): Enable for MVE
22964 and move to vec-common.md.
22965 (vec_cmpu<mode><mode>): Likewise.
22966 (vcond<mode><mode>): Likewise.
22967 (vcond<V_cvtto><mode>): Likewise.
22968 (vcondu<mode><v_cmp_result>): Likewise.
22969 (vcond_mask_<mode><v_cmp_result>): Likewise.
22970 * config/arm/unspecs.md (VCMPNEQ_U, VCMPNEQ_S, VCMPEQQ_S)
22971 (VCMPEQQ_N_S, VCMPNEQ_N_S, VCMPEQQ_U, CMPEQQ_N_U, VCMPNEQ_N_U)
22972 (VCMPGEQ_N_S, VCMPGEQ_S, VCMPGTQ_N_S, VCMPGTQ_S, VCMPLEQ_N_S)
22973 (VCMPLEQ_S, VCMPLTQ_N_S, VCMPLTQ_S, VCMPCSQ_N_U, VCMPCSQ_U)
22974 (VCMPHIQ_N_U, VCMPHIQ_U): Remove.
22975 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result): Moved
22977 (vec_cmpu<mode><mode>): Likewise.
22978 (vcond<mode><mode>): Likewise.
22979 (vcond<V_cvtto><mode>): Likewise.
22980 (vcondu<mode><v_cmp_result>): Likewise.
22981 (vcond_mask_<mode><v_cmp_result>): Likewise. Added unsafe math
22984 2021-05-17 liuhongt <hongtao.liu@intel.com>
22987 * config/i386/i386.c (ix86_gimple_fold_builtin): Use
22988 gsi_insert_seq_before instead.
22990 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
22992 * doc/sourcebuild.texi (arm_qbit_ok): Rename into...
22993 (arm_sat_ok): ...this.
22995 2021-05-17 Martin Liska <mliska@suse.cz>
22997 * lto-wrapper.c (merge_flto_options): Factor out a new function.
22998 (merge_and_complain): Use it.
22999 (run_gcc): Merge also linker command line -flto=foo argument
23002 2021-05-16 Christophe Lyon <christophe.lyon@linaro.org>
23004 * config/arm/arm.h (CPP_SPEC): Remove error message about
23005 -mlittle-endian/-mbig-endian conflict.
23007 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
23009 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
23010 __ROP_PROTECT__ if -mrop-protect is selected.
23012 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
23014 * config/rs6000/rs6000-internal.h (rs6000_stack): Add
23015 rop_hash_save_offset and rop_hash_size.
23016 * config/rs6000/rs6000-logue.c (rs6000_stack_info): Compute
23017 rop_hash_size and rop_hash_save_offset.
23018 (debug_stack_info): Dump rop_hash_save_offset and rop_hash_size.
23019 (rs6000_emit_prologue): Emit hashst[p] in prologue.
23020 (rs6000_emit_epilogue): Emit hashchk[p] in epilogue.
23021 * config/rs6000/rs6000.md (unspec): Add UNSPEC_HASHST and
23023 (hashst): New define_insn.
23024 (hashchk): Likewise.
23026 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
23028 * config/rs6000/rs6000.c (rs6000_option_override_internal):
23029 Disable shrink wrap when inserting ROP-protect instructions.
23030 * config/rs6000/rs6000.opt (mrop-protect): New option.
23031 (mprivileged): Likewise.
23032 * doc/invoke.texi: Document mrop-protect and mprivileged.
23034 2021-05-15 Hans-Peter Nilsson <hp@axis.com>
23036 * reorg.c (fill_slots_from_thread): Reinstate code typoed out in
23039 2021-05-15 Martin Jambor <mjambor@suse.cz>
23042 2021-05-13 Martin Jambor <mjambor@suse.cz>
23044 PR tree-optimization/100453
23045 * tree-sra.c (sra_modify_assign): All const base accesses do not
23046 need refreshing, not just those from decl_pool.
23047 (sra_modify_assign): Do not refresh into a const base decl.
23049 2021-05-15 Jakub Jelinek <jakub@redhat.com>
23051 PR rtl-optimization/100342
23052 * regcprop.c (copy_value): When copying a source reg in a wider
23053 mode than it has recorded for the value, adjust recorded destination
23054 mode too or punt if !REG_CAN_CHANGE_MODE_P.
23056 2021-05-14 Jason Merrill <jason@redhat.com>
23058 * intl.h: Add comments.
23060 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23062 * config/aarch64/aarch64-simd.md
23063 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Split into...
23064 (aarch64_sqdmlsl2_lane<mode>_internal): ... This...
23065 (aarch64_sqdmlal2_lane<mode>_internal): ... And this.
23066 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>_internal): Split into ...
23067 (aarch64_sqdmlsl2_laneq<mode>_internal): ... This...
23068 (aarch64_sqdmlal2_laneq<mode>_internal): ... And this.
23069 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>_internal): Split into...
23070 (aarch64_sqdmlsl2_n<mode>_internal): ... This...
23071 (aarch64_sqdmlal2_n<mode>_internal): ... And this.
23073 2021-05-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23076 * config/arm/arm_neon.h (vtst_s8): Replace call to vtst builtin with it's
23077 boolean logic equivalent.
23078 (vtst_s16): Likewise.
23079 (vtst_s32): Likewise.
23080 (vtst_u8): Likewise.
23081 (vtst_u16): Likewise.
23082 (vtst_u32): Likewise.
23083 (vtst_p8): Likewise.
23084 (vtst_p16): Likewise.
23085 (vtstq_s8): Likewise.
23086 (vtstq_s16): Likewise.
23087 (vtstq_s32): Likewise.
23088 (vtstq_u8): Likewise.
23089 (vtstq_u16): Likewise.
23090 (vtstq_u32): Likewise.
23091 (vtstq_p8): Likewise.
23092 (vtstq_p16): Likewise.
23093 * config/arm/arm_neon_builtins.def: Remove entry for vtst.
23094 * config/arm/neon.md (neon_vtst<mode>): Remove pattern.
23096 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23098 * config/aarch64/aarch64-simd.md (aarch64_sqdmlal2<mode>): Merge into...
23099 (aarch64_sqdml<SBINQOPS:as>l2<mode>): ... This.
23100 (aarch64_sqdmlsl2<mode>): Delete.
23101 (aarch64_sqdmlal2_lane<mode>): Merge this...
23102 (aarch64_sqdmlsl2_lane<mode>): ... And this...
23103 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>): ... Into this.
23104 (aarch64_sqdmlal2_laneq<mode>): Merge this...
23105 (aarch64_sqdmlsl2_laneq<mode>): ... And this...
23106 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>): ... Into this.
23107 (aarch64_sqdmlal2_n<mode>): Merge this...
23108 (aarch64_sqdmlsl2_n<mode>): ... And this...
23109 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>): ... Into this.
23111 2021-05-13 Martin Sebor <msebor@redhat.com>
23113 PR middle-end/100574
23114 * builtins.c (access_ref::get_ref): Improve detection of PHIs with
23115 all null arguments.
23117 2021-05-13 Martin Sebor <msebor@redhat.com>
23119 PR tree-optimization/93100
23120 PR middle-end/98583
23121 * tree-ssa-uninit.c (check_defs): Exclude intrinsic functions that
23122 don't modify referenced objects.
23124 2021-05-13 Martin Jambor <mjambor@suse.cz>
23126 PR tree-optimization/100453
23127 * tree-sra.c (sra_modify_assign): All const base accesses do not
23128 need refreshing, not just those from decl_pool.
23129 (sra_modify_assign): Do not refresh into a const base decl.
23131 2021-05-13 Martin Liska <mliska@suse.cz>
23133 * tree-ssa-dom.c: Remove m_simplifier.
23135 2021-05-13 Richard Earnshaw <rearnsha@arm.com>
23138 * config/arm/arm.c (arm_canonicalize_comparison): Correctly
23139 canonicalize DImode inequality comparisons against the
23140 maximum integral value.
23142 2021-05-13 Jakub Jelinek <jakub@redhat.com>
23144 PR tree-optimization/98856
23145 * config/i386/i386.c (ix86_shift_rotate_cost): Add CODE argument.
23146 Expect V2DI and V4DI arithmetic right shifts to be emulated.
23147 (ix86_rtx_costs, ix86_add_stmt_cost): Adjust ix86_shift_rotate_cost
23149 * config/i386/i386-expand.c (expand_vec_perm_2perm_interleave,
23150 expand_vec_perm_2perm_pblendv): New functions.
23151 (ix86_expand_vec_perm_const_1): Use them.
23152 * config/i386/sse.md (ashr<mode>3<mask_name>): Rename to ...
23153 (<mask_codefor>ashr<mode>3<mask_name>): ... this.
23154 (ashr<mode>3): New define_expand with VI248_AVX512BW iterator.
23155 (ashrv4di3): New define_expand.
23156 (ashrv2di3): Change condition to TARGET_SSE2, handle !TARGET_XOP
23157 and !TARGET_AVX512VL expansion.
23159 2021-05-13 Uroš Bizjak <ubizjak@gmail.com>
23162 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Force mode
23163 sizes < 16 to a register when constructing vpcmov pattern.
23164 * config/i386/mmx.md (*xop_pcmov_<mode>): Use MMXMODE124 mode.
23166 2021-05-13 Martin Liska <mliska@suse.cz>
23168 * gcov-io.c (gcov_write_block): Remove.
23169 (gcov_write_words): Likewise.
23170 (gcov_read_words): Re-implement using gcov_read_bytes.
23171 (gcov_allocate): Remove.
23172 (GCOV_BLOCK_SIZE): Likewise.
23173 (struct gcov_var): Remove most of the fields.
23174 (gcov_position): Implement with ftell.
23175 (gcov_rewrite): Remove setting of start and offset fields.
23176 (from_file): Re-format.
23177 (gcov_open): Remove setbuf call. It should not be needed.
23178 (gcov_close): Remove internal buffer handling.
23179 (gcov_magic): Use __builtin_bswap32.
23180 (gcov_write_counter): Use directly gcov_write_unsigned.
23181 (gcov_write_string): Use direct fwrite and do not round
23183 (gcov_seek): Use directly fseek.
23184 (gcov_write_tag): Use gcov_write_unsigned directly.
23185 (gcov_write_length): Likewise.
23186 (gcov_write_tag_length): Likewise.
23187 (gcov_read_bytes): Use directly fread.
23188 (gcov_read_unsigned): Use gcov_read_words.
23189 (gcov_read_counter): Likewise.
23190 (gcov_read_string): Use gcov_read_bytes.
23191 * gcov-io.h (GCOV_WORD_SIZE): Adjust to reflect
23192 that size is not in bytes, but words (4B).
23193 (GCOV_TAG_FUNCTION_LENGTH): Likewise.
23194 (GCOV_TAG_ARCS_LENGTH): Likewise.
23195 (GCOV_TAG_ARCS_NUM): Likewise.
23196 (GCOV_TAG_COUNTER_LENGTH): Likewise.
23197 (GCOV_TAG_COUNTER_NUM): Likewise.
23198 (GCOV_TAG_SUMMARY_LENGTH): Likewise.
23200 2021-05-13 liuhongt <hongtao.liu@intel.com>
23203 * config/i386/sse.md (ssedoublevecmode): Add attribute for
23204 V64QI/V32HI/V16SI/V4DI.
23205 (ssehalfvecmode): Add attribute for V2DI/V2DF.
23206 (*vec_concatv4si_0): Extend to VI124_128.
23207 (*vec_concat<mode>_0): New pre-reload splitter.
23208 * config/i386/predicates.md (movq_parallel): New predicate.
23210 2021-05-13 Alexandre Oliva <oliva@adacore.com>
23212 * targhooks.c (default_zero_call_used_regs): Retry using
23213 successfully-zeroed registers as sources.
23215 2021-05-12 Tobias Burnus <tobias@codesourcery.com>
23217 * omp-low.c (finish_taskreg_scan): Use the proper detach decl.
23219 2021-05-12 Aldy Hernandez <aldyh@redhat.com>
23222 * gimple-range.cc (range_of_builtin_call): Skip out on
23223 processing __builtin_clz when varying.
23225 2021-05-12 Tom de Vries <tdevries@suse.de>
23228 * config/nvptx/nvptx-opts.h (enum ptx_version): New enum.
23229 * config/nvptx/nvptx.c (nvptx_file_start): Print .version according
23230 to ptx_version_option.
23231 * config/nvptx/nvptx.h (TARGET_PTX_6_3): Define.
23232 * config/nvptx/nvptx.md (define_insn "nvptx_shuffle<mode>")
23233 (define_insn "nvptx_vote_ballot"): Use sync variant for
23235 * config/nvptx/nvptx.opt (ptx_version): Add enum.
23236 (mptx): Add option.
23237 * doc/invoke.texi (Nvidia PTX Options): Add mptx item.
23239 2021-05-12 Richard Biener <rguenther@suse.de>
23241 PR tree-optimization/100566
23242 * tree-ssa-sccvn.c (dominated_by_p_w_unex): Properly handle
23243 allow_back for all edge queries.
23245 2021-05-12 liuhongt <hongtao.liu@intel.com>
23248 * config/i386/sse.md (<sse4_1_avx2>_pblendvb): Add
23249 splitters for pblendvb of NOT mask register.
23251 2021-05-12 Richard Biener <rguenther@suse.de>
23253 PR tree-optimization/100519
23254 * tree-ssa-reassoc.c (can_associate_p): Split into...
23255 (can_associate_op_p): ... this
23256 (can_associate_type_p): ... and this.
23257 (is_reassociable_op): Call can_associate_op_p.
23258 (break_up_subtract_bb): Call the appropriate predicates.
23259 (reassociate_bb): Likewise.
23261 2021-05-12 Martin Liska <mliska@suse.cz>
23263 * lto-wrapper.c (merge_and_complain): Merge -flto=arg options.
23264 (run_gcc): Use -flto argument detection for merged
23267 2021-05-12 Martin Liska <mliska@suse.cz>
23269 * lto-wrapper.c (print_lto_docs_link): New function.
23270 (run_gcc): Print warning about missing job server detection
23271 after we know NR of partitions. Do the same for -flto{,=1}.
23272 * opts.c (get_option_html_page): Support -flto option.
23274 2021-05-12 Martin Liska <mliska@suse.cz>
23276 * lto-wrapper.c (get_options_from_collect_gcc_options): Change
23278 (append_option): Remove.
23279 (find_option): Rework to use the vector type.
23280 (remove_option): Remove.
23281 (merge_and_complain): Use vectors for cl_decoded_option data
23283 (append_compiler_options): Likewise.
23284 (append_diag_options): Likewise.
23285 (append_linker_options): Likewise.
23286 (append_offload_options): Likewise.
23287 (compile_offload_image): Likewise.
23288 (compile_images_for_offload_targets): Likewise.
23289 (find_and_merge_options): Likewise.
23290 (run_gcc): Likewise.
23292 2021-05-12 Bernd Edlinger <bernd.edlinger@hotmail.de>
23295 * dwarf2out.c (dwarf2out_finish): Set
23296 have_multiple_function_sections with multi-range text_section.
23298 2021-05-12 Martin Liska <mliska@suse.cz>
23300 PR bootstrap/100560
23301 * Makefile.in: Remove version.h from linker command line.
23303 2021-05-12 Richard Biener <rguenther@suse.de>
23305 PR middle-end/100547
23306 * rtl.h (rtvec_alloc): Make argument size_t.
23307 * rtl.c (rtvec_alloc): Verify the count is less than INT_MAX.
23309 2021-05-12 Jakub Jelinek <jakub@redhat.com>
23311 PR middle-end/100508
23312 * cfgexpand.c (expand_debug_expr): For DEBUG_EXPR_DECL with vector
23313 type, don't reuse DECL_RTL if it has different mode, instead force
23314 creation of a new DEBUG_EXPR.
23316 2021-05-12 Jakub Jelinek <jakub@redhat.com>
23317 Marc Glisse <marc.glisse@inria.fr>
23319 PR tree-optimization/94589
23320 * match.pd ((X & Y) == X -> (X & ~Y) == 0,
23321 (X | Y) == Y -> (X & ~Y) == 0): New GIMPLE simplifications.
23323 2021-05-12 Uroš Bizjak <ubizjak@gmail.com>
23326 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Handle V2SF mode.
23327 * config/i386/mmx.md (MMXMODE124): New mode iterator.
23329 (mmxintvecmode): New mode attribute.
23330 (mmxintvecmodelower): Ditto.
23331 (*mmx_maskcmpv2sf3_comm): New insn pattern.
23332 (*mmx_maskcmpv2sf3): Ditto.
23333 (vec_cmpv2sfv2si): New expander.
23334 (vcond<V2FI:mode>v2si): Ditto.
23335 (mmx_vlendvps): New insn pattern.
23336 (vcond<MMXMODE124:mode><MMXMODEI:mode>): Also handle V2SFmode.
23337 (vcondu<MMXMODE124:mode><MMXMODEI:mode>): Ditto.
23338 (vcond_mask_<mode><mmxintvecmodelower>): Ditto.
23340 2021-05-11 Martin Sebor <msebor@redhat.com>
23342 PR middle-end/21433
23343 * expr.c (expand_expr_real_1): Replace unreachable code with an assert.
23345 2021-05-11 Richard Biener <rguenther@suse.de>
23347 * gimple-fold.c (gimple_fold_call): Do not call
23348 maybe_fold_reference on call arguments or the static chain.
23349 (fold_stmt_1): Do not call maybe_fold_reference on GIMPLE_ASM
23352 2021-05-11 Martin Liska <mliska@suse.cz>
23354 * builtins.def (DEF_HSAIL_BUILTIN): Remove.
23355 (DEF_HSAIL_ATOMIC_BUILTIN): Likewise.
23356 (DEF_HSAIL_SAT_BUILTIN): Likewise.
23357 (DEF_HSAIL_INTR_BUILTIN): Likewise.
23358 (DEF_HSAIL_CVT_ZEROI_SAT_BUILTIN): Likewise.
23359 * doc/frontends.texi: Remove BRIG.
23360 * doc/install.texi: Likewise.
23361 * doc/invoke.texi: Likewise.
23362 * doc/standards.texi: Likewise.
23363 * brig-builtins.def: Removed.
23364 * brig/ChangeLog: Removed.
23365 * brig/Make-lang.in: Removed.
23366 * brig/brig-builtins.h: Removed.
23367 * brig/brig-c.h: Removed.
23368 * brig/brig-lang.c: Removed.
23369 * brig/brigfrontend/brig-arg-block-handler.cc: Removed.
23370 * brig/brigfrontend/brig-atomic-inst-handler.cc: Removed.
23371 * brig/brigfrontend/brig-basic-inst-handler.cc: Removed.
23372 * brig/brigfrontend/brig-branch-inst-handler.cc: Removed.
23373 * brig/brigfrontend/brig-cmp-inst-handler.cc: Removed.
23374 * brig/brigfrontend/brig-code-entry-handler.cc: Removed.
23375 * brig/brigfrontend/brig-code-entry-handler.h: Removed.
23376 * brig/brigfrontend/brig-comment-handler.cc: Removed.
23377 * brig/brigfrontend/brig-control-handler.cc: Removed.
23378 * brig/brigfrontend/brig-copy-move-inst-handler.cc: Removed.
23379 * brig/brigfrontend/brig-cvt-inst-handler.cc: Removed.
23380 * brig/brigfrontend/brig-fbarrier-handler.cc: Removed.
23381 * brig/brigfrontend/brig-function-handler.cc: Removed.
23382 * brig/brigfrontend/brig-function.cc: Removed.
23383 * brig/brigfrontend/brig-function.h: Removed.
23384 * brig/brigfrontend/brig-inst-mod-handler.cc: Removed.
23385 * brig/brigfrontend/brig-label-handler.cc: Removed.
23386 * brig/brigfrontend/brig-lane-inst-handler.cc: Removed.
23387 * brig/brigfrontend/brig-machine.c: Removed.
23388 * brig/brigfrontend/brig-machine.h: Removed.
23389 * brig/brigfrontend/brig-mem-inst-handler.cc: Removed.
23390 * brig/brigfrontend/brig-module-handler.cc: Removed.
23391 * brig/brigfrontend/brig-queue-inst-handler.cc: Removed.
23392 * brig/brigfrontend/brig-seg-inst-handler.cc: Removed.
23393 * brig/brigfrontend/brig-signal-inst-handler.cc: Removed.
23394 * brig/brigfrontend/brig-to-generic.cc: Removed.
23395 * brig/brigfrontend/brig-to-generic.h: Removed.
23396 * brig/brigfrontend/brig-util.cc: Removed.
23397 * brig/brigfrontend/brig-util.h: Removed.
23398 * brig/brigfrontend/brig-variable-handler.cc: Removed.
23399 * brig/brigfrontend/hsa-brig-format.h: Removed.
23400 * brig/brigfrontend/phsa.h: Removed.
23401 * brig/brigspec.c: Removed.
23402 * brig/config-lang.in: Removed.
23403 * brig/gccbrig.texi: Removed.
23404 * brig/lang-specs.h: Removed.
23405 * brig/lang.opt: Removed.
23407 2021-05-11 Richard Biener <rguenther@suse.de>
23410 * ipa-param-manipulation.c
23411 (ipa_param_body_adjustments::modify_call_stmt): Avoid
23412 altering SSA_NAME_DEF_STMT by adjusting the calls LHS
23413 via gimple_call_lhs_ptr.
23415 2021-05-11 Alex Coplan <alex.coplan@arm.com>
23418 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear):
23419 Avoid emitting CFA adjusts on the sp if we have the fp.
23421 2021-05-11 Richard Sandiford <richard.sandiford@arm.com>
23423 * config/aarch64/iterators.md (VMUL_CHANGE_NLANES): Delete.
23424 (VMULD): New iterator.
23425 (VCOND): Handle V4HF and V8HF.
23426 (VCONQ): Fix entry for V2SF.
23427 * config/aarch64/aarch64-simd.md (mul_lane<mode>3): Use VMULD
23428 instead of VMUL. Use a 64-bit vector mode for the indexed operand.
23429 (*aarch64_mul3_elt_<vswap_width_name><mode>): Merge with...
23430 (mul_laneq<mode>3): ...this define_insn. Use VMUL instead of VDQSF.
23431 Use a 128-bit vector mode for the indexed operand. Use stype for
23432 the scheduling type.
23434 2021-05-11 Richard Biener <rguenther@suse.de>
23436 * gimple-fold.c (maybe_fold_reference): Only return
23437 is_gimple_min_invariant values.
23439 2021-05-11 Richard Biener <rguenther@suse.de>
23441 PR middle-end/100509
23442 * gimple-fold.c (fold_gimple_assign): Only call
23443 get_symbol_constant_value on register type symbols.
23445 2021-05-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
23446 Joe Ramsay <joe.ramsay@arm.com>
23449 * config/arm/arm_mve.h (__arm_vstrwq_scatter_offset): Fix wrong arguments.
23450 (__arm_vcmpneq): Remove duplicate definition.
23451 (__arm_vstrwq_scatter_offset_p): Likewise.
23452 (__arm_vmaxq_x): Likewise.
23453 (__arm_vmlsdavaq): Likewise.
23454 (__arm_vmlsdavaxq): Likewise.
23455 (__arm_vmlsdavq_p): Likewise.
23456 (__arm_vmlsdavxq_p): Likewise.
23457 (__arm_vrmlaldavhaq): Likewise.
23458 (__arm_vstrbq_p): Likewise.
23459 (__arm_vstrbq_scatter_offset): Likewise.
23460 (__arm_vstrbq_scatter_offset_p): Likewise.
23461 (__arm_vstrdq_scatter_offset): Likewise.
23462 (__arm_vstrdq_scatter_offset_p): Likewise.
23463 (__arm_vstrdq_scatter_shifted_offset): Likewise.
23464 (__arm_vstrdq_scatter_shifted_offset_p): Likewise.
23466 2021-05-11 Jakub Jelinek <jakub@redhat.com>
23468 PR middle-end/100471
23469 * omp-low.c (lower_omp_task_reductions): For OMP_TASKLOOP, if data
23470 is 0, bypass the reduction loop including
23471 GOMP_taskgroup_reduction_unregister call.
23473 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
23475 * config/rs6000/rs6000.c (struct rs6000_cost_data): New member
23476 costing_for_scalar.
23477 (rs6000_density_test): Early return if costing_for_scalar is true.
23478 (rs6000_init_cost): Init costing_for_scalar of rs6000_cost_data.
23480 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
23482 * doc/tm.texi: Regenerated.
23483 * target.def (init_cost): Add new parameter costing_for_scalar.
23484 * targhooks.c (default_init_cost): Adjust for new parameter.
23485 * targhooks.h (default_init_cost): Likewise.
23486 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
23487 (vect_compute_single_scalar_iteration_cost): Likewise.
23488 (vect_analyze_loop_2): Likewise.
23489 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
23490 (vect_bb_vectorization_profitable_p): Likewise.
23491 * tree-vectorizer.h (init_cost): Likewise.
23492 * config/aarch64/aarch64.c (aarch64_init_cost): Likewise.
23493 * config/i386/i386.c (ix86_init_cost): Likewise.
23494 * config/rs6000/rs6000.c (rs6000_init_cost): Likewise.
23496 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
23498 * config/rs6000/rs6000.c (rs6000_vect_nonmem): Renamed to
23499 vect_nonmem and moved into...
23500 (struct rs6000_cost_data): ...here.
23501 (rs6000_init_cost): Use vect_nonmem of cost_data instead.
23502 (rs6000_add_stmt_cost): Likewise.
23503 (rs6000_finish_cost): Likewise.
23505 2021-05-10 Eric Botcazou <ebotcazou@adacore.com>
23507 * range-op.cc (get_bool_state): Adjust head comment.
23508 (operator_not_equal::op1_range): Fix comment.
23509 (operator_bitwise_xor::op1_range): Remove call to gcc_unreachable.
23511 2021-05-10 Martin Sebor <msebor@redhat.com>
23513 PR middle-end/100425
23514 PR middle-end/100510
23515 * gimple-ssa-warn-alloca.c (pass_walloca::firast_time_p): Rename...
23516 (pass_walloca::xlimit_certain_p): ...to this.
23517 (pass_walloca::gate): Execute for any kind of handled warning.
23518 (pass_walloca::execute): Avoid issuing "maybe" and "unbounded"
23519 warnings when xlimit_certain_p is set.
23521 2021-05-10 Pat Haugen <pthaugen@linux.ibm.com>
23523 * config/rs6000/rs6000.c (rs6000_ira_change_pseudo_allocno_class):
23524 Return ALTIVEC_REGS if that is best_class.
23525 (rs6000_compute_pressure_classes): Add ALTIVEC_REGS.
23527 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
23529 * config/arm/arm.h (CPP_SPEC): Remove error message about
23532 2021-05-10 Martin Jambor <mjambor@suse.cz>
23534 * ipa-prop.h (IPA_NODE_REF): Removed.
23535 (IPA_NODE_REF_GET_CREATE): Likewise.
23536 (IPA_EDGE_REF): Likewise.
23537 (IPA_EDGE_REF_GET_CREATE): Likewise.
23538 (IS_VALID_JUMP_FUNC_INDEX): Likewise.
23539 * ipa-cp.c (print_all_lattices): Replaced IPA_NODE_REF with a direct
23540 use of ipa_node_params_sum.
23541 (ipcp_versionable_function_p): Likewise.
23542 (push_node_to_stack): Likewise.
23543 (pop_node_from_stack): Likewise.
23544 (set_single_call_flag): Replaced two IPA_NODE_REF with one single
23545 direct use of ipa_node_params_sum.
23546 (initialize_node_lattices): Replaced IPA_NODE_REF with a direct use of
23547 ipa_node_params_sum.
23548 (ipa_context_from_jfunc): Replaced IPA_EDGE_REF with a direct use of
23550 (ipcp_verify_propagated_values): Replaced IPA_NODE_REF with a direct
23551 use of ipa_node_params_sum.
23552 (self_recursively_generated_p): Likewise.
23553 (propagate_scalar_across_jump_function): Likewise.
23554 (propagate_context_across_jump_function): Replaced IPA_EDGE_REF with a
23555 direct use of ipa_edge_args_sum, moved the lookup after the early
23556 exit. Replaced IPA_NODE_REF with a direct use of ipa_node_params_sum.
23557 (propagate_bits_across_jump_function): Replaced IPA_NODE_REF with
23558 direct uses of ipa_node_params_sum.
23559 (propagate_vr_across_jump_function): Likewise.
23560 (propagate_aggregate_lattice): Likewise.
23561 (propagate_aggs_across_jump_function): Likewise.
23562 (propagate_constants_across_call): Likewise, also replaced
23563 IPA_EDGE_REF with a direct use of ipa_edge_args_sum.
23564 (good_cloning_opportunity_p): Replaced IPA_NODE_REF with a direct use
23565 of ipa_node_params_sum.
23566 (estimate_local_effects): Likewise.
23567 (add_all_node_vals_to_toposort): Likewise.
23568 (propagate_constants_topo): Likewise.
23569 (ipcp_propagate_stage): Likewise.
23570 (ipcp_discover_new_direct_edges): Likewise.
23571 (calls_same_node_or_its_all_contexts_clone_p): Likewise.
23572 (cgraph_edge_brings_value_p): Likewise (in both overloaded functions).
23573 (get_info_about_necessary_edges): Likewise.
23574 (want_remove_some_param_p): Likewise.
23575 (create_specialized_node): Likewise.
23576 (self_recursive_pass_through_p): Likewise.
23577 (self_recursive_agg_pass_through_p): Likewise.
23578 (find_more_scalar_values_for_callers_subset): Likewise and also
23579 replaced IPA_EDGE_REF with direct uses of ipa_edge_args_sum, in one
23580 case replacing two of those with a single query.
23581 (find_more_contexts_for_caller_subset): Likewise for the
23582 ipa_polymorphic_call_context overload.
23583 (intersect_aggregates_with_edge): Replaced IPA_EDGE_REF with a direct
23584 use of ipa_edge_args_sum. Replaced IPA_NODE_REF with direct uses of
23585 ipa_node_params_sum.
23586 (find_aggregate_values_for_callers_subset): Likewise, also reusing
23587 results of ipa_edge_args_sum->get.
23588 (cgraph_edge_brings_all_scalars_for_node): Replaced IPA_NODE_REF with
23589 direct uses of ipa_node_params_sum, replaced IPA_EDGE_REF with a
23590 direct use of ipa_edge_args_sum.
23591 (cgraph_edge_brings_all_agg_vals_for_node): Likewise, moved node
23592 summary query after the early exit and reused the result later.
23593 (decide_about_value): Replaced IPA_NODE_REF with a direct use of
23594 ipa_node_params_sum.
23595 (decide_whether_version_node): Likewise. Removed re-querying for
23596 summaries after cloning.
23597 (spread_undeadness): Replaced IPA_NODE_REF with a direct use of
23598 ipa_node_params_sum.
23599 (has_undead_caller_from_outside_scc_p): Likewise, reusing results of
23601 (identify_dead_nodes): Likewise.
23602 (ipcp_store_bits_results): Replaced IPA_NODE_REF with direct uses of
23603 ipa_node_params_sum.
23604 (ipcp_store_vr_results): Likewise.
23605 * ipa-fnsummary.c (evaluate_properties_for_edge): Likewise.
23606 (ipa_fn_summary_t::duplicate): Likewise.
23607 (analyze_function_body): Likewise.
23608 (estimate_calls_size_and_time): Likewise.
23609 (ipa_cached_call_context::duplicate_from): Likewise.
23610 (ipa_call_context::equal_to): Likewise.
23611 (remap_edge_params): Likewise.
23612 (ipa_merge_fn_summary_after_inlining): Likewise.
23613 (inline_read_section): Likewise.
23614 * ipa-icf.c (sem_function::param_used_p): Likewise.
23615 * ipa-modref.c (compute_parm_map): Likewise.
23616 (compute_parm_map): Replaced IPA_EDGE_REF with a direct use of
23618 (get_access_for_fnspec): Replaced IPA_NODE_REF with a direct use of
23619 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
23621 * ipa-profile.c (check_argument_count): Likewise.
23622 * ipa-prop.c (ipa_alloc_node_params): Replaced IPA_NODE_REF_GET_CREATE
23623 with a direct use of ipa_node_params_sum.
23624 (ipa_initialize_node_params): Likewise.
23625 (ipa_print_node_jump_functions_for_edge): Replaced IPA_EDGE_REF with a
23626 direct use of ipa_edge_args_sum and reused the query result.
23627 (ipa_compute_jump_functions_for_edge): Replaced IPA_NODE_REF with a
23628 direct use of ipa_node_params_sum and replaced IPA_EDGE_REF with a
23629 direct use of ipa_edge_args_sum.
23630 (ipa_note_param_call): Replaced IPA_NODE_REF with a direct use of
23631 ipa_node_params_sum and reused the result of the query.
23632 (ipa_analyze_node): Likewise.
23633 (ipa_analyze_controlled_uses): Replaced IPA_NODE_REF with a direct use
23634 of ipa_node_params_sum.
23635 (update_jump_functions_after_inlining): Replaced IPA_EDGE_REF with
23636 direct uses of ipa_edge_args_sum.
23637 (update_indirect_edges_after_inlining): Replaced IPA_NODE_REF with
23638 direct uses of ipa_node_params_sum and replaced IPA_EDGE_REF with a
23639 direct use of ipa_edge_args_sum. Removed superficial re-querying the
23641 (propagate_controlled_uses): Replaced IPA_NODE_REF with direct uses of
23642 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
23644 (ipa_propagate_indirect_call_infos): Replaced IPA_EDGE_REF with a
23645 direct use of ipa_edge_args_sum.
23646 (ipa_edge_args_sum_t::duplicate): Replaced IPA_NODE_REF with a direct
23647 use of ipa_node_params_sum.
23648 (ipa_print_node_params): Likewise.
23649 (ipa_write_node_info): Likewise and also replaced IPA_EDGE_REF with
23650 direct uses of ipa_edge_args_sum.
23651 (ipa_read_edge_info): Replaced IPA_EDGE_REF with a direct use of
23653 (ipa_read_node_info): Replaced IPA_NODE_REF with a direct use of
23654 ipa_node_params_sum.
23655 (ipa_prop_write_jump_functions): Likewise. Move variable node to the
23656 scopes where it is used.
23658 2021-05-10 Uroš Bizjak <ubizjak@gmail.com>
23660 * config/i386/i386-expand.c (ix86_expand_sse_movcc)
23661 <case E_V2SImode>: Force op_true to register.
23663 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
23665 * config/arm/iterators.md (MVE_FP_COMPARISONS): New.
23666 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_f<mode>)
23667 (mve_vcmp<mve_cmp_op>q_n_f<mode>): New, merge all vcmp_*f*
23669 (mve_vcmpeqq_f<mode>, mve_vcmpeqq_n_f<mode>, mve_vcmpgeq_f<mode>)
23670 (mve_vcmpgeq_n_f<mode>, mve_vcmpgtq_f<mode>)
23671 (mve_vcmpgtq_n_f<mode>, mve_vcmpleq_f<mode>)
23672 (mve_vcmpleq_n_f<mode>, mve_vcmpltq_f<mode>)
23673 (mve_vcmpltq_n_f<mode>, mve_vcmpneq_f<mode>)
23674 (mve_vcmpneq_n_f<mode>): Remove.
23675 * config/arm/unspecs.md (VCMPEQQ_F, VCMPEQQ_N_F, VCMPGEQ_F)
23676 (VCMPGEQ_N_F, VCMPGTQ_F, VCMPGTQ_N_F, VCMPLEQ_F, VCMPLEQ_N_F)
23677 (VCMPLTQ_F, VCMPLTQ_N_F, VCMPNEQ_F, VCMPNEQ_N_F): Remove.
23679 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
23681 * config/arm/iterators.md (MVE_COMPARISONS): New.
23683 (mve_cmp_type): New.
23684 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_<mode>): New, merge all
23686 (mve_vcmpneq_<mode>, mve_vcmpcsq_n_<mode>, mve_vcmpcsq_<mode>)
23687 (mve_vcmpeqq_n_<mode>, mve_vcmpeqq_<mode>, mve_vcmpgeq_n_<mode>)
23688 (mve_vcmpgeq_<mode>, mve_vcmpgtq_n_<mode>, mve_vcmpgtq_<mode>)
23689 (mve_vcmphiq_n_<mode>, mve_vcmphiq_<mode>, mve_vcmpleq_n_<mode>)
23690 (mve_vcmpleq_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
23691 (mve_vcmpneq_n_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
23692 (mve_vcmpneq_n_<mode>): Remove.
23694 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
23696 * config/arm/arm_mve.h (__arm_vcmp*): Remove 's' suffix.
23697 * config/arm/arm_mve_builtins.def (vcmp*): Remove 's' suffix.
23698 * config/arm/mve.md (mve_vcmp*): Remove 's' suffix in pattern
23701 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
23703 * config/arm/arm_mve_builtins.def (vcmpneq_u): Remove.
23704 (vcmpneq_n_u): Likewise.
23705 (vcmpeqq_u,): Likewise.
23706 (vcmpeqq_n_u): Likewise.
23707 * config/arm/iterators.md (supf): Remove VCMPNEQ_U, VCMPEQQ_U,
23708 VCMPEQQ_N_U and VCMPNEQ_N_U.
23709 * config/arm/mve.md (mve_vcmpneq): Remove <supf> iteration.
23710 (mve_vcmpeqq_n): Likewise.
23711 (mve_vcmpeqq): Likewise.
23712 (mve_vcmpneq_n): Likewise.
23714 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
23716 * config/arm/arm_mve.h (__arm_vcmpeq*u*, __arm_vcmpne*u*): Call
23717 the 's' version of the builtin.
23719 2021-05-10 Richard Biener <rguenther@suse.de>
23721 PR tree-optimization/100492
23722 * tree-loop-distribution.c (find_seed_stmts_for_distribution):
23723 Find nothing when the loop contains an irreducible region.
23725 2021-05-10 Richard Biener <rguenther@suse.de>
23727 PR middle-end/100464
23729 * gimple-fold.c (canonicalize_constructor_val): Do not set
23732 2021-05-10 Richard Biener <rguenther@suse.de>
23734 PR tree-optimization/100434
23735 * tree-ssa-dse.c (initialize_ao_ref_for_dse): Handle
23737 (dse_optimize_stmt): Handle call LHS by dropping the
23738 LHS or the whole call if it doesn't have other
23740 (pass_dse::execute): Adjust.
23742 2021-05-10 Martin Liska <mliska@suse.cz>
23744 * Makefile.in: Add missing genversion rule.
23746 2021-05-10 Alex Coplan <alex.coplan@arm.com>
23749 * config/arm/mve.md (*mve_mov<mode>): Simplify output code. Use
23750 vldrw.u32 and vstrw.32 for V2D[IF]mode loads and stores.
23752 2021-05-10 Martin Liska <mliska@suse.cz>
23754 * builtins.c (is_builtin_name): Use startswith
23755 function instead of strncmp.
23756 * collect2.c (main): Likewise.
23757 (has_lto_section): Likewise.
23758 (scan_libraries): Likewise.
23759 * coverage.c (coverage_checksum_string): Likewise.
23760 (coverage_init): Likewise.
23761 * dwarf2out.c (is_cxx): Likewise.
23762 (gen_compile_unit_die): Likewise.
23763 * gcc-ar.c (main): Likewise.
23764 * gcc.c (init_spec): Likewise.
23765 (read_specs): Likewise.
23766 (execute): Likewise.
23767 (check_live_switch): Likewise.
23768 * genattrtab.c (write_attr_case): Likewise.
23769 (IS_ATTR_GROUP): Likewise.
23770 * gencfn-macros.c (main): Likewise.
23771 * gengtype.c (type_for_name): Likewise.
23772 (gen_rtx_next): Likewise.
23773 (get_file_langdir): Likewise.
23774 (write_local): Likewise.
23775 * genmatch.c (get_operator): Likewise.
23776 (get_operand_type): Likewise.
23777 (expr::gen_transform): Likewise.
23778 * genoutput.c (validate_optab_operands): Likewise.
23779 * incpath.c (add_sysroot_to_chain): Likewise.
23780 * langhooks.c (lang_GNU_C): Likewise.
23781 (lang_GNU_CXX): Likewise.
23782 (lang_GNU_Fortran): Likewise.
23783 (lang_GNU_OBJC): Likewise.
23784 * lto-wrapper.c (run_gcc): Likewise.
23785 * omp-general.c (omp_max_simt_vf): Likewise.
23786 * omp-low.c (omp_runtime_api_call): Likewise.
23787 * opts-common.c (parse_options_from_collect_gcc_options): Likewise.
23788 * read-rtl-function.c (function_reader::read_rtx_operand_r): Likewise.
23789 * real.c (real_from_string): Likewise.
23790 * selftest.c (assert_str_startswith): Likewise.
23791 * timevar.c (timer::validate_phases): Likewise.
23792 * tree.c (get_file_function_name): Likewise.
23793 * ubsan.c (ubsan_use_new_style_p): Likewise.
23794 * varasm.c (default_function_rodata_section): Likewise.
23795 (incorporeal_function_p): Likewise.
23796 (default_section_type_flags): Likewise.
23797 * system.h (startswith): Define startswith.
23799 2021-05-10 Martin Liska <mliska@suse.cz>
23801 * bitmap.h (class auto_bitmap): Remove
23802 __cplusplus >= 201103.
23803 * config/aarch64/aarch64.c: Likewise.
23804 * gimple-ssa-store-merging.c (store_immediate_info::store_immediate_info):
23806 * sbitmap.h: Likewise.
23808 2021-05-10 Martin Liska <mliska@suse.cz>
23810 * Makefile.in: Rename gcov-iov to genversion and depend
23811 on version.h (instead of gcov-iov.h).
23812 * gcov-io.h: Include version.h instread of gcov-iov.h.
23813 * gengtype-state.c (read_state_version): Likewise.
23814 * gcov-iov.c: Moved to...
23815 * genversion.c: ...here.
23816 * lto-streamer.h (LTO_major_version): Define it with
23818 * version.c: Removed.
23819 * version.h: Removed.
23821 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
23823 * config/arc/arc.md (UNSPEC_ARC_DMPYWH): Define.
23824 * config/arc/simdext.md (VCT): Add predicates for iterator
23827 (voptab): Likewise.
23828 (vec_widen_<V_US>mult_hi_v4hi): Change pattern predicate.
23829 (<voptab>v2si3): New patterns.
23831 (reduc_plus_scal_v4hi): Likewise.
23832 (reduc_plus_scal_v2si): Likewise.
23833 (vec_duplicatev2si): Likewise.
23834 (vec_duplicatev4hi): Likewise.
23836 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
23838 * config/arc/simdext.md: Format and cleanup file.
23840 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
23842 * config/arc/simdext.md (movmisalignv2hi): Allow misaligned access
23843 only when munaligned-access option is on.
23844 (movmisalign<mode>): Likewise.
23846 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
23848 * common/config/arc/arc-common.c (arc_handle_option): Remove dot
23850 * config/arc/arc.c (arc_reorg): Remove underscore from string.
23852 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
23854 * config/arc/arc.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
23855 (CTZ_DEFINED_VALUE_AT_ZERO): Likewise.
23856 * config/arc/arc.md (clrsbsi2): Cleanup pattern.
23857 (norm_f): Likewise.
23860 (clzsi2): Use fls instruction when available.
23861 (arc_clzsi2): Likewise.
23863 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
23865 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add r26 and r27.
23867 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
23869 * doc/extend.texi (__builtin_arc_sr): Swap arguments.
23871 2021-05-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
23873 PR middle-end/100467
23874 * toplev.c (compile_file): Call insn_locations_init before
23875 targetm.asm_out.code_end.
23877 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
23880 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
23882 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
23884 2021-05-07 Jakub Jelinek <jakub@redhat.com>
23885 Andrew Stubbs <amd@codesourcery.com>
23888 * builtins.c (try_store_by_multiple_pieces): Use force_operand for
23889 emit_move_insn operands.
23891 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
23893 * cfgexpand.c (expand_gimple_basic_block): Do not inherit a current
23894 location for the outgoing edges of an empty block.
23895 * dwarf2out.c (add_subscript_info): Retrieve the bounds and index
23896 type by means of the get_array_descr_info langhook, if it is set and
23897 returns true. Remove obsolete code dealing with unnamed subtypes.
23899 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
23901 * gimple-range-cache.cc (ssa_block_ranges): Virtualize.
23902 (sbr_vector): Renamed from ssa_block_cache.
23903 (sbr_vector::sbr_vector): Allocate from obstack abd initialize.
23904 (ssa_block_ranges::~ssa_block_ranges): Remove.
23905 (sbr_vector::set_bb_range): Use varying and undefined cached values.
23906 (ssa_block_ranges::set_bb_varying): Remove.
23907 (sbr_vector::get_bb_range): Adjust assert.
23908 (sbr_vector::bb_range_p): Adjust assert.
23909 (~block_range_cache): No freeing loop required.
23910 (block_range_cache::get_block_ranges): Remove.
23911 (block_range_cache::set_bb_range): Inline get_block_ranges.
23912 (block_range_cache::set_bb_varying): Remove.
23913 * gimple-range-cache.h (set_bb_varying): Remove prototype.
23914 * value-range.h (irange_allocator::get_memory): New.
23916 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
23918 * gimple-range-cache.cc (non_null_ref::non_null_deref_p): Search
23919 dominator tree is available and requested.
23920 (ranger_cache::ssa_range_in_bb): Don't search dom tree here.
23921 (ranger_cache::fill_block_cache): Don't search dom tree here either.
23922 * gimple-range-cache.h (non_null_deref_p): Add dom_search param.
23924 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
23926 * gimple-range.cc (gimple_ranger::range_on_exit): Handle block with
23927 only PHI nodes better.
23929 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
23931 * gimple-range-edge.h (gimple_outgoing_range): Rename from
23933 (gcond_edge_range): Export prototype.
23934 * gimple-range-edge.cc (gcond_edge_range): New.
23935 (gimple_outgoing_range::edge_range_p): Use gcond_edge_range.
23936 * gimple-range-gori.h (gori_compute): Use gimple_outgoing_range.
23938 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
23940 * gimple-range-edge.cc (outgoing_range::calc_switch_ranges): Compute
23941 default range into a temp and allocate only what is needed.
23943 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
23945 * range-op.cc (operator_trunc_mod::wi_fold): x % 0 is UNDEFINED.
23947 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
23949 * gimple-range.h (gimple_range_global): Pick up parameter initial
23950 values, and use-before defined locals are UNDEFINED.
23952 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
23954 * doc/extend.texi (scalar_storage_order): Mention effect on pointer
23956 * tree.h (reverse_storage_order_for_component_p): Return false if
23957 the type is a pointer.
23959 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
23961 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
23963 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
23966 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
23967 Handle V8QI, V4HI and V2SI modes.
23968 * config/i386/mmx.md (mmx_pblendvb): New insn pattern.
23969 * config/i386/sse.md (unspec): Move UNSPEC_BLENDV ...
23970 * config/i386/i386.md (unspec): ... here.
23972 2021-05-07 Tobias Burnus <tobias@codesourcery.com>
23973 Tom de Vries <tdevries@suse.de>
23975 * omp-low.c (lower_rec_simd_input_clauses): Set max_vf = 1 if
23976 a truth_value_p reduction variable is nonintegral.
23978 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
23981 * config/i386/i386-expand.c (ix86_use_mask_cmp_p):
23982 Return false for mode sizes < 16.
23984 2021-05-07 Jakub Jelinek <jakub@redhat.com>
23987 * config/i386/mmx.md (*xop_pcmov_<mode>): New define_insn.
23989 2021-05-06 Martin Jambor <mjambor@suse.cz>
23991 * ipa-sra.c (ipa_sra_dump_all_summaries): Dump edge summaries even
23992 when there is no function summary.
23993 (ipa_sra_summarize_function): produce edge summaries even when
23996 2021-05-06 Tom Tromey <tom@tromey.com>
23998 * godump.c (string_hash_eq): Remove.
23999 (go_finish): Use htab_eq_string.
24001 2021-05-06 Tom Tromey <tom@tromey.com>
24003 * gengtype-state.c (read_state): Use htab_eq_string.
24004 (string_eq): Remove.
24006 2021-05-06 Tom Tromey <tom@tromey.com>
24008 * gensupport.c (htab_eq_string): Remove.
24010 2021-05-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
24013 * debug.h (gcc_debug_hooks): Add set_ignored_loc function pointer.
24014 * dwarf2out.h (dw_fde_node::ignored_debug): New data item.
24015 * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Add dummy
24016 set_ignored_loc callbacks.
24017 * debug.c (do_nothing_debug_hooks): Likewise.
24018 * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
24019 * dwarf2out.c (text_section_used, cold_text_section_used): Remove.
24020 (in_text_section_p, last_text_label, last_cold_label,
24021 switch_text_ranges, switch_cold_ranges): New data items.
24022 (dwarf2out_note_section_used): Remove.
24023 (dwarf2out_begin_prologue): Set fde->ignored_debug and
24025 (mark_ignored_debug_section): New helper function.
24026 (dwarf2out_end_epilogue, dwarf2out_switch_text_section): Call
24027 mark_ignored_debug_section.
24028 (dwarf2_debug_hooks): Use dwarf2out_set_ignored_loc.
24029 (dwarf2_lineno_debug_hooks): Use dummy for set_ignored_loc.
24030 (size_of_aranges): Adjust formula for multi-part text ranges size.
24031 (output_aranges): Output multi-part text ranges.
24032 (dwarf2out_set_ignored_loc): New callback function.
24033 (dwarf2out_finish): Output multi-part text ranges.
24034 (dwarf2out_c_finalize): Clear new data items.
24035 * final.c (final_start_function_1): Call set_ignored_loc callback.
24036 (final_scan_insn_1): Likewise.
24037 * ggc-page.c (gt_ggc_mx): New helper function.
24038 * stringpool.c (gt_pch_nx): Likewise.
24040 2021-05-06 Richard Biener <rguenther@suse.de>
24042 * timevar.def (TV_TREE_INSERT_PHI_NODES): Remove.
24043 (TV_TREE_SSA_REWRITE_BLOCKS): Likewise.
24044 (TV_TREE_INTO_SSA): New.
24045 * tree-into-ssa.c (insert_phi_nodes): Do not account separately.
24046 (rewrite_blocks): Likewise.
24047 (pass_data_build_ssa): Account to TV_TREE_INTO_SSA.
24049 2021-05-06 Jakub Jelinek <jakub@redhat.com>
24051 * tree-ssa-phiopt.c (value_replacement, minmax_replacement,
24052 abs_replacement, xor_replacement,
24053 cond_removal_in_popcount_clz_ctz_pattern,
24054 replace_phi_edge_with_variable): Change type of phi argument from
24055 gimple * to gphi *.
24057 2021-05-06 Richard Biener <rguenther@suse.de>
24059 * tree-ssa-loop-split.c (split_loop): Delay updating SSA form.
24060 Output an opt-info message.
24061 (do_split_loop_on_cond): Likewise.
24062 (tree_ssa_split_loops): Update SSA form here.
24064 2021-05-06 Richard Biener <rguenther@suse.de>
24066 * tree-inline.c (tree_function_versioning): Fix DECL_BY_REFERENCE
24067 return variable removal.
24069 2021-05-06 Marius Hillenbrand <mhillen@linux.ibm.com>
24071 * config/s390/s390-builtins.def (O_M5, O1_M5, ...): Remove unused macros.
24072 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
24073 (s390_vec_permi_dbl, s390_vpdi): Use the O3_U2 type for the immediate
24075 * config/s390/s390.c (s390_const_operand_ok): Remove unused
24078 2021-05-06 Jakub Jelinek <jakub@redhat.com>
24080 PR tree-optimization/94589
24081 * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Call
24082 spaceship_replacement.
24083 (cond_only_block_p, spaceship_replacement): New functions.
24085 2021-05-06 Richard Biener <rguenther@suse.de>
24088 * tree-emutls.c (gen_emutls_addr): Pass in whether we're
24089 dealing with a debug use and only query existing addresses
24091 (lower_emutls_1): Avoid splitting out addresses for debug
24092 stmts, reset the debug stmt when we fail to find existing
24094 (lower_emutls_phi_arg): Set wi.stmt.
24096 2021-05-06 Christoph Muellner <cmuellner@gcc.gnu.org>
24099 * config/riscv/riscv.c (riscv_block_move_loop): Use cbranch helper.
24100 * config/riscv/riscv.md (cbranch<mode>4): Generate helpers.
24101 (stack_protect_test): Use cbranch helper.
24103 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
24106 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
24107 always return the establisher frame for __builtin_frame_address (0).
24109 2021-05-05 Ivan Sorokin <vanyacpp@gmail.com>
24112 * config/i386/i386-builtins.c (ix86_cpu_model_type_node): New.
24113 (ix86_cpu_model_var): Likewise.
24114 (ix86_cpu_features2_type_node): Likewise.
24115 (ix86_cpu_features2_var): Likewise.
24116 (fold_builtin_cpu): Cache __cpu_model and __cpu_features2 with
24119 2021-05-05 Martin Sebor <msebor@redhat.com>
24121 * passes.def (pass_warn_printf): Run after SSA.
24123 2021-05-05 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
24125 * config/arm/neon.md (neon_vtst_combine<mode>): New pattern.
24126 * config/arm/predicates.md (minus_one_operand): New predicate.
24128 2021-05-05 Jeff Law <jlaw@tachyum.com>
24130 * config/avr/avr.md: Remove references to CC_STATUS_INIT.
24132 2021-05-05 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
24134 PR rtl-optimization/100263
24135 * postreload.c (move2add_valid_value_p): Ensure register can
24138 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
24140 PR rtl-optimization/100411
24141 * cfgcleanup.c (try_crossjump_to_edge): Also skip end of prologue
24142 and beginning of function markers.
24144 2021-05-05 Jeff Law <jlaw@tachyum.com>
24146 * config/cr16/cr16.h (NOTICE_UPDATE_CC): Remove.
24147 * config/cr16/cr16.c (notice_update_cc): Remove.
24148 * config/cr16/cr16-protos.h (notice_update_cc): Remove.
24150 2021-05-05 Uroš Bizjak <ubizjak@gmail.com>
24153 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
24154 Handle V8QI, V4HI and V2SI modes.
24155 * config/i386/i386.c (ix86_build_const_vector): Handle V2SImode.
24156 (ix86_build_signbit_mask): Ditto.
24157 * config/i386/mmx.md (MMXMODE14): New mode iterator.
24158 (<smaxmin:code><MMXMODE14:mode>3): New expander.
24159 (*mmx_<smaxmin:code><MMXMODE14:mode>3): New insn pattern.
24160 (<umaxmin:code><MMXMODE24:mode>3): New expander.
24161 (*mmx_<umaxmin:code><MMXMODE24:mode>3): New insn pattern.
24162 (vec_cmp<MMXMODEI:mode><MMXMODEI:mode>): New expander.
24163 (vec_cmpu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
24164 (vcond<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
24165 (vcondu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
24166 (vcond_mask_<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
24168 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
24170 * dwarf2out.c (loc_list_from_tree_1) <DECL>: During early DWARF, do
24171 not expand the VALUE_EXPR of variables put in the non-local frame.
24172 * gimplify.c (gimplify_type_sizes) <RECORD_TYPE>: If the type is not
24173 to be ignored for debug info, ensure its variable offsets are not.
24175 2021-05-05 Richard Biener <rguenther@suse.de>
24177 PR tree-optimization/79333
24178 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt):
24179 Fold stmt following SSA edges.
24181 2021-05-05 Richard Biener <rguenther@suse.de>
24183 PR middle-end/100394
24184 * calls.c (expand_call): Preserve possibly throwing calls.
24185 * cfgexpand.c (expand_call_stmt): When a call can throw signal
24186 RTL expansion there are side-effects.
24187 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Simplify,
24188 mark all possibly throwing stmts necessary unless we can elide
24190 * tree-ssa-dse.c (pass_dse::execute): Preserve exceptions unless
24191 -fdelete-dead-exceptions.
24192 * tree.h (DECL_PURE_P): Add note about exceptions.
24194 2021-05-05 Alexandre Oliva <oliva@adacore.com>
24196 * config/i386/vxworks.h (DBX_REGISTER_NUMBER): Make it
24199 2021-05-04 David Edelsohn <dje.gcc@gmail.com>
24201 * config/rs6000/rs6000-call.c (rs6000_output_mi_thunk): Use
24202 get_fnname_from_decl for name of thunk.
24203 * config/rs6000/rs6000.c (rs6000_declare_alias): Use assemble_name
24204 and ASM_OUTPUT_LABEL.
24205 (rs6000_xcoff_declare_function_name): Use assemble_name and
24207 (rs6000_xcoff_declare_object_name): Use ASM_OUTPUT_LABEL.
24208 (rs6000_xcoff_encode_section_info): Don't add mapping class
24209 for aliases. Always add [DS] mapping class to primary
24211 (rs6000_asm_weaken_decl): Don't explicitly add [DS].
24213 2021-05-04 Martin Sebor <msebor@redhat.com>
24215 PR middle-end/100307
24216 * builtins.c (compute_objsize_r): Clear base0 for pointers.
24218 2021-05-04 Jeff Law <jlaw@tachyum.com>
24220 * config/bfin/bfin.h (NOTICE_UPDATE_CC): Remove.
24222 2021-05-04 Segher Boessenkool <segher@kernel.crashing.org>
24224 * caller-save.c: Remove CC0.
24225 * cfgcleanup.c: Remove CC0.
24226 * cfgrtl.c: Remove CC0.
24227 * combine.c: Remove CC0.
24228 * compare-elim.c: Remove CC0.
24229 * conditions.h: Remove CC0.
24230 * config/h8300/h8300.h: Remove CC0.
24231 * config/h8300/h8300-protos.h: Remove CC0.
24232 * config/h8300/peepholes.md: Remove CC0.
24233 * config/i386/x86-tune-sched.c: Remove CC0.
24234 * config/m68k/m68k.c: Remove CC0.
24235 * config/rl78/rl78.c: Remove CC0.
24236 * config/sparc/sparc.c: Remove CC0.
24237 * config/xtensa/xtensa.c: Remove CC0.
24238 (gen_conditional_move): Use pc_rtx instead of cc0_rtx in a piece of
24239 RTL where that is used as a placeholder only.
24240 * cprop.c: Remove CC0.
24241 * cse.c: Remove CC0.
24242 * cselib.c: Remove CC0.
24243 * df-problems.c: Remove CC0.
24244 * df-scan.c: Remove CC0.
24245 * doc/md.texi: Remove CC0. Adjust an example.
24246 * doc/rtl.texi: Remove CC0. Adjust an example.
24247 * doc/tm.texi: Regenerate.
24248 * doc/tm.texi.in: Remove CC0.
24249 * emit-rtl.c: Remove CC0.
24250 * final.c: Remove CC0.
24251 * fwprop.c: Remove CC0.
24252 * gcse-common.c: Remove CC0.
24253 * gcse.c: Remove CC0.
24254 * genattrtab.c: Remove CC0.
24255 * genconfig.c: Remove CC0.
24256 * genemit.c: Remove CC0.
24257 * genextract.c: Remove CC0.
24258 * gengenrtl.c: Remove CC0.
24259 * genrecog.c: Remove CC0.
24260 * haifa-sched.c: Remove CC0.
24261 * ifcvt.c: Remove CC0.
24262 * ira-costs.c: Remove CC0.
24263 * ira.c: Remove CC0.
24264 * jump.c: Remove CC0.
24265 * loop-invariant.c: Remove CC0.
24266 * lra-constraints.c: Remove CC0.
24267 * lra-eliminations.c: Remove CC0.
24268 * optabs.c: Remove CC0.
24269 * postreload-gcse.c: Remove CC0.
24270 * postreload.c: Remove CC0.
24271 * print-rtl.c: Remove CC0.
24272 * read-rtl-function.c: Remove CC0.
24273 * reg-notes.def: Remove CC0.
24274 * reg-stack.c: Remove CC0.
24275 * reginfo.c: Remove CC0.
24276 * regrename.c: Remove CC0.
24277 * reload.c: Remove CC0.
24278 * reload1.c: Remove CC0.
24279 * reorg.c: Remove CC0.
24280 * resource.c: Remove CC0.
24281 * rtl.c: Remove CC0.
24282 * rtl.def: Remove CC0.
24283 * rtl.h: Remove CC0.
24284 * rtlanal.c: Remove CC0.
24285 * sched-deps.c: Remove CC0.
24286 * sched-rgn.c: Remove CC0.
24287 * shrink-wrap.c: Remove CC0.
24288 * simplify-rtx.c: Remove CC0.
24289 * system.h: Remove CC0. Poison NOTICE_UPDATE_CC, CC_STATUS_MDEP_INIT,
24290 CC_STATUS_MDEP, and CC_STATUS.
24291 * target.def: Remove CC0.
24292 * valtrack.c: Remove CC0.
24293 * var-tracking.c: Remove CC0.
24295 2021-05-04 Richard Biener <rguenther@suse.de>
24297 PR tree-optimization/100414
24298 * tree-ssa-phiopt.c (get_non_trapping): Do not compute dominance
24300 (tree_ssa_phiopt_worker): But unconditionally here.
24302 2021-05-04 Tobias Burnus <tobias@codesourcery.com>
24304 * omp-low.c (lower_rec_input_clauses, lower_reduction_clauses): Handle
24305 && and || with floating-point and complex arguments.
24307 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
24309 * tree-inline.c (insert_debug_decl_map): Delete.
24310 (copy_debug_stmt): Minor tweak.
24311 (setup_one_parameter): Do not use a variable if the value is either
24312 a read-only DECL or a non-addressable local variable in the caller.
24313 In this case, insert the debug-only variable in the map manually.
24314 (expand_call_inline): Do not generate a CLOBBER for these values.
24315 * tree-inline.h (debug_map): Minor tweak.
24317 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
24319 * builtins.c (builtin_with_linkage_p): Return true for stp[n]cpy.
24320 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Tidy up.
24322 2021-05-04 Richard Biener <rguenther@suse.de>
24324 PR tree-optimization/100329
24325 * tree-ssa-reassoc.c (can_reassociate_p): Do not reassociate
24327 (insert_stmt_after): Assert we're not running into asm goto.
24329 2021-05-04 Richard Biener <rguenther@suse.de>
24331 PR tree-optimization/100398
24332 * tree-ssa-dse.c (pass_dse::execute): Preserve control
24335 2021-05-04 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
24337 * builtins.c (try_store_by_multiple_pieces): Fix constfun's prototype.
24339 2021-05-04 Alexandre Oliva <oliva@adacore.com>
24341 * builtins.c (try_store_by_multiple_pieces): New.
24342 (expand_builtin_memset_args): Use it. If target_char_cast
24343 fails, proceed as for non-constant val. Pass len's ctz to...
24344 * expr.c (clear_storage_hints): ... this. Try store by
24345 multiple pieces after setmem.
24346 (clear_storage): Adjust.
24347 * expr.h (clear_storage_hints): Likewise.
24348 (try_store_by_multiple_pieces): Declare.
24349 * passes.def: Replace the last copy_prop with ccp.
24351 2021-05-03 Tom de Vries <tdevries@suse.de>
24354 * omp-low.c (lower_rec_input_clauses): Disable SIMT for user-defined
24357 2021-05-03 Richard Biener <rguenther@suse.de>
24359 * tree-ssa-dse.c (dse_classify_store): Track two PHI defs.
24361 2021-05-03 Richard Biener <rguenther@suse.de>
24363 * tree-ssa-dse.c: Do not include domwalk.h but cfganal.h.
24364 (dse_dom_walker): Remove.
24365 (dse_dom_walker::dse_optimize_stmt): Rename...
24366 (dse_optimize_stmt): ... to this, pass in live_bytes sbitmap.
24367 (dse_dom_walker::before_dom_children): Inline ...
24368 (pass_dse::execute): ... here. Perform a reverse program
24371 2021-05-03 H.J. Lu <hjl.tools@gmail.com>
24374 * configure: Regenerated.
24376 2021-05-03 Ilya Leoshkevich <iii@linux.ibm.com>
24379 * config/s390/s390.c (s390_hard_fp_reg_p): New function.
24380 (s390_md_asm_adjust): Handle hard registers.
24382 2021-05-03 Jakub Jelinek <jakub@redhat.com>
24384 PR tree-optimization/100382
24385 * tree-ssa-dse.c: Include tree-eh.h.
24386 (dse_dom_walker::before_dom_children): Don't remove stmts if
24387 stmt_unremovable_because_of_non_call_eh_p is true.
24389 2021-05-02 David Edelsohn <dje.gcc@gmail.com>
24391 * varasm.c (compute_reloc_for_var): Split out from...
24392 (get_variable_section): Use it.
24393 * output.h (compute_reloc_for_var): Declare.
24394 * config/rs6000/rs6000-protos.h
24395 (rs6000_xcoff_asm_output_aligned_decl_common): Change alignment to
24397 * config/rs6000/rs6000.c (rs6000_legitimize_tls_address_aix):
24398 Don't append storage mapping class to symbol.
24399 (rs6000_xcoff_asm_named_section): Add BS and UL mapping classes.
24400 Don't convert TLS BSS to common.
24401 (rs6000_xcoff_unique_section): Don't fall back to select_secton.
24402 (rs6000_xcoff_section_type_flags): Add SECTION_BSS if DECL is
24404 (rs6000_xcoff_asm_globalize_decl_name): Don't strip storage
24406 (rs6000_xcoff_asm_output_aligned_decl_common): Align is unsigned int.
24407 If align is 0 from TLS class, use the same rules as varasm.c
24408 If not common, switch to BSS section manually.
24409 If common, emit appropriate comm or lcomm directive.
24410 (rs6000_xcoff_encode_section_info): Add logic to append all
24411 storage mapping classes.
24412 (rs6000_asm_weaken_decl): Adjust for qualname symbols.
24413 * config/rs6000/xcoff.h (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
24414 rs6000_xcoff_asm_output_aligned_decl_common.
24415 (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
24416 rs6000_xcoff_asm_output_aligned_decl_common.
24417 (ASM_OUTPUT_TLS_COMMON): Use
24418 rs6000_xcoff_asm_output_aligned_decl_common.
24420 2021-05-02 Jakub Jelinek <jakub@redhat.com>
24423 * config/nvptx/nvptx.c (nvptx_sese_pseudo): Use nullptr instead of 0
24424 as first argument of pseudo_node_t constructors.
24426 2021-05-02 Jakub Jelinek <jakub@redhat.com>
24429 * config/i386/t-i386 (TM_H): Add $(srcdir)/config/i386/i386-isa.def.
24431 2021-05-01 Aldy Hernandez <aldyh@redhat.com>
24433 * value-range.cc (DEFINE_INT_RANGE_GC_STUBS): Remove.
24434 (gt_pch_nx (int_range<1> *&)): New.
24435 (gt_ggc_mx (int_range<1> *&)): New.
24436 * value-range.h (class irange): Add GTY support for
24439 2021-05-01 Geng Qi <gengqi@linux.alibaba.com>
24441 * doc/options.texi (Negative): Change either or to both and.
24443 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
24445 * config/aarch64/aarch64-simd-builtins.def: Add
24446 float_ml[as][q]_laneq builtin generator macros.
24447 * config/aarch64/aarch64-simd.md (mul_laneq<mode>3): Define.
24448 (aarch64_float_mla_laneq<mode>): Define.
24449 (aarch64_float_mls_laneq<mode>): Define.
24450 * config/aarch64/arm_neon.h (vmla_laneq_f32): Use RTL builtin
24451 instead of GCC vector extensions.
24452 (vmlaq_laneq_f32): Likewise.
24453 (vmls_laneq_f32): Likewise.
24454 (vmlsq_laneq_f32): Likewise.
24456 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
24458 * config/aarch64/aarch64-simd-builtins.def: Add
24459 float_ml[as]_lane builtin generator macros.
24460 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt<mode>):
24462 (mul_lane<mode>3): This, and re-order arguments.
24463 (aarch64_float_mla_lane<mode>): Define.
24464 (aarch64_float_mls_lane<mode>): Define.
24465 * config/aarch64/arm_neon.h (vmla_lane_f32): Use RTL builtin
24466 instead of GCC vector extensions.
24467 (vmlaq_lane_f32): Likewise.
24468 (vmls_lane_f32): Likewise.
24469 (vmlsq_lane_f32): Likewise.
24471 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
24473 * config/aarch64/aarch64-simd-builtins.def: Add float_ml[as]
24474 builtin generator macros.
24475 * config/aarch64/aarch64-simd.md (aarch64_float_mla<mode>):
24477 (aarch64_float_mls<mode>): Define.
24478 * config/aarch64/arm_neon.h (vmla_f32): Use RTL builtin
24479 instead of relying on GCC vector extensions.
24480 (vmla_f64): Likewise.
24481 (vmlaq_f32): Likewise.
24482 (vmlaq_f64): Likewise.
24483 (vmls_f32): Likewise.
24484 (vmls_f64): Likewise.
24485 (vmlsq_f32): Likewise.
24486 (vmlsq_f64): Likewise.
24487 * config/aarch64/iterators.md: Define VDQF_DF mode iterator.
24489 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
24491 * config/aarch64/aarch64-simd-builtins.def: Add
24492 float_ml[as]_n_builtin generator macros.
24493 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt_from_dup<mode>):
24495 (mul_n<mode>3): This, and re-order arguments.
24496 (aarch64_float_mla_n<mode>): Define.
24497 (aarch64_float_mls_n<mode>): Define.
24498 * config/aarch64/arm_neon.h (vmla_n_f32): Use RTL builtin
24499 instead of inline asm.
24500 (vmlaq_n_f32): Likewise.
24501 (vmls_n_f32): Likewise.
24502 (vmlsq_n_f32): Likewise.
24504 2021-04-30 Jonathan Wright <joanthan.wright@arm.com>
24506 * config/aarch64/aarch64-simd-builtins.def: Add pmull[2]
24507 builtin generator macros.
24508 * config/aarch64/aarch64-simd.md (aarch64_pmullv8qi): Define.
24509 (aarch64_pmull_hiv16qi_insn): Define.
24510 (aarch64_pmull_hiv16qi): Define.
24511 * config/aarch64/arm_neon.h (vmull_high_p8): Use RTL builtin
24512 instead of inline asm.
24513 (vmull_p8): Likewise.
24515 2021-04-30 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
24517 * config/avr/avr.md: Adjust peepholes to match and
24518 generate parallels with clobber of REG_CC.
24519 (mov<mode>_insn): Rename to mov<mode>_insn_split.
24520 (*mov<mode>_insn): Rename to mov<mode>_insn.
24522 2021-04-30 David Edelsohn <dje.gcc@gmail.com>
24524 * varasm.c (use_blocks_for_decl_p): Don't use section anchors
24525 for VAR_DECLs if -fdata-sections enabled.
24527 2021-04-30 Michael Meissner <meissner@linux.ibm.com>
24529 PR bootstrap/100327
24530 * config/rs6000/rs6000.c
24531 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Define.
24532 (rs6000_libgcc_floating_mode_supported_p): New target hook.
24534 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
24536 * tree-ssa-threadbackward.c (class thread_jumps): Split out code
24538 (class back_threader_registry): ...to here...
24539 (class back_threader_profitability): ...and here...
24540 (thread_jumps::thread_through_all_blocks): Remove argument.
24541 (back_threader_registry::back_threader_registry): New.
24542 (back_threader_registry::~back_threader_registry): New.
24543 (back_threader_registry::thread_through_all_blocks): New.
24544 (thread_jumps::profitable_jump_thread_path): Move from here...
24545 (back_threader_profitability::profitable_path_p): ...to here.
24546 (thread_jumps::find_taken_edge): New.
24547 (thread_jumps::convert_and_register_current_path): Move...
24548 (back_threader_registry::register_path): ...to here.
24549 (thread_jumps::register_jump_thread_path_if_profitable): Move...
24550 (thread_jumps::maybe_register_path): ...to here.
24551 (thread_jumps::handle_phi): Call find_taken_edge and
24552 maybe_register_path.
24553 (thread_jumps::handle_assignment): Same.
24554 (thread_jumps::fsm_find_control_statement_thread_paths): Remove
24555 tree argument to handle_phi and handle_assignment.
24556 (thread_jumps::find_jump_threads_backwards): Set m_name. Remove
24557 set of m_speed_p and m_max_threaded_paths.
24558 (pass_thread_jumps::execute): Remove second argument from
24559 find_jump_threads_backwards.
24560 (pass_early_thread_jumps::execute): Same.
24562 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
24564 * tree-ssa-dom.c (class dom_jump_threader_simplifier): New.
24565 (class dom_opt_dom_walker): Initialize some class variables.
24566 (pass_dominator::execute): Pass evrp_range_analyzer and
24567 dom_jump_threader_simplifier to dom_opt_dom_walker.
24568 Adjust for some functions moving into classes.
24569 (simplify_stmt_for_jump_threading): Adjust and move to...
24570 (jump_threader_simplifier::simplify): ...here.
24571 (dom_opt_dom_walker::before_dom_children): Adjust for
24572 m_evrp_range_analyzer.
24573 (dom_opt_dom_walker::after_dom_children): Remove x_vr_values hack.
24574 (test_for_singularity): Place in dom_opt_dom_walker class.
24575 (dom_opt_dom_walker::optimize_stmt): The argument
24576 evrp_range_analyzer is now a class field.
24577 * tree-ssa-threadbackward.c (class thread_jumps): Add m_registry.
24578 (thread_jumps::thread_through_all_blocks): New.
24579 (thread_jumps::convert_and_register_current_path): Use m_registry.
24580 (pass_thread_jumps::execute): Adjust for thread_through_all_blocks
24581 being in the threader class.
24582 (pass_early_thread_jumps::execute): Same.
24583 * tree-ssa-threadedge.c (threadedge_initialize_values): Move...
24584 (jump_threader::jump_threader): ...here.
24585 (threadedge_finalize_values): Move...
24586 (jump_threader::~jump_threader): ...here.
24587 (jump_threader::remove_jump_threads_including): New.
24588 (jump_threader::thread_through_all_blocks): New.
24589 (record_temporary_equivalences_from_phis): Move...
24590 (jump_threader::record_temporary_equivalences_from_phis): ...here.
24591 (record_temporary_equivalences_from_stmts_at_dest): Move...
24592 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
24594 (simplify_control_stmt_condition_1): Move to jump_threader class.
24595 (simplify_control_stmt_condition): Move...
24596 (jump_threader::simplify_control_stmt_condition): ...here.
24597 (thread_around_empty_blocks): Move...
24598 (jump_threader::thread_around_empty_blocks): ...here.
24599 (thread_through_normal_block): Move...
24600 (jump_threader::thread_through_normal_block): ...here.
24601 (thread_across_edge): Move...
24602 (jump_threader::thread_across_edge): ...here.
24603 (thread_outgoing_edges): Move...
24604 (jump_threader::thread_outgoing_edges): ...here.
24605 * tree-ssa-threadedge.h: Move externally facing functings...
24606 (class jump_threader): ...here...
24607 (class jump_threader_simplifier): ...and here.
24608 * tree-ssa-threadupdate.c (struct redirection_data): Remove comment.
24609 (jump_thread_path_allocator::jump_thread_path_allocator): New.
24610 (jump_thread_path_allocator::~jump_thread_path_allocator): New.
24611 (jump_thread_path_allocator::allocate_thread_edge): New.
24612 (jump_thread_path_allocator::allocate_thread_path): New.
24613 (jump_thread_path_registry::jump_thread_path_registry): New.
24614 (jump_thread_path_registry::~jump_thread_path_registry): New.
24615 (jump_thread_path_registry::allocate_thread_edge): New.
24616 (jump_thread_path_registry::allocate_thread_path): New.
24617 (dump_jump_thread_path): Make extern.
24618 (debug (const vec<jump_thread_edge *> &path)): New.
24619 (struct removed_edges): Move to tree-ssa-threadupdate.h.
24620 (struct thread_stats_d): Remove.
24621 (remove_ctrl_stmt_and_useless_edges): Make static.
24622 (lookup_redirection_data): Move...
24623 (jump_thread_path_registry::lookup_redirection_data): ...here.
24624 (ssa_redirect_edges): Make static.
24625 (thread_block_1): Move...
24626 (jump_thread_path_registry::thread_block_1): ...here.
24627 (thread_block): Move...
24628 (jump_thread_path_registry::thread_block): ...here.
24629 (thread_through_loop_header): Move...
24630 (jump_thread_path_registry::thread_through_loop_header): ...here.
24631 (mark_threaded_blocks): Move...
24632 (jump_thread_path_registry::mark_threaded_blocks): ...here.
24633 (debug_path): Move...
24634 (jump_thread_path_registry::debug_path): ...here.
24635 (debug_all_paths): Move...
24636 (jump_thread_path_registry::dump): ..here.
24637 (rewire_first_differing_edge): Move...
24638 (jump_thread_path_registry::rewire_first_differing_edge): ...here.
24639 (adjust_paths_after_duplication): Move...
24640 (jump_thread_path_registry::adjust_paths_after_duplication): ...here.
24641 (duplicate_thread_path): Move...
24642 (jump_thread_path_registry::duplicate_thread_path): ..here.
24643 (remove_jump_threads_including): Move...
24644 (jump_thread_path_registry::remove_jump_threads_including): ...here.
24645 (thread_through_all_blocks): Move to...
24646 (jump_thread_path_registry::thread_through_all_blocks): ...here.
24647 (delete_jump_thread_path): Remove.
24648 (register_jump_thread): Move...
24649 (jump_thread_path_registry::register_jump_thread): ...here.
24650 * tree-ssa-threadupdate.h: Move externally facing functions...
24651 (class jump_thread_path_allocator): ...here...
24652 (class jump_thread_path_registry): ...and here.
24653 (thread_through_all_blocks): Remove.
24654 (struct removed_edges): New.
24655 (register_jump_thread): Remove.
24656 (remove_jump_threads_including): Remove.
24657 (delete_jump_thread_path): Remove.
24658 (remove_ctrl_stmt_and_useless_edges): Remove.
24659 (free_dom_edge_info): New prototype.
24660 * tree-vrp.c: Remove x_vr_values hack.
24661 (class vrp_jump_threader_simplifier): New.
24662 (vrp_jump_threader_simplifier::simplify): New.
24663 (vrp_jump_threader::vrp_jump_threader): Adjust method signature.
24664 Remove m_dummy_cond.
24665 Instantiate m_simplifier and m_threader.
24666 (vrp_jump_threader::thread_through_all_blocks): New.
24667 (vrp_jump_threader::simplify_stmt): Remove.
24668 (vrp_jump_threader::after_dom_children): Do not set m_dummy_cond.
24669 Remove x_vr_values hack.
24670 (execute_vrp): Adjust for thread_through_all_blocks being in a
24673 2021-04-30 Christophe Lyon <christophe.lyon@linaro.org>
24675 * genflags.c (gen_insn): Print failed expansion string.
24677 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
24679 * expr.c (alignment_for_piecewise_move): Call mode_for_size
24680 without limit to MAX_FIXED_MODE_SIZE.
24682 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
24684 PR middle-end/90773
24685 * builtins.c (builtin_memset_gen_str): Don't use return from
24686 simplify_gen_subreg.
24688 2021-04-30 Uroš Bizjak <ubizjak@gmail.com>
24691 * config/i386/i386.md (*add<mode>3_carry_0r): New insn pattern.
24692 (*addsi3_carry_zext_0r): Ditto.
24693 (*sub<mode>3_carry_0): Ditto.
24694 (*subsi3_carry_zext_0r): Ditto.
24695 * config/i386/predicates.md (ix86_carry_flag_unset_operator):
24697 * config/i386/i386.c (ix86_rtx_costs) <case PLUS, case MINUS>:
24698 Also consider ix86_carry_flag_unset_operator to calculate
24699 the cost of adc/sbb insn.
24701 2021-04-30 Roman Zhuykov <zhroma@ispras.ru>
24703 PR rtl-optimization/100225
24704 PR rtl-optimization/84878
24705 * modulo-sched.c (sms_schedule): Use note_stores to skip loops
24706 where we have an instruction which touches (writes) any hard
24707 register from df->regular_block_artificial_uses set.
24708 Allow not-single-set instruction only right before basic block
24711 2021-04-30 Geng Qi <gengqi@linux.alibaba.com>
24713 * config/riscv/riscv.opt (march=,mabi=): Negative itself.
24715 2021-04-30 LevyHsu <admin@levyhsu.com>
24717 * config/riscv/riscv.c (riscv_min_arithmetic_precision): New.
24718 * config/riscv/riscv.h (TARGET_MIN_ARITHMETIC_PRECISION): New.
24719 * config/riscv/riscv.md (addv<mode>4, uaddv<mode>4): New.
24720 (subv<mode>4, usubv<mode>4, mulv<mode>4, umulv<mode>4): New.
24722 2021-04-29 Alexandre Oliva <oliva@adacore.com>
24724 * config.gcc: Merged x86 and x86_64 cpu_type-setting cases.
24726 2021-04-29 Alexandre Oliva <oliva@adacore.com>
24728 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Rename to...
24729 (ASM_OUTPUT_MAX_SKIP_ALIGN): ... this. Enclose in do/while(0).
24730 * config/i386/i386.c: Adjust.
24731 * config/i386/i386.md: Adjust.
24732 * config/i386/darwin.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Drop.
24733 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
24734 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
24735 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
24736 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
24737 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
24738 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
24739 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
24740 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
24741 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
24742 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
24744 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
24746 * config/i386/i386-expand.c (ix86_expand_int_compare):
24747 Swap operands of GTU and LEU comparison to emit carry flag comparison.
24748 * config/i386/i386.md (*add<mode>3_carry_0): Change insn
24749 predicate to allow more combine opportunities with memory operands.
24750 (*sub<mode>3_carry_0): Ditto.
24752 2021-04-29 Richard Sandiford <richard.sandiford@arm.com>
24754 PR rtl-optimization/100303
24755 * rtl-ssa/accesses.cc (function_info::make_use_available): Take a
24756 boolean that indicates whether the use will only be used in
24757 debug instructions. Treat it in the same way that existing
24758 cross-EBB debug references would be handled if so.
24759 (function_info::make_uses_available): Likewise.
24760 * rtl-ssa/functions.h (function_info::make_uses_available): Update
24761 prototype accordingly.
24762 (function_info::make_uses_available): Likewise.
24763 * fwprop.c (try_fwprop_subst): Update call accordingly.
24765 2021-04-29 Jeff Law <jlaw@tachyum.com>
24767 * config/nios2/nios2-protos.h (nios2_fpu_insn_enabled): Move outside
24770 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
24771 Richard Biener <rguenther@suse.de>
24774 * config/i386/i386-builtin.def (IX86_BUILTIN_MASKLOADPD)
24775 (IX86_BUILTIN_MASKLOADPS, IX86_BUILTIN_MASKLOADPD256)
24776 (IX86_BUILTIN_MASKLOADPS256, IX86_BUILTIN_MASKLOADD)
24777 (IX86_BUILTIN_MASKLOADQ, IX86_BUILTIN_MASKLOADD256)
24778 (IX86_BUILTIN_MASKLOADQ256): Move from SPECIAL_ARGS
24779 to PURE_ARGS category.
24780 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
24781 Handle PURE_ARGS category.
24782 * config/i386/i386-expand.c (ix86_expand_builtin): Ditto.
24784 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
24786 * configure.ac: Check for the presence of sys/locking.h header and
24787 for whether _LK_LOCK is supported by _locking.
24788 * configure: Regenerate.
24789 * config.in: Likewise.
24790 * gcov-io.h: Define GCOV_LOCKED_WITH_LOCKING if HOST_HAS_LK_LOCK.
24791 * gcov-io.c (gcov_open): Add support for GCOV_LOCKED_WITH_LOCKING.
24792 * system.h: Include <sys/locking.h> if HAVE_SYS_LOCKING_H.
24794 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
24796 * config/i386/predicates.md (fcmov_comparison_operator):
24797 Do not check for trivial FP comparison operator.
24798 <case GEU, case LTU>: Allow CCGZmode.
24799 <case GTU, case LEU>: Do not allow CCCmode.
24800 (ix86_comparison_operator) <case GTU, case LEU>: Allow only CCmode.
24801 (ix86_carry_flag_operator): Match only LTU and UNLT code.
24802 Do not check for trivial FP comparison operator. Allow CCGZmode.
24804 2021-04-29 Tom de Vries <tdevries@suse.de>
24806 * omp-expand.c (expand_omp_simd): Add step_orig, and replace uses of
24807 fd->loop.step by either step or orig_step.
24809 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
24811 * config/sparc/sparc.c (gen_load_pcrel_sym): Delete.
24812 (load_got_register): Do the PIC dance here.
24813 (sparc_legitimize_tls_address): Simplify.
24814 (sparc_emit_probe_stack_range): Likewise.
24815 (sparc32_initialize_trampoline): Likewise.
24816 (sparc64_initialize_trampoline): Likewise.
24817 * config/sparc/sparc.md (load_pcrel_sym<P:mode>): Add @ marker.
24818 (probe_stack_range<P:mode>): Likewise.
24819 (flush<P:mode>): Likewise.
24820 (tgd_hi22<P:mode>): Likewise.
24821 (tgd_lo10<P:mode>): Likewise.
24822 (tgd_add<P:mode>): Likewise.
24823 (tgd_call<P:mode>): Likewise.
24824 (tldm_hi22<P:mode>): Likewise.
24825 (tldm_lo10<P:mode>): Likewise.
24826 (tldm_add<P:mode>): Likewise.
24827 (tldm_call<P:mode>): Likewise.
24828 (tldo_hix22<P:mode>): Likewise.
24829 (tldo_lox10<P:mode>): Likewise.
24830 (tldo_add<P:mode>): Likewise.
24831 (tie_hi22<P:mode>): Likewise.
24832 (tie_lo10<P:mode>): Likewise.
24833 (tie_add<P:mode>): Likewise.
24834 (tle_hix22<P:mode>): Likewise.
24835 (tle_lox10<P:mode>): Likewise.
24836 (stack_protect_setsi): Rename to...
24837 (stack_protect_set32): ...this.
24838 (stack_protect_setdi): Rename to...
24839 (stack_protect_set64): ...this.
24840 (stack_protect_set): Adjust calls to above.
24841 (stack_protect_testsi): Rename to...
24842 (stack_protect_test32): ...this.
24843 (stack_protect_testdi): Rename to...
24844 (stack_protect_test64): ...this.
24845 (stack_protect_test): Adjust calls to above.
24847 2021-04-29 H.J. Lu <hjl.tools@gmail.com>
24849 PR middle-end/90773
24850 * builtins.c (builtin_memcpy_read_str): Add a dummy argument.
24851 (builtin_strncpy_read_str): Likewise.
24852 (builtin_memset_read_str): Add an argument for the previous RTL
24853 information and generate the new RTL from the previous RTL info.
24854 (builtin_memset_gen_str): Likewise.
24855 * builtins.h (builtin_strncpy_read_str): Update the prototype.
24856 (builtin_memset_read_str): Likewise.
24857 * expr.c (by_pieces_ninsns): If targetm.overlap_op_by_pieces_p()
24858 returns true, round up size and alignment to the widest integer
24859 mode for maximum size.
24860 (pieces_addr::adjust): Add a pointer to by_pieces_prev argument
24861 and pass it to m_constfn.
24862 (op_by_pieces_d): Add m_push and m_overlap_op_by_pieces.
24863 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
24864 initialize m_push. Initialize m_overlap_op_by_pieces with
24865 targetm.overlap_op_by_pieces_p ().
24866 (op_by_pieces_d::run): Pass the previous RTL information to
24867 pieces_addr::adjust and generate overlapping operations if
24868 m_overlap_op_by_pieces is true.
24870 (move_by_pieces_d::move_by_pieces_d): Updated for op_by_pieces_d
24872 (store_by_pieces_d::store_by_pieces_d): Updated for op_by_pieces_d
24874 (can_store_by_pieces): Use by_pieces_constfn on constfun.
24875 (store_by_pieces): Use by_pieces_constfn on constfun. Updated
24876 for op_by_pieces_d change.
24877 (clear_by_pieces_1): Add a dummy argument.
24878 (clear_by_pieces): Updated for op_by_pieces_d change.
24879 (compare_by_pieces_d::compare_by_pieces_d): Likewise.
24880 (string_cst_read_str): Add a dummy argument.
24881 * expr.h (by_pieces_constfn): Add a dummy argument.
24882 (by_pieces_prev): New.
24883 * target.def (overlap_op_by_pieces_p): New target hook.
24884 * config/i386/i386.c (TARGET_OVERLAP_OP_BY_PIECES_P): New.
24885 * doc/tm.texi.in: Add TARGET_OVERLAP_OP_BY_PIECES_P.
24886 * doc/tm.texi: Regenerated.
24888 2021-04-29 Richard Biener <rguenther@suse.de>
24890 PR tree-optimization/100253
24891 * tree-vect-stmts.c (vectorizable_load): Do not assume
24892 element alignment when DR_MISALIGNMENT is -1.
24893 (vectorizable_store): Likewise.
24895 2021-04-29 Jakub Jelinek <jakub@redhat.com>
24898 * config/aarch64/aarch64.c (aarch64_add_offset_1_temporaries): Use
24899 absu_hwi instead of abs_hwi.
24901 2021-04-29 Richard Biener <rguenther@suse.de>
24903 PR middle-end/38474
24904 * tree-ssa-structalias.c (add_graph_edge): Avoid direct
24905 forwarding when indirect forwarding through ESCAPED
24908 2021-04-29 Tom de Vries <tdevries@suse.de>
24911 * internal-fn.c (expand_GOMP_SIMT_ENTER_ALLOC)
24912 (expand_GOMP_SIMT_LAST_LANE, expand_GOMP_SIMT_ORDERED_PRED)
24913 (expand_GOMP_SIMT_VOTE_ANY, expand_GOMP_SIMT_XCHG_BFLY)
24914 (expand_GOMP_SIMT_XCHG_IDX): Ensure target is assigned to.
24916 2021-04-29 Richard Biener <rguenther@suse.de>
24918 PR tree-optimization/99912
24919 * tree-ssa-dse.c (dse_dom_walker::m_need_cfg_cleanup): New.
24920 (dse_dom_walker::todo): Likewise.
24921 (dse_dom_walker::dse_optimize_stmt): Move VDEF check to the
24923 (dse_dom_walker::before_dom_children): Remove trivially
24924 dead SSA defs and schedule CFG cleanup if we removed all
24926 (pass_dse::execute): Get TODO as computed by the DOM walker
24927 and return it. Wipe dominator info earlier.
24929 2021-04-29 Richard Biener <rguenther@suse.de>
24932 * ipa-prop.c (ipcp_modif_dom_walker::before_dom_children):
24933 Track blocks to cleanup EH in new m_need_eh_cleanup.
24934 (ipcp_modif_dom_walker::cleanup_eh): New.
24935 (ipcp_transform_function): Release dominator info before
24938 2021-04-29 Martin Sebor <msebor@redhat.com>
24940 PR middle-end/100250
24941 * attribs.c (attr_access::array_as_string): Avoid dereferencing
24942 a pointer when it's null.
24944 2021-04-29 Martin Sebor <msebor@redhat.com>
24946 * Makefile.in (OBJS): Add ipa-free-lang-data.o.
24947 * ipa-free-lang-data.cc: New file.
24948 * tree.c: Move pass free_lang_data to file above.
24949 (build_array_type_1): Declare extern.
24950 * tree.h (build_array_type_1): Declare.
24952 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
24954 * config/aarch64/aarch64-simd-builtins.def: Modify comment to
24955 make consistent with updated RTL pattern.
24956 * config/aarch64/aarch64-simd.md (aarch64_<sur>qmovn<mode>):
24957 Implement using ss_truncate and us_truncate rather than
24959 * config/aarch64/iterators.md: Remove redundant unspecs and
24960 iterator: UNSPEC_[SU]QXTN and SUQMOVN respectively.
24962 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
24964 * config/aarch64/arm_acle.h (__attribute__): Make intrinsic
24965 attributes consistent with those defined in arm_neon.h.
24967 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
24969 * config/aarch64/arm_fp16.h (__attribute__): Make intrinsic
24970 attributes consistent with those defined in arm_neon.h.
24972 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
24974 * config/aarch64/aarch64-simd-builtins.def: Add
24975 float_trunc_rodd builtin generator macros.
24976 * config/aarch64/aarch64-simd.md (aarch64_float_trunc_rodd_df):
24978 (aarch64_float_trunc_rodd_lo_v2sf): Define.
24979 (aarch64_float_trunc_rodd_hi_v4sf_le): Define.
24980 (aarch64_float_trunc_rodd_hi_v4sf_be): Define.
24981 (aarch64_float_trunc_rodd_hi_v4sf): Define.
24982 * config/aarch64/arm_neon.h (vcvtx_f32_f64): Use RTL builtin
24983 instead of inline asm.
24984 (vcvtx_high_f32_f64): Likewise.
24985 (vcvtxd_f32_f64): Likewise.
24986 * config/aarch64/iterators.md: Add FCVTXN unspec.
24988 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
24990 * config/aarch64/aarch64-simd-builtins.def: Add tbx1 builtin
24992 * config/aarch64/aarch64-simd.md (aarch64_tbx1<mode>):
24994 * config/aarch64/arm_neon.h (vqtbx1_s8): USE RTL builtin
24995 instead of inline asm.
24996 (vqtbx1_u8): Likewise.
24997 (vqtbx1_p8): Likewise.
24998 (vqtbx1q_s8): Likewise.
24999 (vqtbx1q_u8): Likewise.
25000 (vqtbx1q_p8): Likewise.
25001 (vtbx2_s8): Likewise.
25002 (vtbx2_u8): Likewise.
25003 (vtbx2_p8): Likewise.
25005 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
25007 * config/aarch64/aarch64-simd-builtins.def: Add tbl1 builtin
25009 * config/aarch64/arm_neon.h (vqtbl1_p8): Use RTL builtin
25010 instead of inline asm.
25011 (vqtbl1_s8): Likewise.
25012 (vqtbl1_u8): Likewise.
25013 (vqtbl1q_p8): Likewise.
25014 (vqtbl1q_s8): Likewise.
25015 (vqtbl1q_u8): Likewise.
25016 (vtbl1_s8): Likewise.
25017 (vtbl1_u8): Likewise.
25018 (vtbl1_p8): Likewise.
25019 (vtbl2_s8): Likewise.
25020 (vtbl2_u8): Likewise.
25021 (vtbl2_p8): Likewise.
25023 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
25025 * config/aarch64/aarch64-simd-builtins.def: Add polynomial
25026 ssri_n buitin generator macro.
25027 * config/aarch64/arm_neon.h (vsri_n_p8): Use RTL builtin
25028 instead of inline asm.
25029 (vsri_n_p16): Likewise.
25030 (vsri_n_p64): Likewise.
25031 (vsriq_n_p8): Likewise.
25032 (vsriq_n_p16): Likewise.
25033 (vsriq_n_p64): Likewise.
25035 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
25037 * config/aarch64/aarch64-simd-builtins.def: Use VALLP mode
25038 iterator for polynomial ssli_n builtin generator macro.
25039 * config/aarch64/arm_neon.h (vsli_n_p8): Use RTL builtin
25040 instead of inline asm.
25041 (vsli_n_p16): Likewise.
25042 (vsliq_n_p8): Likewise.
25043 (vsliq_n_p16): Likewise.
25044 * config/aarch64/iterators.md: Define VALLP mode iterator.
25046 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
25048 * config/aarch64/aarch64-simd-builtins.def: Use VDQV_L
25049 iterator to generate [su]adalp RTL builtins.
25050 * config/aarch64/aarch64-simd.md: Use VDQV_L iterator in
25051 [su]adalp RTL pattern.
25052 * config/aarch64/arm_neon.h (vpadal_s32): Use RTL builtin
25053 instead of inline asm.
25054 (vpadal_u32): Likewise.
25056 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
25058 * config/aarch64/aarch64-simd-builtins.def: Add [su]addlp
25059 builtin generator macros.
25060 * config/aarch64/aarch64-simd.md (aarch64_<su>addlp<mode>):
25062 * config/aarch64/arm_neon.h (vpaddl_s8): Use RTL builtin
25063 instead of inline asm.
25064 (vpaddl_s16): Likewise.
25065 (vpaddl_s32): Likewise.
25066 (vpaddl_u8): Likewise.
25067 (vpaddl_u16): Likewise.
25068 (vpaddl_u32): Likewise.
25069 (vpaddlq_s8): Likewise.
25070 (vpaddlq_s16): Likewise.
25071 (vpaddlq_s32): Likewise.
25072 (vpaddlq_u8): Likewise.
25073 (vpaddlq_u16): Likewise.
25074 (vpaddlq_u32): Liwewise.
25075 * config/aarch64/iterators.md: Define [SU]ADDLP unspecs with
25076 appropriate attributes.
25078 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
25080 * config/aarch64/aarch64-simd-builtins.def: Use VDQ_I iterator
25081 for aarch64_addp<mode> builtin macro generator.
25082 * config/aarch64/aarch64-simd.md: Use VDQ_I iterator in
25083 aarch64_addp<mode> RTL pattern.
25084 * config/aarch64/arm_neon.h (vpaddq_s8): Use RTL builtin
25085 instead of inline asm.
25086 (vpaddq_s16): Likewise.
25087 (vpaddq_s32): Likewise.
25088 (vpaddq_s64): Likewise.
25089 (vpaddq_u8): Likewise.
25090 (vpaddq_u16): Likewise.
25091 (vpaddq_u32): Likewise.
25092 (vpaddq_u64): Likewise.
25094 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
25096 * config/aarch64/aarch64-simd-builtins.def: Add sq[r]dmulh_n
25097 builtin generator macros.
25098 * config/aarch64/aarch64-simd.md (aarch64_sq<r>dmulh_n<mode>):
25100 * config/aarch64/arm_neon.h (vqdmulh_n_s16): Use RTL builtin
25101 instead of inline asm.
25102 (vqdmulh_n_s32): Likewise.
25103 (vqdmulhq_n_s16): Likewise.
25104 (vqdmulhq_n_s32): Likewise.
25105 (vqrdmulh_n_s16): Likewise.
25106 (vqrdmulh_n_s32): Likewise.
25107 (vqrdmulhq_n_s16): Likewise.
25108 (vqrdmulhq_n_s32): Likewise.
25110 2021-04-28 Tobias Burnus <tobias@codesourcery.com>
25112 * doc/install.texi (--enable-offload-defaulted): Document.
25114 2021-04-28 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
25116 * config/avr/avr-dimode.md: Turn existing patterns into
25117 define_insn_and_split style patterns where the splitter
25118 adds a clobber of the condition code register. Drop "cc"
25119 attribute. Add new patterns to match output of
25121 * config/avr/avr-fixed.md: Likewise.
25122 * config/avr/avr.c (cc_reg_rtx): New.
25123 (avr_parallel_insn_from_insns): Adjust insn count
25124 for removal of set of cc0.
25125 (avr_is_casesi_sequence): Likewise.
25126 (avr_casei_sequence_check_operands): Likewise.
25127 (avr_optimize_casesi): Likewise. Also insert
25128 new insns after jump_insn.
25129 (avr_pass_casesi::avr_rest_of_handle_casesi): Adjust
25130 for removal of set of cc0.
25131 (avr_init_expanders): Initialize cc_reg_rtx.
25132 (avr_regno_reg_class): Handle REG_CC.
25133 (cond_string): Remove usage of CC_OVERFLOW_UNUSABLE.
25134 (avr_notice_update_cc): Remove function.
25135 (ret_cond_branch): Remove usage of CC_OVERFLOW_UNUSABLE.
25136 (compare_condition): Adjust for PARALLEL with
25138 (out_shift_with_cnt): Likewise.
25139 (ashlhi3_out): Likewise.
25140 (ashrhi3_out): Likewise.
25141 (lshrhi3_out): Likewise.
25142 (avr_class_max_nregs): Return single reg for REG_CC.
25143 (avr_compare_pattern): Check for REG_CC instead
25145 (avr_reorg_remove_redundant_compare): Likewise.
25146 (avr_reorg):Adjust for PARALLEL with REG_CC clobber.
25147 (avr_hard_regno_nregs): Return single reg for REG_CC.
25148 (avr_hard_regno_mode_ok): Allow only CCmode for REG_CC.
25149 (avr_md_asm_adjust): Clobber REG_CC.
25150 (TARGET_HARD_REGNO_NREGS): Define.
25151 (TARGET_CLASS_MAX_NREGS): Define.
25152 (TARGET_MD_ASM_ADJUST): Define.
25153 * config/avr/avr.h (FIRST_PSEUDO_REGISTER): Adjust
25155 (enum reg_class): Add CC_REG class.
25156 (NOTICE_UPDATE_CC): Remove.
25157 (CC_OVERFLOW_UNUSABLE): Remove.
25158 (CC_NO_CARRY): Remove.
25159 * config/avr/avr.md: Turn existing patterns into
25160 define_insn_and_split style patterns where the splitter
25161 adds a clobber of the condition code register. Drop "cc"
25162 attribute. Add new patterns to match output of
25164 (sez): Remove unused pattern.
25166 2021-04-28 Richard Earnshaw <rearnsha@arm.com>
25169 * config/arm/arm.c (arm_hard_regno_mode_ok): Only allow VPR to be
25172 2021-04-28 Richard Sandiford <richard.sandiford@arm.com>
25175 * config/aarch64/constraints.md (Utq): Require the address to
25176 be valid for both the element mode and for V2DImode.
25178 2021-04-28 Jakub Jelinek <jakub@redhat.com>
25179 Tobias Burnus <tobias@codesourcery.com>
25181 * configure.ac (OFFLOAD_DEFAULTED): AC_DEFINE if offload-defaulted.
25182 * gcc.c (process_command): New variable.
25183 (driver::maybe_putenv_OFFLOAD_TARGETS): If OFFLOAD_DEFAULTED,
25184 set it if -foffload is defaulted.
25185 * lto-wrapper.c (OFFLOAD_TARGET_DEFAULT_ENV): Define.
25186 (compile_offload_image): If OFFLOAD_DEFAULTED and
25187 OFFLOAD_TARGET_DEFAULT is in the environment, don't fail
25188 if corresponding mkoffload can't be found.
25189 (compile_images_for_offload_targets): Likewise. Free and clear
25190 offload_names if no valid offload is found.
25191 * config.in: Regenerate.
25192 * configure: Regenerate.
25194 2021-04-28 Richard Biener <rguenther@suse.de>
25196 PR tree-optimization/100292
25197 * tree-vect-generic.c (expand_vector_condition): Do not fold
25200 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
25202 * config/rs6000/aix.h (SUBTARGET_DRIVER_SELF_SPECS): New.
25203 * config/rs6000/aix64.opt (m64): New.
25206 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
25208 * config/vax/vax.c (print_operand_address, vax_address_cost_1)
25209 (index_term_p): Handle ASHIFT too.
25211 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
25213 * config/vax/builtins.md (jbb<ccss>i<mode>): Remove operand #3.
25214 (sync_lock_test_and_set<mode>): Adjust accordingly.
25215 (sync_lock_release<mode>): Likewise.
25217 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
25219 * config/vax/vax-protos.h (adjacent_operands_p): Remove
25221 * config/vax/vax.c (adjacent_operands_p): Remove.
25223 2021-04-27 Maciej W. Rozycki <macro@linux-mips.org>
25225 * ifcvt.c (dead_or_predicable) [!IFCVT_MODIFY_TESTS]: Fall
25226 through to the non-conditional execution case if getting the
25227 condition for conditional execution has failed.
25229 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
25231 PR middle-end/100284
25232 * gimple.c (gimple_could_trap_p_1): Remove VEC_COND_EXPR test.
25233 * tree-eh.c (operation_could_trap_p): Handle VEC_COND_EXPR rather
25234 than asserting on it.
25236 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
25238 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): Protect
25239 with TARGET_AIX_OS.
25241 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
25244 * calls.c (precompute_register_parameters): Additionally test
25245 targetm.precompute_tls_p to pre-compute argument.
25246 * config/rs6000/aix.h (TARGET_PRECOMPUTE_TLS_P): Define.
25247 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): New.
25248 * target.def (precompute_tls_p): New.
25249 * doc/tm.texi.in (TARGET_PRECOMPUTE_TLS_P): Add hook documentation.
25250 * doc/tm.texi: Regenerated.
25252 2021-04-27 Jakub Jelinek <jakub@redhat.com>
25255 * config/aarch64/aarch64.c (aarch64_print_operand): Cast -UINTVAL
25256 back to HOST_WIDE_INT.
25258 2021-04-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
25261 * simplify-rtx.c (simplify_context::simplify_subreg): Check the
25262 memory alignment for the outer mode.
25264 2021-04-27 H.J. Lu <hjl.tools@gmail.com>
25266 PR middle-end/90773
25267 * expr.c (op_by_pieces_d::get_usable_mode): New member function.
25268 (op_by_pieces_d::run): Cange a while loop to a do-while loop.
25270 2021-04-27 Alex Coplan <alex.coplan@arm.com>
25273 * config/arm/arm.c (arm_split_compare_and_swap): Fix up codegen
25274 with negative immediates: ensure we expand cbranchsi4_scratch
25275 correctly and ensure we satisfy its constraints.
25276 * config/arm/sync.md
25277 (@atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1): Don't
25278 attempt to tie two output operands together with constraints;
25279 collapse two alternatives.
25280 (@atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): Likewise.
25281 * config/arm/thumb1.md (cbranchsi4_neg_late): New.
25283 2021-04-27 Jakub Jelinek <jakub@redhat.com>
25286 * config/aarch64/predicates.md (aarch64_sub_immediate,
25287 aarch64_plus_immediate): Use -UINTVAL instead of -INTVAL.
25288 * config/aarch64/aarch64.md (casesi, rotl<mode>3): Likewise.
25289 * config/aarch64/aarch64.c (aarch64_print_operand,
25290 aarch64_split_atomic_op, aarch64_expand_subvti): Likewise.
25292 2021-04-27 Jakub Jelinek <jakub@redhat.com>
25294 PR tree-optimization/100239
25295 * tree-vect-generic.c (lower_vec_perm): Don't accept constant
25296 permutations with all indices from the first zero element as vec_shl.
25298 2021-04-27 Jakub Jelinek <jakub@redhat.com>
25300 PR rtl-optimization/100254
25301 * cfgcleanup.c (outgoing_edges_match): Check REG_EH_REGION on
25302 last1 and last2 insns rather than BB_END (bb1) and BB_END (bb2) insns.
25304 2021-04-27 Richard Biener <rguenther@suse.de>
25306 PR tree-optimization/99912
25307 * passes.def: Add comment about new TODO_remove_unused_locals.
25308 * tree-stdarg.c (pass_data_stdarg): Run TODO_remove_unused_locals
25311 2021-04-27 Richard Biener <rguenther@suse.de>
25313 PR tree-optimization/99912
25314 * passes.def (pass_all_optimizations): Add pass_dse before
25315 the first pass_dce, move the first pass_dse before the
25316 pass_dce following pass_pre.
25318 2021-04-27 Jakub Jelinek <jakub@redhat.com>
25320 PR tree-optimization/95527
25321 * generic-match-head.c: Include tm.h.
25322 * gimple-match-head.c: Include tm.h.
25323 * match.pd (CLZ == INTEGER_CST): Don't use
25324 #ifdef CLZ_DEFINED_VALUE_AT_ZERO, only test CLZ_DEFINED_VALUE_AT_ZERO
25325 if clz == CFN_CLZ. Add missing val declaration.
25326 (CTZ cmp CST): New simplifications.
25328 2021-04-27 Jakub Jelinek <jakub@redhat.com>
25330 PR tree-optimization/96696
25331 * expr.c (expand_expr_divmod): New function.
25332 (expand_expr_real_2) <case TRUNC_DIV_EXPR>: Use it for truncations and
25333 divisions. Formatting fixes.
25334 <case MULT_EXPR>: Optimize x / y * y as x - x % y if the latter is
25337 2021-04-27 Martin Jambor <mjambor@suse.cz>
25340 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call):
25341 If removing a call statement LHS SSA name, release it.
25343 2021-04-27 Richard Earnshaw <rearnsha@arm.com>
25346 * config/arm/arm.c (THUMB2_WORK_REGS): Check PIC_OFFSET_TABLE_REGNUM
25347 is valid before including it in the mask.
25349 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
25352 * config/aarch64/aarch64.c (aarch64_comp_type_attributes): Handle
25355 2021-04-27 Richard Biener <rguenther@suse.de>
25357 PR tree-optimization/100051
25358 * tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Add
25359 disambiguator based on access size vs. decl size.
25361 2021-04-27 Richard Biener <rguenther@suse.de>
25363 PR tree-optimization/100278
25364 * tree-ssa-pre.c (compute_avail): Give up when we cannot
25365 adjust TBAA beacuse of mismatching bases.
25367 2021-04-27 Jakub Jelinek <jakub@redhat.com>
25370 * config/i386/i386.md (*<insn><mode>3_mask, *<insn><mode>3_mask_1):
25371 For any_rotate define_insn_split and following splitters, use
25372 SWI iterator instead of SWI48.
25374 2021-04-27 Richard Biener <rguenther@suse.de>
25376 PR tree-optimization/99776
25377 * match.pd (bit_field_ref (ctor)): Relax element extract
25378 type compatibility checks.
25380 2021-04-27 Cui,Lili <lili.cui@intel.com>
25382 * common/config/i386/i386-common.c (processor_names):
25383 Sync processor_names with processor_type.
25384 * config/i386/i386-options.c (processor_cost_table):
25385 Sync processor_cost_table with processor_type.
25387 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
25389 * value-range.cc (irange::irange_set_1bit_anti_range): Add assert.
25390 (irange::set): Call irange_set_1bit_anti_range for handling all
25391 1-bit ranges. Fall through on ~[MIN,MAX].
25393 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
25395 * value-range.cc (irange::legacy_num_pairs): Remove.
25396 (irange::invert): Change gcc_assert to gcc_checking_assert.
25397 * value-range.h (irange::num_pairs): Adjust for a cached
25398 num_pairs(). Also, rename all gcc_assert's to
25399 gcc_checking_assert's.
25401 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
25403 * value-range.cc (irange::operator=): Set m_kind.
25404 (irange::copy_to_legacy): Handle varying and undefined sources
25405 as a legacy copy since they can be easily copied.
25406 (irange::irange_set): Set m_kind.
25407 (irange::irange_set_anti_range): Same.
25408 (irange::set): Rename normalize_min_max to normalize_kind.
25409 (irange::verify_range): Adjust for multi-ranges having the
25411 (irange::irange_union): Set m_kind.
25412 (irange::irange_intersect): Same.
25413 (irange::invert): Same.
25414 * value-range.h (irange::kind): Always return m_kind.
25415 (irange::varying_p): Rename to...
25416 (irange::varying_comptaible_p): ...this.
25417 (irange::undefined_p): Only look at m_kind.
25418 (irange::irange): Always set VR_UNDEFINED if applicable.
25419 (irange::set_undefined): Always set VR_UNDEFINED.
25420 (irange::set_varying): Always set m_kind to VR_VARYING.
25421 (irange::normalize_min_max): Rename to...
25422 (irange::normalize_kind): ...this.
25424 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
25426 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::set_ssa_range_info):
25427 Adjust for constant_p including varying_p.
25428 * tree-vrp.c (vrp_prop::finalize): Same.
25429 (determine_value_range): Same.
25430 * vr-values.c (vr_values::range_of_expr): Same.
25431 * value-range.cc (irange::symbolic_p): Do not check varying_p.
25432 (irange::constant_p): Same.
25434 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
25436 * value-range.cc (irange::legacy_lower_bound): Replace
25437 !undefined_p check with num_ranges > 0.
25438 (irange::legacy_upper_bound): Same.
25439 * value-range.h (irange::type): Same.
25440 (irange::lower_bound): Same.
25441 (irange::upper_bound): Same.
25443 2021-04-26 Richard Biener <rguenther@suse.de>
25445 PR tree-optimization/99956
25446 * gimple-loop-interchange.cc (compute_access_stride):
25447 Try instantiating the access in a shallower loop nest
25448 if instantiating failed.
25449 (compute_access_strides): Pass adjustable loop_nest
25450 to compute_access_stride.
25452 2021-04-26 Christophe Lyon <christophe.lyon@linaro.org>
25454 * doc/sourcebuild.texi (arm_cmse_hw): Document.
25456 2021-04-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25458 * config/aarch64/iterators.md (vwcore): Handle V4BF, V8BF.
25460 2021-04-26 Thomas Schwinge <thomas@codesourcery.com>
25461 Nathan Sidwell <nathan@codesourcery.com>
25462 Tom de Vries <vries@codesourcery.com>
25463 Julian Brown <julian@codesourcery.com>
25464 Kwok Cheung Yeung <kcy@codesourcery.com>
25466 * omp-offload.c (oacc_validate_dims): Implement
25467 '-Wopenacc-parallelism'.
25468 * doc/invoke.texi (-Wopenacc-parallelism): Document.
25470 2021-04-26 Richard Biener <rguenther@suse.de>
25472 * tree-cfg.h (gimplify_build1): Remove.
25473 (gimplify_build2): Likewise.
25474 (gimplify_build3): Likewise.
25475 * tree-cfg.c (gimplify_build1): Move to tree-vect-generic.c.
25476 (gimplify_build2): Likewise.
25477 (gimplify_build3): Likewise.
25478 * tree-vect-generic.c (gimplify_build1): Move from tree-cfg.c.
25480 (gimplify_build2): Likewise.
25481 (gimplify_build3): Likewise.
25482 (tree_vec_extract): Use resimplify with following SSA edges.
25483 (expand_vector_parallel): Avoid passing NULL size/bitpos
25484 to tree_vec_extract.
25485 * expr.c (store_constructor): Deal with zero-element CTORs.
25486 * match.pd (bit_field_ref <vector CTOR>): Make sure to
25487 produce vector constants when possible.
25489 2021-04-26 Richard Biener <rguenther@suse.de>
25491 * tree-complex.c: Include gimple-fold.h.
25492 (expand_complex_addition): Use gimple_build.
25493 (expand_complex_multiplication_components): Likewise.
25494 (expand_complex_multiplication): Likewise.
25495 (expand_complex_div_straight): Likewise.
25496 (expand_complex_div_wide): Likewise.
25497 (expand_complex_division): Likewise.
25498 (expand_complex_conjugate): Likewise.
25499 (expand_complex_comparison): Likewise.
25501 2021-04-26 Richard Biener <rguenther@suse.de>
25503 * tree-ssa-phiopt.c (two_value_replacement): Remove use
25504 of legacy gimplify_buildN API.
25506 2021-04-26 Richard Biener <rguenther@suse.de>
25508 PR tree-optimization/99473
25509 * tree-ssa-phiopt.c (cond_store_replacement): Handle all
25512 2021-04-26 Richard Biener <rguenther@suse.de>
25514 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_builtin):
25515 Use replace_call_with_value.
25517 2021-04-26 Richard Biener <rguenther@suse.de>
25519 * tree-ssa-propagate.h (valid_gimple_rhs_p): Remove.
25520 (update_gimple_call): Likewise.
25521 (update_call_from_tree): Likewise.
25522 * tree-ssa-propagate.c (valid_gimple_rhs_p): Remove.
25523 (valid_gimple_call_p): Likewise.
25524 (move_ssa_defining_stmt_for_defs): Likewise.
25525 (finish_update_gimple_call): Likewise.
25526 (update_gimple_call): Likewise.
25527 (update_call_from_tree): Likewise.
25528 (propagate_tree_value_into_stmt): Use replace_call_with_value.
25529 * gimple-fold.h (update_gimple_call): Declare.
25530 * gimple-fold.c (valid_gimple_rhs_p): Move here from
25531 tree-ssa-propagate.c.
25532 (update_gimple_call): Likewise.
25533 (valid_gimple_call_p): Likewise.
25534 (finish_update_gimple_call): Likewise, and simplify.
25535 (gimplify_and_update_call_from_tree): Implement
25536 update_call_from_tree functionality, avoid excessive
25537 push/pop_gimplify_context.
25538 (gimple_fold_builtin): Use only gimplify_and_update_call_from_tree.
25539 (gimple_fold_call): Likewise.
25540 * gimple-ssa-sprintf.c (try_substitute_return_value): Likewise.
25541 * tree-ssa-ccp.c (ccp_folder::fold_stmt): Likewise.
25542 (pass_fold_builtins::execute): Likewise.
25543 (optimize_stack_restore): Use replace_call_with_value.
25544 * tree-cfg.c (fold_loop_internal_call): Likewise.
25545 * tree-ssa-dce.c (maybe_optimize_arith_overflow): Use
25546 only gimplify_and_update_call_from_tree.
25547 * tree-ssa-strlen.c (handle_builtin_strlen): Likewise.
25548 (handle_builtin_strchr): Likewise.
25549 * tsan.c: Include gimple-fold.h instead of tree-ssa-propagate.h.
25551 2021-04-26 Jakub Jelinek <jakub@redhat.com>
25554 * vmsdbgout.c (ASM_OUTPUT_DEBUG_STRING, vmsdbgout_begin_block,
25555 vmsdbgout_end_block, lookup_filename, vmsdbgout_source_line): Remove
25558 2021-04-25 liuhongt <hongtao.liu@intel.com>
25561 * config/i386/i386-builtin.def (BDESC): Change the icode of
25562 the following builtins to CODE_FOR_nothing.
25563 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
25564 IX86_BUILTIN_PCMPEQB128, IX86_BUILTIN_PCMPEQW128,
25565 IX86_BUILTIN_PCMPEQD128, IX86_BUILTIN_PCMPEQQ,
25566 IX86_BUILTIN_PCMPEQB256, IX86_BUILTIN_PCMPEQW256,
25567 IX86_BUILTIN_PCMPEQD256, IX86_BUILTIN_PCMPEQQ256,
25568 IX86_BUILTIN_PCMPGTB128, IX86_BUILTIN_PCMPGTW128,
25569 IX86_BUILTIN_PCMPGTD128, IX86_BUILTIN_PCMPGTQ,
25570 IX86_BUILTIN_PCMPGTB256, IX86_BUILTIN_PCMPGTW256,
25571 IX86_BUILTIN_PCMPGTD256, IX86_BUILTIN_PCMPGTQ256.
25572 * config/i386/sse.md (avx2_eq<mode>3): Deleted.
25573 (sse2_eq<mode>3): Ditto.
25574 (sse4_1_eqv2di3): Ditto.
25575 (sse2_gt<mode>3): Rename to ..
25576 (*sse2_gt<mode>3): .. this.
25578 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
25581 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
25584 * config/darwin.c (darwin_binds_local_p): Assume that any
25585 public symbol might be interposed for PIC code. Update function
25586 header comment to reflect current Darwin capability.
25588 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
25591 * config/darwin.c (darwin_binds_local_p): Assume that any
25592 public symbol might be interposed for PIC code. Update function
25593 header comment to reflect current Darwin capability.
25595 2021-04-24 Richard Sandiford <richard.sandiford@arm.com>
25597 * doc/sourcebuild.texi: Document no-opts and any-opts target
25600 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
25602 * config/bpf/bpf.h (ASM_OUTPUT_ALIGNED_BSS): Use .type and .lcomm.
25604 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
25606 * config/bpf/bpf.h (FUNCTION_BOUNDARY): Set to 64.
25608 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
25611 * config/i386/i386-options.c (ix86_option_override_internal):
25612 Error out when -m96bit-long-double is used with 64bit targets.
25613 * config/i386/i386.md (*pushxf_rounded): Remove pattern.
25615 2021-04-23 Martin Liska <mliska@suse.cz>
25617 * lto-wrapper.c: Remove FIXME about usage of
25618 hardware_concurrency. The function is not on par with
25621 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
25624 * config/i386/sync.md (FILD_ATOMIC/FIST_ATOMIC FP load peephole2):
25625 Copy operand 3 to operand 4. Use sse_reg_operand
25626 as operand 3 predicate.
25627 (FILD_ATOMIC/FIST_ATOMIC FP load peephole2 with mem blockage): Ditto.
25628 (LDX_ATOMIC/STX_ATOMIC FP load peephole2): Ditto.
25629 (LDX_ATOMIC/LDX_ATOMIC FP load peephole2 with mem blockage): Ditto.
25630 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2):
25631 Copy operand 1 to operand 0.
25632 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2 with mem blockage): Ditto.
25633 (LDX_ATOMIC/STX_ATOMIC FP store peephole2): Ditto.
25634 (LDX_ATOMIC/LDX_ATOMIC FP store peephole2 with mem blockage): Ditto.
25636 2021-04-23 Alex Coplan <alex.coplan@arm.com>
25638 PR rtl-optimization/100230
25639 * early-remat.c (early_remat::sort_candidates): Use delete[]
25640 instead of delete for array allocated with new[].
25642 2021-04-23 Richard Biener <rguenther@suse.de>
25644 * genmatch.c (lower_cond): Remove VEC_COND_EXPR special-casing.
25645 (capture_info::capture_info): Likewise.
25646 (capture_info::walk_match): Likewise.
25647 (expr::gen_transform): Likewise.
25648 (dt_simplify::gen_1): Likewise.
25649 * gimple-match-head.c (maybe_resimplify_conditional_op):
25650 Remove VEC_COND_EXPR special-casing.
25651 (gimple_simplify): Likewise.
25652 * gimple.c (gimple_could_trap_p_1): Adjust.
25653 * tree-ssa-pre.c (compute_avail): Allow VEC_COND_EXPR
25654 to participate in PRE.
25656 2021-04-23 Richard Biener <rguenther@suse.de>
25658 * cfganal.c (connect_infinite_loops_to_exit): First call
25659 add_noreturn_fake_exit_edges.
25660 * ipa-sra.c (process_scan_results): Do not call the now redundant
25661 add_noreturn_fake_exit_edges.
25662 * predict.c (tree_estimate_probability): Likewise.
25663 (rebuild_frequencies): Likewise.
25664 * store-motion.c (one_store_motion_pass): Likewise.
25666 2021-04-23 Richard Biener <rguenther@suse.de>
25668 PR tree-optimization/100222
25669 * predict.c (pass_profile::execute): Remove redundant call to
25670 mark_irreducible_loops.
25671 (report_predictor_hitrates): Likewise.
25673 2021-04-23 Richard Biener <rguenther@suse.de>
25675 * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Avoid
25676 valid_gimple_rhs_p by instead gimplifying to one.
25678 2021-04-23 Richard Biener <rguenther@suse.de>
25680 PR tree-optimization/99971
25681 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
25682 Always use TBAA for loads.
25684 2021-04-23 liuhongt <hongtao.liu@intel.com>
25687 * config/i386/i386-options.c (ix86_option_override_internal):
25688 Clear MASK_AVX256_SPLIT_UNALIGNED_LOAD/STORE in x_target_flags
25689 when X86_TUNE_AVX256_UNALIGNED_LOAD/STORE_OPTIMAL is enabled
25690 by target attribute.
25692 2021-04-23 David Edelsohn <dje.gcc@gmail.com>
25694 * config/rs6000/aix71.h (PREFERRED_DEBUGGING_TYPE): Change to
25696 * config/rs6000/aix72.h (PREFERRED_DEBUGGING_TYPE): Same.
25698 2021-04-22 David Edelsohn <dje.gcc@gmail.com>
25700 * config.gcc (powerpc-ibm-aix6.*): Remove.
25701 * config/rs6000/aix61.h: Delete.
25703 2021-04-22 Martin Liska <mliska@suse.cz>
25705 PR testsuite/100159
25706 PR testsuite/100192
25707 * builtins.c (expand_builtin): Fix typos and missing comments.
25708 * dwarf2out.c (gen_subprogram_die): Likewise.
25709 (gen_struct_or_union_type_die): Likewise.
25711 2021-04-22 Uroš Bizjak <ubizjak@gmail.com>
25714 * config/i386/i386-expand.c (ix86_expand_convert_uns_sidf_sse):
25715 Remove the sign with FE_DOWNWARD, where x - x = -0.0.
25717 2021-04-21 Iain Sandoe <iain@sandoe.co.uk>
25719 * config/i386/darwin.h (TARGET_64BIT): Remove definition
25720 based on TARGET_ISA_64BIT.
25721 (TARGET_64BIT_P): Remove definition based on
25722 TARGET_ISA_64BIT_P().
25724 2021-04-21 Martin Liska <mliska@suse.cz>
25727 2021-04-21 Martin Liska <mliska@suse.cz>
25729 * lto-wrapper.c (cpuset_popcount): Remove.
25730 (init_num_threads): Remove and use hardware_concurrency.
25732 2021-04-21 Martin Liska <mliska@suse.cz>
25735 * main.c (main): Call toplev::finalize in CHECKING_P mode.
25736 * ipa-modref.c (ipa_modref_c_finalize): summaries are NULL
25737 when incremental LTO linking happens.
25739 2021-04-21 Martin Liska <mliska@suse.cz>
25741 * lto-wrapper.c (run_gcc): When -flto=jobserver is used, but the
25742 makeserver cannot be detected, then use -flto=N fallback.
25744 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
25746 * acinclude.m4 (gcc_AC_INITFINI_ARRAY): When cross-compiling,
25747 default to yes for aarch64-linux-gnu.
25748 * configure: Regenerate.
25750 2021-04-21 Martin Liska <mliska@suse.cz>
25752 * lto-wrapper.c (cpuset_popcount): Remove.
25753 (init_num_threads): Remove and use hardware_concurrency.
25755 2021-04-21 Martin Liska <mliska@suse.cz>
25757 * config/i386/i386.c: Remove superfluous || TARGET_MACHO
25758 which remains to be '(... || 0)' and clang complains about it.
25759 * dwarf2out.c (AT_vms_delta): Declare conditionally.
25760 (add_AT_vms_delta): Likewise.
25761 * tree.c (fld_simplified_type): Use rather more common pattern
25762 for disabling of something (#if 0).
25763 (get_tree_code_name): Likewise.
25764 (verify_type_variant): Likewise.
25766 2021-04-21 Martin Liska <mliska@suse.cz>
25768 * config/i386/i386-expand.c (decide_alignment): Use newly named
25769 macro TARGET_CPU_P.
25770 * config/i386/i386.c (ix86_decompose_address): Likewise.
25771 (ix86_address_cost): Likewise.
25772 (ix86_lea_outperforms): Likewise.
25773 (ix86_avoid_lea_for_addr): Likewise.
25774 (ix86_add_stmt_cost): Likewise.
25775 * config/i386/i386.h (TARGET_*): Remove.
25776 (TARGET_CPU_P): New macro.
25777 * config/i386/i386.md: Use newly named macro TARGET_CPU_P.
25778 * config/i386/x86-tune-sched-atom.c (do_reorder_for_imul): Likewise.
25779 (swap_top_of_ready_list): Likewise.
25780 (ix86_atom_sched_reorder): Likewise.
25781 * config/i386/x86-tune-sched-bd.c (ix86_bd_has_dispatch): Likewise.
25782 * config/i386/x86-tune-sched.c (ix86_adjust_cost): Likewise.
25784 2021-04-21 Martin Liska <mliska@suse.cz>
25786 * config/i386/i386-options.c (TARGET_EXPLICIT_NO_SAHF_P):
25788 (SET_TARGET_NO_SAHF): Likewise.
25789 (TARGET_EXPLICIT_PREFETCH_SSE_P): Likewise.
25790 (SET_TARGET_PREFETCH_SSE): Likewise.
25791 (TARGET_EXPLICIT_NO_TUNE_P): Likewise.
25792 (SET_TARGET_NO_TUNE): Likewise.
25793 (TARGET_EXPLICIT_NO_80387_P): Likewise.
25794 (SET_TARGET_NO_80387): Likewise.
25796 * config/i386/i386.h (TARGET_*): Remove.
25797 * opth-gen.awk: Generate new used macros.
25799 2021-04-21 Martin Liska <mliska@suse.cz>
25801 * config/i386/i386.h (PTA_*): Remove.
25802 (enum pta_flag): New.
25803 (DEF_PTA): Generate PTA_* values from i386-isa.def.
25804 * config/i386/i386-isa.def: New file.
25806 2021-04-21 Alex Coplan <alex.coplan@arm.com>
25809 * config/aarch64/aarch64-bti-insert.c (aarch64_bti_j_insn_p): New.
25810 (rest_of_insert_bti): Avoid inserting duplicate bti j insns for
25811 jump table targets.
25813 2021-04-21 H.J. Lu <hjl.tools@gmail.com>
25815 * config.gcc: Install mwaitintrin.h for i[34567]86-*-* and
25816 x86_64-*-* targets.
25817 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_MWAIT_SET):
25819 (OPTION_MASK_ISA2_MWAIT_UNSET): Likewise.
25820 (ix86_handle_option): Handle -mmwait.
25821 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
25822 Replace OPTION_MASK_ISA_SSE3 with OPTION_MASK_ISA2_MWAIT on
25823 __builtin_ia32_monitor and __builtin_ia32_mwait.
25824 * config/i386/i386-options.c (isa2_opts): Add -mmwait.
25825 (ix86_valid_target_attribute_inner_p): Likewise.
25826 (ix86_option_override_internal): Enable mwait/monitor
25827 instructions for -msse3.
25828 * config/i386/i386.h (TARGET_MWAIT): New.
25829 (TARGET_MWAIT_P): Likewise.
25830 * config/i386/i386.opt: Add -mmwait.
25831 * config/i386/mwaitintrin.h: New file.
25832 * config/i386/pmmintrin.h: Include <mwaitintrin.h>.
25833 * config/i386/sse.md (sse3_mwait): Replace TARGET_SSE3 with
25835 (@sse3_monitor_<mode>): Likewise.
25836 * config/i386/x86gprintrin.h: Include <mwaitintrin.h>.
25837 * doc/extend.texi: Document mwait target attribute.
25838 * doc/invoke.texi: Document -mmwait.
25840 2021-04-21 Martin Liska <mliska@suse.cz>
25842 * config/i386/i386-options.c (DEF_ENUM): Remove it.
25843 * config/i386/i386-opts.h (DEF_ENUM): Likewise.
25844 * config/i386/stringop.def (DEF_ENUM): Likewise.
25846 2021-04-21 Martin Liska <mliska@suse.cz>
25848 * tree-cfg.c (gimple_verify_flow_info): Use qD instead
25849 of print_generic_expr.
25851 2021-04-21 Jakub Jelinek <jakub@redhat.com>
25853 PR rtl-optimization/100148
25854 * cprop.c (constprop_register): Use next_nondebug_insn instead of
25857 2021-04-21 Martin Liska <mliska@suse.cz>
25860 * cgraphunit.c (cgraph_node::analyze): Remove duplicate
25861 free_dominance_info calls.
25863 2021-04-21 Richard Biener <rguenther@suse.de>
25865 * gimple-fold.c (maybe_fold_reference): Remove is_lhs
25866 parameter (and assume it to be false).
25867 (fold_gimple_assign): Adjust, remove all callers of
25868 maybe_fold_reference calling it with is_lhs true.
25869 (gimple_fold_call): Likewise.
25870 (fold_stmt_1): Likewise.
25872 2021-04-21 Richard Biener <rguenther@suse.de>
25874 * fold-const.c (pedantic_non_lvalue_loc): Remove.
25875 (fold_binary_loc): Adjust.
25876 (fold_ternary_loc): Likewise.
25878 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
25880 PR middle-end/100130
25881 * varasm.c (get_block_for_decl): Make sure that any use of the
25882 retain attribute matches the section's retain flag.
25883 (switch_to_section): Check for retain mismatches even when
25884 changing sections, but do not warn if the given decl is the
25885 section's named.decl.
25886 (output_object_block): Pass the first decl in the block (if any)
25887 to switch_to_section.
25889 2021-04-20 H.J. Lu <hjl.tools@gmail.com>
25891 * config/i386/i386-c.c (ix86_target_macros_internal): Define
25892 __CRC32__ for -mcrc32.
25893 * config/i386/i386-options.c (ix86_option_override_internal):
25894 Enable crc32 instruction for -msse4.2.
25895 * config/i386/i386.md (sse4_2_crc32<mode>): Remove TARGET_SSE4_2
25897 (sse4_2_crc32di): Likewise.
25898 * config/i386/ia32intrin.h: Use crc32 target option for CRC32
25901 2021-04-20 Segher Boessenkool <segher@kernel.crashing.org>
25904 * config/rs6000/rs6000.c (rs6000_machine_from_flags): Do not consider
25907 2021-04-20 Martin Liska <mliska@suse.cz>
25909 * doc/invoke.texi: Fix typo.
25910 * params.opt: Likewise.
25912 2021-04-20 Martin Liska <mliska@suse.cz>
25914 * doc/invoke.texi: Document new param.
25916 2021-04-19 Andrew MacLeod <amacleod@redhat.com>
25918 PR tree-optimization/100081
25919 * gimple-range-cache.h (ranger_cache): Inherit from gori_compute
25920 rather than gori_compute_cache.
25921 * gimple-range-gori.cc (is_gimple_logical_p): Move to top of file.
25922 (range_def_chain::m_logical_depth): New member.
25923 (range_def_chain::range_def_chain): Initialize m_logical_depth.
25924 (range_def_chain::get_def_chain): Don't build defchains through more
25925 than LOGICAL_LIMIT logical expressions.
25926 * params.opt (param_ranger_logical_depth): New.
25928 2021-04-19 Richard Earnshaw <rearnsha@arm.com>
25931 * config/arm/arm.c (arm_configure_build_target): Do not strip
25932 extended FPU/SIMD feature bits from the target ISA when -mfpu
25933 is specified (partial revert of r11-8168).
25935 2021-04-19 Thomas Schwinge <thomas@codesourcery.com>
25937 * params.opt (-param=openacc-kernels=): Add.
25938 * omp-oacc-kernels-decompose.cc
25939 (pass_omp_oacc_kernels_decompose::gate): Use it.
25940 * doc/invoke.texi (-fopenacc-kernels=@var{mode}): Move...
25941 (--param): ... here, 'openacc-kernels'.
25943 2021-04-19 Martin Liska <mliska@suse.cz>
25946 * gengtype.c (finish_root_table): Align function arguments
25947 in between declaration and definition.
25949 2021-04-19 Eric Botcazou <ebotcazou@adacore.com>
25951 * config/i386/winnt.c (i386_pe_seh_cold_init): Properly deal with
25952 frames larger than the SEH maximum frame size.
25954 2021-04-18 Segher Boessenkool <segher@kernel.crashing.org>
25956 PR rtl-optimization/99927
25957 * combine.c (distribute_notes) [REG_UNUSED]: If the register already
25958 is dead, just drop it.
25960 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
25963 * config/i386/winnt-d.c (TARGET_D_TEMPLATES_ALWAYS_COMDAT): Define.
25964 * doc/tm.texi: Regenerate.
25965 * doc/tm.texi.in (D language and ABI): Add @hook for
25966 TARGET_D_TEMPLATES_ALWAYS_COMDAT.
25968 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
25970 * config/darwin-d.c (darwin_d_handle_target_object_format): New
25972 (darwin_d_register_target_info): New function.
25973 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
25974 * config/dragonfly-d.c (dragonfly_d_handle_target_object_format): New
25976 (dragonfly_d_register_target_info): New function.
25977 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
25978 * config/freebsd-d.c (freebsd_d_handle_target_object_format): New
25980 (freebsd_d_register_target_info): New function.
25981 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
25982 * config/glibc-d.c (glibc_d_handle_target_object_format): New
25984 (glibc_d_register_target_info): New function.
25985 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
25986 * config/i386/i386-d.c (ix86_d_handle_target_object_format): New
25988 (ix86_d_register_target_info): Add ix86_d_handle_target_object_format
25989 as handler for objectFormat key.
25990 * config/i386/winnt-d.c (winnt_d_handle_target_object_format): New
25992 (winnt_d_register_target_info): New function.
25993 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
25994 * config/netbsd-d.c (netbsd_d_handle_target_object_format): New
25996 (netbsd_d_register_target_info): New function.
25997 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
25998 * config/openbsd-d.c (openbsd_d_handle_target_object_format): New
26000 (openbsd_d_register_target_info): New function.
26001 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
26002 * config/pa/pa-d.c (pa_d_handle_target_object_format): New function.
26003 (pa_d_register_target_info): Add pa_d_handle_target_object_format as
26004 handler for objectFormat key.
26005 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_object_format): New
26007 (rs6000_d_register_target_info): Add
26008 rs6000_d_handle_target_object_format as handler for objectFormat key.
26009 * config/sol2-d.c (solaris_d_handle_target_object_format): New
26011 (solaris_d_register_target_info): New function.
26012 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
26014 2021-04-16 Jakub Jelinek <jakub@redhat.com>
26017 * config/aarch64/aarch64.c (aarch64_function_arg_alignment): Change
26018 abi_break argument from bool * to unsigned *, store there the pre-GCC 9
26020 (aarch64_layout_arg, aarch64_gimplify_va_arg_expr): Adjust callers.
26021 (aarch64_function_arg_regno_p): Likewise. Only emit -Wpsabi note if
26022 the old and new alignment after applying MIN/MAX to it is different.
26024 2021-04-16 Tamar Christina <tamar.christina@arm.com>
26027 * config/aarch64/aarch64-sve.md (@aarch64_sve_trn1_conv<mode>): New.
26028 * config/aarch64/aarch64.c (aarch64_expand_sve_const_pred_trn): Use new
26030 * config/aarch64/iterators.md (UNSPEC_TRN1_CONV): New.
26032 2021-04-16 Bill Schmidt <wschmidt@linux.ibm.com>
26034 * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions): Revise
26035 this section and its subsections.
26037 2021-04-16 Jakub Jelinek <jakub@redhat.com>
26040 * config/aarch64/aarch64.md (*neg_asr_si2_extr, *extrsi5_insn_di): New
26041 define_insn patterns.
26043 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
26045 PR rtl-optimization/98689
26046 * reg-notes.def (UNTYPED_CALL): New note.
26047 * combine.c (distribute_notes): Handle it.
26048 * emit-rtl.c (try_split): Likewise.
26049 * rtlanal.c (rtx_properties::try_to_add_insn): Likewise. Assume
26050 that calls with the note implicitly set all return value registers.
26051 * builtins.c (expand_builtin_apply): Add a REG_UNTYPED_CALL
26054 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
26056 PR rtl-optimization/99596
26057 * rtlanal.c (rtx_properties::try_to_add_insn): Don't add global
26058 register accesses for const calls. Assume that pure functions
26059 can only read from global registers. Ignore cases in which
26060 the stack pointer has been marked global.
26062 2021-04-16 Jakub Jelinek <jakub@redhat.com>
26065 * tree-vect-loop.c (vect_transform_loop): Don't remove just
26066 dead scalar .MASK_LOAD calls, but also dead .COND_* calls - replace
26067 them by their last argument.
26069 2021-04-15 Martin Liska <mliska@suse.cz>
26071 * doc/invoke.texi: Other params don't use it, remove it.
26073 2021-04-15 Richard Biener <rguenther@suse.de>
26075 * gimple-builder.h: Add deprecation note.
26077 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
26080 * attribs.h (restrict_type_identity_attributes_to): Declare.
26081 * attribs.c (restrict_type_identity_attributes_to): New function.
26083 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
26086 * attribs.h (affects_type_identity_attributes): Declare.
26087 * attribs.c (remove_attributes_matching): New function.
26088 (affects_type_identity_attributes): Likewise.
26090 2021-04-15 Jakub Jelinek <jakub@redhat.com>
26093 * config/aarch64/aarch64.md (*<LOGICAL:optab>_<SHIFT:optab><mode>3):
26094 Add combine splitters for *<LOGICAL:optab>_ashl<mode>3 with
26095 ZERO_EXTEND, SIGN_EXTEND or AND.
26097 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
26099 PR rtl-optimization/99929
26100 * rtl.h (same_vector_encodings_p): New function.
26101 * cse.c (exp_equiv_p): Check that CONST_VECTORs have the same encoding.
26102 * cselib.c (rtx_equal_for_cselib_1): Likewise.
26103 * jump.c (rtx_renumbered_equal_p): Likewise.
26104 * lra-constraints.c (operands_match_p): Likewise.
26105 * reload.c (operands_match_p): Likewise.
26106 * rtl.c (rtx_equal_p_cb, rtx_equal_p): Likewise.
26108 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
26110 * print-rtl.c (rtx_writer::print_rtx_operand_codes_E_and_V): Print
26111 more information about variable-length CONST_VECTORs.
26113 2021-04-14 Vladimir N. Makarov <vmakarov@redhat.com>
26115 PR rtl-optimization/100066
26116 * lra-constraints.c (split_reg): Check paradoxical_subreg_p for
26117 ordered modes when choosing splitting mode for hard reg.
26119 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
26122 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector_sel):
26124 (aarch64_expand_sve_const_vector): Use it for nelts_per_pattern==2.
26126 2021-04-14 Andreas Krebbel <krebbel@linux.ibm.com>
26128 * config/s390/s390-builtins.def (O_M5, O_M12, ...): Add new macros
26129 for mask operand types.
26130 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
26131 (s390_vec_permi_dbl, s390_vpdi): Use the M5 type for the immediate
26133 (s390_vec_msum_u128, s390_vmslg): Use the M12 type for the
26135 * config/s390/s390.c (s390_const_operand_ok): Check the new
26136 operand types and generate a list of valid values.
26138 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
26140 * doc/tm.texi: Regenerate.
26141 * doc/tm.texi.in (D language and ABI): Add @hook for
26142 TARGET_D_REGISTER_OS_TARGET_INFO.
26144 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
26146 * config/aarch64/aarch64-d.c (aarch64_d_handle_target_float_abi): New
26148 (aarch64_d_register_target_info): New function.
26149 * config/aarch64/aarch64-protos.h (aarch64_d_register_target_info):
26151 * config/aarch64/aarch64.h (TARGET_D_REGISTER_CPU_TARGET_INFO):
26153 * config/arm/arm-d.c (arm_d_handle_target_float_abi): New function.
26154 (arm_d_register_target_info): New function.
26155 * config/arm/arm-protos.h (arm_d_register_target_info): Declare.
26156 * config/arm/arm.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
26157 * config/i386/i386-d.c (ix86_d_handle_target_float_abi): New function.
26158 (ix86_d_register_target_info): New function.
26159 * config/i386/i386-protos.h (ix86_d_register_target_info): Declare.
26160 * config/i386/i386.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
26161 * config/mips/mips-d.c (mips_d_handle_target_float_abi): New function.
26162 (mips_d_register_target_info): New function.
26163 * config/mips/mips-protos.h (mips_d_register_target_info): Declare.
26164 * config/mips/mips.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
26165 * config/pa/pa-d.c (pa_d_handle_target_float_abi): New function.
26166 (pa_d_register_target_info): New function.
26167 * config/pa/pa-protos.h (pa_d_register_target_info): Declare.
26168 * config/pa/pa.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
26169 * config/riscv/riscv-d.c (riscv_d_handle_target_float_abi): New
26171 (riscv_d_register_target_info): New function.
26172 * config/riscv/riscv-protos.h (riscv_d_register_target_info): Declare.
26173 * config/riscv/riscv.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
26174 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_float_abi): New
26176 (rs6000_d_register_target_info): New function.
26177 * config/rs6000/rs6000-protos.h (rs6000_d_register_target_info):
26179 * config/rs6000/rs6000.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
26180 * config/s390/s390-d.c (s390_d_handle_target_float_abi): New function.
26181 (s390_d_register_target_info): New function.
26182 * config/s390/s390-protos.h (s390_d_register_target_info): Declare.
26183 * config/s390/s390.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
26184 * config/sparc/sparc-d.c (sparc_d_handle_target_float_abi): New
26186 (sparc_d_register_target_info): New function.
26187 * config/sparc/sparc-protos.h (sparc_d_register_target_info): Declare.
26188 * config/sparc/sparc.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
26189 * doc/tm.texi: Regenerate.
26190 * doc/tm.texi.in (D language and ABI): Add @hook for
26191 TARGET_D_REGISTER_CPU_TARGET_INFO.
26193 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
26195 * config/i386/i386-d.c (ix86_d_has_stdcall_convention): New function.
26196 * config/i386/i386-protos.h (ix86_d_has_stdcall_convention): Declare.
26197 * config/i386/i386.h (TARGET_D_HAS_STDCALL_CONVENTION): Define.
26198 * doc/tm.texi: Regenerate.
26199 * doc/tm.texi.in (D language and ABI): Add @hook for
26200 TARGET_D_HAS_STDCALL_CONVENTION.
26202 2021-04-14 Richard Biener <rguenther@suse.de>
26204 * tree-cfg.c (verify_gimple_assign_ternary): Verify that
26205 VEC_COND_EXPRs have a gimple_val condition.
26206 * tree-ssa-propagate.c (valid_gimple_rhs_p): VEC_COND_EXPR
26207 can no longer have a GENERIC condition.
26209 2021-04-14 Richard Earnshaw <rearnsha@arm.com>
26212 * config/arm/arm.c (arm_configure_build_target): Strip isa_all_fpbits
26213 from the isa_delta when -mfpu has been used.
26214 (arm_options_perform_arch_sanity_checks): It's the architecture that
26215 lacks an FPU not the processor.
26217 2021-04-13 Richard Biener <rguenther@suse.de>
26219 PR tree-optimization/100053
26220 * tree-ssa-sccvn.c (vn_nary_op_get_predicated_value): Do
26221 not use optimistic dominance queries for backedges to validate
26223 (dominated_by_p_w_unex): Add parameter to ignore executable
26224 state on backedges.
26225 (rpo_elim::eliminate_avail): Adjust.
26227 2021-04-13 Jakub Jelinek <jakub@redhat.com>
26230 * config/aarch64/aarch64.md (*aarch64_bfxil<mode>_extr,
26231 *aarch64_bfxilsi_extrdi): New define_insn patterns.
26233 2021-04-13 Jakub Jelinek <jakub@redhat.com>
26236 * simplify-rtx.c (simplify_immed_subreg): For MODE_COMPOSITE_P
26237 outermode, return NULL if the result doesn't encode back to the
26238 original byte sequence.
26239 (simplify_gen_subreg): Don't create SUBREGs from constants to
26240 MODE_COMPOSITE_P outermode.
26242 2021-04-12 Jakub Jelinek <jakub@redhat.com>
26244 PR rtl-optimization/99905
26245 * combine.c (expand_compound_operation): If pos + len > modewidth,
26246 perform the right shift by pos in inner_mode and then convert to mode,
26247 instead of trying to simplify a shift of rtx with inner_mode by pos
26248 as if it was a shift in mode.
26250 2021-04-12 Jakub Jelinek <jakub@redhat.com>
26253 * combine.c (simplify_and_const_int_1): Don't optimize varop
26254 away if it has side-effects.
26256 2021-04-12 Martin Liska <mliska@suse.cz>
26258 * doc/extend.texi: Escape @smallexample content.
26260 2021-04-12 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
26262 * config/s390/s390.md ("*movdi_31", "*movdi_64"): Add
26263 alternative in order to load a DFP zero.
26265 2021-04-12 Martin Liska <mliska@suse.cz>
26267 * doc/extend.texi: Be more precise in documentation
26268 of symver attribute.
26270 2021-04-12 Martin Liska <mliska@suse.cz>
26273 * gimplify.c (gimplify_expr): Right now, we unpoison all
26274 variables before a goto <dest>. We should not do it if we are
26277 2021-04-12 Cui,Lili <lili.cui@intel.com>
26279 * common/config/i386/cpuinfo.h (get_intel_cpu): Handle
26281 * common/config/i386/i386-common.c (processor_names): Add
26283 (processor_alias_table): Add rocketlake.
26284 * common/config/i386/i386-cpuinfo.h (processor_subtypes): Add
26285 INTEL_COREI7_ROCKETLAKE.
26286 * config.gcc: Add -march=rocketlake.
26287 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
26289 * config/i386/i386-options.c (m_ROCKETLAKE) : Define.
26290 (processor_cost_table): Add rocketlake cost.
26291 * config/i386/i386.h (ix86_size_cost) : Define
26293 (processor_type) : Add PROCESSOR_ROCKETLAKE.
26294 (PTA_ROCKETLAKE): Ditto.
26295 * doc/extend.texi: Add rocketlake.
26296 * doc/invoke.texi: Add rocketlake.
26298 2021-04-12 Cui,Lili <lili.cui@intel.com>
26300 * config/i386/i386.h (PTA_ALDERLAKE): Change alderlake ISA list.
26301 * config/i386/i386-options.c (m_CORE_AVX2): Add m_ALDERLAKE.
26302 * common/config/i386/cpuinfo.h (get_intel_cpu): Add AlderLake model.
26303 * doc/invoke.texi: Change alderlake ISA list.
26305 2021-04-11 Hafiz Abid Qadeer <abidh@codesourcery.com>
26307 PR middle-end/98088
26308 * omp-expand.c (expand_oacc_collapse_init): Update condition in
26311 2021-04-10 H.J. Lu <hjl.tools@gmail.com>
26314 * config/i386/serializeintrin.h (_serialize): Defined as macro.
26316 2021-04-10 Jakub Jelinek <jakub@redhat.com>
26319 * expr.c (expand_expr_addr_expr_1): Test is_global_var rather than
26320 just TREE_STATIC on COMPOUND_LITERAL_EXPR_DECLs.
26322 2021-04-10 Jakub Jelinek <jakub@redhat.com>
26324 PR middle-end/99989
26325 * gimple-ssa-warn-alloca.c
26326 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit to
26327 0 with integer precision unconditionally.
26329 2021-04-10 Jakub Jelinek <jakub@redhat.com>
26331 PR rtl-optimization/98601
26332 * rtlanal.c (rtx_addr_can_trap_p_1): Allow in assert unknown size
26333 not just for BLKmode, but also for VOIDmode. For STRICT_ALIGNMENT
26334 unaligned_mems handle VOIDmode like BLKmode.
26336 2021-04-10 Jan Hubicka <hubicka@ucw.cz>
26339 * tree.c (free_lang_data_in_decl): Do not release body of
26340 declare_variant_alt.
26342 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
26344 * config/aarch64/aarch64.c (aarch64_option_restore): If the
26345 architecture was specified explicitly and the tuning wasn't,
26346 tune for the architecture rather than the configured default CPU.
26348 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
26350 * config/aarch64/aarch64.md (tlsdesc_small_sve_<mode>): Use X30
26351 as the temporary register.
26353 2021-04-09 Martin Liska <mliska@suse.cz>
26355 * doc/extend.texi: Move non-target attributes on the top level.
26357 2021-04-09 Martin Liska <mliska@suse.cz>
26359 * doc/invoke.texi: Document minimum and maximum value of the
26360 argument for both supported compression algorithms.
26362 2021-04-08 David Edelsohn <dje.gcc@gmail.com>
26364 * config/rs6000/rs6000.c (rs6000_xcoff_select_section): Select
26365 TLS BSS before TLS data.
26366 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .comm.
26368 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
26370 * doc/sourcebuild.texi (stdint_types_mbig_endian): Document.
26372 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
26374 * match.pd: Extend vec_cond folds to handle shifts.
26376 2021-04-08 Maciej W. Rozycki <macro@orcam.me.uk>
26378 * config/vax/vax.md: Fix comment for `*bit<mode>' pattern's
26381 2021-04-08 Alex Coplan <alex.coplan@arm.com>
26384 * config/arm/iterators.md (MVE_vecs): New.
26385 (V_elem): Also handle V2DF.
26386 * config/arm/mve.md (*mve_mov<mode>): Rename to ...
26387 (*mve_vdup<mode>): ... this. Remove second alternative since
26388 vec_duplicate of const_int is not canonical RTL, and we don't
26389 want to match symbol_refs.
26390 (*mve_vec_duplicate<mode>): Delete (pattern is redundant).
26392 2021-04-08 Xionghu Luo <luoxhu@linux.ibm.com>
26394 * fold-const.c (fold_single_bit_test): Fix typo.
26395 * print-rtl.c (print_rtx_insn_vec): Call print_rtl_single
26398 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
26400 PR tree-optimization/97513
26401 * tree-vect-slp.c (vect_add_slp_permutation): New function,
26403 (vectorizable_slp_permutation): ...here. Detect cases in which
26404 all VEC_PERM_EXPRs are guaranteed to have the same stepped
26405 permute vector and only generate one permute vector for that case.
26406 Extend that case to handle variable-length vectors.
26408 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
26410 PR tree-optimization/99873
26411 * tree-vect-slp.c (vect_slp_prefer_store_lanes_p): New function.
26412 (vect_build_slp_instance): Don't split store groups that could
26413 use IFN_STORE_LANES.
26415 2021-04-07 Jakub Jelinek <jakub@redhat.com>
26418 * varasm.c (output_constant_pool_contents): Don't strip name encoding
26419 from XSTR (desc->sym, 0) or from label before passing those to
26422 2021-04-07 Richard Biener <rguenther@suse.de>
26424 PR tree-optimization/99954
26425 * tree-loop-distribution.c: Include tree-affine.h.
26426 (generate_memcpy_builtin): Try using tree-affine to prove
26428 (loop_distribution::classify_builtin_ldst): Always classify
26431 2021-04-07 Richard Biener <rguenther@suse.de>
26433 PR tree-optimization/99947
26434 * tree-vect-loop.c (vectorizable_induction): Pre-allocate
26435 steps vector to avoid pushing elements from the reallocated
26438 2021-04-07 Richard Biener <rguenther@suse.de>
26440 * tree-ssa-sccvn.h (print_vn_reference_ops): Declare.
26441 * tree-ssa-pre.c (print_pre_expr): Factor out VN reference operand
26443 * tree-ssa-sccvn.c (print_vn_reference_ops): ... into this new
26445 (debug_vn_reference_ops): New.
26447 2021-04-07 Bin Cheng <bin.cheng@linux.alibaba.com>
26449 PR tree-optimization/98736
26450 * tree-loop-distribution.c
26451 * (loop_distribution::bb_top_order_init):
26452 Compute RPO with programing order preserved by calling function
26453 rev_post_order_and_mark_dfs_back_seme.
26455 2021-04-06 Vladimir N. Makarov <vmakarov@redhat.com>
26458 * lra-constraints.c (split_reg): Don't check paradoxical_subreg_p.
26459 * lra-lives.c (clear_sparseset_regnos, regnos_in_sparseset_p): New
26461 (process_bb_lives): Don't update biggest mode of hard reg for
26462 implicit in multi-register group. Use the new functions for
26463 updating dead_set and unused_set by register notes.
26465 2021-04-06 Xianmiao Qu <xianmiao_qu@c-sky.com>
26467 * config/csky/csky_pipeline_ck802.md : Use insn reservation name
26470 2021-04-06 H.J. Lu <hjl.tools@gmail.com>
26472 * config/i386/x86-tune-costs.h (skylake_memcpy): Updated.
26473 (skylake_memset): Likewise.
26474 (skylake_cost): Change CLEAR_RATIO to 17.
26475 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
26476 Replace m_CANNONLAKE, m_ICELAKE_CLIENT, m_ICELAKE_SERVER,
26477 m_TIGERLAKE and m_SAPPHIRERAPIDS with m_SKYLAKE and m_CORE_AVX512.
26479 2021-04-06 Richard Biener <rguenther@suse.de>
26481 PR tree-optimization/99880
26482 * tree-vect-loop.c (maybe_set_vectorized_backedge_value): Only
26483 set vectorized defs of relevant PHIs.
26485 2021-04-06 Richard Biener <rguenther@suse.de>
26487 PR tree-optimization/99924
26488 * tree-vect-slp.c (vect_bb_partition_graph_r): Do not mark
26489 nodes w/o scalar stmts as visited.
26491 2021-04-06 Alex Coplan <alex.coplan@arm.com>
26494 * config/arm/arm.c (arm_libcall_uses_aapcs_base): Also use base
26495 PCS for [su]fix_optab.
26497 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
26499 * config/darwin.c (machopic_legitimize_pic_address): Check
26500 that the current pic register is one of the hard reg set
26501 before setting liveness.
26503 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
26505 * config/darwin.c (machopic_legitimize_pic_address): Fix
26506 whitespace, remove unused code.
26508 2021-04-03 Jakub Jelinek <jakub@redhat.com>
26510 PR tree-optimization/99882
26511 * gimple-ssa-store-merging.c (bswap_view_convert): Handle val with
26514 2021-04-03 Jakub Jelinek <jakub@redhat.com>
26516 PR rtl-optimization/99863
26517 * dse.c (replace_read): Drop regs_live argument. Instead of
26518 regs_live, use store_insn->fixed_regs_live if non-NULL,
26519 otherwise punt if insns sequence clobbers or sets any hard
26522 2021-04-03 Jakub Jelinek <jakub@redhat.com>
26525 * targhooks.h (default_print_patchable_function_entry_1): Declare.
26526 * targhooks.c (default_print_patchable_function_entry_1): New function,
26527 copied from default_print_patchable_function_entry with an added flags
26529 (default_print_patchable_function_entry): Rewritten into a small
26530 wrapper around default_print_patchable_function_entry_1.
26531 * config/rs6000/rs6000.c (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY):
26533 (rs6000_print_patchable_function_entry): New function.
26535 2021-04-02 Eric Botcazou <ebotcazou@adacore.com>
26537 * doc/invoke.texi (fdelete-dead-exceptions): Minor tweak.
26539 2021-04-01 Jason Merrill <jason@redhat.com>
26542 * common.opt: Document v15 and v16.
26544 2021-04-01 Richard Biener <rguenther@suse.de>
26546 PR tree-optimization/99863
26547 * gimplify.c (gimplify_init_constructor): Recompute vector
26550 2021-04-01 Jakub Jelinek <jakub@redhat.com>
26552 * doc/extend.texi (symver attribute): Fix up syntax errors
26555 2021-04-01 Jakub Jelinek <jakub@redhat.com>
26557 PR tree-optimization/96573
26558 * gimple-ssa-store-merging.c (init_symbolic_number): Handle
26559 also pointer types.
26561 2021-04-01 Richard Biener <rguenther@suse.de>
26563 PR tree-optimization/99856
26564 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Promote
26565 precision to vector element precision.
26567 2021-04-01 Martin Jambor <mjambor@suse.cz>
26569 PR tree-optimization/97009
26570 * tree-sra.c (access_or_its_child_written): New function.
26571 (propagate_subaccesses_from_rhs): Use it instead of a simple grp_write
26574 2021-03-31 Jan Hubicka <hubicka@ucw.cz>
26577 * cif-code.def (USES_COMDAT_LOCAL): Make CIF_FINAL_NORMAL.
26579 2021-03-31 Pat Haugen <pthaugen@linux.ibm.com>
26582 * config/rs6000/altivec.md (xxspltiw_v4si, xxspltiw_v4sf_inst,
26583 xxspltidp_v2df_inst, xxsplti32dx_v4si_inst, xxsplti32dx_v4sf_inst,
26584 xxblend_<mode>, xxpermx_inst, xxeval): Mark prefixed.
26585 * config/rs6000/mma.md (mma_<vvi4i4i8>, mma_<avvi4i4i8>,
26586 mma_<vvi4i4i2>, mma_<avvi4i4i2>, mma_<vvi4i4>, mma_<avvi4i4>,
26587 mma_<pvi4i2>, mma_<apvi4i2>, mma_<vvi4i4i4>, mma_<avvi4i4i4>):
26589 * config/rs6000/rs6000.c (rs6000_final_prescan_insn): Adjust test.
26590 * config/rs6000/rs6000.md (define_attr "maybe_prefixed"): New.
26591 (define_attr "prefixed"): Update initializer.
26593 2021-03-31 Jakub Jelinek <jakub@redhat.com>
26596 * dwarf2out.c (debug_ranges_dwo_section): New variable.
26597 (DW_RANGES_IDX_SKELETON): Define.
26598 (struct dw_ranges): Add begin_entry and end_entry members.
26599 (DEBUG_DWO_RNGLISTS_SECTION): Define.
26600 (add_ranges_num): Adjust r initializer for addition of *_entry
26602 (add_ranges_by_labels): For -gsplit-dwarf and force_direct,
26603 set idx to DW_RANGES_IDX_SKELETON.
26604 (use_distinct_base_address_for_range): New function.
26605 (index_rnglists): Don't set r->idx if it is equal to
26606 DW_RANGES_IDX_SKELETON. Initialize r->begin_entry and
26607 r->end_entry for -gsplit-dwarf if those will be needed by
26609 (output_rnglists): Add DWO argument. If true, switch to
26610 debug_ranges_dwo_section rather than debug_ranges_section.
26611 Adjust l1/l2 label indexes. Only output the offset table when
26612 dwo is true and don't include in there the skeleton range
26613 entry if present. For -gsplit-dwarf, skip ranges that belong
26614 to the other rnglists section. Change return type from void
26615 to bool and return true if there are any range entries for
26616 the other section. For dwarf_split_debug_info use
26617 DW_RLE_startx_endx, DW_RLE_startx_length and DW_RLE_base_addressx
26618 entries instead of DW_RLE_start_end, DW_RLE_start_length and
26619 DW_RLE_base_address. Use use_distinct_base_address_for_range.
26620 (init_sections_and_labels): Initialize debug_ranges_dwo_section
26621 if -gsplit-dwarf and DWARF >= 5. Adjust ranges_section_label
26622 and range_base_label indexes.
26623 (dwarf2out_finish): Call index_rnglists earlier before finalizing
26624 .debug_addr. Never emit DW_AT_rnglists_base attribute. For
26625 -gsplit-dwarf and DWARF >= 5 call output_rnglists up to twice
26626 with different dwo arguments.
26627 (dwarf2out_c_finalize): Clear debug_ranges_dwo_section.
26629 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
26631 PR tree-optimization/98268
26632 * gimple-fold.c (maybe_canonicalize_mem_ref_addr): Call
26633 recompute_tree_invariant_for_addr_expr after successfully
26634 folding a TARGET_MEM_REF that occurs inside an ADDR_EXPR.
26636 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
26638 PR tree-optimization/99726
26639 * tree-data-ref.c (create_intersect_range_checks_index): Bail
26640 out if there is more than one access function SCEV for the loop
26643 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
26645 PR rtl-optimization/97141
26646 PR rtl-optimization/98726
26647 * emit-rtl.c (valid_for_const_vector_p): Return true for
26649 * rtx-vector-builder.h (rtx_vector_builder::step): Return a
26650 poly_wide_int instead of a wide_int.
26651 (rtx_vector_builder::apply_set): Take a poly_wide_int instead
26653 * rtx-vector-builder.c (rtx_vector_builder::apply_set): Likewise.
26654 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Return
26655 false for CONST_VECTORs that cannot be forced to memory.
26656 * config/aarch64/aarch64-simd.md (mov<mode>): If a CONST_VECTOR
26657 is too complex to force to memory, build it up from individual
26660 2021-03-31 Jan Hubicka <jh@suse.cz>
26663 * cgraph.c (cgraph_node::release_body): Fix overactive check.
26665 2021-03-31 Christophe Lyon <christophe.lyon@linaro.org>
26668 * config/arm/vec-common.md (mul<mode>3): Disable on iwMMXT, expect
26671 2021-03-31 H.J. Lu <hjl.tools@gmail.com>
26673 * config/i386/i386-expand.c (expand_set_or_cpymem_via_rep):
26674 For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, don't convert QImode
26676 (decide_alg): For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, use
26677 "rep movsb/stosb" only for known sizes.
26678 * config/i386/i386-options.c (processor_cost_table): Use Ice
26679 Lake cost for Cannon Lake, Ice Lake, Tiger Lake, Sapphire
26680 Rapids and Alder Lake.
26681 * config/i386/i386.h (TARGET_PREFER_KNOWN_REP_MOVSB_STOSB): New.
26682 * config/i386/x86-tune-costs.h (icelake_memcpy): New.
26683 (icelake_memset): Likewise.
26684 (icelake_cost): Likewise.
26685 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
26688 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
26691 * config/aarch64/aarch64.c
26692 (aarch64_vectorize_preferred_vector_alignment): Query the size
26693 of the provided SVE vector; do not assume that all SVE vectors
26694 have the same size.
26696 2021-03-31 Jan Hubicka <jh@suse.cz>
26699 * cgraph.c (cgraph_node::release_body): Remove all callers and
26701 * cgraphclones.c (cgraph_node::materialize_clone): Do not do it here.
26702 * cgraphunit.c (cgraph_node::expand): And here.
26704 2021-03-31 Martin Liska <mliska@suse.cz>
26706 * ipa-modref.c (analyze_ssa_name_flags): Fix coding style
26707 and one negated condition.
26709 2021-03-31 Jakub Jelinek <jakub@redhat.com>
26710 Richard Sandiford <richard.sandiford@arm.com>
26713 * config/aarch64/aarch64.md (*add<mode>3_poly_1): Swap Uai and Uav
26714 constraints on operands[2] and similarly 0 and rk constraints
26715 on operands[1] corresponding to that.
26717 2021-03-31 Jakub Jelinek <jakub@redhat.com>
26720 * configure.ac (HAVE_LD_BROKEN_PE_DWARF5): New AC_DEFINE if PECOFF
26721 linker doesn't support DWARF sections new in DWARF5.
26722 * config/i386/i386-options.c (ix86_option_override_internal): Default
26723 to dwarf_version 4 if HAVE_LD_BROKEN_PE_DWARF5 for TARGET_PECOFF
26725 * config.in: Regenerated.
26726 * configure: Regenerated.
26728 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26731 * config/aarch64/aarch64.c (aarch64_analyze_loop_vinfo): Check for
26732 available issue_info before using it.
26734 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26737 * config/aarch64/aarch64.md (sub<mode>3_compare1_imm): Do not allow zero
26740 2021-03-30 Xionghu Luo <luoxhu@linux.ibm.com>
26743 * config/rs6000/altivec.md (altivec_lvsl_reg): Change to ...
26744 (altivec_lvsl_reg_<mode>): ... this.
26745 (altivec_lvsr_reg): Change to ...
26746 (altivec_lvsr_reg_<mode>): ... this.
26747 * config/rs6000/predicates.md (vec_set_index_operand): New.
26748 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
26749 Enable 32bit variable vec_insert for all TARGET_VSX.
26750 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
26751 Enable 32bit variable vec_insert for p9 and above.
26752 (rs6000_expand_vector_set_var_p8): Rename to ...
26753 (rs6000_expand_vector_set_var_p7): ... this.
26754 (rs6000_expand_vector_set): Use TARGET_VSX and adjust assert
26756 * config/rs6000/vector.md (vec_set<mode>): Use vec_set_index_operand.
26757 * config/rs6000/vsx.md (xl_len_r): Use gen_altivec_lvsl_reg_di and
26758 gen_altivec_lvsr_reg_di.
26760 2021-03-30 H.J. Lu <hjl.tools@gmail.com>
26763 * config/i386/ia32intrin.h (__rdtsc): Defined as macro.
26764 (__rdtscp): Likewise.
26766 2021-03-30 Tamar Christina <tamar.christina@arm.com>
26768 PR tree-optimization/99825
26769 * tree-vect-slp-patterns.c (vect_check_evenodd_blend):
26770 Reject non-mult 2 lanes.
26772 2021-03-30 Richard Earnshaw <rearnsha@arm.com>
26775 * config/arm/arm.c (arm_file_start): Fix emission of
26776 Tag_ABI_VFP_args attribute.
26778 2021-03-30 Richard Biener <rguenther@suse.de>
26780 PR tree-optimization/99824
26781 * stor-layout.c (set_min_and_max_values_for_integral_type):
26782 Assert the precision is within the bounds of
26783 WIDE_INT_MAX_PRECISION.
26784 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
26785 the outermost component ref only to lower the access size
26786 and initialize that from the access type.
26788 2021-03-30 Richard Sandiford <richard.sandiford@arm.com>
26791 * config/aarch64/aarch64.md (mov<mode>): Pass multi-instruction
26792 CONST_INTs to aarch64_expand_mov_immediate when called after RA.
26794 2021-03-30 Mihailo Stojanovic <mihailo.stojanovic@typhoon-hil.com>
26796 * config/aarch64/aarch64.md
26797 (<optab>_trunc<fcvt_target><GPI:mode>2): Set the "arch"
26798 attribute to disambiguate between SIMD and FP variants of the
26801 2021-03-29 Jan Hubicka <hubicka@ucw.cz>
26803 * ipa-modref.c (merge_call_lhs_flags): Correct handling of deref.
26804 (analyze_ssa_name_flags): Fix typo in comment.
26806 2021-03-29 Alex Coplan <alex.coplan@arm.com>
26809 * config/aarch64/aarch64-sve-builtins.cc
26810 (function_builder::add_function): Add placeholder_p argument, use
26811 placeholder decls if this is set.
26812 (function_builder::add_unique_function): Instead of conditionally adding
26813 direct overloads, unconditionally add either a direct overload or a
26815 (function_builder::add_overloaded_function): Set placeholder_p if we're
26816 using C++ overloads. Use the obstack for string storage instead
26817 of relying on the tree nodes.
26818 (function_builder::add_overloaded_functions): Don't return early for
26819 m_direct_overloads: we need to add placeholders.
26820 * config/aarch64/aarch64-sve-builtins.h
26821 (function_builder::add_function): Add placeholder_p argument.
26823 2021-03-29 Richard Biener <rguenther@suse.de>
26825 PR tree-optimization/99807
26826 * tree-vect-slp.c (vect_slp_analyze_node_operations_1): Move
26827 assert below VEC_PERM handling.
26829 2021-03-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26832 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>): Use
26833 aarch64_simd_or_scalar_imm_zero to match zeroes. Remove pattern
26834 matching const_int 0.
26835 (move_lo_quad_internal_be_<mode>): Likewise.
26836 (move_lo_quad_<mode>): Update for the above.
26837 * config/aarch64/iterators.md (VQ_2E): Delete.
26839 2021-03-29 Jakub Jelinek <jakub@redhat.com>
26841 PR tree-optimization/99777
26842 * fold-const.c (extract_muldiv_1): For conversions, punt on casts from
26843 types other than scalar integral types.
26845 2021-03-28 David Edelsohn <dje.gcc@gmail.com>
26847 * config/rs6000/rs6000.c (rs6000_output_dwarf_dtprel): Do not add
26848 XCOFF TLS reloc decorations.
26850 2021-03-28 Gerald Pfeifer <gerald@pfeifer.com>
26852 * doc/analyzer.texi (Analyzer Internals): Update link to
26853 "A Memory Model for Static Analysis of C Programs".
26855 2021-03-26 David Edelsohn <dje.gcc@gmail.com>
26857 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Call function.
26858 * config/rs6000/rs6000-protos.h (rs6000_special_adjust_field_align):
26860 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align): New.
26861 (rs6000_special_round_type_align): Recursively check innermost first
26864 2021-03-26 Jakub Jelinek <jakub@redhat.com>
26867 * dwarf2out.h (struct dw_fde_node): Add rule18 member.
26868 * dwarf2cfi.c (dwarf2out_frame_debug_expr): When handling (set hfp sp)
26869 assignment with drap_reg active, queue reg save for hfp with offset 0
26870 and flush queued reg saves. When handling a push with rule18,
26871 defer queueing reg save for hfp and just assert the offset is 0.
26872 (scan_trace): Assert that fde->rule18 is false.
26874 2021-03-26 Vladimir Makarov <vmakarov@redhat.com>
26877 * ira-costs.c (record_reg_classes): Put case with
26878 CT_RELAXED_MEMORY adjacent to one with CT_MEMORY.
26879 * ira.c (ira_setup_alts): Ditto.
26880 * lra-constraints.c (process_alt_operands): Ditto.
26881 * recog.c (asm_operand_ok): Ditto.
26882 * reload.c (find_reloads): Ditto.
26884 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
26886 * config/aarch64/aarch64-protos.h
26887 (cpu_addrcost_table::post_modify_ld3_st3): New member variable.
26888 (cpu_addrcost_table::post_modify_ld4_st4): Likewise.
26889 * config/aarch64/aarch64.c (generic_addrcost_table): Update
26890 accordingly, using the same costs as for post_modify.
26891 (exynosm1_addrcost_table, xgene1_addrcost_table): Likewise.
26892 (thunderx2t99_addrcost_table, thunderx3t110_addrcost_table):
26893 (tsv110_addrcost_table, qdf24xx_addrcost_table): Likewise.
26894 (a64fx_addrcost_table): Likewise.
26895 (neoversev1_addrcost_table): New.
26896 (neoversev1_tunings): Use neoversev1_addrcost_table.
26897 (aarch64_address_cost): Use the new post_modify costs for CImode
26900 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
26902 * config/aarch64/aarch64.opt
26903 (-param=aarch64-loop-vect-issue-rate-niters=): New parameter.
26904 * doc/invoke.texi: Document it.
26905 * config/aarch64/aarch64-protos.h (aarch64_base_vec_issue_info)
26906 (aarch64_scalar_vec_issue_info, aarch64_simd_vec_issue_info)
26907 (aarch64_advsimd_vec_issue_info, aarch64_sve_vec_issue_info)
26908 (aarch64_vec_issue_info): New structures.
26909 (cpu_vector_cost): Write comments above the variables rather
26911 (cpu_vector_cost::issue_info): New member variable.
26912 * config/aarch64/aarch64.c: Include gimple-pretty-print.h
26913 and tree-ssa-loop-niter.h.
26914 (generic_vector_cost, a64fx_vector_cost, qdf24xx_vector_cost)
26915 (thunderx_vector_cost, tsv110_vector_cost, cortexa57_vector_cost)
26916 (exynosm1_vector_cost, xgene1_vector_cost, thunderx2t99_vector_cost)
26917 (thunderx3t110_vector_cost): Initialize issue_info to null.
26918 (neoversev1_scalar_issue_info, neoversev1_advsimd_issue_info)
26919 (neoversev1_sve_issue_info, neoversev1_vec_issue_info): New structures.
26920 (neoversev1_vector_cost): Use them.
26921 (aarch64_vec_op_count, aarch64_sve_op_count): New structures.
26922 (aarch64_vector_costs::saw_sve_only_op): New member variable.
26923 (aarch64_vector_costs::num_vector_iterations): Likewise.
26924 (aarch64_vector_costs::scalar_ops): Likewise.
26925 (aarch64_vector_costs::advsimd_ops): Likewise.
26926 (aarch64_vector_costs::sve_ops): Likewise.
26927 (aarch64_vector_costs::seen_loads): Likewise.
26928 (aarch64_simd_vec_costs_for_flags): New function.
26929 (aarch64_analyze_loop_vinfo): Initialize num_vector_iterations.
26930 Count the number of predicate operations required by SVE WHILE
26932 (aarch64_comparison_type, aarch64_multiply_add_p): New functions.
26933 (aarch64_sve_only_stmt_p, aarch64_in_loop_reduction_latency): Likewise.
26934 (aarch64_count_ops): Likewise.
26935 (aarch64_add_stmt_cost): Record whether see an SVE operation
26936 that cannot currently be implementing using Advanced SIMD.
26937 Record issue information about the scalar, Advanced SIMD
26938 and (where relevant) SVE versions of a loop.
26939 (aarch64_vec_op_count::dump): New function.
26940 (aarch64_sve_op_count::dump): Likewise.
26941 (aarch64_estimate_min_cycles_per_iter): Likewise.
26942 (aarch64_adjust_body_cost): If issue information is available,
26943 try to compare the issue rates of the various loop implementations
26944 and increase or decrease the vector body cost accordingly.
26946 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
26948 * config/aarch64/aarch64.c (aarch64_detect_vector_stmt_subtype):
26949 Assume a zero cost for induction phis.
26951 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
26953 * config/aarch64/aarch64.c (aarch64_embedded_comparison_type): New
26955 (aarch64_adjust_stmt_cost): Add the costs of embedded scalar and
26956 vector comparisons.
26958 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
26960 * config/aarch64/aarch64.c (aarch64_detect_scalar_stmt_subtype):
26962 (aarch64_add_stmt_cost): Call it.
26964 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
26966 * config/aarch64/aarch64-tuning-flags.def (matched_vector_throughput):
26967 New tuning parameter.
26968 * config/aarch64/aarch64.c (neoversev1_tunings): Use it.
26969 (aarch64_estimated_sve_vq): New function.
26970 (aarch64_vector_costs::analyzed_vinfo): New member variable.
26971 (aarch64_vector_costs::is_loop): Likewise.
26972 (aarch64_vector_costs::unrolled_advsimd_niters): Likewise.
26973 (aarch64_vector_costs::unrolled_advsimd_stmts): Likewise.
26974 (aarch64_record_potential_advsimd_unrolling): New function.
26975 (aarch64_analyze_loop_vinfo, aarch64_analyze_bb_vinfo): Likewise.
26976 (aarch64_add_stmt_cost): Call aarch64_analyze_loop_vinfo or
26977 aarch64_analyze_bb_vinfo on the first use of a costs structure.
26978 Detect whether we're vectorizing a loop for SVE that might be
26979 completely unrolled if it used Advanced SIMD instead.
26980 (aarch64_adjust_body_cost_for_latency): New function.
26981 (aarch64_finish_cost): Call it.
26983 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
26985 * config/aarch64/aarch64.c (aarch64_vector_costs): New structure.
26986 (aarch64_init_cost): New function.
26987 (aarch64_add_stmt_cost): Use aarch64_vector_costs instead of
26988 the default unsigned[3].
26989 (aarch64_finish_cost, aarch64_destroy_cost_data): New functions.
26990 (TARGET_VECTORIZE_INIT_COST): Override.
26991 (TARGET_VECTORIZE_FINISH_COST): Likewise.
26992 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
26994 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
26996 * config/aarch64/aarch64.c (neoversev1_advsimd_vector_cost)
26997 (neoversev1_sve_vector_cost): New cost structures.
26998 (neoversev1_vector_cost): Likewise.
26999 (neoversev1_tunings): Use them. Enable use_new_vector_costs.
27001 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
27003 * config/aarch64/aarch64-protos.h
27004 (sve_vec_cost::scatter_store_elt_cost): New member variable.
27005 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
27006 accordingly, taking the cost from the cost of a scalar_store.
27007 (a64fx_sve_vector_cost): Likewise.
27008 (aarch64_detect_vector_stmt_subtype): Detect scatter stores.
27010 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
27012 * config/aarch64/aarch64-protos.h
27013 (simd_vec_cost::store_elt_extra_cost): New member variable.
27014 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
27015 accordingly, using the vec_to_scalar cost for the new field.
27016 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
27017 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
27018 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
27019 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
27020 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
27021 (thunderx3t110_advsimd_vector_cost): Likewise.
27022 (aarch64_detect_vector_stmt_subtype): Detect single-element stores.
27024 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
27026 * config/aarch64/aarch64-protos.h (simd_vec_cost::ld2_st2_permute_cost)
27027 (simd_vec_cost::ld3_st3_permute_cost): New member variables.
27028 (simd_vec_cost::ld4_st4_permute_cost): Likewise.
27029 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
27030 accordingly, using zero for the new costs.
27031 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
27032 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
27033 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
27034 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
27035 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
27036 (thunderx3t110_advsimd_vector_cost): Likewise.
27037 (aarch64_ld234_st234_vectors): New function.
27038 (aarch64_adjust_stmt_cost): Likewise.
27039 (aarch64_add_stmt_cost): Call aarch64_adjust_stmt_cost if using
27040 the new vector costs.
27042 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
27044 * config/aarch64/aarch64-protos.h (sve_vec_cost): Turn into a
27045 derived class of simd_vec_cost. Add information about CLAST[AB]
27046 and FADDA instructions.
27047 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
27048 accordingly, using the vec_to_scalar costs for the new fields.
27049 (a64fx_sve_vector_cost): Likewise.
27050 (aarch64_reduc_type): New function.
27051 (aarch64_sve_in_loop_reduction_latency): Likewise.
27052 (aarch64_detect_vector_stmt_subtype): Take a vinfo parameter.
27053 Use aarch64_sve_in_loop_reduction_latency to handle SVE reductions
27054 that occur in the loop body.
27055 (aarch64_add_stmt_cost): Update call accordingly.
27057 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
27059 * config/aarch64/aarch64-tuning-flags.def (use_new_vector_costs):
27061 * config/aarch64/aarch64-protos.h (simd_vec_cost): Put comments
27062 above the fields rather than to the right.
27063 (simd_vec_cost::reduc_i8_cost): New member variable.
27064 (simd_vec_cost::reduc_i16_cost): Likewise.
27065 (simd_vec_cost::reduc_i32_cost): Likewise.
27066 (simd_vec_cost::reduc_i64_cost): Likewise.
27067 (simd_vec_cost::reduc_f16_cost): Likewise.
27068 (simd_vec_cost::reduc_f32_cost): Likewise.
27069 (simd_vec_cost::reduc_f64_cost): Likewise.
27070 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
27071 accordingly, using the vec_to_scalar_cost for the new fields.
27072 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
27073 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
27074 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
27075 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
27076 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
27077 (thunderx3t110_advsimd_vector_cost): Likewise.
27078 (aarch64_use_new_vector_costs_p): New function.
27079 (aarch64_simd_vec_costs): New function, split out from...
27080 (aarch64_builtin_vectorization_cost): ...here.
27081 (aarch64_is_reduction): New function.
27082 (aarch64_detect_vector_stmt_subtype): Likewise.
27083 (aarch64_add_stmt_cost): Call aarch64_detect_vector_stmt_subtype if
27084 using the new vector costs.
27086 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
27089 * tree-emutls.c (get_emutls_init_templ_addr): Mark initializer of weak
27090 TLS declarations as public.
27092 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
27094 * config/aarch64/aarch64-d.c (IN_TARGET_CODE): Define.
27095 * config/arm/arm-d.c (IN_TARGET_CODE): Likewise.
27096 * config/i386/i386-d.c (IN_TARGET_CODE): Likewise.
27097 * config/mips/mips-d.c (IN_TARGET_CODE): Likewise.
27098 * config/pa/pa-d.c (IN_TARGET_CODE): Likewise.
27099 * config/riscv/riscv-d.c (IN_TARGET_CODE): Likewise.
27100 * config/rs6000/rs6000-d.c (IN_TARGET_CODE): Likewise.
27101 * config/s390/s390-d.c (IN_TARGET_CODE): Likewise.
27102 * config/sparc/sparc-d.c (IN_TARGET_CODE): Likewise.
27104 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
27107 * config.gcc (*-*-cygwin*): Add winnt-d.o
27108 (*-*-mingw*): Likewise.
27109 * config/i386/cygwin.h (EXTRA_TARGET_D_OS_VERSIONS): New macro.
27110 * config/i386/mingw32.h (EXTRA_TARGET_D_OS_VERSIONS): Likewise.
27111 * config/i386/t-cygming: Add winnt-d.o.
27112 * config/i386/winnt-d.c: New file.
27114 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
27116 * config/freebsd-d.c: Include memmodel.h.
27118 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
27121 * config.gcc (*-*-openbsd*): Add openbsd-d.o.
27122 * config/t-openbsd: Add openbsd-d.o.
27123 * config/openbsd-d.c: New file.
27125 2021-03-25 Stam Markianos-Wright <stam.markianos-wright@arm.com>
27127 PR tree-optimization/96974
27128 * tree-vect-stmts.c (vect_get_vector_types_for_stmt): Replace assert
27129 with graceful exit.
27131 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
27134 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
27138 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
27139 always_inline in system headers.
27141 2021-03-25 Kewen Lin <linkw@linux.ibm.com>
27143 * tree-vect-loop.c (vect_model_reduction_cost): Init inside_cost.
27145 2021-03-25 Jakub Jelinek <jakub@redhat.com>
27148 * tree-core.h (enum operand_equal_flag): Add OEP_ADDRESS_OF_SAME_FIELD.
27149 * fold-const.c (operand_compare::operand_equal_p): Don't compare
27150 field offsets if OEP_ADDRESS_OF_SAME_FIELD.
27152 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
27156 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
27157 always_inline in system headers.
27159 2021-03-25 Richard Biener <rguenther@suse.de>
27161 PR tree-optimization/99746
27162 * tree-vect-slp-patterns.c (complex_pattern::build): Do not mark
27163 the scalar stmt as patterned. Instead set up required things
27166 2021-03-25 Xionghu Luo <luoxhu@linux.ibm.com>
27168 * config/rs6000/rs6000.c (power8_costs): Change l2 cache
27171 2021-03-24 Martin Liska <mliska@suse.cz>
27174 * common/config/i386/i386-common.c (ARRAY_SIZE): Fix off-by-one
27176 * config/i386/i386-options.c (ix86_option_override_internal):
27177 Add run-time assert.
27179 2021-03-24 Martin Jambor <mjambor@suse.cz>
27182 * ipa-cp.c (initialize_node_lattices): Mark as bottom all
27183 parameters with unknown type.
27184 (ipacp_value_safe_for_type): New function.
27185 (propagate_vals_across_arith_jfunc): Verify that the constant type
27186 can be used for a type of the formal parameter.
27187 (propagate_vals_across_ancestor): Likewise.
27188 (propagate_scalar_across_jump_function): Likewise. Pass the type
27189 also to propagate_vals_across_ancestor.
27191 2021-03-24 Christophe Lyon <christophe.lyon@linaro.org>
27194 * config/arm/mve.md (movmisalign<mode>_mve_store): Use Ux
27196 (movmisalign<mode>_mve_load): Likewise.
27198 2021-03-24 Jakub Jelinek <jakub@redhat.com>
27201 * config/arm/vec-common.md (one_cmpl<mode>2, neg<mode>2,
27202 movmisalign<mode>): Disable expanders for TARGET_REALLY_IWMMXT.
27204 2021-03-24 Alexandre Oliva <oliva@adacore.com>
27206 * doc/sourcebuild.texi (sysconf): New effective target.
27208 2021-03-24 Alexandre Oliva <oliva@adacore.com>
27210 * config/i386/predicates.md (reg_or_const_vec_operand): New.
27211 * config/i386/sse.md (ssse3_pshufbv8qi3): Add an expander for
27212 the now *-prefixed insn_and_split, turn the splitter const vec
27213 into an input for the insn, making it an ignored immediate for
27214 non-split cases, and loaded into the scratch register
27217 2021-03-23 Vladimir N. Makarov <vmakarov@redhat.com>
27220 * config/aarch64/constraints.md (Utq, UOb, UOh, UOw, UOd, UOty):
27221 Use define_relaxed_memory_constraint for them.
27223 2021-03-23 Iain Sandoe <iain@sandoe.co.uk>
27226 * config/host-darwin.c (darwin_gt_pch_use_address): Add a
27227 colon to the diagnostic message.
27229 2021-03-23 Ilya Leoshkevich <iii@linux.ibm.com>
27231 * fwprop.c (fwprop_propagation::fwprop_propagation): Look at
27233 (try_fwprop_subst_note): Use set_info instead of insn_info.
27234 (try_fwprop_subst_pattern): Likewise.
27235 (try_fwprop_subst_notes): Likewise.
27236 (try_fwprop_subst): Likewise.
27237 (forward_propagate_subreg): Likewise.
27238 (forward_propagate_and_simplify): Likewise.
27239 (forward_propagate_into): Likewise.
27240 * rtl-ssa/accesses.h (set_info::single_nondebug_use) New
27242 (set_info::single_nondebug_insn_use): Likewise.
27243 (set_info::single_phi_use): Likewise.
27244 * rtl-ssa/member-fns.inl (set_info::single_nondebug_use) New
27246 (set_info::single_nondebug_insn_use): Likewise.
27247 (set_info::single_phi_use): Likewise.
27249 2021-03-23 Christophe Lyon <christophe.lyon@linaro.org>
27251 * doc/sourcebuild.texi (arm_dsp_ok, arm_dsp): Document.
27253 2021-03-23 Jakub Jelinek <jakub@redhat.com>
27256 * config/aarch64/aarch64.c (aarch64_add_offset): Tell
27257 expand_mult to perform an unsigned rather than a signed
27260 2021-03-23 H.J. Lu <hjl.tools@gmail.com>
27263 * config/i386/cpuid.h (__cpuid): Add __volatile__.
27264 (__cpuid_count): Likewise.
27266 2021-03-23 Richard Biener <rguenther@suse.de>
27268 PR tree-optimization/99721
27269 * tree-vect-slp.c (vect_slp_analyze_node_operations):
27270 Make sure we can schedule the node.
27272 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
27274 * config/riscv/riscv.c (riscv_subword): Take endianness into
27275 account when calculating the byte offset.
27277 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
27279 * config/riscv/predicates.md (subreg_lowpart_operator): New predicate
27280 * config/riscv/riscv.md (*addsi3_extended2, *subsi3_extended2)
27281 (*negsi2_extended2, *mulsi3_extended2, *<optab>si3_mask)
27282 (*<optab>si3_mask_1, *<optab>di3_mask, *<optab>di3_mask_1)
27283 (*<optab>si3_extend_mask, *<optab>si3_extend_mask_1): Use
27284 new predicate "subreg_lowpart_operator"
27286 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
27288 * config/riscv/riscv.c (riscv_swap_instruction): New function
27289 to byteswap an SImode rtx containing an instruction.
27290 (riscv_trampoline_init): Byteswap the generated instructions
27293 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
27295 * common/config/riscv/riscv-common.c
27296 (TARGET_DEFAULT_TARGET_FLAGS): Set default endianness.
27297 * config.gcc (riscv32be-*, riscv64be-*): Set
27298 TARGET_BIG_ENDIAN_DEFAULT to 1.
27299 * config/riscv/elf.h (LINK_SPEC): Change -melf* value
27300 depending on default endianness.
27301 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
27302 * config/riscv/linux.h (LINK_SPEC): Likewise.
27303 * config/riscv/riscv.c (TARGET_DEFAULT_TARGET_FLAGS): Set
27304 default endianness.
27305 * config/riscv/riscv.h (DEFAULT_ENDIAN_SPEC): New macro.
27307 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
27309 * config/riscv/elf.h (LINK_SPEC): Pass linker endianness flag.
27310 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
27311 * config/riscv/linux.h (LINK_SPEC): Likewise.
27312 * config/riscv/riscv.h (ASM_SPEC): Pass -mbig-endian and
27314 (BYTES_BIG_ENDIAN): Handle big endian.
27315 (WORDS_BIG_ENDIAN): Define to BYTES_BIG_ENDIAN.
27316 * config/riscv/riscv.opt (-mbig-endian, -mlittle-endian): New
27318 * doc/invoke.texi (-mbig-endian, -mlittle-endian): Document.
27320 2021-03-23 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
27322 * regcprop.c (find_oldest_value_reg): Ask target whether
27323 different mode is fine for replacement register.
27325 2021-03-23 Aldy Hernandez <aldyh@redhat.com>
27327 PR tree-optimization/99296
27328 * value-range.cc (irange::irange_set_1bit_anti_range): New.
27329 (irange::irange_set_anti_range): Call irange_set_1bit_anti_range
27330 * value-range.h (irange::irange_set_1bit_anti_range): New.
27332 2021-03-22 Vladimir N. Makarov <vmakarov@redhat.com>
27335 * config/aarch64/constraints.md (UtQ): Use
27336 define_relaxed_memory_constraint for it.
27337 * doc/md.texi (define_relaxed_memory_constraint): Describe it.
27338 * genoutput.c (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
27339 * genpreds.c (constraint_data): Add bitfield is_relaxed_memory.
27340 (have_relaxed_memory_constraints): New static var.
27341 (relaxed_memory_start, relaxed_memory_end): Ditto.
27342 (add_constraint): Add arg is_relaxed_memory. Check name for
27343 relaxed memory. Set up is_relaxed_memory in constraint_data and
27344 have_relaxed_memory_constraints. Adjust calls.
27345 (choose_enum_order): Process relaxed memory.
27346 (write_tm_preds_h): Ditto.
27347 (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
27348 * gensupport.c (process_rtx): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
27349 * ira-costs.c (record_reg_classes): Process CT_RELAXED_MEMORY.
27350 * ira-lives.c (single_reg_class): Use
27351 insn_extra_relaxed_memory_constraint.
27352 * ira.c (ira_setup_alts): CT_RELAXED_MEMORY.
27353 * lra-constraints.c (valid_address_p): Use
27354 insn_extra_relaxed_memory_constraint instead of other memory
27356 (process_alt_operands): Process CT_RELAXED_MEMORY.
27357 (curr_insn_transform): Use insn_extra_relaxed_memory_constraint.
27358 * recog.c (asm_operand_ok, preprocess_constraints): Process
27360 * reload.c (find_reloads): Ditto.
27361 * rtl.def (DEFINE_RELAXED_MEMORY_CONSTRAINT): New.
27362 * stmt.c (parse_input_constraint): Use
27363 insn_extra_relaxed_memory_constraint.
27365 2021-03-22 Segher Boessenkool <segher@kernel.crashing.org>
27368 * ubsan.c (ubsan_instrument_float_cast): Don't test for unordered if
27371 2021-03-22 Alex Coplan <alex.coplan@arm.com>
27374 * config/arm/arm-protos.h (neon_make_constant): Add generate
27375 argument to guard emitting insns, default to true.
27376 * config/arm/arm.c (arm_legitimate_constant_p_1): Reject
27377 CONST_VECTORs which neon_make_constant can't handle.
27378 (neon_vdup_constant): Add generate argument, avoid emitting
27379 insns if it's not set.
27380 (neon_make_constant): Plumb new generate argument through.
27381 * config/arm/constraints.md (Ui): New. Use it...
27382 * config/arm/mve.md (*mve_mov<mode>): ... here.
27383 * config/arm/vec-common.md (movv8hf): Use neon_make_constant to
27384 synthesize constants.
27386 2021-03-22 Richard Biener <rguenther@suse.de>
27388 * debug.h: Add deprecation warning.
27390 2021-03-22 Richard Biener <rguenther@suse.de>
27392 PR tree-optimization/99694
27393 * tree-ssa-sccvn.c (visit_phi): Ignore edges with the
27396 2021-03-22 Kito Cheng <kito.cheng@sifive.com>
27399 * config/riscv/riscv.c (riscv_expand_block_move): Get RTL value
27400 after type checking.
27402 2021-03-22 Jakub Jelinek <jakub@redhat.com>
27406 * dwarf2out.c (get_full_len): Use get_precision rather than
27408 (add_const_value_attribute): Make sure add_AT_wide argument has
27409 precision prec rather than some very wide one.
27411 2021-03-22 Kewen Lin <linkw@linux.ibm.com>
27413 * config/rs6000/rs6000.md (*rotldi3_insert_sf,
27414 *mov<SFDF:mode><SFDF2:mode>cc_p9, floatsi<mode>2_lfiwax,
27415 floatsi<mode>2_lfiwax_mem, floatunssi<mode>2_lfiwzx,
27416 floatunssi<mode>2_lfiwzx_mem, *floatsidf2_internal,
27417 *floatunssidf2_internal, fix_trunc<mode>si2_stfiwx,
27418 fix_trunc<mode>si2_internal, fixuns_trunc<mode>si2_stfiwx,
27419 *round32<mode>2_fprs, *roundu32<mode>2_fprs,
27420 *fix_trunc<mode>si2_internal): Fix empty split condition.
27421 * config/rs6000/vsx.md (*vsx_le_undo_permute_<mode>,
27422 vsx_reduc_<VEC_reduc_name>_v2df, vsx_reduc_<VEC_reduc_name>_v4sf,
27423 *vsx_reduc_<VEC_reduc_name>_v2df_scalar,
27424 *vsx_reduc_<VEC_reduc_name>_v4sf_scalar): Likewise.
27426 2021-03-22 Xionghu Luo <luoxhu@linux.ibm.com>
27429 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
27430 Convert idx to DImode.
27431 (rs6000_expand_vector_set_var_p8): Likewise.
27433 2021-03-21 Jakub Jelinek <jakub@redhat.com>
27436 * dwarf2out.c (insert_float): Change return type from void to
27437 unsigned, handle GET_MODE_SIZE (mode) == 2 and return element size.
27438 (mem_loc_descriptor, loc_descriptor, add_const_value_attribute):
27441 2021-03-20 H.J. Lu <hjl.tools@gmail.com>
27444 * config/i386/i386.c (construct_container): Check cfun != NULL
27445 before accessing silent_p.
27447 2021-03-20 Ahamed Husni <ahamedhusni73@gmail.com>
27449 * asan.c: Fix typos in comments.
27451 2021-03-20 Vladimir N. Makarov <vmakarov@redhat.com>
27453 PR rtl-optimization/99680
27454 * lra-constraints.c (skip_contraint_modifiers): Rename to skip_constraint_modifiers.
27455 (process_address_1): Check empty constraint before using
27458 2021-03-19 Pat Haugen <pthaugen@linux.ibm.com>
27460 * config/rs6000/rs6000.c (power10_cost): New.
27461 (rs6000_option_override_internal): Set Power10 costs.
27462 (rs6000_issue_rate): Set Power10 issue rate.
27463 * config/rs6000/power10.md: Rewrite for Power10.
27465 2021-03-19 Vladimir N. Makarov <vmakarov@redhat.com>
27468 * lra-constraints.c (process_address_1): Don't use unknown
27469 constraint for address constraint.
27471 2021-03-19 Iain Sandoe <iain@sandoe.co.uk>
27474 * config.gcc (powerpc-*-darwin8): Delete the reference to
27475 the now removed darwin8.h.
27477 2021-03-19 Olivier Hainque <hainque@adacore.com>
27480 * config/vxworksae.h (VX_CPU_PREFIX): Define.
27482 2021-03-19 John David Anglin <danglin@gcc.gnu.org>
27484 * config/pa/pa.c (import_milli): Use memcpy instead of strncpy.
27486 2021-03-19 Tamar Christina <tamar.christina@arm.com>
27488 PR tree-optimization/99656
27489 * tree-vect-slp-patterns.c (linear_loads_p,
27490 complex_add_pattern::matches, is_eq_or_top,
27491 vect_validate_multiplication, complex_mul_pattern::matches,
27492 complex_fms_pattern::matches): Remove complex_perm_kinds_t.
27493 * tree-vectorizer.h: (complex_load_perm_t): Removed.
27494 (slp_tree_to_load_perm_map_t): Use complex_perm_kinds_t instead of
27495 complex_load_perm_t.
27497 2021-03-19 H.J. Lu <hjl.tools@gmail.com>
27500 * config/i386/i386-options.c (ix86_init_machine_status): Set
27502 * config/i386/i386.c (init_cumulative_args): Set silent_p to
27504 (construct_container): Return early for return and argument
27505 errors if silent_p is true.
27506 * config/i386/i386.h (machine_function): Add silent_p.
27508 2021-03-19 Jakub Jelinek <jakub@redhat.com>
27511 * config/arm/constraints.md (Ds): New constraint.
27512 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): Use w,Ds
27513 constraint instead of w,Dm.
27515 2021-03-19 Andrew Stubbs <ams@codesourcery.com>
27517 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Fix quotes
27520 2021-03-19 Eric Botcazou <ebotcazou@adacore.com>
27522 PR middle-end/99641
27523 * fold-const.c (native_encode_initializer) <CONSTRUCTOR>: For an
27524 array type, do the computation of the current position in sizetype.
27526 2021-03-18 Vladimir N. Makarov <vmakarov@redhat.com>
27529 * lra-constraints.c (process_address_1): Use lookup_constraint
27530 only for a single constraint.
27532 2021-03-18 Martin Sebor <msebor@redhat.com>
27534 PR middle-end/99502
27535 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): Rename...
27536 (inbounds_memaccess_p): ...to this. Check the ending offset of
27537 the accessed member.
27539 2021-03-18 Andrew Stubbs <ams@codesourcery.com>
27541 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Add %< and
27542 %> quote markers to error messages.
27543 (gcn_goacc_validate_dims): Likewise.
27544 (gcn_conditional_register_usage): Remove exclaimation mark from error
27546 (gcn_vectorize_vec_perm_const): Ensure perm is fully uninitialized.
27548 2021-03-18 Jan Hubicka <hubicka@ucw.cz>
27550 * config/i386/x86-tune-costs.h (struct processor_costs): Fix costs of
27553 2021-03-18 Sinan Lin <sinan@isrc.iscas.ac.cn>
27554 Kito Cheng <kito.cheng@sifive.com>
27556 * config/riscv/riscv.c (riscv_block_move_straight): Change type
27557 to unsigned HOST_WIDE_INT for parameter and local variable with
27558 HOST_WIDE_INT type.
27559 (riscv_adjust_block_mem): Ditto.
27560 (riscv_block_move_loop): Ditto.
27561 (riscv_expand_block_move): Ditto.
27563 2021-03-18 Nick Clifton <nickc@redhat.com>
27565 * config/v850/v850.c (construct_restore_jr): Increase static
27567 (construct_save_jarl): Likewise.
27568 * config/v850/v850.h (DWARF2_DEBUGGING_INFO): Define.
27570 2021-03-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27572 * config/aarch64/aarch64.c (aarch64_adjust_generic_arch_tuning): Define.
27573 (aarch64_override_options_internal): Use it.
27574 (generic_tunings): Add AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS to
27577 2021-03-17 Sandra Loosemore <sandra@codesourcery.com>
27579 * config/nios2/nios2.c (nios2_custom_check_insns): Clean up
27580 error message format issues.
27581 (nios2_option_override): Likewise.
27582 (nios2_expand_fpu_builtin): Likewise.
27583 (nios2_init_custom_builtins): Adjust to avoid bogus strncpy
27584 truncation warning.
27585 (nios2_expand_custom_builtin): More error message format fixes.
27586 (nios2_expand_rdwrctl_builtin): Likewise.
27587 (nios2_expand_rdprs_builtin): Likewise.
27588 (nios2_expand_eni_builtin): Likewise.
27589 (nios2_expand_builtin): Likewise.
27590 (nios2_register_custom_code): Likewise.
27591 (nios2_valid_target_attribute_rec): Likewise.
27592 (nios2_add_insn_asm): Fix uninitialized variable warning.
27594 2021-03-17 Jan Hubicka <jh@suse.cz>
27596 * config/i386/x86-tune-costs.h (struct processor_costs): Update costs
27597 of gather to match reality.
27598 * config/i386/x86-tune.def (X86_TUNE_USE_GATHER): Enable for znver3.
27600 2021-03-17 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27602 * config/aarch64/aarch64-builtins.c (aarch64_expand_rng_builtin): Use EQ
27603 to compare against CC_REG rather than NE.
27605 2021-03-17 H.J. Lu <hjl.tools@gmail.com>
27608 * config/i386/i386.c (ix86_force_load_from_GOT_p): Support
27609 inline assembly statements.
27610 (ix86_print_operand): Update 'P' handling for -fno-plt.
27612 2021-03-17 Tamar Christina <tamar.christina@arm.com>
27615 * config/aarch64/aarch64.c
27616 (aarch64_simd_clone_compute_vecsize_and_simdlen): Remove unused var.
27618 2021-03-16 Segher Boessenkool <segher@kernel.crashing.org>
27621 * config/rs6000/predicates.md (branch_comparison_operator): Allow
27622 ordered and unordered for CCFPmode, if flag_finite_math_only.
27624 2021-03-16 Jakub Jelinek <jakub@redhat.com>
27627 * config/i386/i386-expand.c (ix86_split_lea_for_addr): Emit a MULT
27628 rather than ASHIFT.
27629 * config/i386/i386.md (mult by 1248 into ashift): New splitter.
27631 2021-03-16 Martin Liska <mliska@suse.cz>
27634 * optc-save-gen.awk: Add flag_ipa_ra to exceptions for
27635 cl_optimization_compare function.
27637 2021-03-16 Ilya Leoshkevich <iii@linux.ibm.com>
27639 * config/s390/s390.c (f_constraint_p): Treat "fv" constraints
27642 2021-03-16 Jakub Jelinek <jakub@redhat.com>
27645 * config/i386/i386.h (struct machine_function): Add
27646 has_explicit_vzeroupper bitfield.
27647 * config/i386/i386-expand.c (ix86_expand_builtin): Set
27648 cfun->machine->has_explicit_vzeroupper when expanding
27649 IX86_BUILTIN_VZEROUPPER.
27650 * config/i386/i386-features.c (rest_of_handle_insert_vzeroupper):
27651 Do the mode switching only when TARGET_VZEROUPPER, expensive
27652 optimizations turned on and not optimizing for size.
27653 (pass_insert_vzeroupper::gate): Enable even when
27654 cfun->machine->has_explicit_vzeroupper is set.
27656 2021-03-16 Jakub Jelinek <jakub@redhat.com>
27659 * config/aarch64/aarch64.c
27660 (aarch64_simd_clone_compute_vecsize_and_simdlen): If not a function
27661 definition, walk TYPE_ARG_TYPES list if non-NULL for argument types
27662 instead of DECL_ARGUMENTS. Ignore types for uniform arguments.
27664 2021-03-15 Richard Biener <rguenther@suse.de>
27666 PR tree-optimization/98834
27667 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle missing
27668 subsetting by truncating the access size.
27670 2021-03-15 Jan Hubicka <hubicka@ucw.cz>
27672 * config/i386/i386-options.c (processor_cost_table): Add znver3_cost.
27673 * config/i386/x86-tune-costs.h (znver3_cost): New gobal variable; copy
27676 2021-03-15 Martin Liska <mliska@suse.cz>
27678 * spellcheck.c: Add missing comma in initialization.
27680 2021-03-14 Uroš Bizjak <ubizjak@gmail.com>
27682 * config/i386/sse.md (*vec_extract<mode>): Merge alternative 0 with
27683 alternative 2 and alternative 1 with alternative 3 using
27684 YW register constraint.
27685 (*vec_extract<PEXTR_MODE12:mode>_zext): Merge alternatives
27686 using YW register constraint.
27687 (*vec_extractv16qi_zext): Ditto.
27688 (*vec_extractv4si): Merge alternatives 4 and 5
27689 using Yw register constraint.
27690 (*ssse3_palignr<mode>_perm): Use Yw instead of v for alternative 3.
27692 2021-03-13 Martin Sebor <msebor@redhat.com>
27694 PR tree-optimization/99489
27695 * builtins.c (gimple_call_alloc_size): Fail gracefully when argument
27696 is not a call statement.
27698 2021-03-13 Jakub Jelinek <jakub@redhat.com>
27700 PR tree-optimization/99544
27701 * match.pd (X + (X << C) -> X * (1 + (1 << C))): Don't simplify
27702 if for vector types multiplication can't be done in type's mode.
27704 2021-03-12 Eric Botcazou <ebotcazou@adacore.com>
27707 * config/sparc/constraints.md (w): Rename to...
27708 (W): ... this and ditch previous implementation.
27709 * config/sparc/sparc.md (*movdi_insn_sp64): Replace W with m.
27710 (*movdf_insn_sp64): Likewise.
27711 (*mov<VM64:mode>_insn_sp64): Likewise.
27712 * config/sparc/sync.md (*atomic_compare_and_swap<mode>_1): Replace
27714 (atomic_compare_and_swap_leon3_1): Likewise.
27715 (*atomic_compare_and_swapdi_v8plus): Likewise.
27716 * config/sparc/sparc.c (memory_ok_for_ldd): Remove useless test on
27717 architecture and add missing address validity check during LRA.
27719 2021-03-12 Tobias Burnus <tobias@codesourcery.com>
27722 * gimplify.c (omp_add_variable): Handle NULL_TREE as size
27723 occuring for assumed-size arrays in use_device_{ptr,addr}.
27725 2021-03-12 Jakub Jelinek <jakub@redhat.com>
27728 * config/i386/constraints.md (YW): New internal constraint.
27729 * config/i386/sse.md (v_Yw): Add V4TI, V2TI, V1TI and TI cases.
27730 (*<sse2_avx2>_<insn><mode>3<mask_name>,
27731 *<sse2_avx2>_uavg<mode>3<mask_name>, *abs<mode>2,
27732 *<s>mul<mode>3_highpart<mask_name>): Use <v_Yw> instead of v in
27734 (<sse2_avx2>_psadbw): Use YW instead of v in constraints.
27735 (*avx2_pmaddwd, *sse2_pmaddwd, *<code>v8hi3, *<code>v16qi3,
27736 avx2_pmaddubsw256, ssse3_pmaddubsw128): Merge last two alternatives
27737 into one, use Yw instead of former x,v.
27738 (ashr<mode>3, <insn><mode>3): Use <v_Yw> instead of x in constraints of
27739 the last alternative.
27740 (<sse2_avx2>_packsswb<mask_name>, <sse2_avx2>_packssdw<mask_name>,
27741 <sse2_avx2>_packuswb<mask_name>, <sse4_1_avx2>_packusdw<mask_name>,
27742 *<ssse3_avx2>_pmulhrsw<mode>3<mask_name>, <ssse3_avx2>_palignr<mode>,
27743 <ssse3_avx2>_pshufb<mode>3<mask_name>): Merge last two alternatives
27744 into one, use <v_Yw> instead of former x,v.
27745 (avx2_interleave_highv32qi<mask_name>,
27746 vec_interleave_highv16qi<mask_name>): Use Yw instead of v in
27747 constraints. Add && <mask_avx512bw_condition> to condition.
27748 (avx2_interleave_lowv32qi<mask_name>,
27749 vec_interleave_lowv16qi<mask_name>,
27750 avx2_interleave_highv16hi<mask_name>,
27751 vec_interleave_highv8hi<mask_name>,
27752 avx2_interleave_lowv16hi<mask_name>, vec_interleave_lowv8hi<mask_name>,
27753 avx2_pshuflw_1<mask_name>, sse2_pshuflw_1<mask_name>,
27754 avx2_pshufhw_1<mask_name>, sse2_pshufhw_1<mask_name>,
27755 avx2_<code>v16qiv16hi2<mask_name>, sse4_1_<code>v8qiv8hi2<mask_name>,
27756 *sse4_1_<code>v8qiv8hi2<mask_name>_1, <sse2_avx2>_<insn><mode>3): Use
27757 Yw instead of v in constraints.
27758 * config/i386/mmx.md (Yv_Yw): New define_mode_attr.
27759 (*mmx_<insn><mode>3, mmx_ashr<mode>3, mmx_<insn><mode>3): Use <Yv_Yw>
27760 instead of Yv in constraints.
27761 (*mmx_<insn><mode>3, *mmx_mulv4hi3, *mmx_smulv4hi3_highpart,
27762 *mmx_umulv4hi3_highpart, *mmx_pmaddwd, *mmx_<code>v4hi3,
27763 *mmx_<code>v8qi3, mmx_pack<s_trunsuffix>swb, mmx_packssdw,
27764 mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
27765 *mmx_uavgv8qi3, *mmx_uavgv4hi3, mmx_psadbw): Use Yw instead of Yv in
27767 (*mmx_pinsrw, *mmx_pinsrb, *mmx_pextrw, *mmx_pextrw_zext, *mmx_pextrb,
27768 *mmx_pextrb_zext): Use YW instead of Yv in constraints.
27769 (*mmx_eq<mode>3, mmx_gt<mode>3): Use x instead of Yv in constraints.
27770 (mmx_andnot<mode>3, *mmx_<code><mode>3): Split last alternative into
27771 two, one with just x, another isa avx512vl with v.
27773 2021-03-12 Martin Liska <mliska@suse.cz>
27775 * doc/invoke.texi: Add missing param documentation.
27777 2021-03-11 David Malcolm <dmalcolm@redhat.com>
27780 * Makefile.in (ANALYZER_OBJS): Add analyzer/feasible-graph.o and
27781 analyzer/trimmed-graph.o.
27782 * doc/analyzer.texi (Analyzer Paths): Rewrite description of
27783 feasibility checking to reflect new implementation.
27784 * doc/invoke.texi (-fdump-analyzer-feasibility): Document new
27786 * shortest-paths.h (shortest_paths::get_shortest_distance): New.
27788 2021-03-11 David Malcolm <dmalcolm@redhat.com>
27790 * digraph.cc (selftest::test_shortest_paths): Update
27791 shortest_paths init for new param. Add test of
27792 SPS_TO_GIVEN_TARGET.
27793 * shortest-paths.h (enum shortest_path_sense): New.
27794 (shortest_paths::shortest_paths): Add "sense" param.
27795 Update for renamings. Generalize to use "sense" param.
27796 (shortest_paths::get_shortest_path): Rename param.
27797 (shortest_paths::m_sense): New field.
27798 (shortest_paths::m_prev): Rename...
27799 (shortest_paths::m_best_edge): ...to this.
27800 (shortest_paths::get_shortest_path): Update for renamings.
27801 Conditionalize flipping of path on sense of traversal.
27803 2021-03-11 David Malcolm <dmalcolm@redhat.com>
27805 * digraph.cc (selftest::test_shortest_paths): Add test coverage
27806 for paths from B and C.
27807 * shortest-paths.h (shortest_paths::shortest_paths): Handle
27808 unreachable nodes, rather than asserting.
27810 2021-03-11 David Edelsohn <dje.gcc@gmail.com>
27813 * config/rs6000/rs6000.c (rs6000_xcoff_file_start): Don't create
27814 xcoff_tbss_section_name.
27815 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .lcomm.
27816 * xcoffout.c (xcoff_tbss_section_name): Delete.
27817 * xcoffout.h (xcoff_tbss_section_name): Delete.
27819 2021-03-11 Richard Biener <rguenther@suse.de>
27821 PR tree-optimization/99523
27822 * tree-cfg.c (dump_function_to_file): Dump SSA names
27823 w/o identifier to the decls section as well, not only those
27824 without a VAR_DECL.
27826 2021-03-11 Jakub Jelinek <jakub@redhat.com>
27829 * ipa-icf-gimple.c (func_checker::compare_gimple_call): For internal
27830 function calls with lhs fail if the lhs don't have compatible types.
27832 2021-03-11 Hans-Peter Nilsson <hp@axis.com>
27834 * config/cris/cris.h (HARD_FRAME_POINTER_REGNUM): Define.
27835 Change FRAME_POINTER_REGNUM to correspond to a new faked
27836 register faked_fp, part of GENNONACR_REGS like faked_ap.
27837 (CRIS_FAKED_REGS_CONTENTS): New helper macro.
27838 (FIRST_PSEUDO_REGISTER, FIXED_REGISTERS, CALL_USED_REGISTERS):
27839 (REG_ALLOC_ORDER, REG_CLASS_CONTENTS, REGNO_OK_FOR_BASE_P)
27840 (ELIMINABLE_REGS, REGISTER_NAMES): Adjust accordingly.
27841 * config/cris/cris.md (CRIS_FP_REGNUM): Renumber to new faked
27843 (CRIS_REAL_FP_REGNUM): New constant.
27844 * config/cris/cris.c (cris_reg_saved_in_regsave_area): Check
27845 for HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM.
27846 (cris_initial_elimination_offset): Handle elimination changes
27847 to HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM
27848 and add one from FRAME_POINTER_REGNUM to
27849 HARD_FRAME_POINTER_REGNUM.
27850 (cris_expand_prologue, cris_expand_epilogue): Emit code for
27851 hard_frame_pointer_rtx instead of frame_pointer_rtx.
27853 2021-03-10 David Edelsohn <dje.gcc@gmail.com>
27856 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Add check for DCmode.
27857 * config/rs6000/rs6000.c (rs6000_special_round_type_align): Same.
27859 2021-03-10 Vladimir N. Makarov <vmakarov@redhat.com>
27862 * lra-constraints.c (process_address_1): Don't check unknown
27863 constraint, use X for empty constraint.
27865 2021-03-10 Alex Coplan <alex.coplan@arm.com>
27867 * config/aarch64/aarch64.c (aarch64_vfp_is_call_or_return_candidate):
27868 Fix typo in comment describing "is_ha" argument.
27870 2021-03-10 John David Anglin <danglin@gcc.gnu.org>
27872 * doc/sourcebuild.texi: Document LRA target selector.
27874 2021-03-10 David Malcolm <dmalcolm@redhat.com>
27876 * doc/ux.texi: Add subsection contrasting interactive versus
27877 batch usage of GCC.
27879 2021-03-10 Joel Hutton <joel.hutton@arm.com>
27882 * tree-vect-stmts.c (vectorizable_store): Fix scatter store mask
27884 (vectorizable_load): Fix gather load mask check condition.
27886 2021-03-10 Richard Biener <rguenther@suse.de>
27888 PR tree-optimization/99510
27889 * tree.c (check_aligned_type): Check that the candidate
27890 has TYPE_USER_ALIGN set instead of matching with the
27893 2021-03-10 Eric Botcazou <ebotcazou@adacore.com>
27895 * config/sparc/sparc.c (sparc_regmode_natural_size): Return 4 for
27896 float and vector integer modes only if the mode is not larger.
27898 2021-03-10 Hans-Peter Nilsson <hp@axis.com>
27900 * config/cris/cris.h (DWARF_FRAME_REGISTERS): Define.
27902 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
27904 * ira.c (ira_setup_alts, ira_get_dup_out_num): Process digital
27906 * ira-lives.c (single_reg_class): Ditto.
27908 2021-03-09 Sebastian Huber <sebastian.huber@embedded-brains.de>
27910 * config.gcc (aarch64-*-rtems*): Include general rtems.h after
27911 the architecture-specific rtems.h.
27912 (aarch64-*-rtems*): Likewise.
27913 (arm*-*-rtems*): Likewise.
27914 (epiphany-*-rtems*): Likewise.
27915 (riscv*-*-rtems*): Likewise.
27917 2021-03-09 Jakub Jelinek <jakub@redhat.com>
27919 PR tree-optimization/99305
27920 * tree-ssa-phiopt.c (conditional_replacement): Test integer_pow2p
27921 before integer_all_onesp instead of vice versa.
27923 2021-03-09 Richard Earnshaw <rearnsha@arm.com>
27925 * common/config/arm/arm-common.c (arm_config_default): Change type
27926 of 'i' to unsigned.
27928 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
27931 * lra-constraints.c (process_address_1): Process constraint 'g'
27932 separately and digital constraints containing more one digit.
27934 2021-03-09 Nick Clifton <nickc@redhat.com>
27936 * config/rx/rx.h (DBX_DEBUGGING_INFO): Define.
27937 (DWARF"_DEBUGGING_INFO): Define.
27939 2021-03-09 Eric Botcazou <ebotcazou@adacore.com>
27942 * calls.c (initialize_argument_information): When the argument
27943 is passed by reference, do not make a copy in a thunk only if
27944 the argument is already in memory. Remove redundant test for
27945 the case of callee copy.
27947 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
27950 * lra-constraints.c (process_address_1): Process 0..9 constraints
27951 in process_address_1.
27953 2021-03-09 Andreas Krebbel <krebbel@linux.ibm.com>
27955 * config/s390/s390.c (struct s390_processor processor_table):
27956 Binutils name string must not be empty.
27958 2021-03-09 Claudiu Zissulescu <claziss@synopsys.com>
27960 * config/arc/arc.c (arc_attr_type): Remove function.
27962 2021-03-09 Martin Liska <mliska@suse.cz>
27965 * config/i386/i386-options.c (ix86_option_override_internal):
27966 Set isa_flags for OPTS argument and not for the global
27969 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
27971 * config/rs6000/predicates.md (ds_form_mem_operand): Check
27974 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
27977 * config/rs6000/predicates.md (ds_form_mem_operand) New
27979 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10) Use
27980 ds_form_mem_operand in ld/lwa patterns.
27981 * config/rs6000/fusion.md: Regenerate file.
27983 2021-03-08 Martin Sebor <msebor@redhat.com>
27985 PR middle-end/98266
27986 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): New function.
27987 (array_bounds_checker::check_array_bounds): Call it.
27989 2021-03-08 Martin Sebor <msebor@redhat.com>
27991 PR middle-end/97631
27992 * tree-ssa-strlen.c (maybe_warn_overflow): Test rawmem.
27993 (handle_builtin_stxncpy_strncat): Rename locals. Determine
27994 destination size from allocation calls. Issue a more appropriate
27996 (handle_builtin_memcpy): Pass true as rawmem to maybe_warn_overflow.
27997 (handle_builtin_memset): Same.
27999 2021-03-08 Peter Bergner <bergner@linux.ibm.com>
28002 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_permute): Add an assert
28003 to ensure we do not have an Altivec style address.
28004 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): Disable if passed
28005 an Altivec style address.
28006 (*vsx_le_perm_store_<mode>): Likewise.
28007 (splitters after *vsx_le_perm_store_<mode>): Likewise.
28008 (vsx_load_<mode>): Disable special expander if passed an Altivec
28010 (vsx_store_<mode>): Likewise.
28012 2021-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28015 * config/aarch64/predicates.md (aarch64_simd_shift_imm_vec_qi): Define.
28016 (aarch64_simd_shift_imm_vec_hi): Likewise.
28017 (aarch64_simd_shift_imm_vec_si): Likewise.
28018 (aarch64_simd_shift_imm_vec_di): Likewise.
28019 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le): Use
28020 predicate from above.
28021 (aarch64_shrn<mode>_insn_be): Likewise.
28022 (aarch64_rshrn<mode>_insn_le): Likewise.
28023 (aarch64_rshrn<mode>_insn_be): Likewise.
28024 (aarch64_shrn2<mode>_insn_le): Likewise.
28025 (aarch64_shrn2<mode>_insn_be): Likewise.
28026 (aarch64_rshrn2<mode>_insn_le): Likewise.
28027 (aarch64_rshrn2<mode>_insn_be): Likewise.
28029 2021-03-08 Vladimir N. Makarov <vmakarov@redhat.com>
28032 * lra-constraints.c (skip_contraint_modifiers): New function.
28033 (process_address_1): Use it before lookup_constraint call.
28035 2021-03-08 Martin Liska <mliska@suse.cz>
28038 * config/i386/i386-options.c (ix86_option_override_internal):
28039 Enable UINTR and HRESET for -march that supports it.
28041 2021-03-08 Ilya Leoshkevich <iii@linux.ibm.com>
28043 * config/s390/s390.c (f_constraint_p): New function.
28044 (s390_md_asm_adjust): Implement TARGET_MD_ASM_ADJUST.
28045 (TARGET_MD_ASM_ADJUST): Likewise.
28047 2021-03-08 Tobias Burnus <tobias@codesourcery.com>
28050 * tree-nested.c (convert_local_reference_stmt): Avoid calling
28051 lookup_field_for_decl for Fortran module (= namespace context).
28053 2021-03-08 Andreas Krebbel <krebbel@linux.ibm.com>
28055 * config/s390/s390.c (s390_expand_vec_compare): Implement <0
28056 comparison with arithmetic right shift.
28057 (s390_expand_vcond): No need for a force_reg anymore.
28058 s390_vec_compare will do it.
28059 * config/s390/vector.md ("vec_cmp<mode><tointvec>"): Accept also
28060 immediate operands.
28062 2021-03-07 Jakub Jelinek <jakub@redhat.com>
28065 * config/i386/constraints.md (Yw): Use SSE_REGS if TARGET_SSE
28066 but TARGET_AVX512BW or TARGET_AVX512VL is not set. Adjust description
28068 * config/i386/sse.md (v_Yw): New define_mode_attr.
28069 (*<insn><mode>3, *mul<mode>3<mask_name>, *avx2_<code><mode>3,
28070 *sse4_1_<code><mode>3<mask_name>): Use <v_Yw> instead of v
28072 * config/i386/mmx.md (mmx_pshufw_1, *vec_dupv4hi): Use Yw instead of
28073 xYw in constraints.
28075 2021-03-06 Julian Brown <julian@codesourcery.com>
28077 * tree-pretty-print.c (dump_generic_node): Emit non-generic
28078 address space info for aggregates.
28080 2021-03-06 Hans-Peter Nilsson <hp@axis.com>
28082 * config/cris/cris.h (MAX_FIXED_MODE_SIZE): Don't define.
28084 2021-03-05 Jakub Jelinek <jakub@redhat.com>
28086 PR middle-end/99322
28087 * tree-cfg.c (bb_to_omp_idx): New variable.
28088 (execute_build_cfg): Release the bb_to_omp_idx vector after
28089 cleanup_tree_cfg returns.
28090 (handle_abnormal_edges): Remove bb_to_omp_idx argument, adjust
28091 for bb_to_omp_idx being a vec<int> instead of pointer to array
28093 (make_edges): Remove bb_to_omp_idx local variable, don't pass
28094 it to handle_abnormal_edges, adjust for bb_to_omp_idx being a
28095 vec<int> instead of pointer to array of ints and don't free/release
28097 (remove_bb): When removing a bb and placing forced label somewhere
28098 else, ensure it is put into the same OpenMP region during cfg
28099 pass if possible or to entry successor as fallback. Unregister
28100 bb from bb_to_omp_idx.
28102 2021-03-05 Vladimir N. Makarov <vmakarov@redhat.com>
28105 * lra-constraints.c (process_address_1): Skip decomposing address
28106 for asm insn operand with unknown constraint.
28108 2021-03-05 Martin Jambor <mjambor@suse.cz>
28111 * cgraph.c (cgraph_edge::set_call_stmt): Do not update all
28112 corresponding speculative edges if we are about to resolve
28113 sepculation. Make edge direct (and so resolve speculations) before
28114 removing it from call_site_hash.
28115 (cgraph_edge::make_direct): Relax the initial assert to allow calling
28116 the function on speculative direct edges.
28118 2021-03-05 Eric Botcazou <ebotcazou@adacore.com>
28120 PR rtl-optimization/99376
28121 * rtlanal.c (nonzero_bits1) <arithmetic operators>: If the number
28122 of low-order zero bits is too large, set the result to 0 directly.
28124 2021-03-04 Jakub Jelinek <jakub@redhat.com>
28126 PR middle-end/93235
28127 * expmed.c (store_bit_field_using_insv): Return false of xop0 is a
28128 SUBREG and a SUBREG to op_mode can't be created.
28130 2021-03-04 Alex Coplan <alex.coplan@arm.com>
28133 * config/aarch64/aarch64-sve-builtins.cc
28134 (function_resolver::require_vector_type): Handle error_mark_node.
28136 2021-03-04 Ilya Leoshkevich <iii@linux.ibm.com>
28138 * cfgexpand.c (expand_asm_loc): Pass new parameter.
28139 (expand_asm_stmt): Likewise.
28140 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Add new
28142 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
28143 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
28144 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
28145 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
28146 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
28147 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
28148 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
28149 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
28150 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
28151 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
28152 * doc/tm.texi (md_asm_adjust): Likewise.
28153 * target.def (md_asm_adjust): Likewise.
28155 2021-03-04 Richard Biener <rguenther@suse.de>
28157 PR middle-end/97855
28158 * tree-pretty-print.c: Poison pp_printf.
28159 (dump_decl_name): Avoid use of pp_printf.
28160 (dump_block_node): Likewise.
28161 (dump_generic_node): Likewise.
28163 2021-03-04 Martin Sebor <msebor@redhat.com>
28165 PR middle-end/96963
28166 PR middle-end/94655
28167 * builtins.c (handle_array_ref): New helper.
28168 (handle_mem_ref): New helper.
28169 (compute_objsize_r): Factor out ARRAY_REF and MEM_REF handling
28170 into new helper functions. Correct a workaround for vectorized
28173 2021-03-03 Pat Haugen <pthaugen@linux.ibm.com>
28175 * config/rs6000/dfp.md (extendddtd2, trunctddd2, *cmp<mode>_internal1,
28176 floatditd2, ftrunc<mode>2, fix<mode>di2, dfp_ddedpd_<mode>,
28177 dfp_denbcd_<mode>, dfp_dxex_<mode>, dfp_diex_<mode>,
28178 *dfp_sgnfcnc_<mode>, dfp_dscli_<mode>, dfp_dscri_<mode>): Update size
28179 attribute for Power10.
28180 * config/rs6000/mma.md (*movoo): Likewise.
28181 * config/rs6000/rs6000.md (define_attr "size"): Add 256.
28182 (define_mode_attr bits): Add DD/TD modes.
28183 * config/rs6000/sync.md (load_quadpti, store_quadpti, load_lockedpti,
28184 store_conditionalpti): Update size attribute for Power10.
28186 2021-03-03 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
28189 * config/sparc/t-sparc (tree-ssanames.o-warn): Don't error for
28190 -Wuninitialized, -Wmaybe-uninitialized.
28191 (wide-int.o-warn): Likewise.
28193 2021-03-03 Richard Earnshaw <rearnsha@arm.com>
28195 * common/config/arm/arm-common.c: Include configargs.h.
28196 (arm_config_default): New function.
28197 (arm_target_mode): Renamed from arm_target_thumb_only. Handle
28198 processors that do not support Thumb. Take into account the
28199 --with-mode configuration setting for selecting the default.
28200 * config/arm/arm.h (OPTION_DEFAULT_SPECS): Remove entry for 'mode'.
28201 (TARGET_MODE_SPEC_FUNCTIONS): Update for function name change.
28203 2021-03-03 Martin Liska <mliska@suse.cz>
28205 PR gcov-profile/97461
28206 * gcov-io.h (GCOV_PREALLOCATED_KVP): Remove.
28208 2021-03-03 Eric Botcazou <ebotcazou@adacore.com>
28211 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
28212 point back the hard frame pointer to its default location when the
28213 frame is larger than SEH_MAX_FRAME_SIZE.
28215 2021-03-03 Jakub Jelinek <jakub@redhat.com>
28218 * config/i386/predicates.md (logic_operator): New define_predicate.
28219 * config/i386/i386.md (mov + mem using comm arith peephole2):
28220 Punt if operands[1] is EXT_REX_SSE_REGNO_P, AVX512BW is not enabled
28221 and the inner mode is [QH]Imode.
28223 2021-03-03 Jakub Jelinek <jakub@redhat.com>
28226 * dwarf2out.c (dw_loc_list_struct): Add end_entry member.
28227 (new_loc_list): Clear end_entry.
28228 (output_loc_list): Only use DW_LLE_startx_length for -gsplit-dwarf
28229 if HAVE_AS_LEB128, otherwise use DW_LLE_startx_endx. Fix comment
28231 (index_location_lists): For dwarf_version >= 5 without HAVE_AS_LEB128,
28232 initialize also end_entry.
28234 2021-03-03 Jakub Jelinek <jakub@redhat.com>
28237 * cfgrtl.c (fixup_partitions): When changing some bbs from hot to cold
28238 partitions, if in non-layout mode after reorder_blocks also move
28239 affected blocks to ensure a single partition transition.
28241 2021-03-03 Jason Merrill <jason@redhat.com>
28244 * cgraphunit.c (process_function_and_variable_attributes): Don't
28245 warn about flatten on an alias if the target also has it.
28246 * cgraph.h (symtab_node::get_alias_target_tree): New.
28248 2021-03-02 David Edelsohn <dje.gcc@gmail.com>
28250 * config/rs6000/rs6000.md (tls_get_tpointer_internal): Prepend
28251 period to symbol name.
28252 (tls_get_addr_internal<mode>): Same.
28254 2021-03-02 David Malcolm <dmalcolm@redhat.com>
28257 * diagnostic-show-locus.c
28258 (selftest::test_one_liner_many_fixits_2): Fix accidental usage of
28261 2021-03-02 Martin Sebor <msebor@redhat.com>
28263 PR middle-end/99276
28264 * builtins.c (warn_for_access): Remove stray warning text.
28266 2021-03-02 Martin Sebor <msebor@redhat.com>
28268 PR middle-end/99295
28269 * doc/extend.texi (attribute malloc): Reword and clarify nonaliasing
28272 2021-03-02 Jakub Jelinek <jakub@redhat.com>
28275 * dwarf2out.c (output_macinfo_op): Use DW_MACRO_*_str* even with
28276 -gdwarf-5 -gstrict-dwarf. For -gsplit-dwarf -gdwarf-5 use
28277 DW_MACRO_*_strx instead of DW_MACRO_*_strp. Handle
28278 DW_MACRO_define_strx and DW_MACRO_undef_strx.
28279 (save_macinfo_strings): Use DW_MACRO_*_str* even with
28280 -gdwarf-5 -gstrict-dwarf. Handle DW_MACRO_define_strx and
28281 DW_MACRO_undef_strx.
28283 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
28285 * config/s390/s390-builtin-types.def (BT_FN_V4SF_V8HI_UINT): New
28287 (BT_FN_V8HI_V8HI_UINT): Likewise.
28288 (BT_FN_V8HI_V4SF_V4SF_UINT): Likewise.
28289 * config/s390/s390-builtins.def (B_NNPA): New macro definition.
28290 (s390_vclfnhs, s390_vclfnls, s390_vcrnfs, s390_vcfn, s390_vcnf):
28291 New builtin definitions.
28292 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Bump
28293 vector extension version.
28294 * config/s390/s390.c (s390_expand_builtin): Check if builtins are
28295 available with current -march level.
28296 * config/s390/s390.md (UNSPEC_NNPA_VCLFNHS_V8HI)
28297 (UNSPEC_NNPA_VCLFNLS_V8HI, UNSPEC_NNPA_VCRNFS_V8HI)
28298 (UNSPEC_NNPA_VCFN_V8HI, UNSPEC_NNPA_VCNF_V8HI): New constants.
28299 * config/s390/vecintrin.h (vec_extend_to_fp32_hi): New macro.
28300 (vec_extend_to_fp32_lo): Likewise.
28301 (vec_round_from_fp32): Likewise.
28302 (vec_convert_to_fp16): Likewise.
28303 (vec_convert_from_fp16): Likewise.
28304 * config/s390/vx-builtins.md (vclfnhs_v8hi): New insn pattern.
28305 (vclfnls_v8hi): Likewise.
28306 (vcrnfs_v8hi): Likewise.
28307 (vcfn_v8hi): Likewise.
28308 (vcnf_v8hi): Likewise.
28310 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
28312 * common/config/s390/s390-common.c (processor_flags_table): New entry.
28313 * config.gcc: Enable arch14 for --with-arch and --with-tune.
28314 * config/s390/driver-native.c (s390_host_detect_local_cpu): Pick
28315 arch14 for unknown CPU models.
28316 * config/s390/s390-opts.h (enum processor_type): Add PROCESSOR_ARCH14.
28317 * config/s390/s390.c (s390_issue_rate): Add case for PROCESSOR_ARCH14.
28318 (s390_get_sched_attrmask): Likewise.
28319 (s390_get_unit_mask): Likewise.
28320 * config/s390/s390.h (enum processor_flags): Add PF_NNPA and PF_ARCH14.
28321 (TARGET_CPU_ARCH14, TARGET_CPU_ARCH14_P, TARGET_CPU_NNPA)
28322 (TARGET_CPU_NNPA_P, TARGET_ARCH14, TARGET_ARCH14_P, TARGET_NNPA)
28323 (TARGET_NNPA_P): New macro definitions.
28324 * config/s390/s390.md ("cpu_facility", "enabled"): Add arch14 and nnpa.
28325 * config/s390/s390.opt: Add PROCESSOR_ARCH14.
28327 2021-03-02 Jakub Jelinek <jakub@redhat.com>
28329 PR middle-end/95757
28330 * tree-vrp.c (register_edge_assert_for): Remove superfluous ()s around
28331 condition. Call register_edge_assert_for_1 for == 0, != 0, == 1 and
28332 != 1 comparisons if name is lhs of a comparison.
28334 2021-03-01 Iain Sandoe <iain@sandoe.co.uk>
28338 * config/darwin-protos.h (darwin_should_restore_cfa_state): New.
28339 * config/darwin.c (darwin_should_restore_cfa_state): New.
28340 * config/darwin.h (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New.
28341 * doc/tm.texi: Regenerated.
28342 * doc/tm.texi.in: Document TARGET_ASM_SHOULD_RESTORE_CFA_STATE.
28343 * dwarf2cfi.c (connect_traces): If the target requests, restore
28344 the CFA expression after a DW_CFA_restore.
28345 * target.def (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New hook.
28347 2021-03-01 Martin Liska <mliska@suse.cz>
28350 * optc-save-gen.awk: Add 4 more exceptions.
28352 2021-03-01 Nathan Sidwell <nathan@acm.org>
28355 * tree.h (TYPE_ALIGN_RAW): New accessor.
28356 (TYPE_ALIGN): Use it.
28358 2021-03-01 Jan Hubicka <jh@suse.cz>
28361 * ipa-fnsummary.c (compute_fn_summary): Fix sanity check.
28363 2021-03-01 Eric Botcazou <ebotcazou@adacore.com>
28366 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
28367 point the hard frame pointer to the SSE register save area instead
28368 of the general register save area. Perform only minimal adjustment
28369 for small frames if it is initially not correctly aligned.
28370 (ix86_expand_prologue): Remove early saves for a SEH target.
28371 * config/i386/winnt.c (struct seh_frame_state): Document constraint.
28373 2021-02-28 Jakub Jelinek <jakub@redhat.com>
28376 * ipa.c (symbol_table::remove_unreachable_nodes): Fix a comment
28377 typo - referneced -> referenced.
28378 * tree.c (component_ref_size): Fix comment typo -
28379 refernce -> reference.
28380 * tree-ssa-alias.c (access_path_may_continue_p): Fix comment typo -
28381 traling -> trailing.
28382 (aliasing_component_refs_p): Fix comment typos -
28383 refernce -> reference and refernece -> reference and
28384 traling -> trailing.
28385 (nonoverlapping_refs_since_match_p): Fix comment typo -
28386 referneces -> references.
28387 * doc/invoke.texi (--param modref-max-bases): Fix a typo -
28388 referneces -> references.
28390 2021-02-27 Iain Sandoe <iain@sandoe.co.uk>
28392 * config/host-darwin.c (darwin_gt_pch_use_address): Modify
28393 diagnostic message to avoid use of a contraction and format
28396 2021-02-27 Jakub Jelinek <jakub@redhat.com>
28399 * gcse.c (gcse_or_cprop_is_too_expensive): Use %wu instead of
28400 HOST_WIDE_INT_PRINT_UNSIGNED in warning format string.
28401 * ipa-devirt.c (ipa_odr_read_section): Use %wd instead of
28402 HOST_WIDE_INT_PRINT_DEC in inform format string. Fix comment
28405 2021-02-26 Richard Biener <rguenther@suse.de>
28407 PR middle-end/99281
28408 * expr.c (store_field): For calls with return-slot optimization
28409 and addressable return type expand the store directly.
28411 2021-02-26 Richard Biener <rguenther@suse.de>
28414 * builtins.c (warn_string_no_nul): Fix diagnostic formatting.
28416 2021-02-26 Peter Bergner <bergner@linux.ibm.com>
28419 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Replace assert
28422 2021-02-26 Aaron Sawdey <acsawdey@linux.ibm.com>
28424 * config.gcc: Add rs6000-pcrel-opt.o.
28425 * config/rs6000/rs6000-pcrel-opt.c: New file.
28426 * config/rs6000/pcrel-opt.md: New file.
28427 * config/rs6000/predicates.md: Add d_form_memory predicate.
28428 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_PCREL_OPT.
28429 * config/rs6000/rs6000-passes.def: Add pass_pcrel_opt.
28430 * config/rs6000/rs6000-protos.h: Add reg_to_non_prefixed(),
28431 pcrel_opt_valid_mem_p(), output_pcrel_opt_reloc(),
28432 and make_pass_pcrel_opt().
28433 * config/rs6000/rs6000.c (reg_to_non_prefixed): Make global.
28434 (rs6000_option_override_internal): Add pcrel-opt.
28435 (rs6000_delegitimize_address): Support pcrel-opt.
28436 (rs6000_opt_masks): Add pcrel-opt.
28437 (pcrel_opt_valid_mem_p): New function.
28438 (reg_to_non_prefixed): Make global.
28439 (rs6000_asm_output_opcode): Reset prepend_p_to_next_insn.
28440 (output_pcrel_opt_reloc): New function.
28441 * config/rs6000/rs6000.md (loads_extern_addr): New attr.
28442 (pcrel_extern_addr): Set loads_extern_addr.
28443 Add include for pcrel-opt.md.
28444 * config/rs6000/rs6000.opt: Add -mpcrel-opt.
28445 * config/rs6000/t-rs6000: Add rules for pcrel-opt.c and
28448 2021-02-26 YunQiang Su <yunqiang.su@cipunited.com>
28451 * config/mips/mips.c (mips_expand_ext_as_unaligned_load):
28452 If TARGET_64BIT and dest is SUBREG, we check the width, if it
28453 equal to SImode, we use SImode operation, just like what we are
28456 2021-02-26 Marek Polacek <polacek@redhat.com>
28458 * builtins.c (warn_for_access): Fix typos.
28460 2021-02-25 Iain Sandoe <iain@sandoe.co.uk>
28462 * config/aarch64/aarch64.md (<optab>_rol<mode>3): Add a '#'
28463 mark in front of the immediate quantity.
28464 (<optab>_rolsi3_uxtw): Likewise.
28466 2021-02-25 Richard Earnshaw <rearnsha@arm.com>
28469 * config/arm/thumb2.md (nonsecure_call_reg_thumb2_fpcxt): New pattern.
28470 (nonsecure_call_value_reg_thumb2_fpcxt): Likewise.
28471 (nonsecure_call_reg_thumb2): Restrict to using r4 for the callee
28472 address and disable when the FPCXT is not available.
28473 (nonsecure_call_value_reg_thumb2): Likewise.
28475 2021-02-25 Nathan Sidwell <nathan@acm.org>
28478 * doc/invoke.texi (flang-info-module-cmi): Renamed option.
28480 2021-02-25 Tamar Christina <tamar.christina@arm.com>
28482 * tree-vect-slp.c (optimize_load_redistribution_1): Abort on NULL nodes.
28484 2021-02-25 Richard Biener <rguenther@suse.de>
28486 PR tree-optimization/99253
28487 * tree-vect-loop.c (check_reduction_path): First compute
28488 code, then verify out-of-loop uses.
28490 2021-02-25 Jakub Jelinek <jakub@redhat.com>
28493 * match.pd ((T)(A) + CST -> (T)(A + CST)): Add :s to convert.
28495 2021-02-25 Jakub Jelinek <jakub@redhat.com>
28497 PR tree-optimization/80635
28498 * tree-vrp.c (vrp_simplify_cond_using_ranges): Also handle
28499 VIEW_CONVERT_EXPR if modes are the same, innerop is integral and
28500 has mode precision.
28502 2021-02-25 Richard Biener <rguenther@suse.de>
28504 * tree-vect-slp.c (optimize_load_redistribution_1): Delay
28505 load_map population.
28506 (vect_match_slp_patterns_2): Revert part of last change.
28507 (vect_analyze_slp): Do not interleave optimize_load_redistribution
28508 with pattern detection but do it afterwards. Dump the
28509 whole SLP graph after pattern recognition and load
28510 redistribution optimization finished.
28512 2021-02-24 Jakub Jelinek <jakub@redhat.com>
28515 * omp-low.c (struct omp_context): Add teams_nested_p and
28516 nonteams_nested_p members.
28517 (scan_omp_target): Diagnose teams nested inside of target with other
28518 directives strictly nested inside of the same target.
28519 (check_omp_nesting_restrictions): Set ctx->teams_nested_p or
28520 ctx->nonteams_nested_p as needed.
28522 2021-02-24 Vladimir N. Makarov <vmakarov@redhat.com>
28524 PR inline-asm/99123
28525 * lra-constraints.c (uses_hard_regs_p): Don't use decompose_mem_address.
28527 2021-02-24 Hans-Peter Nilsson <hp@axis.com>
28529 * config/cris/cris.c (cris_expand_prologue): Set
28530 current_function_static_stack_size, if flag_stack_usage_info.
28532 2021-02-24 Pat Haugen <pthaugen@linux.ibm.com>
28534 * config/rs6000/rs6000.c (next_insn_prefixed_p): Rename.
28535 (rs6000_final_prescan_insn): Adjust.
28536 (rs6000_asm_output_opcode): Likewise.
28538 2021-02-24 Martin Sebor <msebor@redhat.com>
28540 PR middle-end/97172
28541 * attribs.c (attr_access::free_lang_data): Clear attribute arg spec
28542 from function arguments.
28544 2021-02-24 Tamar Christina <tamar.christina@arm.com>
28546 PR tree-optimization/99220
28547 * tree-vect-slp.c (optimize_load_redistribution_1): Remove
28548 node from cache when it's about to be deleted.
28550 2021-02-24 Jakub Jelinek <jakub@redhat.com>
28552 PR tree-optimization/99225
28553 * fold-const.c (fold_binary_loc) <case NE_EXPR>: In (x & (1 << y)) != 0
28554 to ((x >> y) & 1) != 0 simplifications use build_one_cst instead of
28555 build_int_cst (..., 1). Formatting fixes.
28557 2021-02-24 Tamar Christina <tamar.christina@arm.com>
28559 PR tree-optimization/99149
28560 * tree-vect-slp-patterns.c (vect_detect_pair_op): Don't recreate the
28562 (vect_slp_reset_pattern): Remove.
28563 (complex_fma_pattern::matches): Remove call to vect_slp_reset_pattern.
28564 (complex_mul_pattern::build, complex_fma_pattern::build,
28565 complex_fms_pattern::build): Fix ref counts.
28566 * tree-vect-slp.c (vect_free_slp_tree): Undo SLP only pattern relevancy
28567 when node is being deleted.
28568 (vect_match_slp_patterns_2): Correct result of cache hit on patterns.
28569 (vect_schedule_slp): Invalidate SLP_TREE_REPRESENTATIVE of removed
28571 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize value.
28573 2021-02-24 Matthias Klose <doko@ubuntu.com>
28576 2020-12-07 Matthias Klose <doko@ubuntu.com>
28578 * genextract.c (print_header): Undefine ENABLE_RTL_CHECKING
28579 and ENABLE_RTL_FLAG_CHECKING.
28581 2021-02-24 Richard Biener <rguenther@suse.de>
28584 * builtins.c (fold_builtin_next_arg): Avoid NULL arg.
28586 2021-02-23 Peter Bergner <bergner@linux.ibm.com>
28588 * config/rs6000/mma.md (mma_assemble_pair): Rename from this...
28589 (vsx_assemble_pair): ...to this.
28590 (*mma_assemble_pair): Rename from this...
28591 (*vsx_assemble_pair): ...to this.
28592 (mma_disassemble_pair): Rename from this...
28593 (vsx_disassemble_pair): ...to this.
28594 (*mma_disassemble_pair): Rename from this...
28595 (*vsx_disassemble_pair): ...to this.
28596 * config/rs6000/rs6000-builtin.def (BU_MMA_V2, BU_MMA_V3,
28597 BU_COMPAT): New macros.
28598 (mma_assemble_pair): Rename from this...
28599 (vsx_assemble_pair): ...to this.
28600 (mma_disassemble_pair): Rename from this...
28601 (vsx_disassemble_pair): ...to this.
28602 (mma_assemble_pair): New compatibility built-in.
28603 (mma_disassemble_pair): Likewise.
28604 * config/rs6000/rs6000-call.c (struct builtin_compatibility): New.
28605 (RS6000_BUILTIN_COMPAT): Define.
28606 (bdesc_compat): New.
28607 (mma_expand_builtin): Use VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
28608 (rs6000_gimple_fold_mma_builtin): Use MMA_BUILTIN_DISASSEMBLE_PAIR
28609 and VSX_BUILTIN_ASSEMBLE_PAIR.
28610 (rs6000_init_builtins): Register compatibility built-ins.
28611 (mma_init_builtins): Use VSX_BUILTIN_ASSEMBLE_PAIR,
28612 VSX_BUILTIN_ASSEMBLE_PAIR_INTERNAL, VSX_BUILTIN_DISASSEMBLE_PAIR and
28613 VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
28614 * doc/extend.texi (__builtin_mma_assemble_pair): Rename from this...
28615 (__builtin_vsx_assemble_pair): ...to this.
28616 (__builtin_mma_disassemble_pair): Rename from this...
28617 (__builtin_vsx_disassemble_pair): ...to this.
28619 2021-02-23 Martin Liska <mliska@suse.cz>
28622 * ipa-icf.c (sem_variable::merge): Do not merge 2 variables
28623 with different alignment. That leads to an invalid red zone
28624 size allocated in runtime.
28626 2021-02-23 Jakub Jelinek <jakub@redhat.com>
28628 PR tree-optimization/99204
28629 * fold-const.c (fold_read_from_constant_string): Check that
28630 tree_fits_uhwi_p (index) rather than just that index is INTEGER_CST.
28632 2021-02-23 Segher Boessenkool <segher@kernel.crashing.org>
28633 Kewen Lin <linkw@gcc.gnu.org>
28635 * config/rs6000/rs6000.md (*rotl<mode>3_insert_3): Renamed to...
28636 (rotl<mode>3_insert_3): ...this.
28637 (plus_ior_xor): New code_iterator.
28638 (define_split for GPR rl*imi): New splitter.
28639 * config/rs6000/vsx.md (vsx_init_v4si): Use gen_rotldi3_insert_3
28640 for integer merging.
28642 2021-02-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28644 * config/aarch64/aarch64-tuning-flags.def (cse_sve_vl_constants):
28646 * config/aarch64/aarch64.md (add<mode>3): Force CONST_POLY_INT immediates
28647 into a register when the above is enabled.
28648 * config/aarch64/aarch64.c (neoversev1_tunings):
28649 AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
28650 (aarch64_rtx_costs): Use AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
28652 2021-02-22 Hans-Peter Nilsson <hp@axis.com>
28654 * config/cris/cris.c (cris_print_operand) <'T'>: Change
28655 valid operand from is now an addi mult-value to shift-value.
28656 * config/cris/cris.md (*addi): Change expression of scaled
28657 operand from mult to ashift.
28658 * config/cris/cris.md (*addi_reload): New insn_and_split.
28660 2021-02-22 John David Anglin <danglin@gcc.gnu.org>
28663 * config/pa/pa.c (TARGET_ASM_CAN_OUTPUT_MI_THUNK): Define as
28664 hook_bool_const_tree_hwi_hwi_const_tree_true.
28665 (pa_asm_output_mi_thunk): Add support for nonzero vcall_offset.
28667 2021-02-22 Andre Vieira <andre.simoesdiasvieira@arm.com>
28669 PR rtl-optimization/98791
28670 * ira-conflicts.c (process_regs_for_copy): Don't create allocno copies
28671 for unordered modes.
28673 2021-02-22 Martin Liska <mliska@suse.cz>
28675 * tree-inline.c (inline_forbidden_p): Set
28676 inline_forbidden_reason.
28678 2021-02-22 Richard Biener <rguenther@suse.de>
28680 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Dump
28683 2021-02-22 Richard Biener <rguenther@suse.de>
28685 PR tree-optimization/99165
28686 * gimple-ssa-store-merging.c (pass_store_merging::process_store):
28687 Accumulate changed to ret.
28689 2021-02-21 Uros Bizjak <ubizjak@gmail.com>
28692 2020-12-09 Uroš Bizjak <ubizjak@gmail.com>
28694 * config/i386/i386.h (REG_ALLOC_ORDER): Remove
28696 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
28699 * config/s390/vector.md (trunctf<DFP_ALL:mode>2_vr): New
28701 (trunctf<DFP_ALL:mode>2): Likewise.
28702 (trunctdtf2_vr): Likewise.
28703 (trunctdtf2): Likewise.
28704 (extend<DFP_ALL:mode>tf2_vr): Likewise.
28705 (extend<DFP_ALL:mode>tf2): Likewise.
28706 (extendtftd2_vr): Likewise.
28707 (extendtftd2): Likewise.
28709 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
28711 * config/s390/vector.md (*fprx2_to_tf): Rename to fprx2_to_tf,
28712 add memory alternative.
28713 (tf_to_fprx2): New pattern.
28715 2021-02-19 Martin Sebor <msebor@redhat.com>
28718 * attribs.c (init_attr_rdwr_indices): Guard vblist use.
28719 (attr_access::free_lang_data): Remove a spurious test.
28721 2021-02-19 Nathan Sidwell <nathan@acm.org>
28723 * doc/invoke.texi (flang-info-module-read): Document.
28725 2021-02-19 Martin Liska <mliska@suse.cz>
28727 PR translation/99167
28728 * params.opt: Fix typo.
28730 2021-02-19 Richard Biener <rguenther@suse.de>
28732 PR middle-end/99122
28733 * tree-inline.c (inline_forbidden_p): Do not inline functions
28734 with VLA arguments or return value.
28736 2021-02-19 Jakub Jelinek <jakub@redhat.com>
28739 * config/arm/arm.md (*stack_protect_combined_set_insn,
28740 *stack_protect_combined_test_insn): If force_const_mem result
28741 is not valid general operand, force its address into the destination
28744 2021-02-19 Jakub Jelinek <jakub@redhat.com>
28747 * tree-cfg.c (gimple_merge_blocks): If bb a starts with eh landing
28748 pad or non-local label, put FORCED_LABELs from bb b after that label
28749 rather than before it.
28751 2021-02-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
28754 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3): Use
28755 expand_vector_broadcast' to emit the vec_duplicate operand.
28757 2021-02-18 Vladimir N. Makarov <vmakarov@redhat.com>
28759 PR rtl-optimization/96264
28760 * lra-remat.c (reg_overlap_for_remat_p): Check also output insn
28763 2021-02-18 H.J. Lu <hjl.tools@gmail.com>
28766 * varasm.c (get_section): Replace SUPPORTS_SHF_GNU_RETAIN with
28767 looking up the retain attribute.
28768 (resolve_unique_section): Likewise.
28769 (get_variable_section): Likewise.
28770 (switch_to_section): Likewise. Warn when a symbol without the
28771 retain attribute and a symbol with the retain attribute are
28772 placed in the section with the same name, instead of the used
28774 * doc/extend.texi: Document the "retain" attribute.
28776 2021-02-18 Nathan Sidwell <nathan@acm.org>
28779 * doc/invoke.texi (flang-info-include-translate): Document header
28782 2021-02-18 Richard Biener <rguenther@suse.de>
28784 PR middle-end/99122
28785 * ipa-fnsummary.c (analyze_function_body): Set
28786 CIF_FUNCTION_NOT_INLINABLE for VLA parameter calls.
28787 * tree-inline.c (insert_init_debug_bind): Pass NULL for
28788 error_mark_node values.
28789 (force_value_to_type): Do not build V_C_Es for WITH_SIZE_EXPR
28791 (setup_one_parameter): Delay force_value_to_type until when
28794 2021-02-18 Hans-Peter Nilsson <hp@axis.com>
28796 PR tree-optimization/99142
28797 * match.pd (clz cmp 0): Gate replacement on single_use of clz result.
28799 2021-02-18 Jakub Jelinek <jakub@redhat.com>
28801 * wide-int-bitmask.h (wide_int_bitmask::wide_int_bitmask (),
28802 wide_int_bitmask::wide_int_bitmask (uint64_t),
28803 wide_int_bitmask::wide_int_bitmask (uint64_t, uint64_t),
28804 wide_int_bitmask::operator ~ () const,
28805 wide_int_bitmask::operator | (wide_int_bitmask) const,
28806 wide_int_bitmask::operator & (wide_int_bitmask) const): Use constexpr
28808 * config/i386/i386.h (PTA_3DNOW, PTA_3DNOW_A, PTA_64BIT, PTA_ABM,
28809 PTA_AES, PTA_AVX, PTA_BMI, PTA_CX16, PTA_F16C, PTA_FMA, PTA_FMA4,
28810 PTA_FSGSBASE, PTA_LWP, PTA_LZCNT, PTA_MMX, PTA_MOVBE, PTA_NO_SAHF,
28811 PTA_PCLMUL, PTA_POPCNT, PTA_PREFETCH_SSE, PTA_RDRND, PTA_SSE, PTA_SSE2,
28812 PTA_SSE3, PTA_SSE4_1, PTA_SSE4_2, PTA_SSE4A, PTA_SSSE3, PTA_TBM,
28813 PTA_XOP, PTA_AVX2, PTA_BMI2, PTA_RTM, PTA_HLE, PTA_PRFCHW, PTA_RDSEED,
28814 PTA_ADX, PTA_FXSR, PTA_XSAVE, PTA_XSAVEOPT, PTA_AVX512F, PTA_AVX512ER,
28815 PTA_AVX512PF, PTA_AVX512CD, PTA_NO_TUNE, PTA_SHA, PTA_PREFETCHWT1,
28816 PTA_CLFLUSHOPT, PTA_XSAVEC, PTA_XSAVES, PTA_AVX512DQ, PTA_AVX512BW,
28817 PTA_AVX512VL, PTA_AVX512IFMA, PTA_AVX512VBMI, PTA_CLWB, PTA_MWAITX,
28818 PTA_CLZERO, PTA_NO_80387, PTA_PKU, PTA_AVX5124VNNIW, PTA_AVX5124FMAPS,
28819 PTA_AVX512VPOPCNTDQ, PTA_SGX, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES,
28820 PTA_AVX512VBMI2, PTA_VPCLMULQDQ, PTA_AVX512BITALG, PTA_RDPID,
28821 PTA_PCONFIG, PTA_WBNOINVD, PTA_AVX512VP2INTERSECT, PTA_PTWRITE,
28822 PTA_AVX512BF16, PTA_WAITPKG, PTA_MOVDIRI, PTA_MOVDIR64B, PTA_ENQCMD,
28823 PTA_CLDEMOTE, PTA_SERIALIZE, PTA_TSXLDTRK, PTA_AMX_TILE, PTA_AMX_INT8,
28824 PTA_AMX_BF16, PTA_UINTR, PTA_HRESET, PTA_KL, PTA_WIDEKL, PTA_AVXVNNI,
28825 PTA_X86_64_BASELINE, PTA_X86_64_V2, PTA_X86_64_V3, PTA_X86_64_V4,
28826 PTA_CORE2, PTA_NEHALEM, PTA_WESTMERE, PTA_SANDYBRIDGE, PTA_IVYBRIDGE,
28827 PTA_HASWELL, PTA_BROADWELL, PTA_SKYLAKE, PTA_SKYLAKE_AVX512,
28828 PTA_CASCADELAKE, PTA_COOPERLAKE, PTA_CANNONLAKE, PTA_ICELAKE_CLIENT,
28829 PTA_ICELAKE_SERVER, PTA_TIGERLAKE, PTA_SAPPHIRERAPIDS, PTA_ALDERLAKE,
28830 PTA_KNL, PTA_BONNELL, PTA_SILVERMONT, PTA_GOLDMONT, PTA_GOLDMONT_PLUS,
28831 PTA_TREMONT, PTA_KNM): Use constexpr instead of const.
28833 2021-02-18 Jakub Jelinek <jakub@redhat.com>
28835 PR middle-end/99109
28836 * gimple-array-bounds.cc (build_zero_elt_array_type): Rename to ...
28837 (build_printable_array_type): ... this. Add nelts argument. For
28838 overaligned eltype, use TYPE_MAIN_VARIANT (eltype) instead. If
28839 nelts, call build_array_type_nelts.
28840 (array_bounds_checker::check_mem_ref): Use build_printable_array_type
28841 instead of build_zero_elt_array_type and build_array_type_nelts.
28843 2021-02-18 Jakub Jelinek <jakub@redhat.com>
28846 * config/i386/i386.c (distance_non_agu_define): Don't call
28847 extract_insn_cached here.
28848 (ix86_lea_outperforms): Save and restore recog_data around call
28849 to distance_non_agu_define and distance_agu_use.
28850 (ix86_ok_to_clobber_flags): Remove.
28851 (ix86_avoid_lea_for_add): Don't call ix86_ok_to_clobber_flags.
28852 (ix86_avoid_lea_for_addr): Likewise. Adjust function comment.
28853 * config/i386/i386.md (*lea<mode>): Change from define_insn_and_split
28854 into define_insn. Move the splitting to define_peephole2 and
28855 check there using peep2_regno_dead_p if FLAGS_REG is dead.
28857 2021-02-17 Julian Brown <julian@codesourcery.com>
28859 * gimplify.c (gimplify_scan_omp_clauses): Handle ATTACH_DETACH
28862 2021-02-17 Xi Ruoyao <xry111@mengyan1223.wang>
28865 * config/mips/mips.c (mips_symbol_insns): Do not use
28866 MSA_SUPPORTED_MODE_P if mode is MAX_MACHINE_MODE.
28868 2021-02-16 Vladimir N. Makarov <vmakarov@redhat.com>
28870 PR inline-asm/98096
28871 * stmt.c (resolve_operand_name_1): Take inout operands into account
28872 for access to labels by names.
28873 * doc/extend.texi: Describe counting operands for accessing labels.
28875 2021-02-16 Richard Biener <rguenther@suse.de>
28877 PR tree-optimization/38474
28878 * tree-ssa-structalias.c (variable_info::address_taken): New.
28879 (new_var_info): Initialize address_taken.
28880 (process_constraint): Set address_taken.
28881 (solve_constraints): Use the new address_taken flag rather
28882 than is_reg_var for sorting variables.
28883 (dump_constraint): Dump the variable number if the name
28886 2021-02-16 Jakub Jelinek <jakub@redhat.com>
28889 * tree-vect-stmts.c (vectorizable_simd_clone_call): For num_calls != 1
28890 multiply by 4096 and for inbranch by 8192.
28891 * config/i386/i386.c (ix86_simd_clone_usable): For TARGET_AVX512F,
28892 return 3, 2 or 1 for mangle letters 'b', 'c' or 'd'.
28894 2021-02-15 Maya Rashish <coypu@sdf.org>
28896 * config/aarch64/aarch64.c (aarch64_init_builtins):
28897 Call SUBTARGET_INIT_BUILTINS.
28899 2021-02-15 Peter Bergner <bergner@linux.ibm.com>
28901 PR rtl-optimization/98872
28902 * init-regs.c (initialize_uninitialized_regs): Skip initialization
28903 if CONST0_RTX is NULL.
28905 2021-02-15 Richard Sandiford <richard.sandiford@arm.com>
28907 PR rtl-optimization/98863
28908 * rtl-ssa/functions.h (function_info::bb_live_out_info): Delete.
28909 (function_info::build_info): Turn into a declaration, moving the
28910 definition to internals.h.
28911 (function_info::bb_walker): Declare.
28912 (function_info::create_reg_use): Likewise.
28913 (function_info::calculate_potential_phi_regs): Take a build_info
28915 (function_info::place_phis, function_info::create_ebbs): Declare.
28916 (function_info::calculate_ebb_live_in_for_debug): Likewise.
28917 (function_info::populate_backedge_phis): Delete.
28918 (function_info::start_block, function_info::end_block): Declare.
28919 (function_info::populate_phi_inputs): Delete.
28920 (function_info::m_potential_phi_regs): Move information to build_info.
28921 * rtl-ssa/internals.h: New file.
28922 (function_info::bb_phi_info): New class.
28923 (function_info::build_info): Moved from functions.h.
28924 Add a constructor and destructor.
28925 (function_info::build_info::ebb_use): Delete.
28926 (function_info::build_info::ebb_def): Likewise.
28927 (function_info::build_info::bb_live_out): Likewise.
28928 (function_info::build_info::tmp_ebb_live_in_for_debug): New variable.
28929 (function_info::build_info::potential_phi_regs): Likewise.
28930 (function_info::build_info::potential_phi_regs_for_debug): Likewise.
28931 (function_info::build_info::ebb_def_regs): Likewise.
28932 (function_info::build_info::bb_phis): Likewise.
28933 (function_info::build_info::bb_mem_live_out): Likewise.
28934 (function_info::build_info::bb_to_rpo): Likewise.
28935 (function_info::build_info::def_stack): Likewise.
28936 (function_info::build_info::old_def_stack_limit): Likewise.
28937 * rtl-ssa/internals.inl (function_info::build_info::record_reg_def):
28938 Remove the regno argument. Push the previous definition onto the
28939 definition stack where necessary.
28940 * rtl-ssa/accesses.cc: Include internals.h.
28941 * rtl-ssa/changes.cc: Likewise.
28942 * rtl-ssa/blocks.cc: Likewise.
28943 (function_info::build_info::build_info): Define.
28944 (function_info::build_info::~build_info): Likewise.
28945 (function_info::bb_walker): New class.
28946 (function_info::bb_walker::bb_walker): Define.
28947 (function_info::add_live_out_use): Convert a logarithmic-complexity
28948 test into a linear one. Allow the same definition to be passed
28950 (function_info::calculate_potential_phi_regs): Moved from
28951 functions.cc. Take a build_info parameter and store the
28952 information there instead.
28953 (function_info::place_phis): New function.
28954 (function_info::add_entry_block_defs): Update call to record_reg_def.
28955 (function_info::calculate_ebb_live_in_for_debug): New function.
28956 (function_info::add_phi_nodes): Use bb_phis to decide which
28957 registers need phi nodes and initialize ebb_def_regs accordingly.
28958 Do not add degenerate phis here.
28959 (function_info::add_artificial_accesses): Use create_reg_use.
28960 Assert that all definitions are listed in the DF LR sets.
28961 Update call to record_reg_def.
28962 (function_info::record_block_live_out): Record live-out register
28963 values in the phis of successor blocks. Use the live-out set
28964 when processing the last block in an EBB, instead of always
28965 using the live-in sets of successor blocks. AND the live sets
28966 with the set of registers that have been defined in the EBB,
28967 rather than with all potential phi registers. Cope correctly
28968 with branches back to the start of the current EBB.
28969 (function_info::start_block): New function.
28970 (function_info::end_block): Likewise.
28971 (function_info::populate_phi_inputs): Likewise.
28972 (function_info::create_ebbs): Likewise.
28973 (function_info::process_all_blocks): Rewrite into a multi-phase
28975 * rtl-ssa/functions.cc: Include internals.h.
28976 (function_info::calculate_potential_phi_regs): Move to blocks.cc.
28977 (function_info::init_function_data): Remove caller.
28978 * rtl-ssa/insns.cc: Include internals.h
28979 (function_info::create_reg_use): New function. Lazily any
28980 degenerate phis needed by the linear RPO view.
28981 (function_info::record_use): Use create_reg_use. When processing
28982 debug uses, use potential_phi_regs and test it before checking
28983 whether the register is live on entry to the current EBB. Lazily
28984 calculate ebb_live_in_for_debug.
28985 (function_info::record_call_clobbers): Update call to record_reg_def.
28986 (function_info::record_def): Likewise.
28988 2021-02-15 Martin Liska <mliska@suse.cz>
28990 * toplev.c (init_asm_output): Free output of
28991 gen_command_line_string function.
28992 (process_options): Likewise.
28994 2021-02-15 Martin Liska <mliska@suse.cz>
28996 * params.opt: Add 2 missing Param keywords.
28998 2021-02-15 Eric Botcazou <ebotcazou@adacore.com>
29000 * df-core.c (df_worklist_dataflow_doublequeue): Use proper cast.
29002 2021-02-15 Jakub Jelinek <jakub@redhat.com>
29004 PR tree-optimization/99079
29005 * match.pd (A % (pow2pcst << N) -> A & ((pow2pcst << N) - 1)): Remove
29006 useless tree_nop_conversion_p (type, TREE_TYPE (@3)) check. Instead
29007 require both type and TREE_TYPE (@1) to be integral types and either
29008 type having smaller or equal precision, or TREE_TYPE (@1) being
29009 unsigned type, or type being signed type. If TREE_TYPE (@1)
29010 doesn't have wrapping overflow, perform the subtraction of one in
29013 2021-02-14 Jan Hubicka <hubicka@ucw.cz>
29014 Richard Biener <rguether@suse.de>
29017 * ipa-reference.c (ipa_init): Only conditinally initialize
29018 reference_vars_to_consider.
29019 (propagate): Conditionally deninitialize reference_vars_to_consider.
29020 (ipa_reference_write_optimization_summary): Sanity check that
29021 reference_vars_to_consider is not allocated.
29023 2021-02-13 Levy Hsu <admin@levyhsu.com>
29026 * config/riscv/riscv-shorten-memrefs.c (pass_shorten_memrefs): Add
29027 extend parameter to get_si_mem_base_reg declaration.
29028 (get_si_mem_base_reg): Add extend parameter. Set it.
29029 (analyze): Pass extend arg to get_si_mem_base_reg.
29030 (transform): Likewise. Use it when rewriting mems.
29031 * config/riscv/riscv.c (riscv_legitimize_move): Check for subword
29032 loads and emit sign/zero extending load followed by subreg move.
29034 2021-02-13 Jim Wilson <jimw@sifive.com>
29037 * config/riscv/riscv.c (riscv_compressed_lw_address_p): Drop early
29038 exit when !reload_completed. Only perform check for compressed reg
29039 if reload_completed.
29040 (riscv_rtx_costs): In MEM case, when optimizing for size and
29041 shorten memrefs, if not compressible, then increase cost.
29043 2021-02-13 Jakub Jelinek <jakub@redhat.com>
29045 PR rtl-optimization/98439
29046 * recog.c (pass_split_before_regstack::gate): Enable even when
29047 pass_split_before_sched2 is enabled if -fselective-scheduling2 is
29050 2021-02-13 Jakub Jelinek <jakub@redhat.com>
29053 * config/i386/mmx.md (*mmx_pshufd_1): Add a combine splitter for
29054 swap of V2SImode elements in memory into DImode memory rotate by 32.
29056 2021-02-12 Martin Sebor <msebor@redhat.com>
29058 * tree-pretty-print.c (print_generic_expr_to_str): Update comment.
29060 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
29062 * rtl-ssa/accesses.cc (function_info::make_use_available): Use
29063 m_temp_obstack rather than m_obstack to allocate the temporary use.
29065 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
29067 * df-problems.c (df_lr_bb_local_compute): Treat partial definitions
29068 as read-modify operations.
29070 2021-02-12 Richard Biener <rguenther@suse.de>
29072 PR middle-end/38474
29073 * ipa-fnsummary.c (unmodified_parm_1): Only walk when
29074 fbi->aa_walk_budget is bigger than zero. Update
29075 fbi->aa_walk_budget.
29076 (param_change_prob): Likewise.
29077 * ipa-prop.c (detect_type_change_from_memory_writes):
29078 Properly account walk_aliased_vdefs.
29079 (parm_preserved_before_stmt_p): Canonicalize updates.
29080 (parm_ref_data_preserved_p): Likewise.
29081 (parm_ref_data_pass_through_p): Likewise.
29082 (determine_known_aggregate_parts): Account own alias queries.
29084 2021-02-12 Martin Liska <mliska@suse.cz>
29086 * opts-common.c (decode_cmdline_option): Release werror_arg.
29087 * opts.c (gen_producer_string): Release output of
29088 gen_command_line_string.
29090 2021-02-12 Richard Biener <rguenther@suse.de>
29092 PR tree-optimization/38474
29093 * params.opt (-param=max-store-chains-to-track=): New param.
29094 (-param=max-stores-to-track=): Likewise.
29095 * doc/invoke.texi (max-store-chains-to-track): Document.
29096 (max-stores-to-track): Likewise.
29097 * gimple-ssa-store-merging.c (pass_store_merging::m_n_chains):
29099 (pass_store_merging::m_n_stores): Likewise.
29100 (pass_store_merging::terminate_and_process_chain): Update
29101 m_n_stores and m_n_chains.
29102 (pass_store_merging::process_store): Likewise. Terminate
29103 oldest chains if the number of stores or chains get too large.
29104 (imm_store_chain_info::terminate_and_process_chain): Dump
29107 2021-02-11 Eric Botcazou <ebotcazou@adacore.com>
29109 * config/i386/winnt.c (i386_pe_seh_unwind_emit): When switching to
29110 the cold section, emit a nop before the directive if the previous
29111 active instruction can throw.
29113 2021-02-11 Peter Bergner <bergner@linux.ibm.com>
29116 * config/rs6000/predicates.md (mma_assemble_input_operand): Restrict
29117 memory addresses that are legal for quad word accesses.
29119 2021-02-11 Andrea Corallo <andrea.corallo@arm.com>
29122 * config/arm/thumb2.md (*doloop_end_internal): Generate
29123 alternative sequence to handle long range branches.
29125 2021-02-11 Joel Hutton <joel.hutton@arm.com>
29127 PR tree-optimization/98772
29128 * optabs-tree.c (supportable_half_widening_operation): New function
29129 to check for supportable V8QI->V8HI widening patterns.
29130 * optabs-tree.h (supportable_half_widening_operation): New function.
29131 * tree-vect-stmts.c (vect_create_half_widening_stmts): New function
29132 to create promotion stmts for V8QI->V8HI widening patterns.
29133 (vectorizable_conversion): Add case for V8QI->V8HI.
29135 2021-02-11 Richard Biener <rguenther@suse.de>
29137 * sparseset.h (SPARSESET_ELT_BITS): Remove.
29138 (SPARSESET_ELT_TYPE): Use unsigned int.
29139 * fwprop.c: Do not include sparseset.h.
29141 2021-02-10 Jakub Jelinek <jakub@redhat.com>
29144 * varasm.c (declare_weak): For -fsyntax-only, allow even
29145 TREE_ASM_WRITTEN function decls.
29147 2021-02-10 Jakub Jelinek <jakub@redhat.com>
29150 * config/i386/sse.md (fix<fixunssuffix>_truncv2sfv2di2,
29151 <insn>v8qiv8hi2, <insn>v8qiv8si2, <insn>v4qiv4si2, <insn>v4hiv4si2,
29152 <insn>v8qiv8di2, <insn>v4qiv4di2, <insn>v2qiv2di2, <insn>v4hiv4di2,
29153 <insn>v2hiv2di2, <insn>v2siv2di2): Force operands[1] into REG before
29154 calling simplify_gen_subreg on it.
29156 2021-02-10 Martin Liska <mliska@suse.cz>
29158 * config/nvptx/nvptx.c (nvptx_option_override): Use
29159 flag_patchable_function_entry instead of the removed
29160 function_entry_patch_area_size.
29162 2021-02-10 Martin Liska <mliska@suse.cz>
29164 PR tree-optimization/99002
29165 PR tree-optimization/99026
29166 * gimple-if-to-switch.cc (if_chain::is_beneficial): Fix memory
29167 leak when adjacent cases are merged.
29168 * tree-switch-conversion.c (switch_decision_tree::analyze_switch_statement): Use
29170 (make_pass_lower_switch): Remove trailing whitespace.
29171 * tree-switch-conversion.h (release_clusters): New.
29173 2021-02-10 Richard Biener <rguenther@suse.de>
29175 PR rtl-optimization/99054
29176 * cfgrtl.c (rtl-optimization/99054): Return an auto_vec.
29177 (fixup_partitions): Adjust.
29178 (rtl_verify_edges): Likewise.
29180 2021-02-10 Jakub Jelinek <jakub@redhat.com>
29182 PR middle-end/99007
29183 * gimplify.c (gimplify_scan_omp_clauses): For MEM_REF on reductions,
29184 temporarily disable gimplify_ctxp->into_ssa around gimplify_expr
29187 2021-02-10 Richard Biener <rguenther@suse.de>
29190 * ipa-pure-const.c (propagate_malloc): Use an auto_vec<>
29193 2021-02-10 Richard Biener <rguenther@suse.de>
29195 PR tree-optimization/99024
29196 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Only
29197 clear loop->aux if it is associated with the destroyed loop_vinfo.
29199 2021-02-10 Martin Liska <mliska@suse.cz>
29201 PR tree-optimization/99002
29202 * gimple-if-to-switch.cc (find_conditions): Fix memory leak
29205 2021-02-10 Martin Liska <mliska@suse.cz>
29208 * ipa-icf.c (sem_item::add_reference): Fix memory leak when
29209 a reference exists.
29211 2021-02-10 Jakub Jelinek <jakub@redhat.com>
29214 * dwarf2out.c (prune_unused_types_walk): Mark DW_TAG_variable DIEs
29215 at class scope for DWARF5+.
29217 2021-02-09 Eric Botcazou <ebotcazou@adacore.com>
29219 PR rtl-optimization/96015
29220 * reorg.c (skip_consecutive_labels): Minor comment tweaks.
29221 (relax_delay_slots): When deleting a jump to the next active
29222 instruction over a barrier, first delete the barrier if the
29223 jump is the only way to reach the target label.
29225 2021-02-09 Andre Vieira <andre.simoesdiasvieira@arm.com>
29227 * config/aarch64/aarch64-cost-tables.h: Add entries for vect.mul.
29228 * config/aarch64/aarch64.c (aarch64_rtx_mult_cost): Use vect.mul for
29229 vector multiplies and vect.alu for SSRA.
29230 * config/arm/aarch-common-protos.h (struct vector_cost_table): Define
29231 vect.mul cost field.
29232 * config/arm/aarch-cost-tables.h: Add entries for vect.mul.
29233 * config/arm/arm.c: Likewise.
29235 2021-02-09 Richard Biener <rguenther@suse.de>
29237 PR tree-optimization/98863
29238 * tree-ssa-sccvn.h (vn_avail::next_undo): Add.
29239 * tree-ssa-sccvn.c (last_pushed_avail): New global.
29240 (rpo_elim::eliminate_push_avail): Chain pushed avails.
29241 (unwind_state::avail_top): Add.
29242 (do_unwind): Rewrite unwinding of avail entries.
29243 (do_rpo_vn): Initialize last_pushed_avail and
29244 avail_top of the undo state.
29246 2021-02-09 Jakub Jelinek <jakub@redhat.com>
29248 PR middle-end/99004
29249 * calls.c (maybe_warn_rdwr_sizes): Change s0 and s1 type from
29250 const char * to char * and free those pointers after use.
29252 2021-02-09 Richard Biener <rguenther@suse.de>
29254 PR tree-optimization/99017
29255 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Allow
29256 zero vector cost entries.
29258 2021-02-08 Andre Vieira <andre.simoesdiasvieira@arm.com>
29260 PR middle-end/98974
29261 * tree-vect-stmts.c (vectorizable_condition): Remove shadow vec_num
29262 parameter in vectorizable_condition.
29264 2021-02-08 Richard Biener <rguenther@suse.de>
29267 * tree.c (walk_tree_1): Walk VECTOR_CST elements.
29269 2021-02-08 Martin Liska <mliska@suse.cz>
29272 * cfgexpand.c (pass_expand::execute): Parse per-function option
29273 flag_patchable_function_entry and use it.
29274 * common.opt: Remove function_entry_patch_area_size and
29275 function_entry_patch_area_start global variables.
29276 * opts.c (parse_and_check_patch_area): New function.
29277 (common_handle_option): Use it.
29278 * opts.h (parse_and_check_patch_area): New function.
29279 * toplev.c (process_options): Parse and use
29280 function_entry_patch_area_size.
29282 2021-02-08 Martin Sebor <msebor@redhat.com>
29284 * doc/extend.texi (attribute malloc): Correct typos.
29286 2021-02-05 Nathan Sidwell <nathan@acm.org>
29289 * gcc.c (driver::maybe_run_linker): Check for input file
29290 accessibility if not linking.
29292 2021-02-05 Richard Biener <rguenther@suse.de>
29294 PR tree-optimization/98855
29295 * tree-vectorizer.h (add_stmt_cost): New overload.
29296 * tree-vect-slp.c (li_cost_vec_cmp): New.
29297 (vect_bb_slp_scalar_cost): Cost individual loop regions
29298 separately. Account for the scalar instance root stmt.
29300 2021-02-05 Tom de Vries <tdevries@suse.de>
29303 * tree-switch-conversion.c (jump_table_cluster::emit): Add loc
29305 (bit_test_cluster::emit): Reuse location_t for newly created
29307 (switch_decision_tree::try_switch_expansion): Preserve
29309 * tree-switch-conversion.h: Change function signatures.
29311 2021-02-05 Jakub Jelinek <jakub@redhat.com>
29314 * config/i386/i386-options.c (m_NONE, m_ALL): Define.
29315 * config/i386/x86-tune.def (X86_TUNE_BRANCH_PREDICTION_HINTS,
29316 X86_TUNE_PROMOTE_QI_REGS): Use m_NONE instead of 0U.
29317 (X86_TUNE_QIMODE_MATH): Use m_ALL instead of ~0U.
29319 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29321 * config/aarch64/aarch64-simd-builtins.def (get_high): Define builtin.
29322 * config/aarch64/aarch64-simd.md (aarch64_get_high<mode>): Define.
29323 * config/aarch64/arm_neon.h (__GET_HIGH): Delete.
29324 (vget_high_f16): Reimplement using new builtin.
29325 (vget_high_f32): Likewise.
29326 (vget_high_f64): Likewise.
29327 (vget_high_p8): Likewise.
29328 (vget_high_p16): Likewise.
29329 (vget_high_p64): Likewise.
29330 (vget_high_s8): Likewise.
29331 (vget_high_s16): Likewise.
29332 (vget_high_s32): Likewise.
29333 (vget_high_s64): Likewise.
29334 (vget_high_u8): Likewise.
29335 (vget_high_u16): Likewise.
29336 (vget_high_u32): Likewise.
29337 (vget_high_u64): Likewise.
29339 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29341 * config/aarch64/aarch64-simd-builtins.def (get_low): Define builtin.
29342 * config/aarch64/aarch64-simd.md (aarch64_get_low<mode>): Define.
29343 * config/aarch64/arm_neon.h (__GET_LOW): Delete.
29344 (vget_low_f16): Reimplement using new builtin.
29345 (vget_low_f32): Likewise.
29346 (vget_low_f64): Likewise.
29347 (vget_low_p8): Likewise.
29348 (vget_low_p16): Likewise.
29349 (vget_low_p64): Likewise.
29350 (vget_low_s8): Likewise.
29351 (vget_low_s16): Likewise.
29352 (vget_low_s32): Likewise.
29353 (vget_low_s64): Likewise.
29354 (vget_low_u8): Likewise.
29355 (vget_low_u16): Likewise.
29356 (vget_low_u32): Likewise.
29357 (vget_low_u64): Likewise.
29359 2021-02-05 Kito Cheng <kito.cheng@sifive.com>
29361 * gcc.c (print_multilib_info): Check all required argument is provided
29364 2021-02-05 liuhongt <hongtao.liu@intel.com>
29367 * config/i386/i386-expand.c (ix86_expand_sse_cmp): Don't
29368 generate integer mask comparison for 128/256-bits vector when
29369 op_true/op_false is NULL_RTX or CONSTM1_RTX/CONST0_RTX. Also
29370 delete redundant !maskcmp condition.
29371 (ix86_expand_int_vec_cmp): Ditto but no redundant deletion
29373 (ix86_expand_sse_movcc): Delete definition of maskcmp, add the
29374 condition directly to if (maskcmp), add extra check for
29375 cmpmode, it should be MODE_INT.
29376 (ix86_expand_fp_vec_cmp): Pass NULL to ix86_expand_sse_cmp's
29377 parameters op_true/op_false.
29378 (ix86_use_mask_cmp_p): New.
29380 2021-02-05 liuhongt <hongtao.liu@intel.com>
29383 * config/i386/x86-tune.def (X86_TUNE_AVX256_UNALIGNED_LOAD_OPTIMAL):
29384 Remove m_GENERIC from ~list.
29385 (X86_TUNE_AVX256_UNALIGNED_STORE_OPTIMAL): Ditto.
29387 2021-02-04 David Malcolm <dmalcolm@redhat.com>
29390 * diagnostic-show-locus.c (compatible_locations_p): Require
29391 locations in the same macro map to be either both from the
29392 macro definition, or both from the macro arguments.
29394 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
29396 * config/aarch64/aarch64-simd-builtins.def: Add
29397 [su]mull_hi_lane[q] builtin generator macros.
29398 * config/aarch64/aarch64-simd.md
29399 (aarch64_<su>mull_hi_lane<mode>_insn): Define.
29400 (aarch64_<su>mull_hi_lane<mode>): Define.
29401 (aarch64_<su>mull_hi_laneq<mode>_insn): Define.
29402 (aarch64_<su>mull_hi_laneq<mode>): Define.
29403 * config/aarch64/arm_neon.h (vmull_high_lane_s16): Use RTL
29404 builtin instead of inline asm.
29405 (vmull_high_lane_s32): Likewise.
29406 (vmull_high_lane_u16): Likewise.
29407 (vmull_high_lane_u32): Likewise.
29408 (vmull_high_laneq_s16): Likewise.
29409 (vmull_high_laneq_s32): Likewise.
29410 (vmull_high_laneq_u16): Likewise.
29411 (vmull_high_laneq_u32): Liekwise.
29413 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
29415 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_hi_n
29416 builtin generator macros.
29417 * config/aarch64/aarch64-simd.md
29418 (aarch64_<su>mull_hi_n<mode>_insn): Define.
29419 (aarch64_<su>mull_hi_n<mode>): Define.
29420 * config/aarch64/arm_neon.h (vmull_high_n_s16): Use RTL builtin
29421 instead of inline asm.
29422 (vmull_high_n_s32): Likewise.
29423 (vmull_high_n_u16): Likewise.
29424 (vmull_high_n_u32): Likewise.
29426 2021-02-04 Richard Biener <rguenther@suse.de>
29428 PR tree-optimization/98855
29429 * tree-vect-loop.c (vectorizable_phi): Do not cost
29430 single-argument PHIs.
29431 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Likewise.
29432 * tree-vect-stmts.c (vectorizable_bswap): Also perform
29433 costing for SLP operation.
29435 2021-02-04 Martin Liska <mliska@suse.cz>
29437 * doc/extend.texi: Mention -mprefer-vector-width in target
29440 2021-02-03 Martin Sebor <msebor@redhat.com>
29442 PR tree-optimization/98937
29443 * tree-ssa-strlen.c (strlen_dom_walker::~strlen_dom_walker): Define.
29444 Flush pointer_query cache.
29446 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
29448 * config/rs6000/genfusion.pl (gen_2logical): Add missing
29449 fixes based on patch review.
29450 * config/rs6000/fusion.md: Regenerate file.
29452 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
29454 * config/rs6000/t-rs6000: Comment out auto generation of
29457 2021-02-03 Andrew Stubbs <ams@codesourcery.com>
29459 * config/gcn/gcn-opts.h (enum processor_type): Add PROCESSOR_GFX908.
29460 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Add gfx908.
29461 (output_file_start): Add gfx908.
29462 * config/gcn/gcn.opt (gpu_type): Add gfx908.
29463 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Add march=gfx908.
29464 (MULTILIB_DIRNAMES): Add gfx908.
29465 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): New define.
29466 (main): Recognize gfx908.
29467 * config/gcn/t-omp-device: Add gfx908.
29469 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
29471 * config/aarch64/aarch64-simd-builtins.def: Add
29472 [su]mlsl_hi_lane[q] builtin macro generators.
29473 * config/aarch64/aarch64-simd.md
29474 (aarch64_<su>mlsl_hi_lane<mode>_insn): Define.
29475 (aarch64_<su>mlsl_hi_lane<mode>): Define.
29476 (aarch64_<su>mlsl_hi_laneq<mode>_insn): Define.
29477 (aarch64_<su>mlsl_hi_laneq<mode>): Define.
29478 * config/aarch64/arm_neon.h (vmlsl_high_lane_s16): Use RTL
29479 builtin instead of inline asm.
29480 (vmlsl_high_lane_s32): Likewise.
29481 (vmlsl_high_lane_u16): Likewise.
29482 (vmlsl_high_lane_u32): Likewise.
29483 (vmlsl_high_laneq_s16): Likewise.
29484 (vmlsl_high_laneq_s32): Likewise.
29485 (vmlsl_high_laneq_u16): Likewise.
29486 (vmlsl_high_laneq_u32): Likewise.
29487 (vmlal_high_laneq_u32): Likewise.
29489 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
29491 * config/aarch64/aarch64-simd-builtins.def: Add
29492 [su]mlal_hi_lane[q] builtin generator macros.
29493 * config/aarch64/aarch64-simd.md
29494 (aarch64_<su>mlal_hi_lane<mode>_insn): Define.
29495 (aarch64_<su>mlal_hi_lane<mode>): Define.
29496 (aarch64_<su>mlal_hi_laneq<mode>_insn): Define.
29497 (aarch64_<su>mlal_hi_laneq<mode>): Define.
29498 * config/aarch64/arm_neon.h (vmlal_high_lane_s16): Use RTL
29499 builtin instead of inline asm.
29500 (vmlal_high_lane_s32): Likewise.
29501 (vmlal_high_lane_u16): Likewise.
29502 (vmlal_high_lane_u32): Likewise.
29503 (vmlal_high_laneq_s16): Likewise.
29504 (vmlal_high_laneq_s32): Likewise.
29505 (vmlal_high_laneq_u16): Likewise.
29506 (vmlal_high_laneq_u32): Likewise.
29508 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
29510 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_hi_n
29511 builtin generator macros.
29512 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_hi_n<mode>_insn):
29514 (aarch64_<su>mlsl_hi_n<mode>): Define.
29515 * config/aarch64/arm_neon.h (vmlsl_high_n_s16): Use RTL builtin
29516 instead of inline asm.
29517 (vmlsl_high_n_s32): Likewise.
29518 (vmlsl_high_n_u16): Likewise.
29519 (vmlsl_high_n_u32): Likewise.
29521 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
29523 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_hi_n
29524 builtin generator macros.
29525 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_hi_n<mode>_insn):
29527 (aarch64_<su>mlal_hi_n<mode>): Define.
29528 * config/aarch64/arm_neon.h (vmlal_high_n_s16): Use RTL builtin
29529 instead of inline asm.
29530 (vmlal_high_n_s32): Likewise.
29531 (vmlal_high_n_u16): Likewise.
29532 (vmlal_high_n_u32): Likewise.
29534 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
29536 * config/aarch64/aarch64-simd-builtins.def: Add RTL builtin
29538 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal_hi<mode>):
29540 (aarch64_<su>mlal_hi<mode>_insn): This.
29541 (aarch64_<su>mlal_hi<mode>): Define.
29542 * config/aarch64/arm_neon.h (vmlal_high_s8): Use RTL builtin
29543 instead of inline asm.
29544 (vmlal_high_s16): Likewise.
29545 (vmlal_high_s32): Likewise.
29546 (vmlal_high_u8): Likewise.
29547 (vmlal_high_u16): Likewise.
29548 (vmlal_high_u32): Likewise.
29550 2021-02-03 Ilya Leoshkevich <iii@linux.ibm.com>
29552 * lra-spills.c (remove_pseudos): Call lra_update_insn_recog_data()
29553 after calling alter_subreg() on a (mem).
29555 2021-02-03 Martin Liska <mliska@suse.cz>
29558 * lto-streamer-out.c (produce_lto_section): Fill up missing
29560 * lto-streamer.h (struct lto_section): Add _padding field.
29562 2021-02-03 Richard Biener <rguenther@suse.de>
29564 * lto-streamer.c (lto_get_section_name): Free temporary
29566 * tree-loop-distribution.c
29567 (loop_distribution::merge_dep_scc_partitions): Free edge data.
29569 2021-02-03 Jakub Jelinek <jakub@redhat.com>
29571 PR middle-end/97487
29572 * ifcvt.c (noce_can_force_operand): New function.
29573 (noce_emit_move_insn): Use it.
29574 (noce_try_sign_mask): Likewise. Formatting fix.
29576 2021-02-03 Jakub Jelinek <jakub@redhat.com>
29578 PR middle-end/97971
29579 * lra-constraints.c (process_alt_operands): For inline asm, don't call
29580 fatal_insn, but instead return false.
29582 2021-02-03 Jakub Jelinek <jakub@redhat.com>
29584 PR tree-optimization/98287
29585 * config/i386/mmx.md (<insn><mode>3): For shifts don't enable expander
29588 2021-02-03 Tamar Christina <tamar.christina@arm.com>
29590 PR tree-optimization/98928
29591 * tree-vect-loop.c (vect_analyze_loop_2): Change
29592 STMT_VINFO_SLP_VECT_ONLY to STMT_VINFO_SLP_VECT_ONLY_PATTERN.
29593 * tree-vect-slp-patterns.c (complex_pattern::build): Likewise.
29594 * tree-vectorizer.h (STMT_VINFO_SLP_VECT_ONLY_PATTERN): New.
29595 (class _stmt_vec_info): Add slp_vect_pattern_only_p.
29597 2021-02-02 Richard Biener <rguenther@suse.de>
29599 * gimple-loop-interchange.cc (prepare_data_references):
29601 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
29602 * tree-ssa-loop-im.c (hoist_memory_references): Likewise.
29603 * tree-vect-stmts.c (vectorizable_condition): Do not
29605 (vectorizable_comparison): Likewise.
29607 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29609 * config/aarch64/aarch64-simd-builtins.def (ursqrte): Define builtin.
29610 * config/aarch64/aarch64-simd.md (aarch64_ursqrte<mode>): New pattern.
29611 * config/aarch64/arm_neon.h (vrsqrte_u32): Reimplement using builtin.
29612 (vrsqrteq_u32): Likewise.
29614 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29616 * config/aarch64/aarch64-simd-builtins.def (sqxtun2): Define builtin.
29617 * config/aarch64/aarch64-simd.md (aarch64_sqxtun2<mode>_le): Define.
29618 (aarch64_sqxtun2<mode>_be): Likewise.
29619 (aarch64_sqxtun2<mode>): Likewise.
29620 * config/aarch64/arm_neon.h (vqmovun_high_s16): Reimplement using builtin.
29621 (vqmovun_high_s32): Likewise.
29622 (vqmovun_high_s64): Likewise.
29623 * config/aarch64/iterators.md (UNSPEC_SQXTUN2): Define.
29625 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29627 * config/aarch64/aarch64-simd-builtins.def (bfdot_lane, bfdot_laneq): Use
29629 (bfmlalb_lane, bfmlalt_lane, bfmlalb_lane_q, bfmlalt_lane_q): Use FP flags.
29631 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29633 * config/aarch64/aarch64-simd-builtins.def (fcmla_lane0, fcmla_lane90,
29634 fcmla_lane180, fcmla_lane270, fcmlaq_lane0, fcmlaq_lane90, fcmlaq_lane180,
29635 fcmlaq_lane270, scvtf, ucvtf, fcvtzs, fcvtzu, scvtfsi, scvtfdi, ucvtfsi,
29636 ucvtfdi, fcvtzshf, fcvtzuhf, fmlal_lane_low, fmlsl_lane_low,
29637 fmlal_laneq_low, fmlsl_laneq_low, fmlalq_lane_low, fmlslq_lane_low,
29638 fmlalq_laneq_low, fmlslq_laneq_low, fmlal_lane_high, fmlsl_lane_high,
29639 fmlal_laneq_high, fmlsl_laneq_high, fmlalq_lane_high, fmlslq_lane_high,
29640 fmlalq_laneq_high, fmlslq_laneq_high): Use FP flags.
29642 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29644 * config/aarch64/aarch64-builtins.c (FLAG_LOAD): Define.
29645 * config/aarch64/aarch64-simd-builtins.def (ld1x2, ld2, ld3, ld4, ld2r,
29646 ld3r, ld4r, ld1, ld1x3, ld1x4): Use LOAD flags.
29648 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29650 * config/aarch64/aarch64-simd-builtins.def (combine, zip1, zip2,
29651 uzp1, uzp2, trn1, trn2, simd_bsl): Use AUTO_FP flags.
29653 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29655 * config/aarch64/aarch64-simd-builtins.def (clrsb, clz, ctz, popcount,
29656 vec_smult_lane_, vec_smlal_lane_, vec_smult_laneq_, vec_smlal_laneq_,
29657 vec_umult_lane_, vec_umlal_lane_, vec_umult_laneq_, vec_umlal_laneq_,
29658 ashl, sshl, ushl, srshl, urshl, sdot_lane, udot_lane, sdot_laneq,
29659 udot_laneq, usdot_lane, usdot_laneq, sudot_lane, sudot_laneq, ashr,
29660 ashr_simd, lshr, lshr_simd, srshr_n, urshr_n, ssra_n, usra_n, srsra_n,
29661 ursra_n, sshll_n, ushll_n, sshll2_n, ushll2_n, ssri_n, usri_n, ssli_n,
29662 ssli_n, usli_n, bswap, rbit, simd_bsl, eor3q, rax1q, xarq, bcaxq): Use
29663 NONE builtin flags.
29665 2021-02-02 Jakub Jelinek <jakub@redhat.com>
29667 PR tree-optimization/98848
29668 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Punt if
29669 STMT_VINFO_DEF_TYPE (last_stmt_info) is vect_reduction_def.
29671 2021-02-02 Kito Cheng <kito.cheng@sifive.com>
29674 * expr.c: Check mode before calling store_expr.
29676 2021-02-02 Christophe Lyon <christophe.lyon@linaro.org>
29678 * config/arm/iterators.md (supf): Remove VORNQ_S and VORNQ_U.
29680 * config/arm/mve.md (mve_vornq_s<mode>): New entry for vorn
29681 instruction using expression ior.
29682 (mve_vornq_u<mode>): New expander.
29683 (mve_vornq_f<mode>): Use ior code instead of unspec.
29684 * config/arm/unspecs.md (VORNQ_S, VORNQ_U, VORNQ_F): Remove.
29686 2021-02-02 Alexandre Oliva <oliva@adacore.com>
29688 * tree-nested.c (convert_nonlocal_reference_op): Move
29689 current_function_decl restore after re-gimplification.
29690 (convert_local_reference_op): Likewise.
29692 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29694 * config/aarch64/aarch64-simd-builtins.def (rshrn, rshrn2):
29696 * config/aarch64/aarch64-simd.md (aarch64_rshrn<mode>_insn_le):
29698 (aarch64_rshrn<mode>_insn_be): Likewise.
29699 (aarch64_rshrn<mode>): Likewise.
29700 (aarch64_rshrn2<mode>_insn_le): Likewise.
29701 (aarch64_rshrn2<mode>_insn_be): Likewise.
29702 (aarch64_rshrn2<mode>): Likewise.
29703 * config/aarch64/aarch64.md (unspec): Add UNSPEC_RSHRN.
29704 * config/aarch64/arm_neon.h (vrshrn_high_n_s16): Reimplement
29706 (vrshrn_high_n_s32): Likewise.
29707 (vrshrn_high_n_s64): Likewise.
29708 (vrshrn_high_n_u16): Likewise.
29709 (vrshrn_high_n_u32): Likewise.
29710 (vrshrn_high_n_u64): Likewise.
29711 (vrshrn_n_s16): Likewise.
29712 (vrshrn_n_s32): Likewise.
29713 (vrshrn_n_s64): Likewise.
29714 (vrshrn_n_u16): Likewise.
29715 (vrshrn_n_u32): Likewise.
29716 (vrshrn_n_u64): Likewise.
29718 2021-02-01 Sergei Trofimovich <siarheit@google.com>
29720 PR tree-optimization/98499
29721 * ipa-modref.c (analyze_ssa_name_flags): treat RVO
29722 conservatively and assume all possible side-effects.
29724 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29726 * config/aarch64/aarch64-simd-builtins.def (vec_unpacks_hi,
29727 vec_unpacku_hi_): Define builtins.
29728 * config/aarch64/arm_neon.h (vmovl_high_s8): Reimplement using
29730 (vmovl_high_s16): Likewise.
29731 (vmovl_high_s32): Likewise.
29732 (vmovl_high_u8): Likewise.
29733 (vmovl_high_u16): Likewise.
29734 (vmovl_high_u32): Likewise.
29736 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29738 * config/aarch64/aarch64-simd-builtins.def (sabdl, uabdl):
29740 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl<mode>): New
29742 * config/aarch64/aarch64.md (unspec): Define UNSPEC_SABDL,
29744 * config/aarch64/arm_neon.h (vabdl_s8): Reimplemet using
29746 (vabdl_s16): Likewise.
29747 (vabdl_s32): Likewise.
29748 (vabdl_u8): Likewise.
29749 (vabdl_u16): Likewise.
29750 (vabdl_u32): Likewise.
29751 * config/aarch64/iterators.md (ABDL): New int iterator.
29752 (sur): Handle UNSPEC_SABDL, UNSPEC_UABDL.
29754 2021-02-01 Martin Sebor <msebor@redhat.com>
29756 * tree.h (BLOCK_VARS): Add comment.
29757 (BLOCK_SUBBLOCKS): Same.
29758 (BLOCK_SUPERCONTEXT): Same.
29759 (BLOCK_ABSTRACT_ORIGIN): Same.
29760 (inlined_function_outer_scope_p): Same.
29762 2021-02-01 Martin Sebor <msebor@redhat.com>
29764 PR middle-end/97172
29765 * attribs.c (attr_access::free_lang_data): Define new function.
29766 * attribs.h (attr_access::free_lang_data): Declare new function.
29768 2021-02-01 Richard Biener <rguenther@suse.de>
29770 * vec.h (auto_vec::auto_vec): Add memory stat parameters
29772 * bitmap.h (auto_bitmap::auto_bitmap): Likewise.
29774 2021-02-01 Tamar Christina <tamar.christina@arm.com>
29776 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>,
29777 aarch64_<su>mlsl<mode>, aarch64_<su>mlsl_n<mode>): Flip mult operands.
29779 2021-02-01 Richard Biener <rguenther@suse.de>
29781 PR rtl-optimization/98863
29782 * config/i386/i386-features.c (convert_scalars_to_vector):
29783 Set DF_RD_PRUNE_DEAD_DEFS.
29785 2021-01-31 Eric Botcazou <ebotcazou@adacore.com>
29787 * system.h (SIZE_MAX): Define if not already defined.
29789 2021-01-30 Aaron Sawdey <acsawdey@linux.ibm.com>
29791 * config/rs6000/genfusion.pl (gen_2logical): New function to
29792 generate patterns for logical-logical fusion.
29793 * config/rs6000/fusion.md: Regenerated patterns.
29794 * config/rs6000/rs6000-cpus.def: Add
29795 OPTION_MASK_P10_FUSION_2LOGICAL.
29796 * config/rs6000/rs6000.c (rs6000_option_override_internal):
29797 Enable logical-logical fusion for p10.
29798 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2logical.
29800 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
29802 * config/rs6000/rs6000.opt: Add periods to new AIX options.
29804 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
29806 * config/rs6000/rs6000.opt (mabi=vec-extabi): New.
29807 (mabi=vec-default): New.
29808 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
29809 __EXTABI__ for AIX Vector extended ABI.
29810 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Print AIX Vector
29812 (conditional_register_usage): If AIX vec_extabi enabled, vs20-vs31
29814 * doc/invoke.texi (PowerPC mabi): Add AIX vec-extabi and vec-default.
29816 2021-01-30 Jakub Jelinek <jakub@redhat.com>
29818 * config/i386/i386-features.c (remove_partial_avx_dependency): Clear
29819 DF_DEFER_INSN_RESCAN after calling df_process_deferred_rescans.
29821 2021-01-29 Vladimir N. Makarov <vmakarov@redhat.com>
29824 * lra-constraints.c (in_class_p): Don't narrow class only for REG
29827 2021-01-29 Will Schmidt <will_schmidt@vnet.ibm.com>
29829 * config/rs6000/rs6000-call.c (rs6000_expand_binup_builtin): Add
29830 clauses for CODE_FOR_vsx_xvcvuxddp_scale and
29831 CODE_FOR_vsx_xvcvsxddp_scale to the parameter checking code.
29833 2021-01-29 Andrew MacLeod <amacleod@redhat.com>
29835 PR tree-optimization/98866
29836 * gimple-range-gori.h (gori_compute:set_range_invariant): New.
29837 * gimple-range-gori.cc (gori_map::set_range_invariant): New.
29838 (gori_map::m_maybe_invariant): Rename from all_outgoing.
29839 (gori_map::gori_map): Rename all_outgoing to m_maybe_invariant.
29840 (gori_map::is_export_p): Ditto.
29841 (gori_map::calculate_gori): Ditto.
29842 (gori_compute::set_range_invariant): New.
29843 * gimple-range.cc (gimple_ranger::range_of_stmt): Set range
29844 invariant for pointers evaluating to [1, +INF].
29846 2021-01-29 Richard Biener <rguenther@suse.de>
29848 PR rtl-optimization/98863
29849 * config/i386/i386-features.c (remove_partial_avx_dependency):
29850 Do not perform DF analysis.
29851 (pass_data_remove_partial_avx_dependency): Remove
29854 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
29856 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_n
29857 builtin generator macros.
29858 * config/aarch64/aarch64-simd.md (aarch64_<su>mull_n<mode>):
29860 * config/aarch64/arm_neon.h (vmull_n_s16): Use RTL builtin
29861 instead of inline asm.
29862 (vmull_n_s32): Likewise.
29863 (vmull_n_u16): Likewise.
29864 (vmull_n_u32): Likewise.
29866 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29868 * config/aarch64/aarch64-simd-builtins.def (sabdl2, uabdl2):
29870 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
29872 (aarch64_<sur>abdl2<mode>): ... This.
29873 (<sur>sadv16qi): Adjust use of above.
29874 * config/aarch64/arm_neon.h (vabdl_high_s8): Reimplement using
29876 (vabdl_high_s16): Likewise.
29877 (vabdl_high_s32): Likewise.
29878 (vabdl_high_u8): Likewise.
29879 (vabdl_high_u16): Likewise.
29880 (vabdl_high_u32): Likewise.
29882 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29884 * config/aarch64/aarch64-simd-builtins.def (sabal2): Define
29886 (uabal2): Likewise.
29887 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal2<mode>): New
29889 * config/aarch64/aarch64.md (unspec): Add UNSPEC_SABAL2 and
29891 * config/aarch64/arm_neon.h (vabal_high_s8): Reimplement using
29893 (vabal_high_s16): Likewise.
29894 (vabal_high_s32): Likewise.
29895 (vabal_high_u8): Likewise.
29896 (vabal_high_u16): Likewise.
29897 (vabal_high_u32): Likewise.
29898 * config/aarch64/iterators.md (ABAL2): New mode iterator.
29899 (sur): Handle UNSPEC_SABAL2, UNSPEC_UABAL2.
29901 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29903 * config/aarch64/aarch64-simd-builtins.def (sabal): Define
29906 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal<mode>_4):
29908 (aarch64_<sur>abal<mode>): ... This
29909 (<sur>sadv16qi): Adust use of the above.
29910 * config/aarch64/arm_neon.h (vabal_s8): Reimplement using
29912 (vabal_s16): Likewise.
29913 (vabal_s32): Likewise.
29914 (vabal_u8): Likewise.
29915 (vabal_u16): Likewise.
29916 (vabal_u32): Likewise.
29918 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29920 * config/aarch64/aarch64-simd-builtins.def (saddlv, uaddlv):
29922 * config/aarch64/aarch64-simd.md (aarch64_<su>addlv<mode>):
29924 * config/aarch64/arm_neon.h (vaddlv_s8): Reimplement using
29926 (vaddlv_s16): Likewise.
29927 (vaddlv_u8): Likewise.
29928 (vaddlv_u16): Likewise.
29929 (vaddlvq_s8): Likewise.
29930 (vaddlvq_s16): Likewise.
29931 (vaddlvq_s32): Likewise.
29932 (vaddlvq_u8): Likewise.
29933 (vaddlvq_u16): Likewise.
29934 (vaddlvq_u32): Likewise.
29935 (vaddlv_s32): Likewise.
29936 (vaddlv_u32): Likewise.
29937 * config/aarch64/iterators.md (VDQV_L): New mode iterator.
29938 (unspec): Add UNSPEC_SADDLV, UNSPEC_UADDLV.
29939 (Vwstype): New mode attribute.
29941 (VWIDE_S): Likewise.
29942 (USADDLV): New int iterator.
29943 (su): Handle UNSPEC_SADDLV, UNSPEC_UADDLV.
29945 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
29947 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_lane[q]
29948 builtin generator macros.
29949 * config/aarch64/aarch64-simd.md (aarch64_vec_<su>mlsl_lane<Qlane>):
29951 * config/aarch64/arm_neon.h (vmlsl_lane_s16): Use RTL builtin
29952 instead of inline asm.
29953 (vmlsl_lane_s32): Likewise.
29954 (vmlsl_lane_u16): Likewise.
29955 (vmlsl_lane_u32): Likewise.
29956 (vmlsl_laneq_s16): Likewise.
29957 (vmlsl_laneq_s32): Likewise.
29958 (vmlsl_laneq_u16): Likewise.
29959 (vmlsl_laneq_u32): Likewise.
29961 2021-01-29 Richard Biener <rguenther@suse.de>
29963 * doc/invoke.texi (--param max-gcse-memory): Document unit
29965 * gcse.c (gcse_or_cprop_is_too_expensive): Adjust.
29966 * params.opt (--param max-gcse-memory): Adjust default and
29967 document unit of size.
29969 2021-01-29 Richard Biener <rguenther@suse.de>
29971 PR rtl-optimization/98863
29972 * gcse.c (gcse_or_cprop_is_too_expensive): Use unsigned
29973 HOST_WIDE_INT for the memory estimate.
29975 2021-01-29 Bin Cheng <bin.cheng@linux.alibaba.com>
29976 Richard Biener <rguenther@suse.de>
29978 PR tree-optimization/97627
29979 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
29980 Do not analyze fake edges.
29982 2021-01-29 Richard Biener <rguenther@suse.de>
29984 PR rtl-optimization/98144
29985 * df.h (df_mir_bb_info): Add con_visited member.
29986 * df-problems.c (df_mir_alloc): Initialize con_visited,
29987 do not fully populate IN and OUT.
29988 (df_mir_reset): Likewise.
29989 (df_mir_confluence_0): Set con_visited.
29990 (df_mir_confluence_n): Properly handle implicitely
29991 fully populated IN and OUT as designated by con_visited
29992 and update con_visited accordingly.
29994 2021-01-29 Jakub Jelinek <jakub@redhat.com>
29997 * config/arm/vec-common.md (mve_vshlq_<supf><mode>,
29998 vashl<mode>3, vashr<mode>3, vlshr<mode>3): Add
29999 && !TARGET_REALLY_IWMMXT to conditions.
30001 2021-01-29 Jakub Jelinek <jakub@redhat.com>
30004 * cfgbuild.c (find_bb_boundaries): Reset debug_insn when seeing
30007 2021-01-28 Marek Polacek <polacek@redhat.com>
30010 * stor-layout.c (finalize_type_size): If we reset TYPE_USER_ALIGN in
30011 the main variant, maybe reset it in its variants too.
30012 * tree.c (check_base_type): Return true only if TYPE_USER_ALIGN match.
30013 (check_aligned_type): Check if TYPE_USER_ALIGN match.
30015 2021-01-28 Christophe Lyon <christophe.lyon@linaro.org>
30018 * config/arm/arm.c (arm_rtx_costs_internal): Adjust cost of vector
30019 of constant zero for comparisons.
30021 2021-01-28 Michael Meissner <meissner@linux.ibm.com>
30023 * config/rs6000/rs6000.c (rs6000_mangle_decl_assembler_name): Add
30024 support for mapping built-in function names for long double
30025 built-in functions if long double is IEEE 128-bit.
30027 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
30029 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_n
30030 builtin generator macros.
30031 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_n<mode>):
30033 * config/aarch64/arm_neon.h (vmlsl_n_s16): Use RTL builtin
30034 instead of inline asm.
30035 (vmlsl_n_s32): Likewise.
30036 (vmlsl_n_u16): Likewise.
30037 (vmlsl_n_u32): Likewise.
30039 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
30041 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_n
30042 builtin generator macros.
30043 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>):
30045 * config/aarch64/arm_neon.h (vmlal_n_s16): Use RTL builtin
30046 instead of inline asm.
30047 (vmlal_n_s32): Likewise.
30048 (vmlal_n_u16): Likewise.
30049 (vmlal_n_u32): Likewise.
30051 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30053 * config/aarch64/aarch64-simd-builtins.def (shrn2): Define
30055 * config/aarch64/aarch64-simd.md (aarch64_shrn2<mode>_insn_le):
30057 (aarch64_shrn2<mode>_insn_be): Likewise.
30058 (aarch64_shrn2<mode>): Likewise.
30059 * config/aarch64/arm_neon.h (vshrn_high_n_s16): Reimlplement
30061 (vshrn_high_n_s32): Likewise.
30062 (vshrn_high_n_s64): Likewise.
30063 (vshrn_high_n_u16): Likewise.
30064 (vshrn_high_n_u32): Likewise.
30065 (vshrn_high_n_u64): Likewise.
30067 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30069 * config/aarch64/aarch64-simd-builtins.def (shrn): Define
30071 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le):
30073 (aarch64_shrn<mode>_insn_be): Likewise.
30074 (aarch64_shrn<mode>): Likewise.
30075 * config/aarch64/arm_neon.h (vshrn_n_s16): Reimplement using
30077 (vshrn_n_s32): Likewise.
30078 (vshrn_n_s64): Likewise.
30079 (vshrn_n_u16): Likewise.
30080 (vshrn_n_u32): Likewise.
30081 (vshrn_n_u64): Likewise.
30082 * config/aarch64/iterators.md (vn_mode): New mode attribute.
30084 2021-01-28 Richard Biener <rguenther@suse.de>
30086 PR rtl-optimization/80960
30087 * dse.c (check_mem_read_rtx): Call get_addr on the
30090 2021-01-28 Xionghu Luo <luoxhu@linux.ibm.com>
30091 David Edelsohn <dje.gcc@gmail.com>
30094 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
30095 Don't generate VIEW_CONVERT_EXPR for fcode ALTIVEC_BUILTIN_VEC_INSERT
30097 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
30099 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Remove the
30100 wrapper call rs6000_expand_vector_set_var for cleanup. Call
30101 rs6000_expand_vector_set_var_p9 and rs6000_expand_vector_set_var_p8
30103 (rs6000_expand_vector_set_var): Delete.
30104 (rs6000_expand_vector_set_var_p9): Make static.
30105 (rs6000_expand_vector_set_var_p8): Make static.
30107 2021-01-28 Xing GUO <higuoxing@gmail.com>
30109 * common/config/riscv/riscv-common.c
30110 (riscv_subset_list::parsing_subset_version): Fix -march option parsing
30111 when `p` extension exists.
30113 2021-01-27 Vladimir N. Makarov <vmakarov@redhat.com>
30115 PR rtl-optimization/97684
30116 * ira.c (ira): Call ira_set_pseudo_classes before
30117 update_equiv_regs when it is necessary.
30119 2021-01-27 Jakub Jelinek <jakub@redhat.com>
30122 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Use
30123 %w0, %w1 and %2 instead of %0, %1 and %2.
30125 2021-01-27 Aaron Sawdey <acsawdey@linux.ibm.com>
30127 * config/rs6000/genfusion.pl: New script to generate
30128 define_insn_and_split patterns so combine can arrange fused
30129 instructions next to each other.
30130 * config/rs6000/fusion.md: New file, generated fused instruction
30131 patterns for combine.
30132 * config/rs6000/predicates.md (const_m1_to_1_operand): New predicate.
30133 (non_update_memory_operand): New predicate.
30134 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_P10_FUSION and
30135 OPTION_MASK_P10_FUSION_LD_CMPI to ISA_3_1_MASKS_SERVER and
30137 * config/rs6000/rs6000-protos.h (address_is_non_pfx_d_or_x): Add
30139 * config/rs6000/rs6000.c (rs6000_option_override_internal):
30140 Automatically set OPTION_MASK_P10_FUSION and
30141 OPTION_MASK_P10_FUSION_LD_CMPI if target is power10.
30142 (rs600_opt_masks): Allow -mpower10-fusion
30143 in function attributes.
30144 (address_is_non_pfx_d_or_x): New function.
30145 * config/rs6000/rs6000.h: Add MASK_P10_FUSION.
30146 * config/rs6000/rs6000.md: Include fusion.md.
30147 * config/rs6000/rs6000.opt: Add -mpower10-fusion
30148 and -mpower10-fusion-ld-cmpi.
30149 * config/rs6000/t-rs6000: Add dependencies involving fusion.md.
30151 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
30153 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal
30154 builtin generator macros.
30155 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal<mode>):
30157 (aarch64_<su>mlal<mode>): This.
30158 * config/aarch64/arm_neon.h (vmlal_s8): Use RTL builtin
30159 instead of inline asm.
30160 (vmlal_s16): Likewise.
30161 (vmlal_s32): Likewise.
30162 (vmlal_u8): Likewise.
30163 (vmlal_u16): Likewise.
30164 (vmlal_u32): Likewise.
30166 2021-01-27 Richard Biener <rguenther@suse.de>
30168 PR tree-optimization/98854
30169 * tree-vect-slp.c (vect_build_slp_tree_2): Also build
30170 PHIs from scalars when the number of CTORs matches the
30171 number of children.
30173 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
30175 * config/aarch64/aarch64-simd-builtins.def: Add mls_n builtin
30177 * config/aarch64/aarch64-simd.md (*aarch64_mls_elt_merge<mode>):
30179 (aarch64_mls_n<mode>): This.
30180 * config/aarch64/arm_neon.h (vmls_n_s16): Use RTL builtin
30182 (vmls_n_s32): Likewise.
30183 (vmls_n_u16): Likewise.
30184 (vmls_n_u32): Likewise.
30185 (vmlsq_n_s16): Likewise.
30186 (vmlsq_n_s32): Likewise.
30187 (vmlsq_n_u16): Likewise.
30188 (vmlsq_n_u32): Likewise.
30190 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
30192 * config/aarch64/aarch64-simd-builtins.def: Add mls builtin
30194 * config/aarch64/arm_neon.h (vmls_s8): Use RTL builtin rather
30196 (vmls_s16): Likewise.
30197 (vmls_s32): Likewise.
30198 (vmls_u8): Likewise.
30199 (vmls_u16): Likewise.
30200 (vmls_u32): Likewise.
30201 (vmlsq_s8): Likewise.
30202 (vmlsq_s16): Likewise.
30203 (vmlsq_s32): Likewise.
30204 (vmlsq_u8): Likewise.
30205 (vmlsq_u16): Likewise.
30206 (vmlsq_u32): Likewise.
30208 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
30210 * config/aarch64/aarch64-simd-builtins.def: Add mla_n builtin
30212 * config/aarch64/aarch64-simd.md (*aarch64_mla_elt_merge<mode>):
30214 (aarch64_mla_n<mode>): This.
30215 * config/aarch64/arm_neon.h (vmla_n_s16): Use RTL builtin
30217 (vmla_n_s32): Likewise.
30218 (vmla_n_u16): Likewise.
30219 (vmla_n_u32): Likewise.
30220 (vmlaq_n_s16): Likewise.
30221 (vmlaq_n_s32): Likewise.
30222 (vmlaq_n_u16): Likewise.
30223 (vmlaq_n_u32): Likewise.
30225 2021-01-27 liuhongt <hongtao.liu@intel.com>
30228 * config/i386/sse.md (sse2_gt<mode>3): Drop !TARGET_XOP in condition.
30229 (*sse2_eq<mode>3): Ditto.
30231 2021-01-27 Jakub Jelinek <jakub@redhat.com>
30233 * tree-pass.h (PROP_trees): Rename to ...
30234 (PROP_gimple): ... this.
30235 * cfgexpand.c (pass_data_expand): Replace PROP_trees with PROP_gimple.
30236 * passes.c (execute_function_dump, execute_function_todo,
30237 execute_one_ipa_transform_pass, execute_one_pass): Likewise.
30238 * varpool.c (ctor_for_folding): Likewise.
30240 2021-01-27 Jakub Jelinek <jakub@redhat.com>
30242 PR tree-optimization/97260
30243 * varpool.c: Include tree-pass.h.
30244 (ctor_for_folding): In GENERIC return DECL_INITIAL for TREE_READONLY
30245 non-TREE_SIDE_EFFECTS automatic variables.
30247 2021-01-26 Paul Fee <paul.f.fee@gmail.com>
30249 * doc/cpp.texi (__cplusplus): Document value for -std=c++23
30251 * doc/invoke.texi: Document -std=c++23 and -std=gnu++23.
30252 * dwarf2out.c (highest_c_language): Recognise C++20 and C++23.
30253 (gen_compile_unit_die): Recognise C++23.
30255 2021-01-26 Jakub Jelinek <jakub@redhat.com>
30258 * dwarf2asm.c (dw2_assemble_integer): Cast DWARF2_ADDR_SIZE to int
30261 2021-01-26 Jakub Jelinek <jakub@redhat.com>
30264 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
30265 Use UINTVAL (shft_amnt) and UINTVAL (mask) instead of INTVAL (shft_amnt)
30266 and INTVAL (mask). Add && INTVAL (mask) > 0 condition.
30268 2021-01-26 Richard Biener <rguenther@suse.de>
30270 * gimple-pretty-print.c (dump_binary_rhs): Handle
30271 VEC_WIDEN_{PLUS,MINUS}_{LO,HI}_EXPR.
30273 2021-01-26 Richard Biener <rguenther@suse.de>
30275 PR middle-end/98726
30276 * tree.h (vector_cst_int_elt): Remove.
30277 * tree.c (vector_cst_int_elt): Use poly_wide_int for computations,
30280 2021-01-26 Andrew Stubbs <ams@codesourcery.com>
30282 * config/gcn/gcn.c (gcn_expand_reduc_scalar): Use move instructions
30283 for V64DFmode min/max reductions.
30285 2021-01-26 Jakub Jelinek <jakub@redhat.com>
30287 * dwarf2asm.c (dw2_assemble_integer): Handle size twice as large
30288 as DWARF2_ADDR_SIZE if x is not a scalar int by emitting it as
30289 two halves, one with x and the other with const0_rtx, ordered
30290 depending on endianity.
30292 2021-01-26 Alexandre Oliva <oliva@adacore.com>
30294 * gimplify.c (gimplify_decl_expr): Skip asan marking calls for
30295 temporaries not seen in binding block, and not about to be
30296 added as gimple variables.
30298 2021-01-25 Martin Sebor <msebor@redhat.com>
30301 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust warning text.
30303 2021-01-25 Martin Liska <mliska@suse.cz>
30305 * value-prof.c (get_nth_most_common_value): Use %s instead
30308 2021-01-25 Jakub Jelinek <jakub@redhat.com>
30311 * configure.ac (HAVE_AS_GDWARF_5_DEBUG_FLAG): Only define if
30312 readelf -wi is able to read the emitted .debug_info back.
30313 * configure: Regenerated.
30315 2021-01-25 Martin Liska <mliska@suse.cz>
30317 PR gcov-profile/98739
30318 * common.opt: Add missing sign symbol.
30319 * value-prof.c (get_nth_most_common_value): Restore handling
30320 of PROFILE_REPRODUCIBILITY_PARALLEL_RUNS and
30321 PROFILE_REPRODUCIBILITY_MULTITHREADED.
30323 2021-01-25 Richard Biener <rguenther@suse.de>
30325 PR middle-end/98807
30326 * tree.c (vector_element_bits): Always use precision of
30327 the element type for boolean vectors.
30329 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
30331 * config/rtems.h (STARTFILE_SPEC): Remove qnolinkcmds.
30332 (ENDFILE_SPEC): Evaluate qnolinkcmds.
30334 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
30336 * config/rtems.h (STARTFILE_SPEC): Remove nostdlib and
30337 nostartfiles handling since this is already done by
30338 LINK_COMMAND_SPEC. Evaluate qnolinkcmds.
30339 (ENDFILE_SPEC): Remove nostdlib and nostartfiles handling since this
30340 is already done by LINK_COMMAND_SPEC.
30341 (LIB_SPECS): Remove nostdlib and nodefaultlibs handling since
30342 this is already done by LINK_COMMAND_SPEC. Remove qnolinkcmds
30345 2021-01-25 Jakub Jelinek <jakub@redhat.com>
30348 * fold-const-call.c (host_size_t_cst_p): Renamed to ...
30349 (size_t_cst_p): ... this. Check and store unsigned HOST_WIDE_INT
30350 value rather than host size_t.
30351 (fold_const_call): Change type of s2 from size_t to
30352 unsigned HOST_WIDE_INT. Use size_t_cst_p instead of
30353 host_size_t_cst_p. For strncmp calls, pass MIN (s2, SIZE_MAX)
30354 instead of s2 as last argument.
30356 2021-01-25 Tamar Christina <tamar.christina@arm.com>
30358 * config/arm/iterators.md (rotsplit1, rotsplit2, conj_op, fcmac1,
30359 VCMLA_OP, VCMUL_OP): New.
30360 * config/arm/mve.md (mve_vcmlaq<mve_rot><mode>): Support vec_dup 0.
30361 * config/arm/neon.md (cmul<conj_op><mode>3): New.
30362 * config/arm/unspecs.md (UNSPEC_VCMLA_CONJ, UNSPEC_VCMLA180_CONJ,
30363 UNSPEC_VCMUL_CONJ): New.
30364 * config/arm/vec-common.md (cmul<conj_op><mode>3, arm_vcmla<rot><mode>,
30365 cml<fcmac1><conj_op><mode>4): New.
30367 2021-01-23 Jakub Jelinek <jakub@redhat.com>
30370 * config/rs6000/mmintrin.h (__m64): Add __may_alias__ attribute.
30372 2021-01-22 Jonathan Wright <jonathan.wright@arm.com>
30374 * config/aarch64/aarch64-simd-builtins.def: Add mla builtin
30376 * config/aarch64/arm_neon.h (vmla_s8): Use RTL builtin rather
30378 (vmla_s16): Likewise.
30379 (vmla_s32): Likewise.
30380 (vmla_u8): Likewise.
30381 (vmla_u16): Likewise.
30382 (vmla_u32): Likewise.
30383 (vmlaq_s8): Likewise.
30384 (vmlaq_s16): Likewise.
30385 (vmlaq_s32): Likewise.
30386 (vmlaq_u8): Likewise.
30387 (vmlaq_u16): Likewise.
30388 (vmlaq_u32): Likewise.
30390 2021-01-22 David Malcolm <dmalcolm@redhat.com>
30392 * doc/invoke.texi (GCC_EXTRA_DIAGNOSTIC_OUTPUT): Add @findex
30395 2021-01-22 Jakub Jelinek <jakub@redhat.com>
30398 * dwarf2out.c (output_file_names): For -gdwarf-5, if there are no
30399 filenames to emit, still emit the required 0 index directory and
30400 filename entries that match DW_AT_comp_dir and DW_AT_name of the
30403 2021-01-22 Marek Polacek <polacek@redhat.com>
30406 * doc/invoke.texi: Update C++ ABI Version 15 description.
30408 2021-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30410 PR tree-optimization/98766
30411 * tree-ssa-math-opts.c (convert_mult_to_fma): Use maybe_le when
30412 comparing against type size with param_avoid_fma_max_bits.
30414 2021-01-22 Richard Biener <rguenther@suse.de>
30416 PR middle-end/98793
30417 * tree.c (vector_element_bits): Key single-bit bool vector on
30418 integer mode rather than not vector mode.
30420 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
30423 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
30424 Generate ARRAY_REF(VIEW_CONVERT_EXPR) for P8 and later
30426 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): Update
30427 to call different path for P8 and P9.
30428 (rs6000_expand_vector_set_var_p9): New function.
30429 (rs6000_expand_vector_set_var_p8): New function.
30431 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
30435 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
30436 Ajdust variable index vec_insert from address dereference to
30437 ARRAY_REF(VIEW_CONVERT_EXPR) tree expression.
30438 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
30440 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): New function.
30442 2021-01-22 Martin Liska <mliska@suse.cz>
30444 PR gcov-profile/98739
30445 * profile.c (compute_value_histograms): Drop time profile for
30446 -fprofile-reproducible=multithreaded.
30448 2021-01-22 Nathan Sidwell <nathan@acm.org>
30450 * gcc.c (process_command): Don't check OPT_SPECIAL_input_file
30453 2021-01-22 Richard Biener <rguenther@suse.de>
30455 PR middle-end/98773
30456 * tree-data-ref.c (initalize_matrix_A): Revert previous
30457 change, retaining failing on HOST_WIDE_INT_MIN CHREC_RIGHT.
30459 2021-01-22 Jakub Jelinek <jakub@redhat.com>
30461 PR tree-optimization/90248
30462 * match.pd (X cmp 0.0 ? 1.0 : -1.0 -> copysign(1, +-X),
30463 X cmp 0.0 ? -1.0 : +1.0 -> copysign(1, -+X)): Remove
30465 (X * (X cmp 0.0 ? 1.0 : -1.0) -> +-abs(X),
30466 X * (X cmp 0.0 ? -1.0 : 1.0) -> +-abs(X)): New simplifications.
30468 2021-01-22 Jakub Jelinek <jakub@redhat.com>
30470 PR tree-optimization/98255
30471 * tree-dfa.c (get_ref_base_and_extent): For ARRAY_REFs, sign
30472 extend index - low_bound from sizetype's precision rather than index
30474 (get_addr_base_and_unit_offset_1): Likewise.
30475 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Likewise.
30476 * gimple-fold.c (fold_const_aggregate_ref_1): Likewise.
30478 2021-01-22 Richard Biener <rguenther@suse.de>
30480 PR tree-optimization/98786
30481 * tree-ssa-phiopt.c (factor_out_conditional_conversion): Avoid
30482 adding new uses of abnormals. Verify we deal with a conditional
30485 2021-01-22 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
30488 * optc-save-gen.awk: Add arm_fp16_format to checked_options.
30490 2021-01-22 liuhongt <hongtao.liu@intel.com>
30494 * config/i386/sse.md (VI_128_256): New mode iterator.
30495 (*avx_cmp<mode>3_1, *avx_cmp<mode>3_2, *avx_cmp<mode>3_3,
30496 *avx_cmp<mode>3_4, *avx2_eq<mode>3, *avx2_pcmp<mode>3_1,
30497 *avx2_pcmp<mode>3_2, *avx2_gt<mode>3): New
30498 define_insn_and_split to lower avx512 vector comparison to avx
30499 version when dest is vector.
30500 (*<avx512>_cmp<mode>3,*<avx512>_cmp<mode>3,*<avx512>_ucmp<mode>3):
30501 define_insn_and_split for negating the comparison result.
30502 * config/i386/predicates.md (float_vector_all_ones_operand):
30504 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Use
30505 general NOT operator without UNSPEC_MASKOP.
30507 2021-01-21 Vladimir N. Makarov <vmakarov@redhat.com>
30509 PR rtl-optimization/98777
30510 * lra-int.h (lra_pmode_pseudo): New extern.
30511 * lra.c (lra_pmode_pseudo): New global.
30513 * lra-eliminations.c (eliminate_regs_in_insn): Use it.
30515 2021-01-21 Ilya Leoshkevich <iii@linux.ibm.com>
30517 * fwprop.c (fwprop_propagation::classify_result): Allow
30518 (subreg (mem)) simplifications.
30520 2021-01-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30522 * config/aarch64/aarch64-simd.md (aarch64_sqdml<SBINQOPS:as>l<mode>):
30524 (aarch64_sqdmlal<mode>): ... This...
30525 (aarch64_sqdmlsl<mode>): ... And this.
30526 (aarch64_sqdml<SBINQOPS:as>l_lane<mode>): Split into...
30527 (aarch64_sqdmlal_lane<mode>): ... This...
30528 (aarch64_sqdmlsl_lane<mode>): ... And this.
30529 (aarch64_sqdml<SBINQOPS:as>l_laneq<mode>): Split into...
30530 (aarch64_sqdmlsl_laneq<mode>): ... This...
30531 (aarch64_sqdmlal_laneq<mode>): ... And this.
30532 (aarch64_sqdml<SBINQOPS:as>l_n<mode>): Split into...
30533 (aarch64_sqdmlsl_n<mode>): ... This...
30534 (aarch64_sqdmlal_n<mode>): ... And this.
30535 (aarch64_sqdml<SBINQOPS:as>l2<mode>_internal): Split into...
30536 (aarch64_sqdmlal2<mode>_internal): ... This...
30537 (aarch64_sqdmlsl2<mode>_internal): ... And this.
30539 2021-01-21 Christophe Lyon <christophe.lyon@linaro.org>
30541 * config/arm/arm_mve.h (__arm_vcmpneq_s8): Fix return type.
30543 2021-01-21 Andrea Corallo <andrea.corallo@arm.com>
30546 * doc/sourcebuild.texi (arm_thumb2_no_arm_v8_1_lob): Document.
30548 2021-01-21 liuhongt <hongtao.liu@intel.com>
30550 PR rtl-optimization/98694
30551 * regcprop.c (copy_value): If SRC had been assigned a mode
30552 narrower than the copy, we can't link DEST into the chain even
30553 they have same hard_regno_nregs(i.e. HImode/SImode in i386
30556 2021-01-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30558 * config/aarch64/aarch64-simd.md (aarch64_get_lane<mode>):
30559 Convert to define_insn_and_split. Split into simple move when moving
30562 2021-01-20 Segher Boessenkool <segher@kernel.crashing.org>
30564 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_store): Change assert.
30565 Adjust comment. Simplify code.
30567 2021-01-20 Jakub Jelinek <jakub@redhat.com>
30570 * dwarf2out.c (reset_indirect_string): Also reset indirect strings
30571 with DW_FORM_line_strp form.
30572 (prune_unused_types_update_strings): Don't add into debug_str_hash
30573 indirect strings with DW_FORM_line_strp form.
30574 (adjust_name_comp_dir): New function.
30575 (dwarf2out_finish): Call it on CU DIEs after resetting
30576 debug_line_str_hash.
30578 2021-01-20 Vladimir N. Makarov <vmakarov@redhat.com>
30580 PR rtl-optimization/98722
30581 * lra-eliminations.c (eliminate_regs_in_insn): Check that target
30582 has no 3-op add insn to transform insns containing two pluses.
30584 2021-01-20 Richard Biener <rguenther@suse.de>
30586 * hwint.h (add_hwi): New function.
30587 (mul_hwi): Likewise.
30588 * tree-data-ref.c (initialize_matrix_A): Properly translate
30589 tree constants and avoid HOST_WIDE_INT_MIN.
30590 (lambda_matrix_row_add): Avoid undefined integer overflow
30591 and return true on such overflow.
30592 (lambda_matrix_right_hermite): Handle overflow from
30593 lambda_matrix_row_add gracefully. Simplify previous fix.
30594 (analyze_subscript_affine_affine): Likewise.
30596 2021-01-20 Eugene Rozenfeld <erozen@microsoft.com>
30598 PR tree-optimization/96674
30599 * match.pd: New patterns: x < y || y == XXX_MIN --> x <= y - 1
30600 x >= y && y != XXX_MIN --> x > y - 1
30602 2021-01-20 Richard Sandiford <richard.sandiford@arm.com>
30604 PR tree-optimization/98535
30605 * tree-vect-slp.c (duplicate_and_interleave): Use quick_grow_cleared.
30606 If the high and low permutes are the same, remove the high permutes
30607 from the working set and only continue with the low ones.
30609 2021-01-20 Jakub Jelinek <jakub@redhat.com>
30611 PR tree-optimization/98721
30612 * builtins.c (access_ref::inform_access): Don't assume
30613 SSA_NAME_IDENTIFIER must be non-NULL. Print messages about
30614 object whenever allocfn is NULL, rather than only when DECL_P
30615 is true. Use %qE instead of %qD for that. Formatting fixes.
30617 2021-01-20 Richard Biener <rguenther@suse.de>
30619 PR tree-optimization/98758
30620 * tree-data-ref.c (int_divides_p): Use lambda_int arguments.
30621 (lambda_matrix_right_hermite): Avoid undefinedness with
30622 signed integer abs and multiplication.
30623 (analyze_subscript_affine_affine): Use lambda_int.
30625 2021-01-20 David Malcolm <dmalcolm@redhat.com>
30628 * dwarf2out.c (output_line_info): Rename static variable
30629 "generation", moving it out of the function to...
30630 (output_line_info_generation): New.
30631 (init_sections_and_labels): Likewise, renaming the variable to...
30632 (init_sections_and_labels_generation): New.
30633 (dwarf2out_c_finalize): Reset the new variables.
30635 2021-01-19 Martin Sebor <msebor@redhat.com>
30637 PR middle-end/98664
30638 * tree-ssa-live.c (remove_unused_scope_block_p): Keep scopes for
30639 all functions, even if they're not declared artificial or inline.
30640 * tree.c (tree_inlined_location): Use macro expansion location
30641 only if scope traversal fails to expose one.
30643 2021-01-19 Richard Sandiford <richard.sandiford@arm.com>
30645 PR rtl-optimization/92294
30646 * alias.c (compare_base_symbol_refs): Take an extra parameter
30647 and add the distance between two symbols to it. Enshrine in
30648 comments that -1 means "either 0 or 1, but we can't tell
30649 which at compile time".
30650 (memrefs_conflict_p): Update call accordingly.
30651 (rtx_equal_for_memref_p): Likewise. Take the distance between symbols
30654 2021-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30656 * config/aarch64/aarch64-simd-builtins.def (sqshl, uqshl,
30657 sqrshl, uqrshl, sqadd, uqadd, sqsub, uqsub, suqadd, usqadd, sqmovn,
30658 uqmovn, sqxtn2, uqxtn2, sqabs, sqneg, sqdmlal, sqdmlsl, sqdmlal_lane,
30659 sqdmlsl_lane, sqdmlal_laneq, sqdmlsl_laneq, sqdmlal_n, sqdmlsl_n,
30660 sqdmlal2, sqdmlsl2, sqdmlal2_lane, sqdmlsl2_lane, sqdmlal2_laneq,
30661 sqdmlsl2_laneq, sqdmlal2_n, sqdmlsl2_n, sqdmull, sqdmull_lane,
30662 sqdmull_laneq, sqdmull_n, sqdmull2, sqdmull2_lane, sqdmull2_laneq,
30663 sqdmull2_n, sqdmulh, sqrdmulh, sqdmulh_lane, sqdmulh_laneq,
30664 sqrdmulh_lane, sqrdmulh_laneq, sqshrun_n, sqrshrun_n, sqshrn_n,
30665 uqshrn_n, sqrshrn_n, uqrshrn_n, sqshlu_n, sqshl_n, uqshl_n, sqrdmlah,
30666 sqrdmlsh, sqrdmlah_lane, sqrdmlsh_lane, sqrdmlah_laneq, sqrdmlsh_laneq,
30667 sqmovun): Use NONE flags.
30669 2021-01-19 Richard Biener <rguenther@suse.de>
30672 * ipa-modref.c (analyze_stmt): Only record a summary for a
30675 2021-01-19 Richard Biener <rguenther@suse.de>
30677 PR middle-end/98638
30678 * tree-ssanames.c (fini_ssanames): Zero SSA_NAME_DEF_STMT.
30680 2021-01-19 Daniel Hellstrom <daniel@gaisler.com>
30682 * config/sparc/rtemself.h (TARGET_OS_CPP_BUILTINS): Add
30683 built-in define __FIX_LEON3FT_TN0018.
30685 2021-01-19 Richard Biener <rguenther@suse.de>
30688 * tree-inline.c (tree_function_versioning): Set input_location
30689 to UNKNOWN_LOCATION throughout the function.
30691 2021-01-19 Tobias Burnus <tobias@codesourcery.com>
30694 * omp-low.c (lower_omp_target): Handle nonpointer is_device_ptr.
30696 2021-01-19 Martin Jambor <mjambor@suse.cz>
30699 * ipa-sra.c (ssa_name_only_returned_p): New parameter fun. Check
30700 whether non-call exceptions allow removal of a statement.
30701 (isra_analyze_call): Pass the appropriate function to
30702 ssa_name_only_returned_p.
30704 2021-01-19 Geng Qi <gengqi@linux.alibaba.com>
30706 * config/riscv/arch-canonicalize (longext_sort): New function for
30707 sorting 'multi-letter'.
30708 * config/riscv/multilib-generator: Adjusting the loop of 'alt' in
30709 'alts'. The 'arch' may not be the first of 'alts'.
30710 (_expand_combination): Add underline for the 'ext' without '*'.
30711 This is because, a single-letter extension can always be treated well
30712 with a '_' prefix, but it cannot be separated out if it is appended
30715 2021-01-18 Vladimir N. Makarov <vmakarov@redhat.com>
30718 * ira.c (ira): Skip abnormal critical edge splitting.
30720 2021-01-18 Jakub Jelinek <jakub@redhat.com>
30722 PR tree-optimization/98727
30723 * tree-ssa-math-opts.c (match_arith_overflow): Fix up computation of
30724 second .MUL_OVERFLOW operand for signed multiplication with overflow
30725 checking if the second operand of multiplication is not constant.
30727 2021-01-18 David Edelsohn <dje.gcc@gmail.com>
30729 * doc/invoke.texi (-gdwarf): TPF defaults to version 2 and AIX
30730 defaults to version 4.
30732 2021-01-18 David Malcolm <dmalcolm@redhat.com>
30734 * attribs.h (fndecl_dealloc_argno): New decl.
30735 * builtins.c (call_dealloc_argno): Split out second half of
30737 (fndecl_dealloc_argno): New.
30738 * doc/extend.texi (Common Function Attributes): Document the
30739 interaction between the analyzer and the malloc attribute.
30740 * doc/invoke.texi (Static Analyzer Options): Likewise.
30742 2021-01-17 David Edelsohn <dje.gcc@gmail.com>
30744 * config/rs6000/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Override
30745 dwarf_version to 4.
30746 * config/rs6000/aix72.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
30748 2021-01-17 Martin Jambor <mjambor@suse.cz>
30751 * cgraph.c (clone_of_p): Check also former_clone_of as we climb
30754 2021-01-17 Mark Wielaard <mark@klomp.org>
30756 * common.opt (gdwarf-): Init(5).
30757 * doc/invoke.texi (-gdwarf): Document default to 5.
30759 2021-01-16 Kwok Cheung Yeung <kcy@codesourcery.com>
30761 * builtin-types.def
30762 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT): Rename
30764 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR):
30765 ...this. Add extra argument.
30766 * gimplify.c (omp_default_clause): Ensure that event handle is
30767 firstprivate in a task region.
30768 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_DETACH.
30769 (gimplify_adjust_omp_clauses): Likewise.
30770 * omp-builtins.def (BUILT_IN_GOMP_TASK): Change function type to
30771 BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR.
30772 * omp-expand.c (expand_task_call): Add GOMP_TASK_FLAG_DETACH to flags
30773 if detach clause specified. Add detach argument when generating
30775 * omp-low.c (scan_sharing_clauses): Setup data environment for detach
30777 (finish_taskreg_scan): Move field for variable containing the event
30778 handle to the front of the struct.
30779 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_DETACH. Fix
30781 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
30782 OMP_CLAUSE_DETACH clause.
30783 (convert_local_omp_clauses): Handle OMP_CLAUSE_DETACH clause.
30784 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_DETACH.
30785 * tree.c (omp_clause_num_ops): Add entry for OMP_CLAUSE_DETACH.
30787 (omp_clause_code_name): Add entry for OMP_CLAUSE_DETACH. Fix
30789 (walk_tree_1): Handle OMP_CLAUSE_DETACH.
30791 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
30793 * config/nios2/t-rtems: Reset all MULTILIB_* variables. Shorten
30794 multilib directory names. Use MULTILIB_REQUIRED instead of
30795 MULTILIB_EXCEPTIONS. Add -mhw-mul -mhw-mulx -mhw-div
30796 -mcustom-fpu-cfg=fph2 multilib.
30798 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
30800 * config/nios2/nios2.c (NIOS2_FPU_CONFIG_NUM): Adjust value.
30801 (nios2_init_fpu_configs): Provide register values for new
30802 -mcustom-fpu-cfg=fph2 option variant.
30803 * doc/invoke.texi (-mcustom-fpu-cfg=fph2): Document new option
30806 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
30808 * config/nios2/nios2.c (nios2_custom_check_insns): Remove
30809 custom instruction warnings.
30811 2021-01-16 Jakub Jelinek <jakub@redhat.com>
30813 PR tree-optimization/96669
30814 * match.pd ((CST << x) & 1 -> x == 0): New simplification.
30816 2021-01-16 Jakub Jelinek <jakub@redhat.com>
30818 PR tree-optimization/96271
30819 * passes.def: Pass false argument to first two pass_cd_dce
30820 instances and true to last instance. Add comment that
30821 last instance rewrites no longer addressed locals.
30822 * tree-ssa-dce.c (pass_cd_dce): Add update_address_taken_p member and
30824 (pass_cd_dce::set_pass_param): New method.
30825 (pass_cd_dce::execute): Return TODO_update_address_taken from
30826 last cd_dce instance.
30828 2021-01-15 Carl Love <cel@us.ibm.com>
30830 * config/rs6000/altivec.h (vec_mulh, vec_div, vec_dive, vec_mod):
30832 * config/rs6000/altivec.md (VIlong): Move define to file vsx.md.
30833 * config/rs6000/rs6000-builtin.def (DIVES_V4SI, DIVES_V2DI,
30834 DIVEU_V4SI, DIVEU_V2DI, DIVS_V4SI, DIVS_V2DI, DIVU_V4SI,
30835 DIVU_V2DI, MODS_V2DI, MODS_V4SI, MODU_V2DI, MODU_V4SI,
30836 MULHS_V2DI, MULHS_V4SI, MULHU_V2DI, MULHU_V4SI, MULLD_V2DI):
30837 Add builtin define.
30838 (MULH, DIVE, MOD): Add new BU_P10_OVERLOAD_2 definitions.
30839 * config/rs6000/rs6000-call.c (VSX_BUILTIN_VEC_DIV,
30840 VSX_BUILTIN_VEC_DIVE, P10_BUILTIN_VEC_MOD, P10_BUILTIN_VEC_MULH):
30841 New overloaded definitions.
30842 (builtin_function_type) [P10V_BUILTIN_DIVEU_V4SI,
30843 P10V_BUILTIN_DIVEU_V2DI, P10V_BUILTIN_DIVU_V4SI,
30844 P10V_BUILTIN_DIVU_V2DI, P10V_BUILTIN_MODU_V2DI,
30845 P10V_BUILTIN_MODU_V4SI, P10V_BUILTIN_MULHU_V2DI,
30846 P10V_BUILTIN_MULHU_V4SI]: Add case
30847 statement for builtins.
30848 * config/rs6000/rs6000.md (bits): Add new attribute sizes V4SI, V2DI.
30849 * config/rs6000/vsx.md (VIlong): Moved from config/rs6000/altivec.md.
30850 (UNSPEC_VDIVES, UNSPEC_VDIVEU): New unspec definitions.
30851 (vsx_mul_v2di): Add if TARGET_POWER10 statement.
30852 (vsx_udiv_v2di): Add if TARGET_POWER10 statement.
30853 (dives_<mode>, diveu_<mode>, div<mode>3, uvdiv<mode>3,
30854 mods_<mode>, modu_<mode>, mulhs_<mode>, mulhu_<mode>, mulv2di3):
30855 Add define_insn, mode is VIlong.
30856 * doc/extend.texi (vec_mulh, vec_mul, vec_div, vec_dive, vec_mod):
30857 Add builtin descriptions.
30859 2021-01-15 Eric Botcazou <ebotcazou@adacore.com>
30861 * final.c (final_start_function_1): Reset force_source_line.
30863 2021-01-15 Jakub Jelinek <jakub@redhat.com>
30865 PR tree-optimization/96669
30866 * match.pd (((1 << A) & 1) != 0 -> A == 0,
30867 ((1 << A) & 1) == 0 -> A != 0): Generalize for 1s replaced by
30868 possibly different power of two constants and to right shift too.
30870 2021-01-15 Jakub Jelinek <jakub@redhat.com>
30872 PR tree-optimization/96681
30873 * match.pd ((x < 0) ^ (y < 0) to (x ^ y) < 0): New simplification.
30874 ((x >= 0) ^ (y >= 0) to (x ^ y) < 0): Likewise.
30875 ((x < 0) ^ (y >= 0) to (x ^ y) >= 0): Likewise.
30876 ((x >= 0) ^ (y < 0) to (x ^ y) >= 0): Likewise.
30878 2021-01-15 Alexandre Oliva <oliva@adacore.com>
30880 * opts.c (gen_command_line_string): Exclude -dumpbase-ext.
30882 2021-01-15 Tamar Christina <tamar.christina@arm.com>
30884 * config/aarch64/aarch64-simd.md (cml<fcmac1><conj_op><mode>4,
30885 cmul<conj_op><mode>3): New.
30886 * config/aarch64/iterators.md (UNSPEC_FCMUL,
30887 UNSPEC_FCMUL180, UNSPEC_FCMLA_CONJ, UNSPEC_FCMLA180_CONJ,
30888 UNSPEC_CMLA_CONJ, UNSPEC_CMLA180_CONJ, UNSPEC_CMUL, UNSPEC_CMUL180,
30889 FCMLA_OP, FCMUL_OP, conj_op, rotsplit1, rotsplit2, fcmac1, sve_rot1,
30890 sve_rot2, SVE2_INT_CMLA_OP, SVE2_INT_CMUL_OP, SVE2_INT_CADD_OP): New.
30891 (rot): Add UNSPEC_FCMUL, UNSPEC_FCMUL180.
30892 (rot_op): Renamed to conj_op.
30893 * config/aarch64/aarch64-sve.md (cml<fcmac1><conj_op><mode>4,
30894 cmul<conj_op><mode>3): New.
30895 * config/aarch64/aarch64-sve2.md (cml<fcmac1><conj_op><mode>4,
30896 cmul<conj_op><mode>3): New.
30898 2021-01-15 David Malcolm <dmalcolm@redhat.com>
30902 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
30903 Escape the tempfile name when constructing the expected output.
30905 2021-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30907 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlsl_hi<mode>):
30909 (aarch64_<su>mlsl_hi<mode>): ... This.
30910 (aarch64_<su>mlsl_hi<mode>): Define.
30911 (*aarch64_<su>mlsl<mode): Rename to...
30912 (aarch64_<su>mlsl<mode): ... This.
30913 * config/aarch64/aarch64-simd-builtins.def (smlsl, umlsl,
30914 smlsl_hi, umlsl_hi): Define builtins.
30915 * config/aarch64/arm_neon.h (vmlsl_high_s8, vmlsl_high_s16,
30916 vmlsl_high_s32, vmlsl_high_u8, vmlsl_high_u16, vmlsl_high_u32,
30917 vmlsl_s8, vmlsl_s16, vmlsl_s32, vmlsl_u8,
30918 vmlsl_u16, vmlsl_u32): Reimplement with builtins.
30920 2021-01-15 Uroš Bizjak <ubizjak@gmail.com>
30922 * config/i386/i386-c.c (ix86_target_macros):
30923 Use cpp_define_formatted for __SIZEOF_FLOAT80__ definition.
30925 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
30928 * config.gcc (aarch64*-*-*): Add aarch64-cc-fusion.o to extra_objs.
30929 * Makefile.in (RTL_SSA_H): New variable.
30930 * config/aarch64/t-aarch64 (aarch64-cc-fusion.o): New rule.
30931 * config/aarch64/aarch64-protos.h (make_pass_cc_fusion): Declare.
30932 * config/aarch64/aarch64-passes.def: Add pass_cc_fusion after
30934 * config/aarch64/aarch64-cc-fusion.cc: New file.
30936 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
30938 * recog.h (insn_change_watermark::~insn_change_watermark): Avoid
30939 calling cancel_changes for changes that no longer exist.
30941 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
30943 * rtl-ssa/functions.h (function_info::ref_defs): Rename to...
30944 (function_info::reg_defs): ...this.
30945 * rtl-ssa/member-fns.inl (function_info::ref_defs): Rename to...
30946 (function_info::reg_defs): ...this.
30948 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
30951 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
30953 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
30956 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
30959 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
30961 2021-01-15 Richard Biener <rguenther@suse.de>
30963 PR tree-optimization/96376
30964 * tree-vect-stmts.c (get_load_store_type): Disregard alignment
30965 for VMAT_INVARIANT.
30967 2021-01-15 Martin Liska <mliska@suse.cz>
30969 * doc/install.texi: Document that some tests need pytest module.
30970 * doc/sourcebuild.texi: Likewise.
30972 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
30975 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
30977 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
30979 * config/arm/mve.md (mve_vshrq_n_s<mode>_imm): New entry.
30980 (mve_vshrq_n_u<mode>_imm): Likewise.
30981 * config/arm/neon.md (vashr<mode>3, vlshr<mode>3): Move to ...
30982 * config/arm/vec-common.md: ... here.
30984 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
30986 * config/arm/mve.md (mve_vshlq_<supf><mode>): Move to
30988 * config/arm/neon.md (vashl<mode>3): Delete.
30989 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): New.
30990 (vasl<mode>3): New expander.
30992 2021-01-15 Richard Biener <rguenther@suse.de>
30994 PR tree-optimization/98685
30995 * tree-vect-slp.c (vect_schedule_slp_node): Refactor handling
30996 of vector extern defs.
30998 2021-01-14 David Malcolm <dmalcolm@redhat.com>
31001 * diagnostic.c (diagnostic_kind_text): Break out this array
31003 (diagnostic_build_prefix): ...here.
31004 (fancy_abort): Detect when diagnostic_initialize has not yet been
31005 called and fall back to a minimal implementation of printing the
31006 ICE, rather than segfaulting in internal_error.
31008 2021-01-14 David Malcolm <dmalcolm@redhat.com>
31010 * diagnostic.c (diagnostic_initialize): Eliminate
31011 parseable_fixits_p in favor of initializing extra_output_kind from
31012 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
31013 (convert_column_unit): New function, split out from...
31014 (diagnostic_converted_column): ...this.
31015 (print_parseable_fixits): Add "column_unit" and "tabstop" params.
31016 Use them to call convert_column_unit on the column values.
31017 (diagnostic_report_diagnostic): Eliminate conditional on
31018 parseable_fixits_p in favor of a switch statement on
31019 extra_output_kind, passing the appropriate values to the new
31020 params of print_parseable_fixits.
31021 (selftest::test_print_parseable_fixits_none): Update for new
31022 params of print_parseable_fixits.
31023 (selftest::test_print_parseable_fixits_insert): Likewise.
31024 (selftest::test_print_parseable_fixits_remove): Likewise.
31025 (selftest::test_print_parseable_fixits_replace): Likewise.
31026 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
31028 (selftest::diagnostic_c_tests): Call it.
31029 * diagnostic.h (enum diagnostics_extra_output_kind): New.
31030 (diagnostic_context::parseable_fixits_p): Delete field in favor
31032 (diagnostic_context::extra_output_kind): ...this new field.
31033 * doc/invoke.texi (Environment Variables): Add
31034 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
31035 * opts.c (common_handle_option): Update handling of
31036 OPT_fdiagnostics_parseable_fixits for change to diagnostic_context
31039 2021-01-14 Tamar Christina <tamar.christina@arm.com>
31041 * tree-vect-slp-patterns.c (class complex_operations_pattern,
31042 complex_operations_pattern::matches,
31043 complex_operations_pattern::recognize,
31044 complex_operations_pattern::build): New.
31045 (slp_patterns): Use it.
31047 2021-01-14 Tamar Christina <tamar.christina@arm.com>
31049 * internal-fn.def (COMPLEX_FMS, COMPLEX_FMS_CONJ): New.
31050 * optabs.def (cmls_optab, cmls_conj_optab): New.
31051 * doc/md.texi: Document them.
31052 * tree-vect-slp-patterns.c (class complex_fms_pattern,
31053 complex_fms_pattern::matches, complex_fms_pattern::recognize,
31054 complex_fms_pattern::build): New.
31056 2021-01-14 Tamar Christina <tamar.christina@arm.com>
31058 * internal-fn.def (COMPLEX_FMA, COMPLEX_FMA_CONJ): New.
31059 * optabs.def (cmla_optab, cmla_conj_optab): New.
31060 * doc/md.texi: Document them.
31061 * tree-vect-slp-patterns.c (vect_match_call_p,
31062 class complex_fma_pattern, vect_slp_reset_pattern,
31063 complex_fma_pattern::matches, complex_fma_pattern::recognize,
31064 complex_fma_pattern::build): New.
31066 2021-01-14 Tamar Christina <tamar.christina@arm.com>
31068 * internal-fn.def (COMPLEX_MUL, COMPLEX_MUL_CONJ): New.
31069 * optabs.def (cmul_optab, cmul_conj_optab): New.
31070 * doc/md.texi: Document them.
31071 * tree-vect-slp-patterns.c (vect_match_call_complex_mla,
31072 vect_normalize_conj_loc, is_eq_or_top, vect_validate_multiplication,
31073 vect_build_combine_node, class complex_mul_pattern,
31074 complex_mul_pattern::matches, complex_mul_pattern::recognize,
31075 complex_mul_pattern::build): New.
31077 2021-01-14 Tamar Christina <tamar.christina@arm.com>
31079 * tree-vect-slp.c (optimize_load_redistribution_1): New.
31080 (optimize_load_redistribution, vect_is_slp_load_node): New.
31081 (vect_match_slp_patterns): Use it.
31083 2021-01-14 Tamar Christina <tamar.christina@arm.com>
31085 * tree-vect-slp-patterns.c (complex_add_pattern::build):
31088 2021-01-14 Thomas Schwinge <thomas@codesourcery.com>
31090 * config/gcn/mkoffload.c (main): Create an offload image only in
31091 64-bit configurations.
31093 2021-01-14 H.J. Lu <hjl.tools@gmail.com>
31096 * config/i386/i386-options.c (ix86_option_override_internal):
31097 Issue an error for -fcf-protection with CF_BRANCH when compiling
31098 for 32-bit non-TARGET_CMOV targets.
31100 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
31103 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p):
31104 Remove declaration and initialization of shadow variable "ret".
31105 (ix86_option_override_internal): Remove delcaration of
31106 shadow variable "i". Redeclare shadowed variable to unsigned.
31107 * common/config/i386/i386-common.c (pta_size): Redeclare to unsigned.
31108 * config/i386/i386-builtins.c (get_builtin_code_for_version):
31109 Update for redeclaration.
31110 * config/i386/i386.h (pta_size): Ditto.
31112 2021-01-14 Richard Biener <rguenther@suse.de>
31114 PR tree-optimization/98674
31115 * tree-data-ref.c (base_supports_access_fn_components_p): New.
31116 (initialize_data_dependence_relation): For two bases without
31117 possible access fns resort to type size equality when determining
31118 shape compatibility.
31120 2021-01-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
31123 * config/arm/arm_neon.h: Replace calls to __builtin_vcge* by
31124 <=, >= operators in vcle and vcge intrinsics respectively.
31125 * config/arm/arm_neon_builtins.def: Remove entry for
31128 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
31131 * config/i386/i386-options.c (ix86_function_specific_save):
31132 Remove redundant assignment to opts->x_ix86_branch_cost.
31133 * config/i386/i386.c (ix86_prefetch_sse):
31134 Rename from x86_prefetch_sse. Update all uses.
31135 * config/i386/i386.h: Update for rename.
31136 * config/i386/i386-options.h: Ditto.
31138 2021-01-14 Jakub Jelinek <jakub@redhat.com>
31141 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3,
31142 *sse4_1_zero_extendv4hiv4si2_3, *sse4_1_zero_extendv2siv2di2_3):
31143 Use Bm instead of m for non-avx. Add isa attribute.
31145 2021-01-14 Jakub Jelinek <jakub@redhat.com>
31147 PR tree-optimization/96688
31148 * match.pd (~(X >> Y) -> ~X >> Y): New simplification if
31149 ~X can be simplified.
31151 2021-01-14 Richard Sandiford <richard.sandiford@arm.com>
31153 * tree-vect-stmts.c (vect_model_load_cost): Account for unused
31154 IFN_LOAD_LANES results.
31156 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
31158 * config/aarch64/aarch64-simd.md (aarch64_<su>xtl<mode>):
31160 (aarch64_xtn<mode>): Likewise.
31161 * config/aarch64/aarch64-simd-builtins.def (sxtl, uxtl, xtn):
31164 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
31166 (vmovl_s16): Likewise.
31167 (vmovl_s32): Likewise.
31168 (vmovl_u8): Likewise.
31169 (vmovl_u16): Likewise.
31170 (vmovl_u32): Likewise.
31171 (vmovn_s16): Likewise.
31172 (vmovn_s32): Likewise.
31173 (vmovn_s64): Likewise.
31174 (vmovn_u16): Likewise.
31175 (vmovn_u32): Likewise.
31176 (vmovn_u64): Likewise.
31178 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
31180 * config/aarch64/aarch64-simd.md (aarch64_<su>qxtn2<mode>_le):
31182 (aarch64_<su>qxtn2<mode>_be): Likewise.
31183 (aarch64_<su>qxtn2<mode>): Likewise.
31184 * config/aarch64/aarch64-simd-builtins.def (sqxtn2, uqxtn2):
31186 * config/aarch64/iterators.md (SAT_TRUNC): Define code_iterator.
31187 (su): Handle ss_truncate and us_truncate.
31188 * config/aarch64/arm_neon.h (vqmovn_high_s16): Reimplement using
31190 (vqmovn_high_s32): Likewise.
31191 (vqmovn_high_s64): Likewise.
31192 (vqmovn_high_u16): Likewise.
31193 (vqmovn_high_u32): Likewise.
31194 (vqmovn_high_u64): Likewise.
31196 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
31198 * config/aarch64/aarch64-simd.md (aarch64_xtn2<mode>_le):
31200 (aarch64_xtn2<mode>_be): Likewise.
31201 (aarch64_xtn2<mode>): Likewise.
31202 * config/aarch64/aarch64-simd-builtins.def (xtn2): Define
31204 * config/aarch64/arm_neon.h (vmovn_high_s16): Reimplement using
31206 (vmovn_high_s32): Likewise.
31207 (vmovn_high_s64): Likewise.
31208 (vmovn_high_u16): Likewise.
31209 (vmovn_high_u32): Likewise.
31210 (vmovn_high_u64): Likewise.
31212 2021-01-13 Stafford Horne <shorne@gmail.com>
31214 * config/or1k/or1k.h (ASM_PREFERRED_EH_DATA_FORMAT): New macro.
31216 2021-01-13 Stafford Horne <shorne@gmail.com>
31218 * config/or1k/linux.h (TARGET_ASM_FILE_END): Define macro.
31220 2021-01-13 Stafford Horne <shorne@gmail.com>
31222 * config/or1k/or1k.h (TARGET_CPU_CPP_BUILTINS): Add builtin
31223 define for __or1k_hard_float__.
31225 2021-01-13 Stafford Horne <shorne@gmail.com>
31227 * config/or1k/or1k.h (NO_PROFILE_COUNTERS): Define as 1.
31228 (PROFILE_HOOK): Define to call _mcount.
31229 (FUNCTION_PROFILER): Change from abort to no-op.
31231 2021-01-13 Jakub Jelinek <jakub@redhat.com>
31233 PR tree-optimization/96691
31234 * match.pd ((~X | C) ^ D -> (X | C) ^ (~D ^ C),
31235 (~X & C) ^ D -> (X & C) ^ (D ^ C)): New simplifications if
31236 (~D ^ C) or (D ^ C) can be simplified.
31238 2021-01-13 Richard Biener <rguenther@suse.de>
31240 PR tree-optimization/92645
31241 * match.pd (BIT_FIELD_REF to conversion): Delay canonicalization
31242 until after vector lowering.
31244 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
31246 * config/aarch64/aarch64-sve.md (fnma<mode>4): Extend from SVE_FULL_I
31248 (@aarch64_pred_fnma<mode>, cond_fnma<mode>, *cond_fnma<mode>_2)
31249 (*cond_fnma<mode>_4, *cond_fnma<mode>_any): Likewise.
31251 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
31253 * config/aarch64/aarch64-sve.md (fma<mode>4): Extend from SVE_FULL_I
31255 (@aarch64_pred_fma<mode>, cond_fma<mode>, *cond_fma<mode>_2)
31256 (*cond_fma<mode>_4, *cond_fma<mode>_any): Likewise.
31258 2021-01-13 Richard Biener <rguenther@suse.de>
31260 PR tree-optimization/92645
31261 * tree-vect-slp.c (vect_build_slp_tree_1): Relax supported
31262 BIT_FIELD_REF argument.
31263 (vect_build_slp_tree_2): Record the desired vector type
31264 on the external vector def.
31265 (vectorizable_slp_permutation): Handle required punning
31266 of existing vector defs.
31268 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
31270 * rtl-ssa/accesses.h (def_lookup): Fix order of comparison results.
31272 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
31274 * config/sh/sh.md (movsf_ie): Remove operands[2] test.
31276 2021-01-13 Samuel Thibault <samuel.thibault@ens-lyon.org>
31278 * config.gcc [$target == *-*-gnu*]: Enable
31279 'default_gnu_indirect_function'.
31281 2021-01-13 Jakub Jelinek <jakub@redhat.com>
31284 * optabs.c (expand_vec_perm_const): Don't force v0 and v1 into
31285 registers before calling targetm.vectorize.vec_perm_const, only after
31287 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const): Handle
31288 two argument permutation when one operand is zero vector and only
31289 after that force operands into registers.
31290 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_1): New
31291 define_insn_and_split pattern.
31292 (*avx512bw_zero_extendv32qiv32hi2_1): Likewise.
31293 (*avx512f_zero_extendv16hiv16si2_1): Likewise.
31294 (*avx2_zero_extendv8hiv8si2_1): Likewise.
31295 (*avx512f_zero_extendv8siv8di2_1): Likewise.
31296 (*avx2_zero_extendv4siv4di2_1): Likewise.
31297 * config/mips/mips.c (mips_vectorize_vec_perm_const): Force operands
31299 * config/arm/arm.c (arm_vectorize_vec_perm_const): Likewise.
31300 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Likewise.
31301 * config/ia64/ia64.c (ia64_vectorize_vec_perm_const): Likewise.
31302 * config/aarch64/aarch64.c (aarch64_vectorize_vec_perm_const): Likewise.
31303 * config/rs6000/rs6000.c (rs6000_vectorize_vec_perm_const): Likewise.
31304 * config/gcn/gcn.c (gcn_vectorize_vec_perm_const): Likewise. Use std::swap.
31306 2021-01-13 Martin Liska <mliska@suse.cz>
31308 PR tree-optimization/98455
31309 * gimple-if-to-switch.cc (condition_info::record_phi_mapping):
31310 Record also virtual PHIs.
31311 (pass_if_to_switch::execute): Return TODO_cleanup_cfg only
31314 2021-01-13 Jonathan Wakely <jwakely@redhat.com>
31316 * doc/invoke.texi (C++ Modules): Fix typos.
31318 2021-01-13 Richard Biener <rguenther@suse.de>
31320 PR tree-optimization/98640
31321 * tree-ssa-sccvn.c (visit_nary_op): Do not try to
31322 handle plus or minus from a truncated operand to be
31325 2021-01-13 Jakub Jelinek <jakub@redhat.com>
31328 * config/i386/i386.md (*btr<mode>_1, *btr<mode>_2): New
31329 define_insn_and_split patterns.
31330 (splitter after *btr<mode>_2): New splitter.
31332 2021-01-13 Martin Liska <mliska@suse.cz>
31335 * cgraphunit.c (analyze_functions): Remove dead code.
31337 2021-01-13 Qian Jianhua <qianjh@cn.fujitsu.com>
31339 * config/aarch64/aarch64-cost-tables.h (a64fx_extra_costs): New.
31340 * config/aarch64/aarch64.c (a64fx_addrcost_table): New.
31341 (a64fx_regmove_cost, a64fx_vector_cost): New.
31342 (a64fx_tunings): Use the new added cost tables.
31344 2021-01-13 Jakub Jelinek <jakub@redhat.com>
31347 * config/i386/predicates.md (pmovzx_parallel): New predicate.
31348 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): New
31349 define_insn_and_split pattern.
31350 (*sse4_1_zero_extendv4hiv4si2_3): Likewise.
31351 (*sse4_1_zero_extendv2siv2di2_3): Likewise.
31353 2021-01-13 Julian Brown <julian@codesourcery.com>
31355 * config/gcn/gcn.c (gcn_conditional_register_usage): Remove dead code
31356 to fix v0 register.
31358 2021-01-13 Julian Brown <julian@codesourcery.com>
31360 * config/gcn/gcn.c (gcn_md_reorg): Fix case where EXEC reg is live
31363 2021-01-13 Julian Brown <julian@codesourcery.com>
31365 * config/gcn/gcn-valu.md (recip<mode>2<exec>, recip<mode>2): Use unspec
31366 for reciprocal-approximation instructions.
31367 (div<mode>3): Use fused multiply-accumulate operations for reciprocal
31368 refinement and division result.
31369 * config/gcn/gcn.md (UNSPEC_RCP): New unspec constant.
31371 2021-01-13 Julian Brown <julian@codesourcery.com>
31373 * config/gcn/gcn-valu.md (subdf): Rename to...
31376 2021-01-12 Martin Liska <mliska@suse.cz>
31378 * gcov.c (source_info::debug): Fix printf format for 32-bit hosts.
31380 2021-01-12 Andrea Corallo <andrea.corallo@arm.com>
31382 * function-abi.h: Fix typo.
31384 2021-01-12 Christophe Lyon <christophe.lyon@linaro.org>
31388 * config/arm/arm.h (ARM_HAVE_NEON_V8QI_LDST): New macro.
31389 (ARM_HAVE_NEON_V16QI_LDST, ARM_HAVE_NEON_V4HI_LDST): Likewise.
31390 (ARM_HAVE_NEON_V8HI_LDST, ARM_HAVE_NEON_V2SI_LDST): Likewise.
31391 (ARM_HAVE_NEON_V4SI_LDST, ARM_HAVE_NEON_V4HF_LDST): Likewise.
31392 (ARM_HAVE_NEON_V8HF_LDST, ARM_HAVE_NEON_V4BF_LDST): Likewise.
31393 (ARM_HAVE_NEON_V8BF_LDST, ARM_HAVE_NEON_V2SF_LDST): Likewise.
31394 (ARM_HAVE_NEON_V4SF_LDST, ARM_HAVE_NEON_DI_LDST): Likewise.
31395 (ARM_HAVE_NEON_V2DI_LDST): Likewise.
31396 (ARM_HAVE_V8QI_LDST, ARM_HAVE_V16QI_LDST): Likewise.
31397 (ARM_HAVE_V4HI_LDST, ARM_HAVE_V8HI_LDST): Likewise.
31398 (ARM_HAVE_V2SI_LDST, ARM_HAVE_V4SI_LDST, ARM_HAVE_V4HF_LDST): Likewise.
31399 (ARM_HAVE_V8HF_LDST, ARM_HAVE_V4BF_LDST, ARM_HAVE_V8BF_LDST): Likewise.
31400 (ARM_HAVE_V2SF_LDST, ARM_HAVE_V4SF_LDST, ARM_HAVE_DI_LDST): Likewise.
31401 (ARM_HAVE_V2DI_LDST): Likewise.
31402 * config/arm/mve.md (*movmisalign<mode>_mve_store): New pattern.
31403 (*movmisalign<mode>_mve_load): New pattern.
31404 * config/arm/neon.md (movmisalign<mode>): Move to ...
31405 * config/arm/vec-common.md: ... here.
31407 2021-01-12 Vladimir N. Makarov <vmakarov@redhat.com>
31410 * lra-eliminations.c (eliminate_regs_in_insn): Add transformation
31411 of pattern 'plus (plus (hard reg, const), pseudo)'.
31413 2021-01-12 Richard Biener <rguenther@suse.de>
31415 PR tree-optimization/98550
31416 * tree-vect-slp.c (vect_record_max_nunits): Check whether
31417 the group size is a multiple of the vector element count.
31418 (vect_build_slp_tree_1): When we need to fail because
31419 the vector type choosen causes unrolling do so lazily
31420 without affecting matches only at the end to guide group splitting.
31422 2021-01-12 Martin Liska <mliska@suse.cz>
31425 * optc-save-gen.awk: Compare also n_target_save vars with
31428 2021-01-12 Martin Liska <mliska@suse.cz>
31430 * gcov.c (source_info::debug): New.
31431 (print_usage): Add --debug (-D) option.
31432 (process_args): Likewise.
31433 (generate_results): Call src->debug after
31434 accumulate_line_counts.
31435 (read_graph_file): Properly assign id for EXIT_BLOCK.
31436 * profile.c (branch_prob): Dump function body before it is
31439 2021-01-12 Jakub Jelinek <jakub@redhat.com>
31441 PR tree-optimization/98629
31442 * tree-ssa-math-opts.c (arith_overflow_check_p): Don't update use_stmt
31443 unless returning non-zero.
31445 2021-01-12 Jakub Jelinek <jakub@redhat.com>
31447 PR tree-optimization/95731
31448 * tree-ssa-reassoc.c (optimize_range_tests_cmp_bitwise): Also optimize
31449 x < 0 && y < 0 && z < 0 into (x | y | z) < 0 for signed x, y, z.
31450 (optimize_range_tests): Call optimize_range_tests_cmp_bitwise
31451 only after optimize_range_tests_var_bound.
31453 2021-01-12 Jakub Jelinek <jakub@redhat.com>
31455 * configure.ac: Ensure c/Make-lang.in comes first in @all_lang_makefrags@.
31456 * configure: Regenerated.
31458 2021-01-12 liuhongt <hongtao.liu@intel.com>
31461 * config/i386/i386-builtins.h (BUILTIN_DESC_SWAP_OPERANDS):
31463 * config/i386/i386-expand.c (ix86_expand_sse_comi): Delete
31466 2021-01-12 Alexandre Oliva <oliva@adacore.com>
31468 * ssa-iterators.h (end_imm_use_stmt_traverse): Forward
31470 (auto_end_imm_use_stmt_traverse): New struct.
31471 (FOR_EACH_IMM_USE_STMT): Use it.
31472 (BREAK_FROM_IMM_USE_STMT, RETURN_FROM_IMM_USE_STMT): Remove,
31474 * gimple-ssa-strength-reduction.c: ... here, ...
31475 * graphite-scop-detection.c: ... here, ...
31476 * ipa-modref.c, ipa-pure-const.c, ipa-sra.c: ... here, ...
31477 * tree-predcom.c, tree-ssa-ccp.c: ... here, ...
31478 * tree-ssa-dce.c, tree-ssa-dse.c: ... here, ...
31479 * tree-ssa-loop-ivopts.c, tree-ssa-math-opts.c: ... here, ...
31480 * tree-ssa-phiprop.c, tree-ssa.c: ... here, ...
31481 * tree-vect-slp.c: ... and here, ...
31482 * doc/tree-ssa.texi: ... and the example here.
31484 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
31486 * config/aarch64/aarch64-sve.md (sdiv_pow2<mode>3): Extend from
31487 SVE_FULL_I to SVE_I. Generate an UNSPEC_PRED_X.
31488 (*sdiv_pow2<mode>3): New pattern.
31489 (@cond_<sve_int_op><mode>): Extend from SVE_FULL_I to SVE_I.
31490 Wrap the ASRD in an UNSPEC_PRED_X.
31491 (*cond_<sve_int_op><mode>_2): Likewise. Replace the UNSPEC_PRED_X
31492 predicate with a constant PTRUE, if it isn't already.
31493 (*cond_<sve_int_op><mode>_z): Replace with...
31494 (*cond_<sve_int_op><mode>_any): ...this new pattern.
31496 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
31498 * config/aarch64/aarch64-sve.md (*cond_bic<mode>_2): Extend from
31499 SVE_FULL_I to SVE_I.
31500 (*cond_bic<mode>_any): Likewise.
31502 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
31504 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart)
31505 (@aarch64_pred_<MUL_HIGHPART:optab><mode>): Extend from SVE_FULL_I
31508 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
31510 * config/aarch64/aarch64-sve.md (<su>abd<mode>_3): Extend from
31511 SVE_FULL_I to SVE_I.
31512 (*aarch64_cond_<su>abd<mode>_2): Likewise.
31513 (*aarch64_cond_<su>abd<mode>_any): Likewise.
31514 (@aarch64_pred_<su>abd<mode>): Likewise. Use UNSPEC_PRED_X
31515 for the max and min but not for the minus.
31516 (*aarch64_cond_<su>abd<mode>_3): New pattern.
31518 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
31520 * config/aarch64/iterators.md (SVE_24I): New iterator.
31521 * config/aarch64/aarch64-sve.md (*aarch64_adr<mode>_shift): Extend from
31522 SVE_FULL_SDI to SVE_24I. Use containers rather than elements.
31524 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
31526 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_BINARY:optab><mode>)
31527 (*cond_<SVE_INT_BINARY:optab><mode>_2): Extend from SVE_FULL_I
31529 (*cond_<SVE_INT_BINARY:optab><mode>_3): Likewise.
31530 (*cond_<SVE_INT_BINARY:optab><mode>_any): Likewise.
31531 (*cond_<SVE_INT_BINARY:optab><mode>_2_const): Likewise.
31532 (*cond_<SVE_INT_BINARY:optab><mode>_any_const): Likewise.
31534 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
31536 * config/aarch64/aarch64-sve.md (<SVE_INT_BINARY_IMM:optab><mode>3)
31537 (@aarch64_pred_<SVE_INT_BINARY_IMM:optab><mode>)
31538 (*post_ra_<SVE_INT_BINARY_IMM:optab><mode>3): Extend from SVE_FULL_I
31541 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
31543 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3)
31544 (v<ASHIFT:optab><mode>3, @aarch64_pred_<optab><mode>)
31545 (*post_ra_v<ASHIFT:optab><mode>3): Extend from SVE_FULL_I to SVE_I.
31547 2021-01-11 Martin Liska <mliska@suse.cz>
31550 * symtab-clones.h (clone_info::release): Release
31551 symtab::m_clones with ggc_delete as it's a GGC memory.
31553 2021-01-11 Matthias Klose <doko@ubuntu.com>
31555 * Makefile.in (LINK_PROGRESS): Show the link target.
31557 2021-01-11 Richard Biener <rguenther@suse.de>
31559 PR tree-optimization/91403
31560 * tree-vect-data-refs.c (vect_analyze_group_access_1): Cap
31561 single-element interleaving group size at 4096 elements.
31563 2021-01-11 Richard Biener <rguenther@suse.de>
31565 PR tree-optimization/98526
31566 * tree-vect-loop.c (vect_model_reduction_cost): Remove costing
31567 of the actual reduction op for the regular case.
31568 (vectorizable_reduction): Cost the stmts
31569 vect_transform_reduction produces here.
31571 2021-01-11 Andreas Krebbel <krebbel@linux.ibm.com>
31573 * tree-ssa-forwprop.c (simplify_vector_constructor): For
31574 big-endian, use UNPACK[_FLOAT]_HI.
31576 2021-01-11 Tamar Christina <tamar.christina@arm.com>
31578 * tree-vect-slp-patterns.c (class complex_pattern,
31579 class complex_add_pattern): Add parameters to matches.
31580 (complex_add_pattern::build): Free memory.
31581 (complex_add_pattern::matches): Move validation end of match.
31582 (complex_add_pattern::recognize): Likewise.
31584 2021-01-11 Tamar Christina <tamar.christina@arm.com>
31586 * tree-vect-slp-patterns.c (linear_loads_p): Fix externals.
31588 2021-01-11 Tamar Christina <tamar.christina@arm.com>
31590 * tree-vect-slp-patterns.c (is_linear_load_p): Fix ambiguity.
31592 2021-01-11 Jakub Jelinek <jakub@redhat.com>
31594 PR tree-optimization/95867
31595 * tree-ssa-math-opts.h: New header.
31596 * tree-ssa-math-opts.c: Include tree-ssa-math-opts.h.
31597 (powi_as_mults): No longer static. Use build_one_cst instead of
31598 build_real. Formatting fix.
31599 * tree-ssa-reassoc.c: Include tree-ssa-math-opts.h.
31600 (attempt_builtin_powi): Handle multiplication reassociation without
31601 powi_fndecl using powi_as_mults.
31602 (reassociate_bb): For integral types don't require
31603 -funsafe-math-optimizations to call attempt_builtin_powi.
31605 2021-01-11 Jakub Jelinek <jakub@redhat.com>
31607 PR tree-optimization/95852
31608 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): Change
31609 mul_stmts parameter type to vec<gimple *> &. Before cond_stmt
31610 allow in the bb any of the stmts in that vector, div_stmt and
31611 up to 3 cast stmts.
31612 (arith_cast_equal_p): New function.
31613 (arith_overflow_check_p): Add cast_stmt argument, handle signed
31614 multiply overflow checks.
31615 (match_arith_overflow): Adjust caller. Handle signed multiply
31618 2021-01-11 Jakub Jelinek <jakub@redhat.com>
31620 PR tree-optimization/95852
31621 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): New function.
31622 (uaddsub_overflow_check_p): Renamed to ...
31623 (arith_overflow_check_p): ... this. Handle also multiplication
31624 with overflow check.
31625 (match_uaddsub_overflow): Renamed to ...
31626 (match_arith_overflow): ... this. Add cfg_changed argument. Handle
31627 also multiplication with overflow check. Adjust function comment.
31628 (math_opts_dom_walker::after_dom_children): Adjust callers. Call
31629 match_arith_overflow also for MULT_EXPR.
31631 2021-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
31633 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
31634 __builtin_convertvector.
31635 (vmovl_s16): Likewise.
31636 (vmovl_s32): Likewise.
31637 (vmovl_u8): Likewise.
31638 (vmovl_u16): Likewise.
31639 (vmovl_u32): Likewise.
31640 (vmovn_s16): Likewise.
31641 (vmovn_s32): Likewise.
31642 (vmovn_s64): Likewise.
31643 (vmovn_u16): Likewise.
31644 (vmovn_u32): Likewise.
31645 (vmovn_u64): Likewise.
31647 2021-01-11 Martin Liska <mliska@suse.cz>
31649 * gimple-if-to-switch.cc (struct condition_info): Use auto_var.
31650 (if_chain::is_beneficial): Delete clusters
31651 (find_conditions): Make second argument of conditions_in_bbs a
31652 pointer so that we control over it's lifetime.
31653 (pass_if_to_switch::execute): Delete them.
31655 2021-01-11 Kewen Lin <linkw@linux.ibm.com>
31657 * ira.c (move_unallocated_pseudos): Check other_reg and skip if
31660 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
31662 * config/vax/vax.md (cc): Remove mode attribute.
31663 (subst_<cc>, subst_f<cc>): Rename to...
31664 (subst_<mode>, subst_f<VAXccnz:mode>): ... these respectively.
31665 (*cbranch<VAXint:mode>4_<VAXcc:mode>): Update for `cc' removal.
31666 (*cbranch<VAXfp:mode>4_<VAXccnz:mode>): Likewise.
31667 (*branch_<mode>, *branch_<mode>_reversed): Likewise.
31669 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
31671 * config/vax/vax.md (subst_f<cc>): Add mode to operands and
31672 `const_double_zero'.
31674 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
31676 * config/pdp11/pdp11.md (PDPfp): New mode iterator.
31677 (fcc_cc, fcc_ccnz): Use it. Add mode to `const_double_zero' and
31680 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
31682 * genemit.c (gen_exp) <CONST_DOUBLE>: Handle `const_double_zero'
31684 * read-rtl.c (rtx_reader::read_rtx_code): Handle machine mode
31685 with `const_double_zero'.
31686 * doc/rtl.texi (Constant Expression Types): Document it.
31688 2021-01-09 Jakub Jelinek <jakub@redhat.com>
31691 * tree-cfg.c (verify_gimple_assign_binary): Allow lhs of
31692 POINTER_DIFF_EXPR to be any integral type.
31694 2021-01-09 Jakub Jelinek <jakub@redhat.com>
31696 PR rtl-optimization/98603
31697 * function.c (instantiate_virtual_regs_in_insn): For asm goto
31698 with impossible constraints, drop all SETs, CLOBBERs, drop PARALLEL
31699 if any, set ASM_OPERANDS mode to VOIDmode and change
31700 ASM_OPERANDS_OUTPUT_CONSTRAINT and ASM_OPERANDS_OUTPUT_IDX.
31702 2021-01-09 Alexandre Oliva <oliva@gnu.org>
31705 * final.c (notice_source_line): Narrow down the condition to
31706 skip a line-0 marker.
31708 2021-01-08 Sergei Trofimovich <siarheit@google.com>
31710 * ipa-modref.c (merge_call_side_effects): Fix
31711 linebreak split by reordering two print calls.
31713 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
31715 * config/s390/vector.md (*tf_to_fprx2_0): Rename from
31716 "*mov_tf_to_fprx2_0" for consistency, fix constraint.
31717 (*tf_to_fprx2_1): Rename from "*mov_tf_to_fprx2_1" for
31718 consistency, fix constraint.
31720 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
31722 * config/s390/s390-c.c (s390_def_or_undef_macro): Accept
31723 callables instead of mask values.
31724 (struct target_flag_set_p): New predicate.
31725 (s390_cpu_cpp_builtins_internal): Define or undefine
31726 __LONG_DOUBLE_VX__ macro.
31728 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
31731 * config/i386/i386.c (x86_function_profiler): Use R10 and R11
31732 to call mcount in large model with PIC for NO_PROFILE_COUNTERS
31735 2021-01-08 Richard Biener <rguenther@suse.de>
31737 * tree-ssa-sccvn.c (pass_fre::execute): Reset the SCEV hash table.
31739 2021-01-08 Richard Biener <rguenther@suse.de>
31741 * tree-vect-slp.c (scalar_stmts_to_slp_tree_map_t): Fix.
31742 (vect_build_slp_tree): On cache hit release the matched
31743 scalar stmts vector.
31744 * tree-vect-stmts.c (vectorizable_store): Properly free
31745 vec_oprnds before possibly gathering them again.
31747 2021-01-08 Richard Biener <rguenther@suse.de>
31749 PR tree-optimization/98544
31750 * tree-vect-slp.c (vect_optimize_slp): Always materialize
31751 permutes at a permute node.
31753 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
31756 * config/i386/i386.c (x86_function_profiler): Use R10 to call
31757 mcount in large model. Sorry for large model with PIC.
31759 2021-01-08 Jakub Jelinek <jakub@redhat.com>
31762 * config/i386/i386.opt (ix86_cmodel, ix86_incoming_stack_boundary_arg,
31763 ix86_pmode, ix86_preferred_stack_boundary_arg, ix86_regparm,
31764 ix86_veclibabi_type): Remove x_ prefix, use TargetVariable instead of
31765 TargetSave and initialize for variables with enum types.
31766 (mfentry, mstack-protector-guard-reg=, mstack-protector-guard-offset=,
31767 mstack-protector-guard-symbol=): Add Save.
31768 * config/i386/i386-options.c (ix86_function_specific_save,
31769 ix86_function_specific_restore): Don't save or restore x_ix86_cmodel,
31770 x_ix86_incoming_stack_boundary_arg, x_ix86_pmode,
31771 x_ix86_preferred_stack_boundary_arg, x_ix86_regparm,
31772 x_ix86_veclibabi_type.
31774 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
31776 * config/aarch64/aarch64-sve.md (*cnot<mode>): Extend from
31777 SVE_FULL_I to SVE_I.
31778 (*cond_cnot<mode>_2, *cond_cnot<mode>_any): Likewise.
31780 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
31782 * config/aarch64/aarch64-sve.md (*cond_uxt<mode>_2): Extend from
31783 SVE_FULL_I to SVE_I.
31784 (*cond_uxt<mode>_any): Likewise.
31786 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
31788 * config/aarch64/iterators.md (Vwhalf): New iterator.
31789 * config/aarch64/aarch64-simd.md (aarch64_<sur>adalp<mode>_3):
31791 (aarch64_<sur>adalp<mode>): ... This. Make more
31793 (<sur>sadv16qi): Adjust callsite of the above.
31794 * config/aarch64/aarch64-simd-builtins.def (sadalp, uadalp): New
31796 * config/aarch64/arm_neon.h (vpadal_s8): Reimplement using
31798 (vpadal_s16): Likewise.
31799 (vpadal_u8): Likewise.
31800 (vpadal_u16): Likewise.
31801 (vpadalq_s8): Likewise.
31802 (vpadalq_s16): Likewise.
31803 (vpadalq_s32): Likewise.
31804 (vpadalq_u8): Likewise.
31805 (vpadalq_u16): Likewise.
31806 (vpadalq_u32): Likewise.
31808 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
31810 * config/aarch64/aarch64-simd.md (aarch64_<su>abd<mode>_3):
31812 (aarch64_<su>abd<mode>): ... This.
31813 (<sur>sadv16qi): Adjust callsite of the above.
31814 * config/aarch64/aarch64-simd-builtins.def (sabd, uabd): Define
31816 * config/aarch64/arm_neon.h (vabd_s8): Reimplement using
31818 (vabd_s16): Likewise.
31819 (vabd_s32): Likewise.
31820 (vabd_u8): Likewise.
31821 (vabd_u16): Likewise.
31822 (vabd_u32): Likewise.
31823 (vabdq_s8): Likewise.
31824 (vabdq_s16): Likewise.
31825 (vabdq_s32): Likewise.
31826 (vabdq_u8): Likewise.
31827 (vabdq_u16): Likewise.
31828 (vabdq_u32): Likewise.
31830 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
31832 * config/aarch64/aarch64-simd-builtins.def (saba, uaba): Define
31834 * config/aarch64/arm_neon.h (vaba_s8): Implement using builtin.
31835 (vaba_s16): Likewise.
31836 (vaba_s32): Likewise.
31837 (vaba_u8): Likewise.
31838 (vaba_u16): Likewise.
31839 (vaba_u32): Likewise.
31840 (vabaq_s8): Likewise.
31841 (vabaq_s16): Likewise.
31842 (vabaq_s32): Likewise.
31843 (vabaq_u8): Likewise.
31844 (vabaq_u16): Likewise.
31845 (vabaq_u32): Likewise.
31847 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
31849 * config/aarch64/aarch64-simd.md (aba<mode>_3): Rename to...
31850 (aarch64_<su>aba<mode>): ... This. Handle uaba as well.
31851 Change RTL pattern to match.
31853 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
31855 * common/config/riscv/riscv-common.c (riscv_current_subset_list): New.
31856 * config/riscv/riscv-c.c (riscv-subset.h): New.
31857 (INCLUDE_STRING): Define.
31858 (riscv_cpu_cpp_builtins): Add new style architecture extension
31860 * config/riscv/riscv-subset.h (riscv_subset_list::begin): New.
31861 (riscv_subset_list::end): New.
31862 (riscv_current_subset_list): New.
31864 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
31866 * common/config/riscv/riscv-common.c (RISCV_DONT_CARE_VERSION):
31867 Move to riscv-subset.h.
31868 (struct riscv_subset_t): Ditto.
31869 (class riscv_subset_list): Ditto.
31870 * config/riscv/riscv-subset.h (RISCV_DONT_CARE_VERSION): Move
31871 from riscv-common.c.
31872 (struct riscv_subset_t): Ditto.
31873 (class riscv_subset_list): Ditto.
31874 * config/riscv/t-riscv ($(common_out_file)): Add file
31877 2021-01-07 Jakub Jelinek <jakub@redhat.com>
31880 * config/i386/i386.md (*bmi_blsi_<mode>_cmp, *bmi_blsi_<mode>_ccno):
31881 New define_insn patterns.
31883 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
31885 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_UNARY:optab><mode>)
31886 (*cond_<SVE_INT_UNARY:optab><mode>_2): Extend from SVE_FULL_I to SVE_I.
31887 (*cond_<SVE_INT_UNARY:optab><mode>_any): Likewise.
31889 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
31891 PR tree-optimization/98560
31892 * internal-fn.def (IFN_VCONDU, IFN_VCONDEQ): Use type vec_cond.
31893 * internal-fn.c (vec_cond_mask_direct): Get the data mode from
31895 (vec_cond_direct): Likewise argument 2.
31896 (vec_condu_direct, vec_condeq_direct): Delete.
31897 (expand_vect_cond_optab_fn): Rename to...
31898 (expand_vec_cond_optab_fn): ...this, replacing old macro.
31899 (expand_vec_condu_optab_fn, expand_vec_condeq_optab_fn): Delete.
31900 (expand_vect_cond_mask_optab_fn): Rename to...
31901 (expand_vec_cond_mask_optab_fn): ...this, replacing old macro.
31902 (direct_vec_cond_mask_optab_supported_p): Treat the optab as a
31904 (direct_vec_cond_optab_supported_p): Likewise.
31905 (direct_vec_condu_optab_supported_p): Delete.
31906 (direct_vec_condeq_optab_supported_p): Delete.
31907 * gimple-isel.cc: Include internal-fn.h.
31908 (gimple_expand_vec_cond_expr): Check that IFN_VCONDEQ is supported
31911 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
31913 PR tree-optimization/98560
31914 * gimple-isel.cc (gimple_expand_vec_cond_expr): If we fail to use
31915 IFN_VCOND{,U,EQ}, fall back on IFN_VCOND_MASK.
31917 2021-01-07 Uroš Bizjak <ubizjak@gmail.com>
31919 * config/i386/i386.md (insn): Merge from plusminus_insn, shift_insn,
31920 rotate_insn and optab code attributes.
31921 Update all uses to merged code attribute.
31922 * config/i386/sse.md: Update all uses to merged code attribute.
31923 * config/i386/mmx.md: Update all uses to merged code attribute.
31925 2021-01-07 Jakub Jelinek <jakub@redhat.com>
31927 PR tree-optimization/98568
31928 * gimple-ssa-store-merging.c (bswap_view_convert): New function.
31929 (bswap_replace): Use it.
31931 2021-01-06 Vladimir N. Makarov <vmakarov@redhat.com>
31933 PR rtl-optimization/97978
31934 * lra-int.h (lra_hard_reg_split_p): New external.
31935 * lra.c (lra_hard_reg_split_p): New global.
31936 (lra): Set up lra_hard_reg_split_p after splitting a hard reg.
31937 * lra-assigns.c (lra_assign): Don't check allocation correctness
31938 after hard reg splitting.
31940 2021-01-06 Martin Sebor <msebor@redhat.com>
31943 * builtins.c (new_delete_mismatch_p): New overload.
31944 (new_delete_mismatch_p (tree, tree)): Call it.
31946 2021-01-06 Alexandre Oliva <oliva@adacore.com>
31948 * Makefile.in (T_GLIMITS_H): New.
31949 (stmp-int-hdrs): Depend on it, use it.
31950 * config/t-vxworks (T_GLIMITS_H): Override it.
31951 (vxw-glimits.h): New.
31953 2021-01-06 Richard Biener <rguenther@suse.de>
31955 PR tree-optimization/98513
31956 * value-range.cc (intersect_ranges): Compare the upper bounds
31957 for the expected relation.
31959 2021-01-06 Gerald Pfeifer <gerald@pfeifer.com>
31962 2020-12-28 Gerald Pfeifer <gerald@pfeifer.com>
31964 * doc/standards.texi (HSAIL): Remove section.
31966 2021-01-05 Samuel Thibault <samuel.thibault@ens-lyon.org>
31968 * configure: Re-generate.
31970 2021-01-05 Jakub Jelinek <jakub@redhat.com>
31972 * doc/invoke.texi (-std=c++20): Adjust for the publication of
31973 ISO 14882:2020 standard.
31974 * doc/standards.texi: Likewise.
31976 2021-01-05 Jakub Jelinek <jakub@redhat.com>
31978 PR tree-optimization/94802
31979 * expr.h (maybe_optimize_sub_cmp_0): Declare.
31980 * expr.c: Include tree-pretty-print.h and flags.h.
31981 (maybe_optimize_sub_cmp_0): New function.
31982 (do_store_flag): Use it.
31983 * cfgexpand.c (expand_gimple_cond): Likewise.
31985 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
31987 * mux-utils.h (pointer_mux::m_ptr): Tweak description of contents.
31988 * rtlanal.c (simple_regno_set): Tweak description to clarify the
31991 2021-01-05 Richard Biener <rguenther@suse.de>
31993 PR tree-optimization/98516
31994 * tree-vect-slp.c (vect_optimize_slp): Permute the incoming
31995 lanes when materializing on a VEC_PERM node.
31996 (vectorizable_slp_permutation): Dump the permute properly.
31998 2021-01-05 Richard Biener <rguenther@suse.de>
32000 * tree-vect-slp.c (vect_slp_region): Move debug counter
32001 to cover individual subgraphs.
32003 2021-01-05 Richard Biener <rguenther@suse.de>
32005 PR tree-optimization/98428
32006 * tree-vect-slp.c (vect_build_slp_tree_1): Properly reject
32007 vector lane extracts for loop vectorization.
32009 2021-01-05 Jakub Jelinek <jakub@redhat.com>
32011 PR tree-optimization/98514
32012 * tree-ssa-reassoc.c (bb_rank): Change type from long * to
32014 (operand_rank): Change type from hash_map<tree, long> to
32015 hash_map<tree, int64_t>.
32016 (phi_rank): Change return type from long to int64_t.
32017 (loop_carried_phi): Change block_rank variable type from long to
32019 (propagate_rank): Change return type, rank parameter type and
32020 op_rank variable type from long to int64_t.
32021 (find_operand_rank): Change return type from long to int64_t
32022 and change slot variable type from long * to int64_t *.
32023 (insert_operand_rank): Change rank parameter type from long to
32025 (get_rank): Change return type and rank variable type from long to
32026 int64_t. Use PRId64 instead of ld to print the rank.
32027 (init_reassoc): Change rank variable type from long to int64_t
32028 and adjust correspondingly bb_rank and operand_rank initialization.
32030 2021-01-05 Jakub Jelinek <jakub@redhat.com>
32032 PR tree-optimization/96928
32033 * tree-ssa-phiopt.c (xor_replacement): New function.
32034 (tree_ssa_phiopt_worker): Call it.
32036 2021-01-05 Jakub Jelinek <jakub@redhat.com>
32038 PR tree-optimization/96930
32039 * match.pd ((A / (1 << B)) -> (A >> B)): If A is extended
32040 from narrower value which has the same type as 1 << B, perform
32041 the right shift on the narrower value followed by extension.
32043 2021-01-05 Jakub Jelinek <jakub@redhat.com>
32045 PR tree-optimization/96239
32046 * gimple-ssa-store-merging.c (maybe_optimize_vector_constructor): New
32048 (get_status_for_store_merging): Don't return BB_INVALID for blocks
32049 with potential bswap optimizable CONSTRUCTORs.
32050 (pass_store_merging::execute): Optimize vector CONSTRUCTORs with bswap
32053 2021-01-05 Richard Biener <rguenther@suse.de>
32055 PR tree-optimization/98381
32056 * tree.c (vector_element_bits): Properly compute bool vector
32058 * tree-vect-loop.c (vectorizable_live_operation): Properly
32059 compute the last lane bit offset.
32061 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
32064 * config/i386/sse.md (sse_cvtps2pi): Redefine as define_insn_and_split.
32065 Clear the top 64 bytes of the input XMM register.
32066 (sse_cvttps2pi): Ditto.
32068 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
32071 * config/i386/xopintrin.h (_mm256_cmov_si256): New.
32073 2021-01-05 H.J. Lu <hjl.tools@gmail.com>
32076 * config/i386/xmmintrin.h (_mm_extract_pi16): Cast to unsigned
32079 2021-01-05 Claudiu Zissulescu <claziss@synopsys.com>
32081 * config/arc/arc.md (maddsidi4_split): Use ACC_REG_FIRST.
32082 (umaddsidi4_split): Likewise.
32084 2021-01-05 liuhongt <hongtao.liu@intel.com>
32087 * config/i386/sse.md (*sse2_pmovskb_zexthisi): New
32088 define_insn_and_split for zero_extend of subreg HI of pmovskb
32090 (*sse2_pmovskb_zexthisi): Add new combine splitters for
32091 zero_extend of not of subreg HI of pmovskb result.
32093 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
32096 * explow.c (convert_memory_address_addr_space_1): Handle UNSPECs
32098 * config/aarch64/aarch64.c (aarch64_expand_mov_immediate): Use
32099 convert_memory_address to convert symbolic immediates to ptr_mode
32100 before forcing them to memory.
32102 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
32104 PR rtl-optimization/97144
32105 * recog.c (constrain_operands): Initialize matching_operand
32106 for each alternative, rather than only doing it once.
32108 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
32110 PR rtl-optimization/98403
32111 * rtl-ssa/changes.cc (function_info::finalize_new_accesses): Explain
32112 why we don't remove call clobbers.
32113 (function_info::apply_changes_to_insn): Don't attempt to add
32114 call clobbers here.
32116 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
32118 PR tree-optimization/98371
32119 * tree-vect-loop.c (vect_reanalyze_as_main_loop): New function.
32120 (vect_analyze_loop): If an epilogue loop appears to be cheaper
32121 than the main loop, re-analyze it as a main loop before adopting
32124 2021-01-05 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
32127 * configure.ac (NETLIBS): Determine using AX_LIB_SOCKET_NSL.
32128 * aclocal.m4, configure: Regenerate.
32129 * Makefile.in (NETLIBS): Define.
32130 (BACKEND): Remove $(CODYLIB).
32132 2021-01-05 Jakub Jelinek <jakub@redhat.com>
32134 PR rtl-optimization/98334
32135 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
32136 Optimize (X - 1) * Y + Y to X * Y or (X + 1) * Y - Y to X * Y.
32138 2021-01-05 Bernd Edlinger <bernd.edlinger@hotmail.de>
32140 * tree-inline.c (expand_call_inline): Restore input_location.
32141 Return result from recursive call.
32143 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
32145 PR tree-optimization/95401
32146 * config/aarch64/aarch64-sve-builtins.cc
32147 (gimple_folder::load_store_cookie): Use bits rather than bytes
32148 for the alignment argument to IFN_MASK_LOAD and IFN_MASK_STORE.
32149 * gimple-fold.c (gimple_fold_mask_load_store_mem_ref): Likewise.
32150 * tree-vect-stmts.c (vectorizable_store): Likewise.
32151 (vectorizable_load): Likewise.
32153 2021-01-04 Richard Biener <rguenther@suse.de>
32155 PR tree-optimization/98308
32156 * tree-vect-stmts.c (vectorizable_load): Set invariant mask
32159 2021-01-04 Jakub Jelinek <jakub@redhat.com>
32161 PR tree-optimization/95771
32162 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Handle types
32163 with precision smaller than int's precision and types with precision
32164 twice as large as long long. Formatting fixes.
32166 2021-01-04 Richard Biener <rguenther@suse.de>
32168 PR tree-optimization/98464
32169 * tree-ssa-sccvn.c (vn_valueize_for_srt): Rename from ...
32170 (vn_valueize_wrapper): ... this. Temporarily adjust vn_context_bb.
32171 (process_bb): Adjust.
32173 2021-01-04 Matthew Malcomson <matthew.malcomson@arm.com>
32176 * doc/invoke.texi (-fsanitize=address): Fix wording describing
32177 clash with -fsanitize=hwaddress.
32179 2021-01-04 Richard Biener <rguenther@suse.de>
32181 PR tree-optimization/98282
32182 * tree-ssa-sccvn.c (vn_get_stmt_kind): Classify tcc_reference on
32183 invariants as VN_NARY.
32185 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
32188 * config/aarch64/aarch64-simd.md (aarch64_combine<mode>): Accept
32189 aarch64_simd_reg_or_zero for operand 2. Use the combinez patterns
32190 to handle zero operands.
32192 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
32194 * config/aarch64/aarch64.c (offset_6bit_signed_scaled_p): New function.
32195 (offset_6bit_unsigned_scaled_p): Fix typo in comment.
32196 (aarch64_sve_prefetch_operand_p): Accept MUL VLs in the range
32199 2021-01-04 Richard Biener <rguenther@suse.de>
32201 PR tree-optimization/98393
32202 * tree-vect-slp.c (vect_build_slp_tree): Properly zero matches
32203 when hitting the limit.
32205 2021-01-04 Richard Biener <rguenther@suse.de>
32207 PR tree-optimization/98291
32208 * tree-vect-loop.c (vectorizable_reduction): Bypass
32209 associativity check for SLP reductions with VF 1.
32211 2021-01-04 Jakub Jelinek <jakub@redhat.com>
32213 PR tree-optimization/96782
32214 * match.pd (x == ~x -> false, x != ~x -> true): New simplifications.
32216 2021-01-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
32218 * collect-utils.c (collect_execute): Check dumppfx.
32219 * collect2.c (maybe_run_lto_and_relink, do_link): Pass atsuffix
32220 to collect_execute.
32221 (do_link): Add new parameter atsuffix.
32222 (main): Handle -dumpdir option. Skip one argument for
32223 -o, -isystem and -B options.
32224 * gcc.c (make_at_file): New helper function.
32225 (close_at_file): Use it.
32227 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
32229 * config/darwin.h (MIN_LD64_NO_COAL_SECTS): Adjust.
32230 Amend handling for LD64_VERSION fallback defaults.
32232 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
32234 * config.gcc: Compute default version information
32235 from the configured target. Likewise defaults for
32237 * config/darwin10.h: Removed.
32238 * config/darwin12.h: Removed.
32239 * config/darwin9.h: Removed.
32240 * config/rs6000/darwin8.h: Removed.
32242 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
32244 * config/darwin9.h (ASM_OUTPUT_ALIGNED_COMMON): Delete.
32246 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
32248 * config/darwin9.h (STACK_CHECK_STATIC_BUILTIN): Move from here..
32249 * config/darwin.h (STACK_CHECK_STATIC_BUILTIN): .. to here.
32251 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
32253 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move from
32255 * config/darwin.h (LINK_GCC_C_SEQUENCE_SPEC): ... to here.
32257 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
32259 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move the spec
32260 for the Darwin10 unwinder stub from here ...
32261 * config/darwin.h (LINK_COMMAND_SPEC_A): ... to here.
32263 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
32265 * config/darwin.h (DSYMUTIL_SPEC): Default to DWARF
32266 (ASM_DEBUG_SPEC):Only define if the assembler supports
32268 (PREFERRED_DEBUGGING_TYPE): Default to DWARF.
32269 (DARWIN_PREFER_DWARF): Define.
32270 * config/darwin9.h (PREFERRED_DEBUGGING_TYPE): Remove.
32271 (DARWIN_PREFER_DWARF): Likewise
32272 (DSYMUTIL_SPEC): Likewise.
32273 (COLLECT_RUN_DSYMUTIL): Likewise.
32274 (ASM_DEBUG_SPEC): Likewise.
32275 (ASM_DEBUG_OPTION_SPEC): Likewise.
32277 2021-01-02 Jan Hubicka <jh@suse.cz>
32279 * cfg.c (free_block): ggc_free bb.
32281 2021-01-01 Jakub Jelinek <jakub@redhat.com>
32283 * gcc.c (process_command): Update copyright notice dates.
32284 * gcov-dump.c (print_version): Ditto.
32285 * gcov.c (print_version): Ditto.
32286 * gcov-tool.c (print_version): Ditto.
32287 * gengtype.c (create_file): Ditto.
32288 * doc/cpp.texi: Bump @copying's copyright year.
32289 * doc/cppinternals.texi: Ditto.
32290 * doc/gcc.texi: Ditto.
32291 * doc/gccint.texi: Ditto.
32292 * doc/gcov.texi: Ditto.
32293 * doc/install.texi: Ditto.
32294 * doc/invoke.texi: Ditto.
32296 2021-01-01 Jakub Jelinek <jakub@redhat.com>
32298 * ChangeLog-2020: Rotate ChangeLog. New file.
32301 Copyright (C) 2021 Free Software Foundation, Inc.
32303 Copying and distribution of this file, with or without modification,
32304 are permitted in any medium without royalty provided the copyright
32305 notice and this notice are preserved.