1 2021-10-02 Aldy Hernandez <aldyh@redhat.com>
3 PR tree-optimization/102563
4 * range-op.cc (operator_lshift::op1_range): Do not clobber
7 2021-10-02 Martin Liska <mliska@suse.cz>
9 * toplev.c (toplev::main): save_decoded_options[0] is program
10 name and so it should be skipped.
12 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
14 PR tree-optimization/102546
15 * range-op.cc (operator_lshift::op1_range): Teach range-ops that
16 X << Y is non-zero implies X is also non-zero.
18 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
20 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
22 * config/aarch64/aarch64-tune.md: Regenerate.
23 * doc/invoke.texi: Update docs.
25 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
27 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
29 * config/aarch64/aarch64-tune.md: Regenerate.
30 * doc/invoke.texi: Update docs.
32 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
34 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
36 * config/aarch64/aarch64-tune.md: Regenerate.
37 * doc/invoke.texi: Update docs.
39 2021-10-01 Martin Sebor <msebor@redhat.com>
42 * doc/invoke.texi (-Waddress): Update.
43 * gengtype.c (write_types): Avoid -Waddress.
44 * poly-int.h (POLY_SET_COEFF): Avoid using null.
46 2021-10-01 John David Anglin <danglin@gcc.gnu.org>
49 * config/pa/pa.c (pa_option_override): Default to dwarf version 4
52 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
54 * config/aarch64/aarch64.h (AARCH64_FL_V9): Update value.
56 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
58 * gimple-range-path.cc (path_range_query::compute_ranges): Use
60 * gimple-range-path.h (class path_range_query): Remove shadowed
62 (path_range_query::get_path_oracle): New.
64 2021-10-01 Jakub Jelinek <jakub@redhat.com>
65 Richard Biener <rguenther@suse.de>
68 * doc/invoke.texi (-fsanitize=integer-divide-by-zero): Remove
69 INT_MIN / -1 division detection from here ...
70 (-fsanitize=signed-integer-overflow): ... and add it here.
72 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
74 * config/aarch64/aarch64-arches.def (AARCH64_ARCH): Added
76 * config/aarch64/aarch64.h (AARCH64_FL_V9): New.
77 (AARCH64_FL_FOR_ARCH9): New flags for Armv9-A.
78 (AARCH64_ISA_V9): New ISA flag.
79 * doc/invoke.texi: Update docs.
81 2021-10-01 Martin Liska <mliska@suse.cz>
83 * toplev.c (toplev::main): Save decoded optimization options.
84 * toplev.h (save_opt_decoded_options): New.
85 * doc/extend.texi: Be more clear about optimize and target
88 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
90 * explow.c: Include langhooks.h.
91 (set_stack_check_libfunc): Build a proper function type.
93 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
96 * config/i386/i386.c (legitimate_pic_address_disp_p): For PE-COFF do
97 not return true for external weak function symbols in medium model.
99 2021-10-01 Jakub Jelinek <jakub@redhat.com>
101 * tree.h (OMP_CLAUSE_ORDER_REPRODUCIBLE): Define.
102 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_ORDER>: Print
103 reproducible: for OMP_CLAUSE_ORDER_REPRODUCIBLE.
104 * omp-general.c (omp_extract_for_data): If OMP_CLAUSE_ORDER is seen
105 without OMP_CLAUSE_ORDER_UNCONSTRAINED, overwrite sched_kind to
106 OMP_CLAUSE_SCHEDULE_STATIC.
108 2021-10-01 Richard Biener <rguenther@suse.de>
111 * tree-inline.c (setup_one_parameter): Avoid substituting
112 an invariant into contexts where a GIMPLE register is not valid.
114 2021-09-30 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
116 * config/arm/arm-cpus.in: Add Cortex-R52+ CPU.
117 * config/arm/arm-tables.opt: Regenerate.
118 * config/arm/arm-tune.md: Regenerate.
119 * doc/invoke.texi: Update docs.
121 2021-09-30 Uroš Bizjak <ubizjak@gmail.com>
124 * config/i386/i386.md
125 (sign_extend:WIDE (any_logic:NARROW (memory, immediate)) splitters):
128 2021-09-30 Tobias Burnus <tobias@codesourcery.com>
130 * omp-low.c (omp_runtime_api_call): Add omp_aligned_{,c}alloc and
131 omp_{c,re}alloc, fix omp_alloc/omp_free.
133 2021-09-30 Martin Liska <mliska@suse.cz>
135 * defaults.h (ASM_OUTPUT_ASCII): Do not hide global variable
136 asm_out_file and stream directly to MYFILE.
138 2021-09-30 Richard Biener <rguenther@suse.de>
140 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
141 Restore and fix condition under which we apply npeel to
142 the DRs misalignment value.
144 2021-09-30 Richard Biener <rguenther@suse.de>
146 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
147 Fix npeel check for variable amount of peeling.
149 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
151 * lto-wrapper.c (run_gcc): Plug snprintf overflow.
153 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
155 * gimple-range.cc (gimple_ranger::debug): New.
156 * gimple-range.h (class gimple_ranger): Add debug.
158 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
161 * tree-vrp.c (hybrid_threader::~hybrid_threader): Free m_query.
163 2021-09-29 Indu Bhagat <indu.bhagat@oracle.com>
166 * btfout.c (GTY): Add GTY (()) albeit for cosmetic only purpose.
167 (btf_finalize): Empty the hash_map btf_var_ids.
169 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
171 * tree-vrp.c (thread_through_all_blocks): Return bool.
172 (execute_vrp_threader): Return TODO_* flags.
173 (pass_data_vrp_threader): Set todo_flags_finish to 0.
175 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
177 * timevar.def (TV_TREE_VRP_THREADER): New.
178 * tree-vrp.c: Use TV_TREE_VRP_THREADER for VRP threader pass.
180 2021-09-29 David Faust <david.faust@oracle.com>
182 * config.gcc (bpf-*-*): Do not overwrite extra_headers.
184 2021-09-29 Jonathan Wright <jonathan.wright@arm.com>
186 * config/aarch64/aarch64-builtins.c (TYPES_BINOP_PPU): Define
187 new type qualifier enum.
188 (TYPES_TERNOP_SSSU): Likewise.
189 (TYPES_TERNOP_PPPU): Likewise.
190 * config/aarch64/aarch64-simd-builtins.def: Define PPU, SSU,
191 PPPU and SSSU builtin generator macros for qtbl1 and qtbx1
193 * config/aarch64/arm_neon.h (vqtbl1_p8): Use type-qualified
194 builtin and remove casts.
195 (vqtbl1_s8): Likewise.
196 (vqtbl1q_p8): Likewise.
197 (vqtbl1q_s8): Likewise.
198 (vqtbx1_s8): Likewise.
199 (vqtbx1_p8): Likewise.
200 (vqtbx1q_s8): Likewise.
201 (vqtbx1q_p8): Likewise.
202 (vtbl1_p8): Likewise.
203 (vtbl2_p8): Likewise.
204 (vtbx2_p8): Likewise.
206 2021-09-29 Richard Biener <rguenther@suse.de>
208 * tree-vect-data-refs.c (vect_dr_misalign_for_aligned_access):
210 (vect_update_misalignment_for_peel): Use it to update
211 misaligned to the value necessary for an aligned access.
212 (vect_get_peeling_costs_all_drs): Likewise.
213 (vect_enhance_data_refs_alignment): Likewise.
215 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
217 * config/aarch64/aarch64.c (aarch64_expand_cpymem): Count number of
218 emitted operations and adjust heuristic for code size.
220 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
222 * config/aarch64/aarch64.c (aarch64_expand_setmem): Count number of
223 emitted operations and adjust heuristic for code size.
225 2021-09-29 Jakub Jelinek <jakub@redhat.com>
228 * gimplify.c (gimplify_scan_omp_clauses): Use omp_check_private even
229 in OMP_SCOPE clauses, not just on worksharing construct clauses.
231 2021-09-28 Geng Qi <gengqi@linux.alibaba.com>
233 * config/riscv/riscv.md (mulv<mode>4): Call gen_smul<mode>3_highpart.
234 (<u>mulditi3): Call <su>muldi3_highpart.
235 (<u>muldi3_highpart): Rename to <su>muldi3_highpart.
236 (<u>mulsidi3): Call <su>mulsi3_highpart.
237 (<u>mulsi3_highpart): Rename to <su>mulsi3_highpart.
239 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
241 * config/darwin.h (DSYMUTIL_SPEC): Recognize D sources.
243 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
245 * config/rs6000/darwin.h (FIXED_R13): Add for PPC64.
246 (FIRST_SAVED_GP_REGNO): Save from R13 even when it is one
249 2021-09-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
251 * config/aarch64/aarch64.h (AARCH64_FL_LS64): Define
252 (AARCH64_FL_V8_7): Likewise.
253 (AARCH64_FL_FOR_ARCH8_7): Likewise.
254 * config/aarch64/aarch64-arches.def (armv8.7-a): Define.
255 * config/aarch64/aarch64-option-extensions.def (ls64): Define.
256 * doc/invoke.texi: Document the above.
258 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
260 * dbgcnt.c (dbg_cnt_counter): New.
261 * dbgcnt.h (dbg_cnt_counter): New.
262 * dumpfile.c (dump_options): Add entry for TDF_THREADING.
263 * dumpfile.h (enum dump_flag): Add TDF_THREADING.
264 * gimple-range-path.cc (DEBUG_SOLVER): Use TDF_THREADING.
265 * tree-ssa-threadupdate.c (dump_jump_thread_path): Dump out
268 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
270 * cfgcleanup.c (pass_jump::execute): Check
271 flag_expensive_optimizations.
272 (pass_jump_after_combine::gate): Same.
273 * doc/invoke.texi (-fthread-jumps): Enable for -O1.
274 * opts.c (default_options_table): Enable -fthread-jumps at -O1.
275 * tree-ssa-threadupdate.c
276 (fwd_jt_path_registry::remove_jump_threads_including): Bail unless
279 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
281 * tree-ssa-reassoc.c (biased_names): New global.
282 (propagate_bias_p): New function.
283 (loop_carried_phi): Remove.
284 (propagate_rank): Propagate bias along single uses.
285 (get_rank): Update biased_names when needed.
287 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
289 * passes.def (pass_reassoc): Rename parameter to early_p.
290 * tree-ssa-reassoc.c (reassoc_bias_loop_carried_phi_ranks_p):
292 (phi_rank): Don't bias loop-carried phi ranks
293 before vectorization pass.
294 (execute_reassoc): Add bias_loop_carried_phi_ranks_p parameter.
295 (pass_reassoc::pass_reassoc): Add bias_loop_carried_phi_ranks_p
297 (pass_reassoc::set_param): Set bias_loop_carried_phi_ranks_p
299 (pass_reassoc::execute): Pass bias_loop_carried_phi_ranks_p to
301 (pass_reassoc::bias_loop_carried_phi_ranks_p): New member.
303 2021-09-28 Jakub Jelinek <jakub@redhat.com>
306 * config/i386/i386.c (standard_80387_constant_p): Don't recognize
307 special 80387 instruction XFmode constants if flag_rounding_math.
309 2021-09-28 Richard Biener <rguenther@suse.de>
311 PR tree-optimization/100112
312 * tree-ssa-sccvn.c (visit_reference_op_load): Record the
313 referece into the hashtable twice in case last_vuse is
314 different from the original vuse on the stmt.
316 2021-09-28 Jakub Jelinek <jakub@redhat.com>
319 * gimplify.c (gimplify_adjust_omp_clauses_1): Don't call the
320 omp_finish_clause langhook on implicitly added OMP_CLAUSE_PRIVATE
321 clauses on SIMD constructs.
323 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
325 PR tree-optimization/102511
326 * gimple-range-path.cc (path_range_query::range_on_path_entry):
327 Return VARYING when nothing found.
329 2021-09-28 Hongyu Wang <hongyu.wang@intel.com>
332 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): Add
334 (VALID_SSE2_REG_VHF_MODE): Add V4HFmode and V2HFmode.
335 (VALID_MMX_REG_MODE): Add V4HFmode.
336 (SSE_REG_MODE_P): Replace VALID_AVX512FP16_REG_MODE with
337 vector mode condition.
338 * config/i386/i386.c (classify_argument): Parse V4HF/V2HF
340 (function_arg_32): Add V4HFmode.
341 (function_arg_advance_32): Likewise.
342 * config/i386/i386.md (mode): Add V4HF/V2HF.
343 (MODE_SIZE): Likewise.
344 * config/i386/mmx.md (MMXMODE): Add V4HF mode.
345 (V_32): Add V2HF mode.
346 (VHF_32_64): New mode iterator.
347 (*mov<mode>_internal): Adjust sse alternatives to support
349 (*mov<mode>_internal): Adjust sse alternatives to support
351 (<insn><mode>3): New define_insn for add/sub/mul/div.
353 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
355 * tree-ssa-threadbackward.c (pass_thread_jumps::gate): Check
357 (pass_early_thread_jumps::gate): Same.
358 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
359 Return if !flag_thread_jumps.
360 * tree-ssa-threadupdate.c
361 (jt_path_registry::register_jump_thread): Assert that
362 flag_thread_jumps is true.
364 2021-09-28 liuhongt <hongtao.liu@intel.com>
367 (simplify_context::simplify_binary_operation_1): Relax
368 condition of simplifying (vec_concat:M (vec_select op0
369 index0)(vec_select op1 index1)) to allow different modes
370 between op0 and M, but have same inner mode.
372 2021-09-28 liuhongt <hongtao.liu@intel.com>
374 * config/i386/i386-expand.c (emit_reduc_half): Handle
375 V8HF/V16HF/V32HFmode.
376 * config/i386/sse.md (REDUC_SSE_PLUS_MODE): Add V8HF.
377 (REDUC_SSE_SMINMAX_MODE): Ditto.
378 (REDUC_PLUS_MODE): Add V16HF and V32HF.
379 (REDUC_SMINMAX_MODE): Ditto.
381 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
383 * gimple-range-path.cc
384 (path_range_query::precompute_ranges_in_block): Rename to...
385 (path_range_query::compute_ranges_in_block): ...this.
386 (path_range_query::precompute_ranges): Rename to...
387 (path_range_query::compute_ranges): ...this.
388 (path_range_query::precompute_relations): Rename to...
389 (path_range_query::compute_relations): ...this.
390 (path_range_query::precompute_phi_relations): Rename to...
391 (path_range_query::compute_phi_relations): ...this.
392 * gimple-range-path.h: Rename precompute* to compute*.
393 * tree-ssa-threadbackward.c
394 (back_threader::find_taken_edge_switch): Same.
395 (back_threader::find_taken_edge_cond): Same.
396 * tree-ssa-threadedge.c
397 (hybrid_jt_simplifier::compute_ranges_from_state): Same.
398 (hybrid_jt_state::register_equivs_stmt): Inline...
399 * tree-ssa-threadedge.h: ...here.
401 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
403 * tree-vrp.c (lhs_of_dominating_assert): Remove.
404 (class vrp_jt_state): Remove.
405 (class vrp_jt_simplifier): Remove.
406 (vrp_jt_simplifier::simplify): Remove.
407 (class vrp_jump_threader): Remove.
408 (vrp_jump_threader::vrp_jump_threader): Remove.
409 (vrp_jump_threader::~vrp_jump_threader): Remove.
410 (vrp_jump_threader::before_dom_children): Remove.
411 (vrp_jump_threader::after_dom_children): Remove.
413 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
415 * passes.def (pass_vrp_threader): New.
416 * tree-pass.h (make_pass_vrp_threader): Add make_pass_vrp_threader.
417 * tree-ssa-threadedge.c (hybrid_jt_state::register_equivs_stmt): New.
418 (hybrid_jt_simplifier::hybrid_jt_simplifier): New.
419 (hybrid_jt_simplifier::simplify): New.
420 (hybrid_jt_simplifier::compute_ranges_from_state): New.
421 * tree-ssa-threadedge.h (class hybrid_jt_state): New.
422 (class hybrid_jt_simplifier): New.
423 * tree-vrp.c (execute_vrp): Remove ASSERT_EXPR based jump
425 (class hybrid_threader): New.
426 (hybrid_threader::hybrid_threader): New.
427 (hybrid_threader::~hybrid_threader): New.
428 (hybrid_threader::before_dom_children): New.
429 (hybrid_threader::after_dom_children): New.
430 (execute_vrp_threader): New.
431 (class pass_vrp_threader): New.
432 (make_pass_vrp_threader): New.
434 2021-09-27 Martin Liska <mliska@suse.cz>
436 * output.h (enum section_flag): New.
437 (SECTION_FORGET): Remove.
438 (SECTION_ENTSIZE): Make it (1UL << 8) - 1.
439 (SECTION_STYLE_MASK): Define it based on other enum
441 * varasm.c (switch_to_section): Remove unused handling of
444 2021-09-27 Martin Liska <mliska@suse.cz>
446 * common.opt: Add new variable flag_default_complex_method.
447 * opts.c (finish_options): Handle flags related to
448 x_flag_complex_method.
449 * toplev.c (process_options): Remove option handling related
450 to flag_complex_method.
452 2021-09-27 Richard Biener <rguenther@suse.de>
455 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid using
456 type_for_size, instead use int_mode_for_size.
458 2021-09-27 Andrew Pinski <apinski@marvell.com>
461 * gimplify.c (gimplify_save_expr): Return early
462 if the type of val is error_mark_node.
464 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
466 * tree-ssanames.c (ssa_name_has_boolean_range): Use
469 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
471 * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Remove
473 * tree-ssa-dom.c (cprop_operand): Convert to range_query API.
474 (cprop_into_stmt): Same.
475 (dom_opt_dom_walker::optimize_stmt): Same.
477 2021-09-27 Richard Biener <rguenther@suse.de>
479 PR tree-optimization/97351
480 PR tree-optimization/97352
481 PR tree-optimization/82426
482 * tree-vectorizer.h (dr_misalignment): Add vector type
484 (aligned_access_p): Likewise.
485 (known_alignment_for_access_p): Likewise.
486 (vect_supportable_dr_alignment): Likewise.
487 (vect_known_alignment_in_bytes): Likewise. Refactor.
488 (DR_MISALIGNMENT): Remove.
489 (vect_update_shared_vectype): Likewise.
490 * tree-vect-data-refs.c (dr_misalignment): Refactor, handle
491 a vector type with larger alignment requirement and apply
492 the negative step adjustment here.
493 (vect_calculate_target_alignment): Remove.
494 (vect_compute_data_ref_alignment): Get explicit vector type
495 argument, do not apply a negative step alignment adjustment
497 (vect_slp_analyze_node_alignment): Re-analyze alignment
498 when we re-visit the DR with a bigger desired alignment but
499 keep more precise results from smaller alignments.
500 * tree-vect-slp.c (vect_update_shared_vectype): Remove.
501 (vect_slp_analyze_node_operations_1): Do not update the
502 shared vector type on stmts.
503 * tree-vect-stmts.c (vect_analyze_stmt): Push/pop the
504 vector type of an SLP node to the representative stmt-info.
505 (vect_transform_stmt): Likewise.
507 2021-09-27 liuhongt <hongtao.liu@intel.com>
510 2021-09-09 liuhongt <hongtao.liu@intel.com>
513 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
514 (reduc_plus_scal_v4sf): .. this, New define_expand.
515 (reduc_plus_scal_v2df): .. and this, New define_expand.
517 2021-09-26 liuhongt <hongtao.liu@intel.com>
519 * doc/extend.texi (Half-Precision): Remove storage only
520 description for _Float16 w/o avx512fp16.
522 2021-09-25 Dimitar Dimitrov <dimitar@dinux.eu>
524 * config/pru/constraints.md (Rrio): New constraint.
525 * config/pru/predicates.md (regio_operand): New predicate.
526 * config/pru/pru-pragma.c (pru_register_pragmas): Register
527 the __regio_symbol address space.
528 * config/pru/pru-protos.h (pru_symref2ioregno): Declaration.
529 * config/pru/pru.c (pru_symref2ioregno): New helper function.
530 (pru_legitimate_address_p): Remove.
531 (pru_addr_space_legitimate_address_p): Use the address space
533 (pru_nongeneric_pointer_addrspace): New helper function.
534 (pru_insert_attributes): New function to validate __regio_symbol
536 (TARGET_INSERT_ATTRIBUTES): New macro.
537 (TARGET_LEGITIMATE_ADDRESS_P): Remove.
538 (TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P): New macro.
539 * config/pru/pru.h (enum reg_class): Add REGIO_REGS class.
540 * config/pru/pru.md (*regio_readsi): New pattern to read I/O
542 (*regio_nozext_writesi): New pattern to write to I/O registers.
543 (*regio_zext_write_r30<EQS0:mode>): Ditto.
544 * doc/extend.texi: Document the new PRU Named Address Space.
546 2021-09-24 Patrick Palka <ppalka@redhat.com>
550 * real.c (encode_ieee_double): Avoid unwanted sign extension.
551 (encode_ieee_quad): Likewise.
553 2021-09-24 Vladimir Makarov <vmakarov@redhat.com>
555 PR rtl-optimization/102147
556 * ira-build.c (ira_conflict_vector_profitable_p): Make
557 profitability calculation independent of host compiler pointer and
560 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
562 * gimple-range-path.cc (path_range_query::path_range_query):
563 Move debugging header...
564 (path_range_query::precompute_ranges): ...here.
565 (path_range_query::internal_range_of_expr): Do not call
566 range_on_path_entry if NAME is defined in the current block.
568 2021-09-24 Richard Biener <rguenther@suse.de>
570 * cfghooks.c (verify_flow_info): Verify unallocated BB and
571 edge flags are not set.
573 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
575 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
577 (jt_path_registry::register_jump_thread): Call
578 cancel_invalid_paths.
579 * tree-ssa-threadupdate.h (class jt_path_registry): Add
580 cancel_invalid_paths.
582 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
584 PR tree-optimization/102400
585 * tree-ssa-sccvn.c (vn_reference_insert_pieces): Initialize
586 result_vdef to zero value.
588 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
590 PR tree-optimization/102451
591 * tree-ssa-dse.c (delete_dead_or_redundant_call): Record bb of stmt
594 2021-09-24 Hongyu Wang <hongyu.wang@intel.com>
596 * config/i386/sse.md (cond_<insn><mode>): Extend to support
598 (cond_mul<mode>): Likewise.
599 (cond_div<mode>): Likewise.
600 (cond_<code><mode>): Likewise.
601 (cond_fma<mode>): Likewise.
602 (cond_fms<mode>): Likewise.
603 (cond_fnma<mode>): Likewise.
604 (cond_fnms<mode>): Likewise.
606 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
608 PR tree-optimization/102463
609 * gimple-range-fold.cc (fold_using_range::relation_fold_and_or): If
610 there is no range-ops handler, don't look for a relation.
612 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
614 * gimple-range-cache.cc (ranger_cache::ranger_cache): Take
615 non-executable_edge flag as parameter.
616 * gimple-range-cache.h (ranger_cache): Adjust prototype.
617 * gimple-range-gori.cc (gori_compute::gori_compute): Take
618 non-executable_edge flag as parameter.
619 (gori_compute::outgoing_edge_range_p): Check new flag.
620 * gimple-range-gori.h (gori_compute): Adjust prototype.
621 * gimple-range.cc (gimple_ranger::gimple_ranger): Create new flag.
622 (gimple_ranger::range_on_edge): Check new flag.
623 * gimple-range.h (gimple_ranger::non_executable_edge_flag): New.
624 * gimple-ssa-evrp.c (rvrp_folder): Pass ranger flag to simplifer.
625 (hybrid_folder::hybrid_folder): Set ranger non-executable flag value.
626 (hybrid_folder::fold_stmt): Set flag value in the simplifer.
627 * vr-values.c (simplify_using_ranges::set_and_propagate_unexecutable):
628 Use not_executable flag if provided inmstead of EDGE_EXECUTABLE.
629 (simplify_using_ranges::simplify_switch_using_ranges): Clear
630 EDGE_EXECUTABLE like it originally did.
631 (simplify_using_ranges::cleanup_edges_and_switches): Clear any
632 NON_EXECUTABLE flags.
633 (simplify_using_ranges::simplify_using_ranges): Adjust.
634 * vr-values.h (class simplify_using_ranges): Adjust.
635 (simplify_using_ranges::set_range_query): Add non-executable flag param.
637 2021-09-23 Bill Schmidt <wschmidt@linux.ibm.com>
640 * config/rs6000/rs6000-call.c (rs6000_aggregate_candidate): Detect
641 zero-width bit fields and return indicator.
642 (rs6000_discover_homogeneous_aggregate): Diagnose when the
643 presence of a zero-width bit field changes parameter passing in
646 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
648 * gimple-range-fold.cc (fold_using_range::range_of_phi):
649 Remove dominator check.
651 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
653 * gimple-range-path.cc (path_range_query::precompute_relations):
654 Hoist edge calculations before using EDGE_SUCC.
656 2021-09-23 Jonathan Wakely <jwakely@redhat.com>
658 * configure.ac: Fix --with-multilib-list description.
659 * configure: Regenerate.
661 2021-09-23 Richard Biener <rguenther@suse.de>
663 PR tree-optimization/102448
664 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
665 Clear alignment info copied from DR_PTR_INFO.
667 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
669 * config/i386/i386-expand.c (ix86_use_mask_cmp_p): Enable
671 * config/i386/sse.md (sseintvecmodelower): Add HF vector modes.
672 (<avx512>_store<mode>_mask): Extend to support HF vector modes.
673 (vec_cmp<mode><avx512fmaskmodelower>): Likewise.
674 (vcond_mask_<mode><avx512fmaskmodelower>): Likewise.
675 (vcond<mode><mode>): New expander.
676 (vcond<mode><sseintvecmodelower>): Likewise.
677 (vcond<sseintvecmodelower><mode>): Likewise.
678 (vcondu<mode><sseintvecmodelower>): Likewise.
680 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
682 * config/i386/sse.md (extend<ssePHmodelower><mode>2):
684 (extendv4hf<mode>2): Likewise.
685 (extendv2hfv2df2): Likewise.
686 (trunc<mode><ssePHmodelower>2): Likewise.
687 (avx512fp16_vcvt<castmode>2ph_<mode>): Rename to ...
688 (trunc<mode>v4hf2): ... this, and drop constraints.
689 (avx512fp16_vcvtpd2ph_v2df): Rename to ...
690 (truncv2dfv2hf2): ... this, and likewise.
692 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
694 * config/i386/sse.md (float<floatunssuffix><mode><ssePHmodelower>2):
696 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>):
698 (float<floatunssuffix><mode>v4hf2): ... this, and drop constraints.
699 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Rename to ...
700 (float<floatunssuffix>v2div2hf2): ... this, and likewise.
702 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
704 * config/i386/i386.md (fix<fixunssuffix>_trunchf<mode>2): New expander.
705 (fixuns_trunchfhi2): Likewise.
706 (*fixuns_trunchfsi2zext): New define_insn.
707 * config/i386/sse.md (ssePHmodelower): New mode_attr.
708 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
709 New expander for same element vector fix_truncate.
710 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
711 Likewise for V4HF to V4SI/V4DI fix_truncate.
712 (fix<fixunssuffix>_truncv2hfv2di2):
713 Likeise for V2HF to V2DI fix_truncate.
715 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
717 * config/i386/i386.md (<code>hf3): New expander.
719 2021-09-23 liuhongt <hongtao.liu@intel.com>
721 * config/i386/sse.md (FMAMODEM): extend to handle FP16.
722 (VFH_SF_AVX512VL): Extend to handle HFmode.
723 (VF_SF_AVX512VL): Deleted.
725 2021-09-23 liuhongt <hongtao.liu@intel.com>
727 * config/i386/i386.md (rinthf2): New expander.
728 (nearbyinthf2): New expander.
730 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
732 * tree-ssa-dom.c (class dom_jump_threader_simplifier): Rename...
733 (class dom_jt_state): ...this and provide virtual overrides.
734 (dom_jt_state::register_equiv): New.
735 (class dom_jt_simplifier): Rename from
736 dom_jump_threader_simplifier.
737 (dom_jump_threader_simplifier::simplify): Rename...
738 (dom_jt_simplifier::simplify): ...to this.
739 (pass_dominator::execute): Use dom_jt_simplifier and
741 * tree-ssa-threadedge.c (jump_threader::jump_threader):
743 (jt_state::register_equivs_stmt): Abstract out...
744 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
746 (jump_threader::thread_around_empty_blocks): Update state.
747 (jump_threader::thread_through_normal_block): Same.
748 (jt_state::jt_state): Remove.
749 (jt_state::push): Remove pass specific bits. Keep block vector
751 (jt_state::append_path): New.
752 (jt_state::pop): Remove pass specific bits.
753 (jt_state::register_equiv): Same.
754 (jt_state::record_ranges_from_stmt): Same.
755 (jt_state::register_equivs_on_edge): Same. Rename...
756 (jt_state::register_equivs_edge): ...to this.
757 (jt_state::dump): New.
758 (jt_state::debug): New.
759 (jump_threader_simplifier::simplify): Remove.
760 (jt_state::get_path): New.
761 * tree-ssa-threadedge.h (class jt_simplifier): Make into a base
762 class. Expose common functionality as virtual methods.
763 (class jump_threader_simplifier): Same. Rename...
764 (class jt_simplifier): ...to this.
765 * tree-vrp.c (class vrp_jump_threader_simplifier): Rename...
766 (class vrp_jt_simplifier): ...to this. Provide pass specific
768 (class vrp_jt_state): New.
769 (vrp_jump_threader_simplifier::simplify): Rename...
770 (vrp_jt_simplifier::simplify): ...to this. Inline code from
771 what used to be the base class.
772 (vrp_jump_threader::vrp_jump_threader): Use vrp_jt_state and
775 2021-09-22 Tobias Burnus <tobias@codesourcery.com>
778 * doc/invoke.texi (-Wno-missing-include-dirs.): Document Fortran
781 2021-09-22 Roger Sayle <roger@nextmovesoftware.com>
782 Richard Biener <rguenther@suse.de>
784 * match.pd (negation simplifications): Implement some negation
785 folding transformations from fold-const.c's fold_negate_expr.
786 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Add a SIMPLIFY
787 argument, to control whether the op should be simplified prior
788 to looking up/assigning a value number.
789 (vn_nary_build_or_lookup): Update call to vn_nary_build_or_lookup_1.
790 (vn_nary_simplify): Likewise.
791 (visit_nary_op): Likewise, but when constructing a NEGATE_EXPR
792 now call vn_nary_build_or_lookup_1 disabling simplification.
794 2021-09-22 Jiufu Guo <guojiufu@linux.ibm.com>
796 PR tree-optimization/102087
797 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
798 Update bound/cmp/control for niter.
800 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
802 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
803 Move check for non-empty BB here.
804 (fur_source::register_outgoing_edges): ...from here.
806 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
808 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
809 Remove call to improve_range_with_equivs.
810 (path_range_query::improve_range_with_equivs): Remove
811 * gimple-range-path.h: Remove improve_range_with_equivs.
813 2021-09-22 dianhong xu <dianhong.xu@intel.com>
815 * config/i386/avx512fp16intrin.h:
816 (_mm512_mask_blend_ph): New intrinsic.
817 (_mm512_permutex2var_ph): Ditto.
818 (_mm512_permutexvar_ph): Ditto.
819 * config/i386/avx512fp16vlintrin.h:
820 (_mm256_mask_blend_ph): New intrinsic.
821 (_mm256_permutex2var_ph): Ditto.
822 (_mm256_permutexvar_ph): Ditto.
823 (_mm_mask_blend_ph): Ditto.
824 (_mm_permutex2var_ph): Ditto.
825 (_mm_permutexvar_ph): Ditto.
827 2021-09-22 dianhong xu <dianhong.xu@intel.com>
829 * config/i386/avx512fp16intrin.h: Add new intrinsics.
830 (_mm512_conj_pch): New intrinsic.
831 (_mm512_mask_conj_pch): Ditto.
832 (_mm512_maskz_conj_pch): Ditto.
833 * config/i386/avx512fp16vlintrin.h: Add new intrinsics.
834 (_mm256_conj_pch): New intrinsic.
835 (_mm256_mask_conj_pch): Ditto.
836 (_mm256_maskz_conj_pch): Ditto.
837 (_mm_conj_pch): Ditto.
838 (_mm_mask_conj_pch): Ditto.
839 (_mm_maskz_conj_pch): Ditto.
841 2021-09-22 dianhong xu <dianhong.xu@intel.com>
843 * config/i386/avx512fp16intrin.h (_MM512_REDUCE_OP): New macro
844 (_mm512_reduce_add_ph): New intrinsic.
845 (_mm512_reduce_mul_ph): Ditto.
846 (_mm512_reduce_min_ph): Ditto.
847 (_mm512_reduce_max_ph): Ditto.
848 * config/i386/avx512fp16vlintrin.h
849 (_MM256_REDUCE_OP/_MM_REDUCE_OP): New macro.
850 (_mm256_reduce_add_ph): New intrinsic.
851 (_mm256_reduce_mul_ph): Ditto.
852 (_mm256_reduce_min_ph): Ditto.
853 (_mm256_reduce_max_ph): Ditto.
854 (_mm_reduce_add_ph): Ditto.
855 (_mm_reduce_mul_ph): Ditto.
856 (_mm_reduce_min_ph): Ditto.
857 (_mm_reduce_max_ph): Ditto.
859 2021-09-22 dianhong xu <dianhong.xu@intel.com>
861 * config/i386/avx512fp16intrin.h (__m512h_u, __m256h_u,
862 __m128h_u): New typedef.
863 (_mm512_load_ph): New intrinsic.
864 (_mm256_load_ph): Ditto.
865 (_mm_load_ph): Ditto.
866 (_mm512_loadu_ph): Ditto.
867 (_mm256_loadu_ph): Ditto.
868 (_mm_loadu_ph): Ditto.
869 (_mm512_store_ph): Ditto.
870 (_mm256_store_ph): Ditto.
871 (_mm_store_ph): Ditto.
872 (_mm512_storeu_ph): Ditto.
873 (_mm256_storeu_ph): Ditto.
874 (_mm_storeu_ph): Ditto.
875 (_mm512_abs_ph): Ditto.
876 * config/i386/avx512fp16vlintrin.h
878 (_mm256_abs_ph): Ditto.
880 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
882 * config/s390/tpf.md (prologue_tpf, epilogue_tpf): Add cc clobber.
884 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
887 * config/s390/s390.c (s390_expand_insv): Emit a normal move if it
888 is actually a full copy of the source operand into the target.
889 Don't emit a strict low part move if source and target mode match.
891 2021-09-22 Jakub Jelinek <jakub@redhat.com>
894 * omp-expand.c (expand_omp_single): If region->exit is NULL,
895 assert region->entry is GIMPLE_OMP_SCOPE region and return.
897 2021-09-22 Jakub Jelinek <jakub@redhat.com>
899 * tree.h (OMP_CLAUSE_ALLOCATE_ALIGN): Define.
900 * tree.c (omp_clause_num_ops): Change number of OMP_CLAUSE_ALLOCATE
901 arguments from 2 to 3.
902 * tree-pretty-print.c (dump_omp_clause): Print allocator() around
903 allocate clause allocator and print align if present.
904 * omp-low.c (scan_sharing_clauses): Force allocate_map entry even
905 for omp_default_mem_alloc if align modifier is present. If align
906 modifier is present, use TREE_LIST to encode both allocator and
908 (lower_private_allocate, lower_rec_input_clauses, create_task_copyfn):
909 Handle align modifier on allocator clause if present.
911 2021-09-22 liuhongt <hongtao.liu@intel.com>
913 * config/i386/i386.md (define_attr "isa"): Add
915 (define_attr "enabled"): Correspond fma_or_avx512vl to
916 TARGET_FMA || TARGET_AVX512VL.
917 * config/i386/mmx.md (fmav2sf4): Extend to AVX512 fma.
922 2021-09-22 liuhongt <hongtao.liu@intel.com>
924 * config/i386/i386.md (cstorehf3): New define_expand.
926 2021-09-22 liuhongt <hongtao.liu@intel.com>
928 * config/i386/i386.md (<rounding_insn>hf2): New expander.
929 (sse4_1_round<mode>2): Extend from MODEF to MODEFH.
930 * config/i386/sse.md (*sse4_1_round<ssescalarmodesuffix>):
931 Extend from VF_128 to VFH_128.
933 2021-09-22 liuhongt <hongtao.liu@intel.com>
935 * config/i386/i386-features.c (i386-features.c): Handle
937 * config/i386/i386.md (sqrthf2): New expander.
938 (*sqrthf2): New define_insn.
940 (*<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
943 2021-09-22 liuhongt <hongtao.liu@intel.com>
945 * config/i386/avx512fp16intrin.h (_mm_mask_fcmadd_sch):
947 (_mm_mask3_fcmadd_sch): Likewise.
948 (_mm_maskz_fcmadd_sch): Likewise.
949 (_mm_fcmadd_sch): Likewise.
950 (_mm_mask_fmadd_sch): Likewise.
951 (_mm_mask3_fmadd_sch): Likewise.
952 (_mm_maskz_fmadd_sch): Likewise.
953 (_mm_fmadd_sch): Likewise.
954 (_mm_mask_fcmadd_round_sch): Likewise.
955 (_mm_mask3_fcmadd_round_sch): Likewise.
956 (_mm_maskz_fcmadd_round_sch): Likewise.
957 (_mm_fcmadd_round_sch): Likewise.
958 (_mm_mask_fmadd_round_sch): Likewise.
959 (_mm_mask3_fmadd_round_sch): Likewise.
960 (_mm_maskz_fmadd_round_sch): Likewise.
961 (_mm_fmadd_round_sch): Likewise.
962 (_mm_fcmul_sch): Likewise.
963 (_mm_mask_fcmul_sch): Likewise.
964 (_mm_maskz_fcmul_sch): Likewise.
965 (_mm_fmul_sch): Likewise.
966 (_mm_mask_fmul_sch): Likewise.
967 (_mm_maskz_fmul_sch): Likewise.
968 (_mm_fcmul_round_sch): Likewise.
969 (_mm_mask_fcmul_round_sch): Likewise.
970 (_mm_maskz_fcmul_round_sch): Likewise.
971 (_mm_fmul_round_sch): Likewise.
972 (_mm_mask_fmul_round_sch): Likewise.
973 (_mm_maskz_fmul_round_sch): Likewise.
974 * config/i386/i386-builtin.def: Add corresponding new builtins.
976 (avx512fp16_fmaddcsh_v8hf_maskz<round_expand_name>): New expander.
977 (avx512fp16_fcmaddcsh_v8hf_maskz<round_expand_name>): Ditto.
978 (avx512fp16_fma_<complexopname>sh_v8hf<mask_scalarcz_name><round_scalarcz_name>):
980 (avx512fp16_<complexopname>sh_v8hf_mask<round_name>): Ditto.
981 (avx512fp16_<complexopname>sh_v8hf<mask_scalarc_name><round_scalarcz_name>):
983 * config/i386/subst.md (mask_scalarcz_name): New.
984 (mask_scalarc_name): Ditto.
985 (mask_scalarc_operand3): Ditto.
986 (mask_scalarcz_operand4): Ditto.
987 (round_scalarcz_name): Ditto.
988 (round_scalarc_mask_operand3): Ditto.
989 (round_scalarcz_mask_operand4): Ditto.
990 (round_scalarc_mask_op3): Ditto.
991 (round_scalarcz_mask_op4): Ditto.
992 (round_scalarcz_constraint): Ditto.
993 (round_scalarcz_nimm_predicate): Ditto.
994 (mask_scalarcz): Ditto.
995 (mask_scalarc): Ditto.
996 (round_scalarcz): Ditto.
998 2021-09-22 liuhongt <hongtao.liu@intel.com>
1000 * config/i386/avx512fp16intrin.h (_mm512_fcmadd_pch):
1002 (_mm512_mask_fcmadd_pch): Likewise.
1003 (_mm512_mask3_fcmadd_pch): Likewise.
1004 (_mm512_maskz_fcmadd_pch): Likewise.
1005 (_mm512_fmadd_pch): Likewise.
1006 (_mm512_mask_fmadd_pch): Likewise.
1007 (_mm512_mask3_fmadd_pch): Likewise.
1008 (_mm512_maskz_fmadd_pch): Likewise.
1009 (_mm512_fcmadd_round_pch): Likewise.
1010 (_mm512_mask_fcmadd_round_pch): Likewise.
1011 (_mm512_mask3_fcmadd_round_pch): Likewise.
1012 (_mm512_maskz_fcmadd_round_pch): Likewise.
1013 (_mm512_fmadd_round_pch): Likewise.
1014 (_mm512_mask_fmadd_round_pch): Likewise.
1015 (_mm512_mask3_fmadd_round_pch): Likewise.
1016 (_mm512_maskz_fmadd_round_pch): Likewise.
1017 (_mm512_fcmul_pch): Likewise.
1018 (_mm512_mask_fcmul_pch): Likewise.
1019 (_mm512_maskz_fcmul_pch): Likewise.
1020 (_mm512_fmul_pch): Likewise.
1021 (_mm512_mask_fmul_pch): Likewise.
1022 (_mm512_maskz_fmul_pch): Likewise.
1023 (_mm512_fcmul_round_pch): Likewise.
1024 (_mm512_mask_fcmul_round_pch): Likewise.
1025 (_mm512_maskz_fcmul_round_pch): Likewise.
1026 (_mm512_fmul_round_pch): Likewise.
1027 (_mm512_mask_fmul_round_pch): Likewise.
1028 (_mm512_maskz_fmul_round_pch): Likewise.
1029 * config/i386/avx512fp16vlintrin.h (_mm_fmadd_pch):
1031 (_mm_mask_fmadd_pch): Likewise.
1032 (_mm_mask3_fmadd_pch): Likewise.
1033 (_mm_maskz_fmadd_pch): Likewise.
1034 (_mm256_fmadd_pch): Likewise.
1035 (_mm256_mask_fmadd_pch): Likewise.
1036 (_mm256_mask3_fmadd_pch): Likewise.
1037 (_mm256_maskz_fmadd_pch): Likewise.
1038 (_mm_fcmadd_pch): Likewise.
1039 (_mm_mask_fcmadd_pch): Likewise.
1040 (_mm_mask3_fcmadd_pch): Likewise.
1041 (_mm_maskz_fcmadd_pch): Likewise.
1042 (_mm256_fcmadd_pch): Likewise.
1043 (_mm256_mask_fcmadd_pch): Likewise.
1044 (_mm256_mask3_fcmadd_pch): Likewise.
1045 (_mm256_maskz_fcmadd_pch): Likewise.
1046 (_mm_fmul_pch): Likewise.
1047 (_mm_mask_fmul_pch): Likewise.
1048 (_mm_maskz_fmul_pch): Likewise.
1049 (_mm256_fmul_pch): Likewise.
1050 (_mm256_mask_fmul_pch): Likewise.
1051 (_mm256_maskz_fmul_pch): Likewise.
1052 (_mm_fcmul_pch): Likewise.
1053 (_mm_mask_fcmul_pch): Likewise.
1054 (_mm_maskz_fcmul_pch): Likewise.
1055 (_mm256_fcmul_pch): Likewise.
1056 (_mm256_mask_fcmul_pch): Likewise.
1057 (_mm256_maskz_fcmul_pch): Likewise.
1058 * config/i386/i386-builtin-types.def (V8HF_FTYPE_V8HF_V8HF_V8HF,
1059 V8HF_FTYPE_V16HF_V16HF_V16HF, V16HF_FTYPE_V16HF_V16HF_V16HF_UQI,
1060 V32HF_FTYPE_V32HF_V32HF_V32HF_INT,
1061 V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT): Add new builtin types.
1062 * config/i386/i386-builtin.def: Add new builtins.
1063 * config/i386/i386-expand.c: Handle new builtin types.
1064 * config/i386/subst.md (SUBST_CV): New.
1065 (maskc_name): Ditto.
1066 (maskc_operand3): Ditto.
1068 (sdc_maskz_name): Ditto.
1069 (sdc_mask_op4): Ditto.
1070 (sdc_mask_op5): Ditto.
1071 (sdc_mask_mode512bit_condition): Ditto.
1073 (round_maskc_operand3): Ditto.
1074 (round_sdc_mask_operand4): Ditto.
1075 (round_maskc_op3): Ditto.
1076 (round_sdc_mask_op4): Ditto.
1077 (round_saeonly_sdc_mask_operand5): Ditto.
1078 * config/i386/sse.md (unspec): Add complex fma unspecs.
1079 (avx512fmaskcmode): New.
1080 (UNSPEC_COMPLEX_F_C_MA): Ditto.
1081 (UNSPEC_COMPLEX_F_C_MUL): Ditto.
1082 (complexopname): Ditto.
1083 (<avx512>_fmaddc_<mode>_maskz<round_expand_name>): New expander.
1084 (<avx512>_fcmaddc_<mode>_maskz<round_expand_name>): Ditto.
1085 (fma_<complexopname>_<mode><sdc_maskz_name><round_name>): New
1087 (<avx512>_<complexopname>_<mode>_mask<round_name>): Ditto.
1088 (<avx512>_<complexopname>_<mode><maskc_name><round_name>): Ditto.
1090 2021-09-22 Kewen Lin <linkw@linux.ibm.com>
1092 * config/rs6000/rs6000.opt (rs6000-density-pct-threshold,
1093 rs6000-density-size-threshold, rs6000-density-penalty,
1094 rs6000-density-load-pct-threshold,
1095 rs6000-density-load-num-threshold): New parameter.
1096 * config/rs6000/rs6000.c (rs6000_density_test): Adjust with
1097 corresponding parameters.
1099 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1101 * gimple-range-path.cc (path_range_query::defined_outside_path):
1103 (path_range_query::range_on_path_entry): New.
1104 (path_range_query::internal_range_of_expr): Resolve unknowns
1106 (path_range_query::improve_range_with_equivs): New.
1107 (path_range_query::ssa_range_in_phi): Resolve unknowns with
1109 * gimple-range-path.h (class path_range_query): Add
1110 defined_outside_path, range_on_path_entry, and
1111 improve_range_with_equivs.
1113 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1115 * gimple-range-path.cc (path_range_query::add_to_imports): New.
1116 (path_range_query::add_copies_to_imports): New.
1117 (path_range_query::precompute_ranges): Call
1118 add_copies_to_imports.
1119 * gimple-range-path.h (class path_range_query): Add prototypes
1120 for add_copies_to_imports and add_to_imports.
1122 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1124 * gimple-range-path.cc (path_range_query::range_defined_in_block):
1125 Remove useless code.
1127 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1129 * gimple-range-fold.h (class fur_source): Make oracle protected.
1130 * gimple-range-path.cc (path_range_query::path_range_query): Add
1131 resolve argument. Initialize oracle.
1132 (path_range_query::~path_range_query): Delete oracle.
1133 (path_range_query::range_of_stmt): Adapt to use relations.
1134 (path_range_query::precompute_ranges): Pre-compute relations.
1135 (class jt_fur_source): New
1136 (jt_fur_source::jt_fur_source): New.
1137 (jt_fur_source::register_relation): New.
1138 (jt_fur_source::query_relation): New.
1139 (path_range_query::precompute_relations): New.
1140 (path_range_query::precompute_phi_relations): New.
1141 * gimple-range-path.h (path_range_query): Add resolve argument.
1142 Add oracle, precompute_relations, precompute_phi_relations.
1143 * tree-ssa-threadbackward.c (back_threader::back_threader): Pass
1144 resolve argument to solver.
1146 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1148 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
1149 Rename postfold_gcond_edges to register_outgoing_edges and
1151 (fold_using_range::postfold_gcond_edges): Rename...
1152 (fur_source::register_outgoing_edges): ...to this.
1153 * gimple-range-fold.h (postfold_gcond_edges): Rename to
1154 register_outgoing_edges and move to fur_source.
1156 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1158 * gimple-range-fold.cc (fold_using_range::range_of_phi): Check
1159 dom_info_available_p.
1161 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1163 * gimple-range-cache.cc (non_null_ref::non_null_ref): Use create
1164 and quick_grow_cleared instead of safe_grow_cleared.
1166 2021-09-21 Thomas Schwinge <thomas@codesourcery.com>
1169 * omp-oacc-neuter-broadcast.cc (oacc_do_neutering): Evaluate
1172 2021-09-21 Richard Earnshaw <rearnsha@arm.com>
1174 * configure.ac: Detect when the assembler supports new-style
1175 architecture extensions.
1176 * common/config/arm/arm-common.c (arm_rewrite_mcpu): Return
1177 the full CPU string if the assembler can grok it.
1178 (arm_rewrite_march): Likewise but for the architecture.
1179 * config.in: Regenerate.
1180 * configure: Regenerate.
1182 2021-09-21 Richard Biener <rguenther@suse.de>
1184 PR tree-optimization/102421
1185 * tree-vect-loop.c (vect_dissolve_slp_only_groups): Copy and
1186 adjust alignment info.
1188 2021-09-21 Kewen Lin <linkw@linux.ibm.com>
1190 * ipa-fnsummary.c (ipa_fn_summary_write): Remove inconsistent
1191 bitfield stream out.
1193 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
1195 * gimple-range-fold.cc (fold_using_range::range_of_phi): Ignore
1196 undefined edges, apply an equivalence if appropriate.
1197 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Return
1198 UNDEFINED if EDGE_EXECUTABLE is not set.
1199 * gimple-range.cc (gimple_ranger::gimple_ranger): Set all edges
1200 as EXECUTABLE upon startup.
1201 (gimple_ranger::range_on_edge): Return UNDEFINED for edges without
1202 EDGE_EXECUTABLE set.
1203 * vr-values.c (set_and_propagate_unexecutable): New.
1204 (simplify_using_ranges::fold_cond): Call set_and_propagate.
1205 (simplify_using_ranges::simplify_switch_using_ranges): Ditto.
1206 * vr-values.h: Add prototype.
1208 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
1210 * value-relation.cc (equiv_oracle::register_initial_def): New.
1211 (equiv_oracle::register_relation): Call register_initial_def.
1212 (equiv_oracle::add_equiv_to_block): New. Split register_relation.
1213 (relation_oracle::register_stmt): Check def block of PHI arguments.
1214 * value-relation.h (equiv_oracle): Add new prototypes.
1216 2021-09-20 Matthias Kretz <m.kretz@gsi.de>
1218 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
1219 Define __RECIPROCAL_MATH__, __NO_SIGNED_ZEROS__,
1220 __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__, and
1221 __ROUNDING_MATH__ according to their corresponding flags.
1222 * doc/cpp.texi: Document __RECIPROCAL_MATH__,
1223 __NO_SIGNED_ZEROS__, __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__,
1224 and __ROUNDING_MATH__.
1226 2021-09-20 Richard Biener <rguenther@suse.de>
1228 * tree-vect-stmts.c (vectorizable_load): Use the vectype
1231 2021-09-20 Richard Biener <rguenther@suse.de>
1233 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
1234 Do not compute alignment of the vectorized access here.
1236 2021-09-20 Richard Biener <rguenther@suse.de>
1238 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
1239 Store -1 for runtime alias peeling iterations.
1241 2021-09-20 Richard Biener <rguenther@suse.de>
1243 * config.gcc: Obsolete hppa[12]*-*-hpux10* and hppa[12]*-*-hpux11*.
1245 2021-09-20 Thomas Schwinge <thomas@codesourcery.com>
1247 * input.c (string_concat_db::record_string_concatenation)
1248 (string_concat_db::get_string_concatenation): Skip for
1249 'RESERVED_LOCATION_P'.
1251 2021-09-20 Richard Biener <rguenther@suse.de>
1253 PR tree-optimization/65206
1254 * tree-data-ref.h (struct data_reference): Add alt_indices,
1256 * tree-data-ref.c (free_data_ref): Release alt_indices.
1257 (dr_analyze_indices): Work on struct indices and get DR_REF as tree.
1258 (create_data_ref): Adjust.
1259 (initialize_data_dependence_relation): Split into head
1260 and tail. When the base objects fail to match up try
1261 again with pointer-based analysis of indices.
1262 * tree-vectorizer.c (vec_info_shared::check_datarefs): Do
1263 not compare the lazily computed alternate set of indices.
1265 2021-09-20 Iain Sandoe <iain@sandoe.co.uk>
1267 * gcc.c: Test for execute OK when we find the
1268 programs for assembler linker and dsymutil and those
1269 were specified at configure-time.
1271 2021-09-19 Martin Sebor <msebor@redhat.com>
1273 PR middle-end/102403
1274 * gimple-predicate-analysis.cc (predicate::init_from_control_deps):
1275 Correct a function pre/postcondition.
1277 2021-09-19 Martin Sebor <msebor@redhat.com>
1279 PR middle-end/102243
1280 * tree-ssa-strlen.c (get_range): Handle null cfun.
1282 2021-09-19 Iain Sandoe <iain@sandoe.co.uk>
1284 * config/darwin.h (LINK_COMMAND_SPEC_A): Use Darwin10
1285 unwinder shim as a convenience library.
1287 2021-09-19 Andrew Pinski <apinski@marvell.com>
1289 * doc/install.texi: Add note about
1290 binutils 2.35 is required for LTO usage.
1292 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
1294 * tree-ssa-threadbackward.c
1295 (back_threader_registry::register_path): Use push_edge.
1296 * tree-ssa-threadedge.c
1297 (jump_threader::thread_around_empty_blocks): Same.
1298 (jump_threader::thread_through_normal_block): Same.
1299 (jump_threader::thread_across_edge): Same. Also, use auto_bitmap.
1301 * tree-ssa-threadupdate.c
1302 (jt_path_registry::allocate_thread_edge): Remove.
1303 (jt_path_registry::push_edge): New.
1304 (dump_jump_thread_path): Make static.
1305 * tree-ssa-threadupdate.h (allocate_thread_edge): Remove.
1308 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
1310 * gimple-range-path.cc (path_range_query::path_range_query): Add
1312 (path_range_query::dump): Remove extern declaration of dump_ranger.
1313 * gimple-range-trace.cc (dump_ranger): Add DEBUG_FUNCTION marker.
1314 * gimple-range-trace.h (dump_ranger): Add prototype.
1316 2021-09-19 John Ericson <git@JohnEricson.me>
1318 * gcc.c (find_a_program): New function, factored out of...
1319 (find_a_file): Here.
1320 (execute): Use find_a_program when looking for programs rather
1323 2021-09-19 Matwey V. Kornilov <matwey.kornilov@gmail.com>
1325 * config/avr/avr-mcus.def: Add atmega324pb.
1326 * doc/avr-mmcu.texi: Corresponding changes.
1328 2021-09-19 Roger Sayle <roger@nextmovesoftware.com>
1331 * match.pd (cmp @0 REAL_CST@1): When @0 is also REAL_CST, apply
1332 the same transformations as to @1. For comparisons against NaN,
1333 don't check HONOR_SNANS but confirm that neither operand is a
1336 2021-09-19 Benjamin Peterson <benjamin@locrian.net>
1338 * attribs.c (make_unique_name): Delete.
1339 * attribs.h (make_unique_name): Delete.
1341 2021-09-19 Andrew Pinski <apinski@marvell.com>
1343 * lra-constraints.c (check_and_process_move): Assert
1344 that dclass and sclass are greater than or equal to NO_REGS.
1346 2021-09-18 Jakub Jelinek <jakub@redhat.com>
1348 * tree.h (OMP_CLAUSE_ORDER_UNCONSTRAINED): Define.
1349 * tree-pretty-print.c (dump_omp_clause): Print unconstrained:
1350 for OMP_CLAUSE_ORDER_UNCONSTRAINED.
1352 2021-09-18 liuhongt <hongtao.liu@intel.com>
1354 * config/i386/i386-features.c (remove_partial_avx_dependency):
1355 Restrict TARGET_USE_VECTOR_FP_CONVERTS and
1356 TARGET_USE_VECTOR_CONVERTS to conversion instructions only.
1358 2021-09-18 Jakub Jelinek <jakub@redhat.com>
1360 * gimplify.c (omp_default_clause): For C/C++ default({,first}private),
1361 if file/namespace scope variable doesn't have predetermined sharing,
1362 treat it as if there was default(none).
1364 2021-09-18 liuhongt <hongtao.liu@intel.com>
1366 * config/i386/avx512fp16intrin.h (_mm_fmadd_sh):
1368 (_mm_mask_fmadd_sh): Likewise.
1369 (_mm_mask3_fmadd_sh): Likewise.
1370 (_mm_maskz_fmadd_sh): Likewise.
1371 (_mm_fmadd_round_sh): Likewise.
1372 (_mm_mask_fmadd_round_sh): Likewise.
1373 (_mm_mask3_fmadd_round_sh): Likewise.
1374 (_mm_maskz_fmadd_round_sh): Likewise.
1375 (_mm_fnmadd_sh): Likewise.
1376 (_mm_mask_fnmadd_sh): Likewise.
1377 (_mm_mask3_fnmadd_sh): Likewise.
1378 (_mm_maskz_fnmadd_sh): Likewise.
1379 (_mm_fnmadd_round_sh): Likewise.
1380 (_mm_mask_fnmadd_round_sh): Likewise.
1381 (_mm_mask3_fnmadd_round_sh): Likewise.
1382 (_mm_maskz_fnmadd_round_sh): Likewise.
1383 (_mm_fmsub_sh): Likewise.
1384 (_mm_mask_fmsub_sh): Likewise.
1385 (_mm_mask3_fmsub_sh): Likewise.
1386 (_mm_maskz_fmsub_sh): Likewise.
1387 (_mm_fmsub_round_sh): Likewise.
1388 (_mm_mask_fmsub_round_sh): Likewise.
1389 (_mm_mask3_fmsub_round_sh): Likewise.
1390 (_mm_maskz_fmsub_round_sh): Likewise.
1391 (_mm_fnmsub_sh): Likewise.
1392 (_mm_mask_fnmsub_sh): Likewise.
1393 (_mm_mask3_fnmsub_sh): Likewise.
1394 (_mm_maskz_fnmsub_sh): Likewise.
1395 (_mm_fnmsub_round_sh): Likewise.
1396 (_mm_mask_fnmsub_round_sh): Likewise.
1397 (_mm_mask3_fnmsub_round_sh): Likewise.
1398 (_mm_maskz_fnmsub_round_sh): Likewise.
1399 * config/i386/i386-builtin-types.def
1400 (V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT): New builtin type.
1401 * config/i386/i386-builtin.def: Add new builtins.
1402 * config/i386/i386-expand.c: Handle new builtin type.
1403 * config/i386/sse.md (fmai_vmfmadd_<mode><round_name>):
1404 Ajdust to support FP16.
1405 (fmai_vmfmsub_<mode><round_name>): Ditto.
1406 (fmai_vmfnmadd_<mode><round_name>): Ditto.
1407 (fmai_vmfnmsub_<mode><round_name>): Ditto.
1408 (*fmai_fmadd_<mode>): Ditto.
1409 (*fmai_fmsub_<mode>): Ditto.
1410 (*fmai_fnmadd_<mode><round_name>): Ditto.
1411 (*fmai_fnmsub_<mode><round_name>): Ditto.
1412 (avx512f_vmfmadd_<mode>_mask<round_name>): Ditto.
1413 (avx512f_vmfmadd_<mode>_mask3<round_name>): Ditto.
1414 (avx512f_vmfmadd_<mode>_maskz<round_expand_name>): Ditto.
1415 (avx512f_vmfmadd_<mode>_maskz_1<round_name>): Ditto.
1416 (*avx512f_vmfmsub_<mode>_mask<round_name>): Ditto.
1417 (avx512f_vmfmsub_<mode>_mask3<round_name>): Ditto.
1418 (*avx512f_vmfmsub_<mode>_maskz_1<round_name>): Ditto.
1419 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
1420 (*avx512f_vmfnmsub_<mode>_mask3<round_name>): Ditto.
1421 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
1422 (*avx512f_vmfnmadd_<mode>_mask<round_name>): Renamed to ...
1423 (avx512f_vmfnmadd_<mode>_mask<round_name>) ... this, and
1424 adjust to support FP16.
1425 (avx512f_vmfnmadd_<mode>_mask3<round_name>): Ditto.
1426 (avx512f_vmfnmadd_<mode>_maskz_1<round_name>): Ditto.
1427 (avx512f_vmfnmadd_<mode>_maskz<round_expand_name>): New
1430 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
1432 * config/i386/sse.md (avx512fmaskmodelower): Extend to support
1434 (maskload<mode><avx512fmaskmodelower>): Ditto.
1435 (maskstore<mode><avx512fmaskmodelower>): Ditto.
1437 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
1439 * config/i386/i386-expand.c (ix86_expand_fp_absneg_operator):
1441 (ix86_expand_copysign): Ditto.
1442 (ix86_expand_xorsign): Ditto.
1443 * config/i386/i386.c (ix86_build_const_vector): Handle HF vector
1445 (ix86_build_signbit_mask): Ditto.
1446 (ix86_can_change_mode_class): Ditto.
1447 * config/i386/i386.md
1448 (SSEMODEF): Add HFmode.
1449 (ssevecmodef): Ditto.
1450 (<code>hf2): New define_expand.
1451 (*<code>hf2_1): New define_insn_and_split.
1452 (copysign<mode>): Extend to support HFmode under AVX512FP16.
1453 (xorsign<mode>): Ditto.
1454 * config/i386/sse.md (VFB): New mode iterator.
1455 (VFB_128_256): Ditto.
1457 (sseintvecmode2): Support HF vector mode.
1458 (<code><mode>2): Use new mode iterator.
1459 (*<code><mode>2): Ditto.
1460 (copysign<mode>3): Ditto.
1461 (xorsign<mode>3): Ditto.
1462 (<code><mode>3<mask_name>): Ditto.
1463 (<code><mode>3<mask_name>): Ditto.
1464 (<sse>_andnot<mode>3<mask_name>): Adjust for HF vector mode.
1465 (<sse>_andnot<mode>3<mask_name>): Ditto.
1466 (*<code><mode>3<mask_name>): Ditto.
1467 (*<code><mode>3<mask_name>): Ditto.
1469 2021-09-18 liuhongt <hongtao.liu@intel.com>
1471 * config/i386/avx512fp16intrin.h (_mm512_mask_fmadd_ph):
1473 (_mm512_mask3_fmadd_ph): Likewise.
1474 (_mm512_maskz_fmadd_ph): Likewise.
1475 (_mm512_fmadd_round_ph): Likewise.
1476 (_mm512_mask_fmadd_round_ph): Likewise.
1477 (_mm512_mask3_fmadd_round_ph): Likewise.
1478 (_mm512_maskz_fmadd_round_ph): Likewise.
1479 (_mm512_fnmadd_ph): Likewise.
1480 (_mm512_mask_fnmadd_ph): Likewise.
1481 (_mm512_mask3_fnmadd_ph): Likewise.
1482 (_mm512_maskz_fnmadd_ph): Likewise.
1483 (_mm512_fnmadd_round_ph): Likewise.
1484 (_mm512_mask_fnmadd_round_ph): Likewise.
1485 (_mm512_mask3_fnmadd_round_ph): Likewise.
1486 (_mm512_maskz_fnmadd_round_ph): Likewise.
1487 (_mm512_fmsub_ph): Likewise.
1488 (_mm512_mask_fmsub_ph): Likewise.
1489 (_mm512_mask3_fmsub_ph): Likewise.
1490 (_mm512_maskz_fmsub_ph): Likewise.
1491 (_mm512_fmsub_round_ph): Likewise.
1492 (_mm512_mask_fmsub_round_ph): Likewise.
1493 (_mm512_mask3_fmsub_round_ph): Likewise.
1494 (_mm512_maskz_fmsub_round_ph): Likewise.
1495 (_mm512_fnmsub_ph): Likewise.
1496 (_mm512_mask_fnmsub_ph): Likewise.
1497 (_mm512_mask3_fnmsub_ph): Likewise.
1498 (_mm512_maskz_fnmsub_ph): Likewise.
1499 (_mm512_fnmsub_round_ph): Likewise.
1500 (_mm512_mask_fnmsub_round_ph): Likewise.
1501 (_mm512_mask3_fnmsub_round_ph): Likewise.
1502 (_mm512_maskz_fnmsub_round_ph): Likewise.
1503 * config/i386/avx512fp16vlintrin.h (_mm256_fmadd_ph):
1505 (_mm256_mask_fmadd_ph): Likewise.
1506 (_mm256_mask3_fmadd_ph): Likewise.
1507 (_mm256_maskz_fmadd_ph): Likewise.
1508 (_mm_fmadd_ph): Likewise.
1509 (_mm_mask_fmadd_ph): Likewise.
1510 (_mm_mask3_fmadd_ph): Likewise.
1511 (_mm_maskz_fmadd_ph): Likewise.
1512 (_mm256_fnmadd_ph): Likewise.
1513 (_mm256_mask_fnmadd_ph): Likewise.
1514 (_mm256_mask3_fnmadd_ph): Likewise.
1515 (_mm256_maskz_fnmadd_ph): Likewise.
1516 (_mm_fnmadd_ph): Likewise.
1517 (_mm_mask_fnmadd_ph): Likewise.
1518 (_mm_mask3_fnmadd_ph): Likewise.
1519 (_mm_maskz_fnmadd_ph): Likewise.
1520 (_mm256_fmsub_ph): Likewise.
1521 (_mm256_mask_fmsub_ph): Likewise.
1522 (_mm256_mask3_fmsub_ph): Likewise.
1523 (_mm256_maskz_fmsub_ph): Likewise.
1524 (_mm_fmsub_ph): Likewise.
1525 (_mm_mask_fmsub_ph): Likewise.
1526 (_mm_mask3_fmsub_ph): Likewise.
1527 (_mm_maskz_fmsub_ph): Likewise.
1528 (_mm256_fnmsub_ph): Likewise.
1529 (_mm256_mask_fnmsub_ph): Likewise.
1530 (_mm256_mask3_fnmsub_ph): Likewise.
1531 (_mm256_maskz_fnmsub_ph): Likewise.
1532 (_mm_fnmsub_ph): Likewise.
1533 (_mm_mask_fnmsub_ph): Likewise.
1534 (_mm_mask3_fnmsub_ph): Likewise.
1535 (_mm_maskz_fnmsub_ph): Likewise.
1536 * config/i386/i386-builtin.def: Add corresponding new builtins.
1537 * config/i386/sse.md
1538 (<avx512>_fmadd_<mode>_maskz<round_expand_name>): Adjust to
1539 support HF vector modes.
1540 (<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name><round_name>):
1542 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
1543 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
1544 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
1545 (<avx512>_fmadd_<mode>_mask<round_name>): Ditto.
1546 (<avx512>_fmadd_<mode>_mask3<round_name>): Ditto.
1547 (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Ditto.
1548 (<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name><round_name>):
1550 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
1551 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
1552 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
1553 (<avx512>_fmsub_<mode>_mask<round_name>): Ditto.
1554 (<avx512>_fmsub_<mode>_mask3<round_name>): Ditto.
1555 (<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name><round_name>):
1557 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
1558 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
1559 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
1560 (<avx512>_fnmadd_<mode>_mask<round_name>): Ditto.
1561 (<avx512>_fnmadd_<mode>_mask3<round_name>): Ditto.
1562 (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Ditto.
1563 (<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name><round_name>):
1565 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
1566 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
1567 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
1568 (<avx512>_fnmsub_<mode>_mask<round_name>): Ditto.
1569 (<avx512>_fnmsub_<mode>_mask3<round_name>): Ditto.
1571 2021-09-18 liuhongt <hongtao.liu@intel.com>
1573 * config/i386/avx512fp16intrin.h (_mm512_fmaddsub_ph):
1575 (_mm512_mask_fmaddsub_ph): Likewise.
1576 (_mm512_mask3_fmaddsub_ph): Likewise.
1577 (_mm512_maskz_fmaddsub_ph): Likewise.
1578 (_mm512_fmaddsub_round_ph): Likewise.
1579 (_mm512_mask_fmaddsub_round_ph): Likewise.
1580 (_mm512_mask3_fmaddsub_round_ph): Likewise.
1581 (_mm512_maskz_fmaddsub_round_ph): Likewise.
1582 (_mm512_mask_fmsubadd_ph): Likewise.
1583 (_mm512_mask3_fmsubadd_ph): Likewise.
1584 (_mm512_maskz_fmsubadd_ph): Likewise.
1585 (_mm512_fmsubadd_round_ph): Likewise.
1586 (_mm512_mask_fmsubadd_round_ph): Likewise.
1587 (_mm512_mask3_fmsubadd_round_ph): Likewise.
1588 (_mm512_maskz_fmsubadd_round_ph): Likewise.
1589 * config/i386/avx512fp16vlintrin.h (_mm256_fmaddsub_ph):
1591 (_mm256_mask_fmaddsub_ph): Likewise.
1592 (_mm256_mask3_fmaddsub_ph): Likewise.
1593 (_mm256_maskz_fmaddsub_ph): Likewise.
1594 (_mm_fmaddsub_ph): Likewise.
1595 (_mm_mask_fmaddsub_ph): Likewise.
1596 (_mm_mask3_fmaddsub_ph): Likewise.
1597 (_mm_maskz_fmaddsub_ph): Likewise.
1598 (_mm256_fmsubadd_ph): Likewise.
1599 (_mm256_mask_fmsubadd_ph): Likewise.
1600 (_mm256_mask3_fmsubadd_ph): Likewise.
1601 (_mm256_maskz_fmsubadd_ph): Likewise.
1602 (_mm_fmsubadd_ph): Likewise.
1603 (_mm_mask_fmsubadd_ph): Likewise.
1604 (_mm_mask3_fmsubadd_ph): Likewise.
1605 (_mm_maskz_fmsubadd_ph): Likewise.
1606 * config/i386/i386-builtin.def: Add corresponding new builtins.
1607 * config/i386/sse.md (VFH_SF_AVX512VL): New mode iterator.
1608 * (<avx512>_fmsubadd_<mode>_maskz<round_expand_name>): New expander.
1609 * (<avx512>_fmaddsub_<mode>_maskz<round_expand_name>): Use
1611 * (<sd_mask_codefor>fma_fmaddsub_<mode><sd_maskz_name><round_name>):
1613 * (<avx512>_fmaddsub_<mode>_mask<round_name>): Ditto.
1614 * (<avx512>_fmaddsub_<mode>_mask3<round_name>): Ditto.
1615 * (<sd_mask_codefor>fma_fmsubadd_<mode><sd_maskz_name><round_name>):
1617 * (<avx512>_fmsubadd_<mode>_mask<round_name>): Ditto.
1618 * (<avx512>_fmsubadd_<mode>_mask3<round_name>): Ditto.
1620 2021-09-18 liuhongt <hongtao.liu@intel.com>
1623 * config/i386/i386.c (ix86_print_operand): Handle
1624 V8HF/V16HF/V32HFmode.
1625 * config/i386/i386.h (VALID_BCST_MODE_P): Add HFmode.
1626 * config/i386/sse.md (avx512bcst): Remove.
1628 2021-09-17 Martin Sebor <msebor@redhat.com>
1630 * Makefile.in (OBJS): Add gimple-predicate-analysis.o.
1631 * tree-ssa-uninit.c (max_phi_args): Move to gimple-predicate-analysis.
1632 (MASK_SET_BIT, MASK_TEST_BIT, MASK_EMPTY): Same.
1633 (check_defs): Add comment.
1634 (can_skip_redundant_opnd): Update comment.
1635 (compute_uninit_opnds_pos): Adjust to namespace change.
1636 (find_pdom): Move to gimple-predicate-analysis.cc.
1638 (struct uninit_undef_val_t): New.
1639 (is_non_loop_exit_postdominating): Move to gimple-predicate-analysis.cc.
1640 (find_control_equiv_block): Same.
1641 (MAX_NUM_CHAINS, MAX_CHAIN_LEN, MAX_POSTDOM_CHECK): Same.
1642 (MAX_SWITCH_CASES): Same.
1643 (compute_control_dep_chain): Same.
1644 (find_uninit_use): Use predicate analyzer.
1645 (struct pred_info): Move to gimple-predicate-analysis.
1646 (convert_control_dep_chain_into_preds): Same.
1647 (find_predicates): Same.
1648 (collect_phi_def_edges): Same.
1649 (warn_uninitialized_phi): Use predicate analyzer.
1650 (find_def_preds): Move to gimple-predicate-analysis.
1651 (dump_pred_info): Same.
1652 (dump_pred_chain): Same.
1653 (dump_predicates): Same.
1654 (destroy_predicate_vecs): Remove.
1655 (execute_late_warn_uninitialized): New.
1656 (get_cmp_code): Move to gimple-predicate-analysis.
1657 (is_value_included_in): Same.
1658 (value_sat_pred_p): Same.
1659 (find_matching_predicate_in_rest_chains): Same.
1660 (is_use_properly_guarded): Same.
1661 (prune_uninit_phi_opnds): Same.
1662 (find_var_cmp_const): Same.
1663 (use_pred_not_overlap_with_undef_path_pred): Same.
1664 (pred_equal_p): Same.
1665 (is_neq_relop_p): Same.
1666 (is_neq_zero_form_p): Same.
1667 (pred_expr_equal_p): Same.
1668 (is_pred_expr_subset_of): Same.
1669 (is_pred_chain_subset_of): Same.
1670 (is_included_in): Same.
1671 (is_superset_of): Same.
1673 (simplify_pred): Same.
1674 (simplify_preds_2): Same.
1675 (simplify_preds_3): Same.
1676 (simplify_preds_4): Same.
1677 (simplify_preds): Same.
1679 (push_to_worklist): Same.
1680 (get_pred_info_from_cmp): Same.
1681 (is_degenerated_phi): Same.
1682 (normalize_one_pred_1): Same.
1683 (normalize_one_pred): Same.
1684 (normalize_one_pred_chain): Same.
1685 (normalize_preds): Same.
1686 (can_one_predicate_be_invalidated_p): Same.
1687 (can_chain_union_be_invalidated_p): Same.
1688 (uninit_uses_cannot_happen): Same.
1689 (pass_late_warn_uninitialized::execute): Define.
1690 * gimple-predicate-analysis.cc: New file.
1691 * gimple-predicate-analysis.h: New file.
1693 2021-09-17 Julian Brown <julian@codesourcery.com>
1695 * config/gcn/gcn.c (gimple.h): Include.
1696 (gcn_fork_join): Emit barrier for worker-level joins.
1697 * omp-oacc-neuter-broadcast.cc (find_local_vars_to_propagate): Add
1698 writes_gang_private bitmap parameter. Set bit for blocks
1699 containing gang-private variable writes.
1700 (worker_single_simple): Don't emit barrier after predicated block.
1701 (worker_single_copy): Don't emit barrier if we're not broadcasting
1702 anything and the block contains no gang-private writes.
1703 (neuter_worker_single): Don't predicate blocks that only contain
1704 NOPs or internal marker functions. Pass has_gang_private_write
1705 argument to worker_single_copy.
1706 (oacc_do_neutering): Add writes_gang_private bitmap handling.
1708 2021-09-17 Julian Brown <julian@codesourcery.com>
1710 * config/gcn/gcn-protos.h
1711 (gcn_goacc_create_worker_broadcast_record): Update prototype.
1712 * config/gcn/gcn-tree.c (gcn_goacc_get_worker_red_decl): Use
1713 preallocated block of LDS memory. Do not cache/share decls for
1714 reduction temporaries between invocations.
1715 (gcn_goacc_reduction_teardown): Unshare VAR on second use.
1716 (gcn_goacc_create_worker_broadcast_record): Add OFFSET parameter
1717 and return temporary LDS space at that offset. Return pointer in
1719 * config/gcn/gcn.c (acc_lds_size, gang_private_hwm, lds_allocs):
1721 (ACC_LDS_SIZE): Define as acc_lds_size.
1722 (gcn_init_machine_status): Don't initialise lds_allocated,
1723 lds_allocs, reduc_decls fields of machine function struct.
1724 (gcn_option_override): Handle default size for gang-private
1725 variables and -mgang-private-size option.
1726 (gcn_expand_prologue): Use LDS_SIZE instead of LDS_SIZE-1 when
1727 initialising M0_REG.
1728 (gcn_shared_mem_layout): New function.
1729 (gcn_print_lds_decl): Update comment. Use global lds_allocs map and
1730 gang_private_hwm variable.
1731 (TARGET_GOACC_SHARED_MEM_LAYOUT): Define target hook.
1732 * config/gcn/gcn.h (machine_function): Remove lds_allocated,
1733 lds_allocs, reduc_decls. Add reduction_base, reduction_limit.
1734 * config/gcn/gcn.opt (gang_private_size_opt): New global.
1735 (mgang-private-size=): New option.
1736 * doc/tm.texi.in (TARGET_GOACC_SHARED_MEM_LAYOUT): Place
1738 * doc/tm.texi: Regenerate.
1739 * omp-oacc-neuter-broadcast.cc (targhooks.h, diagnostic-core.h):
1741 (build_sender_ref): Handle sender_decl being pointer.
1742 (worker_single_copy): Add PLACEMENT and ISOLATE_BROADCASTS
1743 parameters. Pass placement argument to
1744 create_worker_broadcast_record hook invocations. Handle
1745 sender_decl being pointer and isolate_broadcasts inserting extra
1747 (blk_offset_map_t): Add typedef.
1748 (neuter_worker_single): Add BLK_OFFSET_MAP parameter. Pass
1749 preallocated range to worker_single_copy call.
1750 (dfs_broadcast_reachable_1): New function.
1751 (idx_decl_pair_t, used_range_vec_t): New typedefs.
1752 (sort_size_descending): New function.
1753 (addr_range): New class.
1754 (splay_tree_compare_addr_range, splay_tree_free_key)
1755 (first_fit_range, merge_ranges_1, merge_ranges): New functions.
1756 (execute_omp_oacc_neuter_broadcast): Rename to...
1757 (oacc_do_neutering): ... this. Add BOUNDS_LO, BOUNDS_HI
1758 parameters. Arrange layout of shared memory for broadcast
1760 (execute_omp_oacc_neuter_broadcast): New function.
1761 (pass_omp_oacc_neuter_broadcast::gate): Remove num_workers==1
1762 handling from here. Enable pass for all OpenACC routines in order
1763 to call shared memory-layout hook.
1764 * target.def (create_worker_broadcast_record): Add OFFSET
1766 (shared_mem_layout): New hook.
1768 2021-09-17 Julian Brown <julian@codesourcery.com>
1769 Thomas Schwinge <thomas@codesourcery.com>
1771 * omp-oacc-neuter-broadcast.cc
1772 (pass_omp_oacc_neuter_broadcast::gate): Disable if num_workers is
1774 (execute_omp_oacc_neuter_broadcast): Adjust.
1776 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
1778 * value-relation.cc (class equiv_chain): Move to header file.
1779 (path_oracle::path_oracle): New.
1780 (path_oracle::~path_oracle): New.
1781 (path_oracle::register_relation): New.
1782 (path_oracle::query_relation): New.
1783 (path_oracle::reset_path): New.
1784 (path_oracle::dump): New.
1785 * value-relation.h (class equiv_chain): Move to here.
1786 (class path_oracle): New.
1788 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
1790 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a DOM
1792 * gimple-range-fold.cc (fur_depend::register_relation): Use
1793 register_stmt/edge routines.
1794 * value-relation.cc (equiv_chain::find): Relocate from equiv_oracle.
1795 (equiv_oracle::equiv_oracle): Create self equivalence cache.
1796 (equiv_oracle::~equiv_oracle): Release same.
1797 (equiv_oracle::equiv_set): Return entry from self equiv cache if there
1798 are no equivalences.
1799 (equiv_oracle::find_equiv_block): Move list find to equiv_chain.
1800 (equiv_oracle::register_relation): Rename from register_equiv.
1801 (relation_chain_head::find_relation): Relocate from dom_oracle.
1802 (relation_oracle::register_stmt): New.
1803 (relation_oracle::register_edge): New.
1804 (dom_oracle::*): Rename from relation_oracle.
1805 (dom_oracle::register_relation): Adjust to call equiv_oracle.
1806 (dom_oracle::set_one_relation): Split from register_relation.
1807 (dom_oracle::register_transitives): Consolidate 2 methods.
1808 (dom_oracle::find_relation_block): Move core to relation_chain.
1809 (dom_oracle::query_relation): Rename from find_relation_dom and adjust.
1810 * value-relation.h (class relation_oracle): New pure virtual base.
1811 (class equiv_oracle): Inherit from relation_oracle and adjust.
1812 (class dom_oracle): Rename from old relation_oracle and adjust.
1814 2021-09-17 Martin Sebor <msebor@redhat.com>
1816 PR middle-end/102200
1817 * pointer-query.cc (access_ref::inform_access): Handle MIN/MAX_EXPR.
1818 (handle_min_max_size): Change argument. Store original SSA_NAME for
1819 operands to potentially distinct (sub)objects.
1820 (compute_objsize_r): Adjust call to the above.
1822 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
1824 * config/rs6000/rs6000.c (rs6000-builtins.h): New include.
1825 (rs6000_new_builtin_vectorized_function): New function.
1826 (rs6000_new_builtin_md_vectorized_function): Likewise.
1827 (rs6000_builtin_vectorized_function): Call
1828 rs6000_new_builtin_vectorized_function.
1829 (rs6000_builtin_md_vectorized_function): Call
1830 rs6000_new_builtin_md_vectorized_function.
1832 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
1834 * config/rs6000/rs6000-builtin-new.def (ASSEMBLE_ACC): Add mmaint flag.
1835 (ASSEMBLE_PAIR): Likewise.
1836 (BUILD_ACC): Likewise.
1837 (DISASSEMBLE_ACC): Likewise.
1838 (DISASSEMBLE_PAIR): Likewise.
1839 (PMXVBF16GER2): Likewise.
1840 (PMXVBF16GER2NN): Likewise.
1841 (PMXVBF16GER2NP): Likewise.
1842 (PMXVBF16GER2PN): Likewise.
1843 (PMXVBF16GER2PP): Likewise.
1844 (PMXVF16GER2): Likewise.
1845 (PMXVF16GER2NN): Likewise.
1846 (PMXVF16GER2NP): Likewise.
1847 (PMXVF16GER2PN): Likewise.
1848 (PMXVF16GER2PP): Likewise.
1849 (PMXVF32GER): Likewise.
1850 (PMXVF32GERNN): Likewise.
1851 (PMXVF32GERNP): Likewise.
1852 (PMXVF32GERPN): Likewise.
1853 (PMXVF32GERPP): Likewise.
1854 (PMXVF64GER): Likewise.
1855 (PMXVF64GERNN): Likewise.
1856 (PMXVF64GERNP): Likewise.
1857 (PMXVF64GERPN): Likewise.
1858 (PMXVF64GERPP): Likewise.
1859 (PMXVI16GER2): Likewise.
1860 (PMXVI16GER2PP): Likewise.
1861 (PMXVI16GER2S): Likewise.
1862 (PMXVI16GER2SPP): Likewise.
1863 (PMXVI4GER8): Likewise.
1864 (PMXVI4GER8PP): Likewise.
1865 (PMXVI8GER4): Likewise.
1866 (PMXVI8GER4PP): Likewise.
1867 (PMXVI8GER4SPP): Likewise.
1868 (XVBF16GER2): Likewise.
1869 (XVBF16GER2NN): Likewise.
1870 (XVBF16GER2NP): Likewise.
1871 (XVBF16GER2PN): Likewise.
1872 (XVBF16GER2PP): Likewise.
1873 (XVF16GER2): Likewise.
1874 (XVF16GER2NN): Likewise.
1875 (XVF16GER2NP): Likewise.
1876 (XVF16GER2PN): Likewise.
1877 (XVF16GER2PP): Likewise.
1878 (XVF32GER): Likewise.
1879 (XVF32GERNN): Likewise.
1880 (XVF32GERNP): Likewise.
1881 (XVF32GERPN): Likewise.
1882 (XVF32GERPP): Likewise.
1883 (XVF64GER): Likewise.
1884 (XVF64GERNN): Likewise.
1885 (XVF64GERNP): Likewise.
1886 (XVF64GERPN): Likewise.
1887 (XVF64GERPP): Likewise.
1888 (XVI16GER2): Likewise.
1889 (XVI16GER2PP): Likewise.
1890 (XVI16GER2S): Likewise.
1891 (XVI16GER2SPP): Likewise.
1892 (XVI4GER8): Likewise.
1893 (XVI4GER8PP): Likewise.
1894 (XVI8GER4): Likewise.
1895 (XVI8GER4PP): Likewise.
1896 (XVI8GER4SPP): Likewise.
1897 (XXMFACC): Likewise.
1898 (XXMTACC): Likewise.
1899 (XXSETACCZ): Likewise.
1900 (ASSEMBLE_PAIR_V): Likewise.
1901 (BUILD_PAIR): Likewise.
1902 (DISASSEMBLE_PAIR_V): Likewise.
1905 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_mma_builtin):
1906 Handle RS6000_BIF_LXVP and RS6000_BIF_STXVP.
1907 * config/rs6000/rs6000-gen-builtins.c (attrinfo): Add ismmaint.
1908 (parse_bif_attrs): Handle ismmaint.
1909 (write_decls): Add bif_mmaint_bit and bif_is_mmaint.
1910 (write_bif_static_init): Handle ismmaint.
1912 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
1914 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_builtin): New
1916 (rs6000_gimple_fold_builtin): Call rs6000_gimple_fold_new_builtin.
1917 (rs6000_new_builtin_valid_without_lhs): New function.
1918 (rs6000_gimple_fold_new_mma_builtin): Likewise.
1919 (rs6000_gimple_fold_new_builtin): Likewise.
1921 2021-09-17 Thomas Schwinge <thomas@codesourcery.com>
1923 * hash-table.h (hash_table<Descriptor, Lazy, Allocator>::expand):
1924 Destruct stale Value objects.
1925 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
1928 2021-09-17 Roger Sayle <roger@nextmovesoftware.com>
1931 * match.pd (shift optimizations): Disable recent sign-changing
1932 optimization for shifts by zero, these will be folded later.
1934 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
1936 * config/rs6000/rs6000-builtin-new.def (__builtin_mffsl): Move from
1937 [power9] to [always].
1939 2021-09-17 Richard Biener <rguenther@suse.de>
1941 * tree-vect-stmts.c (vectorizable_load): Do not frob
1944 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
1946 * config/i386/i386-features.c (remove_partial_avx_dependency):
1947 Also check TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY and
1948 and TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY before generating
1950 * config/i386/i386.h (TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY):
1952 (TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
1953 * config/i386/i386.md (SSE FP to FP splitters): Replace
1954 TARGET_SSE_PARTIAL_REG_DEPENDENCY with
1955 TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY.
1956 (SSE INT to FP splitter): Replace TARGET_SSE_PARTIAL_REG_DEPENDENCY
1957 with TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY.
1958 * config/i386/x86-tune.def
1959 (X86_TUNE_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY): New.
1960 (X86_TUNE_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
1962 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
1965 * config/i386/i386-features.c (remove_partial_avx_dependency):
1966 Check TARGET_USE_VECTOR_FP_CONVERTS and TARGET_USE_VECTOR_CONVERTS
1967 before generating vxorps.
1969 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
1971 * config/i386/i386-options.c (processor_cost_table): Use
1972 tremont_cost for Tremont.
1973 * config/i386/x86-tune-costs.h (tremont_memcpy): New.
1974 (tremont_memset): Likewise.
1975 (tremont_cost): Likewise.
1976 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
1979 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
1981 * common/config/i386/i386-common.c: Use Haswell scheduling model
1983 * config/i386/i386.c (ix86_sched_init_global): Prepare for Tremont
1985 * config/i386/x86-tune-sched.c (ix86_issue_rate): Change Tremont
1987 (ix86_adjust_cost): Handle Tremont.
1988 * config/i386/x86-tune.def (X86_TUNE_SSE_PARTIAL_REG_DEPENDENCY):
1990 (X86_TUNE_USE_LEAVE): Likewise.
1991 (X86_TUNE_PUSH_MEMORY): Likewise.
1992 (X86_TUNE_MISALIGNED_MOVE_STRING_PRO_EPILOGUES): Likewise.
1993 (X86_TUNE_USE_CLTD): Likewise.
1994 (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI): Likewise.
1995 (X86_TUNE_AVOID_MFENCE): Likewise.
1996 (X86_TUNE_SSE_TYPELESS_STORES): Likewise.
1997 (X86_TUNE_SSE_LOAD0_BY_PXOR): Likewise.
1998 (X86_TUNE_ACCUMULATE_OUTGOING_ARGS): Disable for Tremont.
1999 (X86_TUNE_FOUR_JUMP_LIMIT): Likewise.
2000 (X86_TUNE_OPT_AGU): Likewise.
2001 (X86_TUNE_AVOID_LEA_FOR_ADDR): Likewise.
2002 (X86_TUNE_AVOID_MEM_OPND_FOR_CMOVE): Likewise.
2003 (X86_TUNE_EXPAND_ABS): Likewise.
2004 (X86_TUNE_SPLIT_MEM_OPND_FOR_FP_CONVERTS): Likewise.
2005 (X86_TUNE_SLOW_PSHUFB): Likewise.
2007 2021-09-17 Eric Botcazou <ebotcazou@adacore.com>
2009 PR rtl-optimization/102306
2010 * combine.c (try_combine): Abort the combination if we are about to
2011 duplicate volatile references.
2013 2021-09-17 liuhongt <hongtao.liu@intel.com>
2015 * config/i386/avx512fp16intrin.h (_mm_undefined_ph):
2017 (_mm256_undefined_ph): Likewise.
2018 (_mm512_undefined_ph): Likewise.
2019 (_mm_cvtsh_h): Likewise.
2020 (_mm256_cvtsh_h): Likewise.
2021 (_mm512_cvtsh_h): Likewise.
2022 (_mm512_castph_ps): Likewise.
2023 (_mm512_castph_pd): Likewise.
2024 (_mm512_castph_si512): Likewise.
2025 (_mm512_castph512_ph128): Likewise.
2026 (_mm512_castph512_ph256): Likewise.
2027 (_mm512_castph128_ph512): Likewise.
2028 (_mm512_castph256_ph512): Likewise.
2029 (_mm512_zextph128_ph512): Likewise.
2030 (_mm512_zextph256_ph512): Likewise.
2031 (_mm512_castps_ph): Likewise.
2032 (_mm512_castpd_ph): Likewise.
2033 (_mm512_castsi512_ph): Likewise.
2034 * config/i386/avx512fp16vlintrin.h (_mm_castph_ps):
2036 (_mm256_castph_ps): Likewise.
2037 (_mm_castph_pd): Likewise.
2038 (_mm256_castph_pd): Likewise.
2039 (_mm_castph_si128): Likewise.
2040 (_mm256_castph_si256): Likewise.
2041 (_mm_castps_ph): Likewise.
2042 (_mm256_castps_ph): Likewise.
2043 (_mm_castpd_ph): Likewise.
2044 (_mm256_castpd_ph): Likewise.
2045 (_mm_castsi128_ph): Likewise.
2046 (_mm256_castsi256_ph): Likewise.
2047 (_mm256_castph256_ph128): Likewise.
2048 (_mm256_castph128_ph256): Likewise.
2049 (_mm256_zextph128_ph256): Likewise.
2051 2021-09-17 liuhongt <hongtao.liu@intel.com>
2053 * config/i386/avx512fp16intrin.h (_mm_cvtsh_ss):
2055 (_mm_mask_cvtsh_ss): Likewise.
2056 (_mm_maskz_cvtsh_ss): Likewise.
2057 (_mm_cvtsh_sd): Likewise.
2058 (_mm_mask_cvtsh_sd): Likewise.
2059 (_mm_maskz_cvtsh_sd): Likewise.
2060 (_mm_cvt_roundsh_ss): Likewise.
2061 (_mm_mask_cvt_roundsh_ss): Likewise.
2062 (_mm_maskz_cvt_roundsh_ss): Likewise.
2063 (_mm_cvt_roundsh_sd): Likewise.
2064 (_mm_mask_cvt_roundsh_sd): Likewise.
2065 (_mm_maskz_cvt_roundsh_sd): Likewise.
2066 (_mm_cvtss_sh): Likewise.
2067 (_mm_mask_cvtss_sh): Likewise.
2068 (_mm_maskz_cvtss_sh): Likewise.
2069 (_mm_cvtsd_sh): Likewise.
2070 (_mm_mask_cvtsd_sh): Likewise.
2071 (_mm_maskz_cvtsd_sh): Likewise.
2072 (_mm_cvt_roundss_sh): Likewise.
2073 (_mm_mask_cvt_roundss_sh): Likewise.
2074 (_mm_maskz_cvt_roundss_sh): Likewise.
2075 (_mm_cvt_roundsd_sh): Likewise.
2076 (_mm_mask_cvt_roundsd_sh): Likewise.
2077 (_mm_maskz_cvt_roundsd_sh): Likewise.
2078 * config/i386/i386-builtin-types.def
2079 (V8HF_FTYPE_V2DF_V8HF_V8HF_UQI_INT,
2080 V8HF_FTYPE_V4SF_V8HF_V8HF_UQI_INT,
2081 V2DF_FTYPE_V8HF_V2DF_V2DF_UQI_INT,
2082 V4SF_FTYPE_V8HF_V4SF_V4SF_UQI_INT): Add new builtin types.
2083 * config/i386/i386-builtin.def: Add corrresponding new builtins.
2084 * config/i386/i386-expand.c: Handle new builtin types.
2085 * config/i386/sse.md (VF48_128): New mode iterator.
2086 (avx512fp16_vcvtsh2<ssescalarmodesuffix><mask_scalar_name><round_saeonly_scalar_name>):
2088 (avx512fp16_vcvt<ssescalarmodesuffix>2sh<mask_scalar_name><round_scalar_name>):
2091 2021-09-17 liuhongt <hongtao.liu@intel.com>
2093 * config/i386/avx512fp16intrin.h (_mm512_cvtph_pd):
2095 (_mm512_mask_cvtph_pd): Likewise.
2096 (_mm512_maskz_cvtph_pd): Likewise.
2097 (_mm512_cvt_roundph_pd): Likewise.
2098 (_mm512_mask_cvt_roundph_pd): Likewise.
2099 (_mm512_maskz_cvt_roundph_pd): Likewise.
2100 (_mm512_cvtxph_ps): Likewise.
2101 (_mm512_mask_cvtxph_ps): Likewise.
2102 (_mm512_maskz_cvtxph_ps): Likewise.
2103 (_mm512_cvtx_roundph_ps): Likewise.
2104 (_mm512_mask_cvtx_roundph_ps): Likewise.
2105 (_mm512_maskz_cvtx_roundph_ps): Likewise.
2106 (_mm512_cvtxps_ph): Likewise.
2107 (_mm512_mask_cvtxps_ph): Likewise.
2108 (_mm512_maskz_cvtxps_ph): Likewise.
2109 (_mm512_cvtx_roundps_ph): Likewise.
2110 (_mm512_mask_cvtx_roundps_ph): Likewise.
2111 (_mm512_maskz_cvtx_roundps_ph): Likewise.
2112 (_mm512_cvtpd_ph): Likewise.
2113 (_mm512_mask_cvtpd_ph): Likewise.
2114 (_mm512_maskz_cvtpd_ph): Likewise.
2115 (_mm512_cvt_roundpd_ph): Likewise.
2116 (_mm512_mask_cvt_roundpd_ph): Likewise.
2117 (_mm512_maskz_cvt_roundpd_ph): Likewise.
2118 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_pd):
2120 (_mm_mask_cvtph_pd): Likewise.
2121 (_mm_maskz_cvtph_pd): Likewise.
2122 (_mm256_cvtph_pd): Likewise.
2123 (_mm256_mask_cvtph_pd): Likewise.
2124 (_mm256_maskz_cvtph_pd): Likewise.
2125 (_mm_cvtxph_ps): Likewise.
2126 (_mm_mask_cvtxph_ps): Likewise.
2127 (_mm_maskz_cvtxph_ps): Likewise.
2128 (_mm256_cvtxph_ps): Likewise.
2129 (_mm256_mask_cvtxph_ps): Likewise.
2130 (_mm256_maskz_cvtxph_ps): Likewise.
2131 (_mm_cvtxps_ph): Likewise.
2132 (_mm_mask_cvtxps_ph): Likewise.
2133 (_mm_maskz_cvtxps_ph): Likewise.
2134 (_mm256_cvtxps_ph): Likewise.
2135 (_mm256_mask_cvtxps_ph): Likewise.
2136 (_mm256_maskz_cvtxps_ph): Likewise.
2137 (_mm_cvtpd_ph): Likewise.
2138 (_mm_mask_cvtpd_ph): Likewise.
2139 (_mm_maskz_cvtpd_ph): Likewise.
2140 (_mm256_cvtpd_ph): Likewise.
2141 (_mm256_mask_cvtpd_ph): Likewise.
2142 (_mm256_maskz_cvtpd_ph): Likewise.
2143 * config/i386/i386-builtin.def: Add corresponding new builtins.
2144 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2145 * config/i386/i386-expand.c: Handle new builtin types.
2146 * config/i386/sse.md
2147 (VF4_128_8_256): New.
2148 (VF48H_AVX512VL): Ditto.
2149 (ssePHmode): Add HF vector modes.
2150 (castmode): Add new convertable modes.
2153 (avx512fp16_vcvt<castmode>2ph_<mode><mask_name><round_name>): Ditto.
2154 (avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
2155 (*avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
2156 (avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
2157 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
2158 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask_1): Ditto.
2159 (avx512fp16_float_extend_ph<mode>2<mask_name><round_saeonly_name>):
2161 (avx512fp16_float_extend_ph<mode>2<mask_name>): Ditto.
2162 (*avx512fp16_float_extend_ph<mode>2_load<mask_name>): Ditto.
2163 (avx512fp16_float_extend_phv2df2<mask_name>): Ditto.
2164 (*avx512fp16_float_extend_phv2df2_load<mask_name>): Ditto.
2166 2021-09-17 liuhongt <hongtao.liu@intel.com>
2168 * config/i386/avx512fp16intrin.h (_mm_cvttsh_i32):
2170 (_mm_cvttsh_u32): Likewise.
2171 (_mm_cvtt_roundsh_i32): Likewise.
2172 (_mm_cvtt_roundsh_u32): Likewise.
2173 (_mm_cvttsh_i64): Likewise.
2174 (_mm_cvttsh_u64): Likewise.
2175 (_mm_cvtt_roundsh_i64): Likewise.
2176 (_mm_cvtt_roundsh_u64): Likewise.
2177 * config/i386/i386-builtin.def: Add corresponding new builtins.
2178 * config/i386/sse.md
2179 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<round_saeonly_name>):
2182 2021-09-17 liuhongt <hongtao.liu@intel.com>
2184 * config/i386/avx512fp16intrin.h (_mm512_cvttph_epi32):
2186 (_mm512_mask_cvttph_epi32): Likewise.
2187 (_mm512_maskz_cvttph_epi32): Likewise.
2188 (_mm512_cvtt_roundph_epi32): Likewise.
2189 (_mm512_mask_cvtt_roundph_epi32): Likewise.
2190 (_mm512_maskz_cvtt_roundph_epi32): Likewise.
2191 (_mm512_cvttph_epu32): Likewise.
2192 (_mm512_mask_cvttph_epu32): Likewise.
2193 (_mm512_maskz_cvttph_epu32): Likewise.
2194 (_mm512_cvtt_roundph_epu32): Likewise.
2195 (_mm512_mask_cvtt_roundph_epu32): Likewise.
2196 (_mm512_maskz_cvtt_roundph_epu32): Likewise.
2197 (_mm512_cvttph_epi64): Likewise.
2198 (_mm512_mask_cvttph_epi64): Likewise.
2199 (_mm512_maskz_cvttph_epi64): Likewise.
2200 (_mm512_cvtt_roundph_epi64): Likewise.
2201 (_mm512_mask_cvtt_roundph_epi64): Likewise.
2202 (_mm512_maskz_cvtt_roundph_epi64): Likewise.
2203 (_mm512_cvttph_epu64): Likewise.
2204 (_mm512_mask_cvttph_epu64): Likewise.
2205 (_mm512_maskz_cvttph_epu64): Likewise.
2206 (_mm512_cvtt_roundph_epu64): Likewise.
2207 (_mm512_mask_cvtt_roundph_epu64): Likewise.
2208 (_mm512_maskz_cvtt_roundph_epu64): Likewise.
2209 (_mm512_cvttph_epi16): Likewise.
2210 (_mm512_mask_cvttph_epi16): Likewise.
2211 (_mm512_maskz_cvttph_epi16): Likewise.
2212 (_mm512_cvtt_roundph_epi16): Likewise.
2213 (_mm512_mask_cvtt_roundph_epi16): Likewise.
2214 (_mm512_maskz_cvtt_roundph_epi16): Likewise.
2215 (_mm512_cvttph_epu16): Likewise.
2216 (_mm512_mask_cvttph_epu16): Likewise.
2217 (_mm512_maskz_cvttph_epu16): Likewise.
2218 (_mm512_cvtt_roundph_epu16): Likewise.
2219 (_mm512_mask_cvtt_roundph_epu16): Likewise.
2220 (_mm512_maskz_cvtt_roundph_epu16): Likewise.
2221 * config/i386/avx512fp16vlintrin.h (_mm_cvttph_epi32):
2223 (_mm_mask_cvttph_epi32): Likewise.
2224 (_mm_maskz_cvttph_epi32): Likewise.
2225 (_mm256_cvttph_epi32): Likewise.
2226 (_mm256_mask_cvttph_epi32): Likewise.
2227 (_mm256_maskz_cvttph_epi32): Likewise.
2228 (_mm_cvttph_epu32): Likewise.
2229 (_mm_mask_cvttph_epu32): Likewise.
2230 (_mm_maskz_cvttph_epu32): Likewise.
2231 (_mm256_cvttph_epu32): Likewise.
2232 (_mm256_mask_cvttph_epu32): Likewise.
2233 (_mm256_maskz_cvttph_epu32): Likewise.
2234 (_mm_cvttph_epi64): Likewise.
2235 (_mm_mask_cvttph_epi64): Likewise.
2236 (_mm_maskz_cvttph_epi64): Likewise.
2237 (_mm256_cvttph_epi64): Likewise.
2238 (_mm256_mask_cvttph_epi64): Likewise.
2239 (_mm256_maskz_cvttph_epi64): Likewise.
2240 (_mm_cvttph_epu64): Likewise.
2241 (_mm_mask_cvttph_epu64): Likewise.
2242 (_mm_maskz_cvttph_epu64): Likewise.
2243 (_mm256_cvttph_epu64): Likewise.
2244 (_mm256_mask_cvttph_epu64): Likewise.
2245 (_mm256_maskz_cvttph_epu64): Likewise.
2246 (_mm_cvttph_epi16): Likewise.
2247 (_mm_mask_cvttph_epi16): Likewise.
2248 (_mm_maskz_cvttph_epi16): Likewise.
2249 (_mm256_cvttph_epi16): Likewise.
2250 (_mm256_mask_cvttph_epi16): Likewise.
2251 (_mm256_maskz_cvttph_epi16): Likewise.
2252 (_mm_cvttph_epu16): Likewise.
2253 (_mm_mask_cvttph_epu16): Likewise.
2254 (_mm_maskz_cvttph_epu16): Likewise.
2255 (_mm256_cvttph_epu16): Likewise.
2256 (_mm256_mask_cvttph_epu16): Likewise.
2257 (_mm256_maskz_cvttph_epu16): Likewise.
2258 * config/i386/i386-builtin.def: Add new builtins.
2259 * config/i386/sse.md
2260 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name><round_saeonly_name>):
2262 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name>): Ditto.
2263 (*avx512fp16_fix<fixunssuffix>_trunc<mode>2_load<mask_name>): Ditto.
2264 (avx512fp16_fix<fixunssuffix>_truncv2di2<mask_name>): Ditto.
2265 (avx512fp16_fix<fixunssuffix>_truncv2di2_load<mask_name>): Ditto.
2267 2021-09-17 liuhongt <hongtao.liu@intel.com>
2269 * config/i386/avx512fp16intrin.h (_mm_cvtsh_i32): New intrinsic.
2270 (_mm_cvtsh_u32): Likewise.
2271 (_mm_cvt_roundsh_i32): Likewise.
2272 (_mm_cvt_roundsh_u32): Likewise.
2273 (_mm_cvtsh_i64): Likewise.
2274 (_mm_cvtsh_u64): Likewise.
2275 (_mm_cvt_roundsh_i64): Likewise.
2276 (_mm_cvt_roundsh_u64): Likewise.
2277 (_mm_cvti32_sh): Likewise.
2278 (_mm_cvtu32_sh): Likewise.
2279 (_mm_cvt_roundi32_sh): Likewise.
2280 (_mm_cvt_roundu32_sh): Likewise.
2281 (_mm_cvti64_sh): Likewise.
2282 (_mm_cvtu64_sh): Likewise.
2283 (_mm_cvt_roundi64_sh): Likewise.
2284 (_mm_cvt_roundu64_sh): Likewise.
2285 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2286 * config/i386/i386-builtin.def: Add corresponding new builtins.
2287 * config/i386/i386-expand.c (ix86_expand_round_builtin):
2288 Handle new builtin types.
2289 * config/i386/sse.md
2290 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix><round_name>):
2292 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix>_2): Likewise.
2293 (avx512fp16_vcvt<floatsuffix>si2sh<rex64namesuffix><round_name>): Likewise.
2295 2021-09-16 Bill Schmidt <wschmidt@linux.ibm.com>
2297 * config/rs6000/rs6000-c.c (rs6000-builtins.h): New include.
2298 (altivec_resolve_new_overloaded_builtin): New forward decl.
2299 (rs6000_new_builtin_type_compatible): New function.
2300 (altivec_resolve_overloaded_builtin): Call
2301 altivec_resolve_new_overloaded_builtin.
2302 (altivec_build_new_resolved_builtin): New function.
2303 (altivec_resolve_new_overloaded_builtin): Likewise.
2304 * config/rs6000/rs6000-call.c (rs6000_new_builtin_is_supported):
2306 * config/rs6000/rs6000-gen-builtins.c (write_decls): Remove _p from
2307 name of rs6000_new_builtin_is_supported.
2309 2021-09-16 Uroš Bizjak <ubizjak@gmail.com>
2311 * config/i386/i386-protos.h (ix86_decompose_address):
2312 Change return type to bool.
2313 * config/i386/i386.c (ix86_decompose_address): Ditto.
2315 2021-09-16 Tobias Burnus <tobias@codesourcery.com>
2318 * config/rs6000/t-rs6000 (build/rs6000-gen-builtins.o, build/rbtree.o):
2319 Added 'build/' to target, use build/%.o rule.
2320 (build/rs6000-gen-builtins$(build_exeext)): Add 'build/' and
2321 '$(build_exeext)' to target and 'build/' for the *.o files.
2322 (rs6000-builtins.c): Update for those changes; run rs6000-gen-builtins
2325 2021-09-16 Martin Jambor <mjambor@suse.cz>
2327 * cgraph.c (cgraph_node::dump): Do not check caller count sums if
2328 the body has been removed. Remove trailing whitespace.
2330 2021-09-16 Richard Biener <rguenther@suse.de>
2332 PR middle-end/102360
2333 * internal-fn.c (expand_DEFERRED_INIT): Make pattern-init
2334 of non-memory more robust.
2336 2021-09-16 Daniel Cederman <cederman@gaisler.com>
2338 * config/sparc/sparc-opts.h (enum sparc_processor_type): Add LEON5
2339 * config/sparc/sparc.c (struct processor_costs): Add LEON5 costs
2340 (leon5_adjust_cost): Increase cost of store with data dependency
2341 on ALU instruction and FPU anti-dependencies.
2342 (sparc_option_override): Add LEON5 costs
2343 (sparc_adjust_cost): Add LEON5 cost adjustments
2344 * config/sparc/sparc.h: Add LEON5
2345 * config/sparc/sparc.md: Include LEON5 scheduling information
2346 * config/sparc/sparc.opt: Add LEON5
2347 * doc/invoke.texi: Add LEON5
2348 * config/sparc/leon5.md: New file.
2350 2021-09-16 Daniel Cederman <cederman@gaisler.com>
2352 * config/sparc/sparc.md (stack_protect_set32): Add NOP to prevent
2353 sensitive sequence for B2BST errata workaround.
2355 2021-09-16 Daniel Cederman <cederman@gaisler.com>
2357 * config/sparc/sparc.c (sparc_do_work_around_errata): Do not begin
2358 functions with atomic instruction in the UT700 errata workaround.
2360 2021-09-16 Daniel Cederman <cederman@gaisler.com>
2362 * config/sparc/sparc.c (next_active_non_empty_insn): New function
2363 that returns next active non empty assembly instruction.
2364 (sparc_do_work_around_errata): Use new function.
2366 2021-09-16 Daniel Cederman <cederman@gaisler.com>
2368 * config/sparc/sparc.c (store_insn_p): Add predicate for store
2370 (load_insn_p): Add predicate for load attributes.
2371 (sparc_do_work_around_errata): Use new predicates.
2373 2021-09-16 Andreas Larsson <andreas@gaisler.com>
2375 * config/sparc/sparc.c (dump_target_flag_bits): Print bit names for
2378 2021-09-16 Martin Liska <mliska@suse.cz>
2380 * config/mips/netbsd.h: Fix typo in name of a macro.
2382 2021-09-16 liuhongt <hongtao.liu@intel.com>
2384 PR middle-end/102080
2385 * match.pd: Check mask type when doing cond_op related gimple
2387 * tree.c (is_truth_type_for): New function.
2388 * tree.h (is_truth_type_for): New declaration.
2390 2021-09-16 liuhongt <hongtao.liu@intel.com>
2392 * config/i386/avx512fp16intrin.h (_mm512_cvtepi32_ph): New
2394 (_mm512_mask_cvtepi32_ph): Likewise.
2395 (_mm512_maskz_cvtepi32_ph): Likewise.
2396 (_mm512_cvt_roundepi32_ph): Likewise.
2397 (_mm512_mask_cvt_roundepi32_ph): Likewise.
2398 (_mm512_maskz_cvt_roundepi32_ph): Likewise.
2399 (_mm512_cvtepu32_ph): Likewise.
2400 (_mm512_mask_cvtepu32_ph): Likewise.
2401 (_mm512_maskz_cvtepu32_ph): Likewise.
2402 (_mm512_cvt_roundepu32_ph): Likewise.
2403 (_mm512_mask_cvt_roundepu32_ph): Likewise.
2404 (_mm512_maskz_cvt_roundepu32_ph): Likewise.
2405 (_mm512_cvtepi64_ph): Likewise.
2406 (_mm512_mask_cvtepi64_ph): Likewise.
2407 (_mm512_maskz_cvtepi64_ph): Likewise.
2408 (_mm512_cvt_roundepi64_ph): Likewise.
2409 (_mm512_mask_cvt_roundepi64_ph): Likewise.
2410 (_mm512_maskz_cvt_roundepi64_ph): Likewise.
2411 (_mm512_cvtepu64_ph): Likewise.
2412 (_mm512_mask_cvtepu64_ph): Likewise.
2413 (_mm512_maskz_cvtepu64_ph): Likewise.
2414 (_mm512_cvt_roundepu64_ph): Likewise.
2415 (_mm512_mask_cvt_roundepu64_ph): Likewise.
2416 (_mm512_maskz_cvt_roundepu64_ph): Likewise.
2417 (_mm512_cvtepi16_ph): Likewise.
2418 (_mm512_mask_cvtepi16_ph): Likewise.
2419 (_mm512_maskz_cvtepi16_ph): Likewise.
2420 (_mm512_cvt_roundepi16_ph): Likewise.
2421 (_mm512_mask_cvt_roundepi16_ph): Likewise.
2422 (_mm512_maskz_cvt_roundepi16_ph): Likewise.
2423 (_mm512_cvtepu16_ph): Likewise.
2424 (_mm512_mask_cvtepu16_ph): Likewise.
2425 (_mm512_maskz_cvtepu16_ph): Likewise.
2426 (_mm512_cvt_roundepu16_ph): Likewise.
2427 (_mm512_mask_cvt_roundepu16_ph): Likewise.
2428 (_mm512_maskz_cvt_roundepu16_ph): Likewise.
2429 * config/i386/avx512fp16vlintrin.h (_mm_cvtepi32_ph): New
2431 (_mm_mask_cvtepi32_ph): Likewise.
2432 (_mm_maskz_cvtepi32_ph): Likewise.
2433 (_mm256_cvtepi32_ph): Likewise.
2434 (_mm256_mask_cvtepi32_ph): Likewise.
2435 (_mm256_maskz_cvtepi32_ph): Likewise.
2436 (_mm_cvtepu32_ph): Likewise.
2437 (_mm_mask_cvtepu32_ph): Likewise.
2438 (_mm_maskz_cvtepu32_ph): Likewise.
2439 (_mm256_cvtepu32_ph): Likewise.
2440 (_mm256_mask_cvtepu32_ph): Likewise.
2441 (_mm256_maskz_cvtepu32_ph): Likewise.
2442 (_mm_cvtepi64_ph): Likewise.
2443 (_mm_mask_cvtepi64_ph): Likewise.
2444 (_mm_maskz_cvtepi64_ph): Likewise.
2445 (_mm256_cvtepi64_ph): Likewise.
2446 (_mm256_mask_cvtepi64_ph): Likewise.
2447 (_mm256_maskz_cvtepi64_ph): Likewise.
2448 (_mm_cvtepu64_ph): Likewise.
2449 (_mm_mask_cvtepu64_ph): Likewise.
2450 (_mm_maskz_cvtepu64_ph): Likewise.
2451 (_mm256_cvtepu64_ph): Likewise.
2452 (_mm256_mask_cvtepu64_ph): Likewise.
2453 (_mm256_maskz_cvtepu64_ph): Likewise.
2454 (_mm_cvtepi16_ph): Likewise.
2455 (_mm_mask_cvtepi16_ph): Likewise.
2456 (_mm_maskz_cvtepi16_ph): Likewise.
2457 (_mm256_cvtepi16_ph): Likewise.
2458 (_mm256_mask_cvtepi16_ph): Likewise.
2459 (_mm256_maskz_cvtepi16_ph): Likewise.
2460 (_mm_cvtepu16_ph): Likewise.
2461 (_mm_mask_cvtepu16_ph): Likewise.
2462 (_mm_maskz_cvtepu16_ph): Likewise.
2463 (_mm256_cvtepu16_ph): Likewise.
2464 (_mm256_mask_cvtepu16_ph): Likewise.
2465 (_mm256_maskz_cvtepu16_ph): Likewise.
2466 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2467 * config/i386/i386-builtin.def: Add corresponding new builtins.
2468 * config/i386/i386-expand.c
2469 (ix86_expand_args_builtin): Handle new builtin types.
2470 (ix86_expand_round_builtin): Ditto.
2471 * config/i386/i386-modes.def: Declare V2HF and V6HF.
2472 * config/i386/sse.md (VI2H_AVX512VL): New.
2474 (sseintvecmode): Add HF vector modes.
2475 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode><mask_name><round_name>):
2477 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
2478 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
2479 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
2480 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
2481 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask_1): Ditto.
2482 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
2483 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
2484 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
2485 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
2486 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask_1): Ditto.
2487 * config/i386/subst.md (round_qq2phsuff): New subst_attr.
2489 2021-09-16 liuhongt <hongtao.liu@intel.com>
2491 * config/i386/avx512fp16intrin.h (_mm512_cvtph_epi32):
2493 (_mm512_mask_cvtph_epi32): Likewise.
2494 (_mm512_maskz_cvtph_epi32): Likewise.
2495 (_mm512_cvt_roundph_epi32): Likewise.
2496 (_mm512_mask_cvt_roundph_epi32): Likewise.
2497 (_mm512_maskz_cvt_roundph_epi32): Likewise.
2498 (_mm512_cvtph_epu32): Likewise.
2499 (_mm512_mask_cvtph_epu32): Likewise.
2500 (_mm512_maskz_cvtph_epu32): Likewise.
2501 (_mm512_cvt_roundph_epu32): Likewise.
2502 (_mm512_mask_cvt_roundph_epu32): Likewise.
2503 (_mm512_maskz_cvt_roundph_epu32): Likewise.
2504 (_mm512_cvtph_epi64): Likewise.
2505 (_mm512_mask_cvtph_epi64): Likewise.
2506 (_mm512_maskz_cvtph_epi64): Likewise.
2507 (_mm512_cvt_roundph_epi64): Likewise.
2508 (_mm512_mask_cvt_roundph_epi64): Likewise.
2509 (_mm512_maskz_cvt_roundph_epi64): Likewise.
2510 (_mm512_cvtph_epu64): Likewise.
2511 (_mm512_mask_cvtph_epu64): Likewise.
2512 (_mm512_maskz_cvtph_epu64): Likewise.
2513 (_mm512_cvt_roundph_epu64): Likewise.
2514 (_mm512_mask_cvt_roundph_epu64): Likewise.
2515 (_mm512_maskz_cvt_roundph_epu64): Likewise.
2516 (_mm512_cvtph_epi16): Likewise.
2517 (_mm512_mask_cvtph_epi16): Likewise.
2518 (_mm512_maskz_cvtph_epi16): Likewise.
2519 (_mm512_cvt_roundph_epi16): Likewise.
2520 (_mm512_mask_cvt_roundph_epi16): Likewise.
2521 (_mm512_maskz_cvt_roundph_epi16): Likewise.
2522 (_mm512_cvtph_epu16): Likewise.
2523 (_mm512_mask_cvtph_epu16): Likewise.
2524 (_mm512_maskz_cvtph_epu16): Likewise.
2525 (_mm512_cvt_roundph_epu16): Likewise.
2526 (_mm512_mask_cvt_roundph_epu16): Likewise.
2527 (_mm512_maskz_cvt_roundph_epu16): Likewise.
2528 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_epi32):
2530 (_mm_mask_cvtph_epi32): Likewise.
2531 (_mm_maskz_cvtph_epi32): Likewise.
2532 (_mm256_cvtph_epi32): Likewise.
2533 (_mm256_mask_cvtph_epi32): Likewise.
2534 (_mm256_maskz_cvtph_epi32): Likewise.
2535 (_mm_cvtph_epu32): Likewise.
2536 (_mm_mask_cvtph_epu32): Likewise.
2537 (_mm_maskz_cvtph_epu32): Likewise.
2538 (_mm256_cvtph_epu32): Likewise.
2539 (_mm256_mask_cvtph_epu32): Likewise.
2540 (_mm256_maskz_cvtph_epu32): Likewise.
2541 (_mm_cvtph_epi64): Likewise.
2542 (_mm_mask_cvtph_epi64): Likewise.
2543 (_mm_maskz_cvtph_epi64): Likewise.
2544 (_mm256_cvtph_epi64): Likewise.
2545 (_mm256_mask_cvtph_epi64): Likewise.
2546 (_mm256_maskz_cvtph_epi64): Likewise.
2547 (_mm_cvtph_epu64): Likewise.
2548 (_mm_mask_cvtph_epu64): Likewise.
2549 (_mm_maskz_cvtph_epu64): Likewise.
2550 (_mm256_cvtph_epu64): Likewise.
2551 (_mm256_mask_cvtph_epu64): Likewise.
2552 (_mm256_maskz_cvtph_epu64): Likewise.
2553 (_mm_cvtph_epi16): Likewise.
2554 (_mm_mask_cvtph_epi16): Likewise.
2555 (_mm_maskz_cvtph_epi16): Likewise.
2556 (_mm256_cvtph_epi16): Likewise.
2557 (_mm256_mask_cvtph_epi16): Likewise.
2558 (_mm256_maskz_cvtph_epi16): Likewise.
2559 (_mm_cvtph_epu16): Likewise.
2560 (_mm_mask_cvtph_epu16): Likewise.
2561 (_mm_maskz_cvtph_epu16): Likewise.
2562 (_mm256_cvtph_epu16): Likewise.
2563 (_mm256_mask_cvtph_epu16): Likewise.
2564 (_mm256_maskz_cvtph_epu16): Likewise.
2565 * config/i386/i386-builtin-types.def: Add new builtin types.
2566 * config/i386/i386-builtin.def: Add new builtins.
2567 * config/i386/i386-expand.c
2568 (ix86_expand_args_builtin): Handle new builtin types.
2569 (ix86_expand_round_builtin): Ditto.
2570 * config/i386/sse.md (sseintconvert): New.
2572 (UNSPEC_US_FIX_NOTRUNC): Ditto.
2573 (sseintconvertsignprefix): Ditto.
2574 (avx512fp16_vcvtph2<sseintconvertsignprefix><sseintconvert>_<mode><mask_name><round_name>):
2577 2021-09-16 liuhongt <hongtao.liu@intel.com>
2579 * config/i386/avx512fp16intrin.h: (_mm_cvtsi16_si128):
2581 (_mm_cvtsi128_si16): Likewise.
2582 (_mm_mask_load_sh): Likewise.
2583 (_mm_maskz_load_sh): Likewise.
2584 (_mm_mask_store_sh): Likewise.
2585 (_mm_move_sh): Likewise.
2586 (_mm_mask_move_sh): Likewise.
2587 (_mm_maskz_move_sh): Likewise.
2588 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2589 * config/i386/i386-builtin.def: Add corresponding new builtins.
2590 * config/i386/i386-expand.c
2591 (ix86_expand_special_args_builtin): Handle new builtin types.
2592 (ix86_expand_vector_init_one_nonzero): Adjust for FP16 target.
2593 * config/i386/sse.md (VI2F): New mode iterator.
2594 (vec_set<mode>_0): Use new mode iterator.
2595 (avx512f_mov<ssescalarmodelower>_mask): Adjust for HF vector mode.
2596 (avx512f_store<mode>_mask): Ditto.
2598 2021-09-16 Kewen Lin <linkw@linux.ibm.com>
2600 * config/rs6000/rs6000.opt (-mtoc-fusion): Remove.
2602 2021-09-15 David Edelsohn <dje.gcc@gmail.com>
2604 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
2605 Proceed if no symbol summary or the symbol alias flag is false.
2607 2021-09-15 Jakub Jelinek <jakub@redhat.com>
2611 * varasm.c (output_constructor_regular_field): Instead of assertion
2612 that array_size_for_constructor result is equal to size of
2613 TREE_TYPE (local->val) in bytes, assert that the type size is greater
2614 or equal to array_size_for_constructor result and use type size as
2617 2021-09-15 Martin Liska <mliska@suse.cz>
2620 * config/i386/vxworks.h: Use new macro TARGET_CPU_P.
2622 2021-09-15 Martin Liska <mliska@suse.cz>
2625 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
2626 Check that we have a symbol summary for a symbol.
2628 2021-09-15 Richard Biener <rguenther@suse.de>
2631 * config/rs6000/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
2632 to inherit from elfos.h
2634 2021-09-15 liuhongt <hongtao.liu@intel.com>
2637 * config/i386/i386-expand.c
2638 (ix86_expand_vector_init_interleave): Use puncklwd to pack 2
2640 (ix86_expand_vector_set): Use blendw instead of pinsrw.
2641 * config/i386/i386.c (ix86_can_change_mode_class): Adjust for
2642 AVX512FP16 which supports 16bit vector load.
2643 * config/i386/sse.md (avx512bw_interleave_highv32hi<mask_name>):
2645 (avx512bw_interleave_high<mode><mask_name>): .. this, and
2646 extend to V32HFmode.
2647 (avx2_interleave_highv16hi<mask_name>): Rename to ..
2648 (avx2_interleave_high<mode><mask_name>): .. this, and extend
2650 (vec_interleave_highv8hi<mask_name>): Rename to ..
2651 (vec_interleave_high<mode><mask_name>): .. this, and extend to V8HFmode.
2652 (<mask_codefor>avx512bw_interleave_lowv32hi<mask_name>):
2654 (<mask_codefor>avx512bw_interleave_low<mode><mask_name>):
2655 this, and extend to V32HFmode.
2656 (avx2_interleave_lowv16hi<mask_name>): Rename to ..
2657 (avx2_interleave_low<mode><mask_name>): .. this, and extend to V16HFmode.
2658 (vec_interleave_lowv8hi<mask_name>): Rename to ..
2659 (vec_interleave_low<mode><mask_name>): .. this, and extend to V8HFmode.
2660 (sse4_1_pblendw): Rename to ..
2661 (sse4_1_pblend<blendsuf>): .. this, and extend to V8HFmode.
2662 (avx2_pblendph): New define_expand.
2663 (<sse2p4_1>_pinsr<ssemodesuffix>): Refactor, use
2664 sseintmodesuffix instead of ssemodesuffix.
2665 (blendsuf): New mode attr.
2667 2021-09-15 Richard Biener <rguenther@suse.de>
2669 * tree-vectorizer.h (dr_misalignment): Move out of line.
2670 (dr_target_alignment): New.
2671 (DR_TARGET_ALIGNMENT): Wrap dr_target_alignment.
2672 (set_dr_target_alignment): New.
2673 (SET_DR_TARGET_ALIGNMENT): Wrap set_dr_target_alignment.
2674 * tree-vect-data-refs.c (dr_misalignment): Compute and
2675 return the group members misalignment.
2676 (vect_compute_data_ref_alignment): Use SET_DR_TARGET_ALIGNMENT.
2677 (vect_analyze_data_refs_alignment): Compute alignment only
2678 for the first element of a DR group.
2679 (vect_slp_analyze_node_alignment): Likewise.
2681 2021-09-15 Hongyu Wang <hongyu.wang@intel.com>
2683 * config/i386/avx512fp16intrin.h: Adjust all builtin calls.
2684 * config/i386/avx512fp16vlintrin.h: Likewise.
2685 * config/i386/i386-builtin.def: Adjust builtin name and
2686 enumeration to match AVX512F style.
2688 2021-09-15 Richard Biener <rguenther@suse.de>
2690 PR tree-optimization/102318
2691 * tree-vect-loop.c (vect_transform_cycle_phi): Revert
2692 previous change and do the mode conversion separately from
2693 the sign conversion.
2695 2021-09-15 Hongtao Liu <hongtao.liu@intel.com>
2696 Peter Cordes <peter@cordes.ca>
2699 * config/i386/sse.md (extract_suf): Add V8SF/V8SI/V4DF/V4DI.
2700 (*vec_extract<mode><ssescalarmodelower>_valign): Output
2701 vextract{i,f}{32x4,64x2} instruction when byte_offset % 16 ==
2704 2021-09-15 Richard Biener <rguenther@suse.de>
2706 * config.gcc: Remove vax-*-openbsd* configuration.
2708 2021-09-15 Richard Biener <rguenther@suse.de>
2710 * config.gcc: Remove m68k-openbsd.
2712 2021-09-15 Max Filippov <jcmvbkbc@gmail.com>
2715 * config/xtensa/t-xtensa (TM_H): Add include/xtensa-config.h.
2717 2021-09-14 Peter Bergner <bergner@linux.ibm.com>
2719 * config/rs6000/mma.md (unspec): Delete UNSPEC_MMA_XXSETACCZ.
2720 (unspecv): Add UNSPECV_MMA_XXSETACCZ.
2721 (*mma_xxsetaccz): Delete.
2722 (mma_xxsetaccz): Change to define_insn. Remove operand 1.
2723 Use UNSPECV_MMA_XXSETACCZ. Update comment.
2724 * config/rs6000/rs6000.c (rs6000_rtx_costs): Use UNSPECV_MMA_XXSETACCZ.
2726 2021-09-14 Iain Sandoe <iain@sandoe.co.uk>
2728 * Makefile.in: Remove variables related to applying no-PIE
2729 to the exes on $build.
2730 * configure: Regenerate.
2731 * configure.ac: Remove configuration related to applying
2732 no-PIE to the exes on $build.
2734 2021-09-14 Claudiu Zissulescu <claziss@synopsys.com>
2736 * config/arc/arc.md (doloop_end): Add missing mode.
2737 (loop_end): Likewise.
2739 2021-09-14 Jakub Jelinek <jakub@redhat.com>
2741 * gimplify.c (goa_stabilize_expr): Add depth argument, propagate
2742 it to recursive calls, for depth above 7 just gimplify or return.
2743 Perform a test even for MODIFY_EXPR, ADDR_EXPR, COMPOUND_EXPR with
2744 __builtin_clear_padding and TARGET_EXPR.
2745 (gimplify_omp_atomic): Adjust goa_stabilize_expr callers.
2747 2021-09-14 liuhongt <hongtao.liu@intel.com>
2749 * config/i386/avx512fp16intrin.h (_mm_fpclass_sh_mask):
2751 (_mm_mask_fpclass_sh_mask): Likewise.
2752 (_mm512_mask_fpclass_ph_mask): Likewise.
2753 (_mm512_fpclass_ph_mask): Likewise.
2754 (_mm_getexp_sh): Likewise.
2755 (_mm_mask_getexp_sh): Likewise.
2756 (_mm_maskz_getexp_sh): Likewise.
2757 (_mm512_getexp_ph): Likewise.
2758 (_mm512_mask_getexp_ph): Likewise.
2759 (_mm512_maskz_getexp_ph): Likewise.
2760 (_mm_getexp_round_sh): Likewise.
2761 (_mm_mask_getexp_round_sh): Likewise.
2762 (_mm_maskz_getexp_round_sh): Likewise.
2763 (_mm512_getexp_round_ph): Likewise.
2764 (_mm512_mask_getexp_round_ph): Likewise.
2765 (_mm512_maskz_getexp_round_ph): Likewise.
2766 (_mm_getmant_sh): Likewise.
2767 (_mm_mask_getmant_sh): Likewise.
2768 (_mm_maskz_getmant_sh): Likewise.
2769 (_mm512_getmant_ph): Likewise.
2770 (_mm512_mask_getmant_ph): Likewise.
2771 (_mm512_maskz_getmant_ph): Likewise.
2772 (_mm_getmant_round_sh): Likewise.
2773 (_mm_mask_getmant_round_sh): Likewise.
2774 (_mm_maskz_getmant_round_sh): Likewise.
2775 (_mm512_getmant_round_ph): Likewise.
2776 (_mm512_mask_getmant_round_ph): Likewise.
2777 (_mm512_maskz_getmant_round_ph): Likewise.
2778 * config/i386/avx512fp16vlintrin.h (_mm_mask_fpclass_ph_mask):
2780 (_mm_fpclass_ph_mask): Likewise.
2781 (_mm256_mask_fpclass_ph_mask): Likewise.
2782 (_mm256_fpclass_ph_mask): Likewise.
2783 (_mm256_getexp_ph): Likewise.
2784 (_mm256_mask_getexp_ph): Likewise.
2785 (_mm256_maskz_getexp_ph): Likewise.
2786 (_mm_getexp_ph): Likewise.
2787 (_mm_mask_getexp_ph): Likewise.
2788 (_mm_maskz_getexp_ph): Likewise.
2789 (_mm256_getmant_ph): Likewise.
2790 (_mm256_mask_getmant_ph): Likewise.
2791 (_mm256_maskz_getmant_ph): Likewise.
2792 (_mm_getmant_ph): Likewise.
2793 (_mm_mask_getmant_ph): Likewise.
2794 (_mm_maskz_getmant_ph): Likewise.
2795 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2796 * config/i386/i386-builtin.def: Add corresponding new builtins.
2797 * config/i386/i386-expand.c
2798 (ix86_expand_args_builtin): Handle new builtin types.
2799 (ix86_expand_round_builtin): Ditto.
2800 * config/i386/sse.md (vecmemsuffix): Add HF vector modes.
2801 (<avx512>_getexp<mode><mask_name><round_saeonly_name>): Adjust
2802 to support HF vector modes.
2803 (avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name):
2805 (avx512dq_fpclass<mode><mask_scalar_merge_name>): Ditto.
2806 (avx512dq_vmfpclass<mode><mask_scalar_merge_name>): Ditto.
2807 (<avx512>_getmant<mode><mask_name><round_saeonly_name>): Ditto.
2808 (avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2811 2021-09-14 liuhongt <hongtao.liu@intel.com>
2813 * config/i386/avx512fp16intrin.h (_mm512_reduce_ph):
2815 (_mm512_mask_reduce_ph): Likewise.
2816 (_mm512_maskz_reduce_ph): Likewise.
2817 (_mm512_reduce_round_ph): Likewise.
2818 (_mm512_mask_reduce_round_ph): Likewise.
2819 (_mm512_maskz_reduce_round_ph): Likewise.
2820 (_mm_reduce_sh): Likewise.
2821 (_mm_mask_reduce_sh): Likewise.
2822 (_mm_maskz_reduce_sh): Likewise.
2823 (_mm_reduce_round_sh): Likewise.
2824 (_mm_mask_reduce_round_sh): Likewise.
2825 (_mm_maskz_reduce_round_sh): Likewise.
2826 (_mm512_roundscale_ph): Likewise.
2827 (_mm512_mask_roundscale_ph): Likewise.
2828 (_mm512_maskz_roundscale_ph): Likewise.
2829 (_mm512_roundscale_round_ph): Likewise.
2830 (_mm512_mask_roundscale_round_ph): Likewise.
2831 (_mm512_maskz_roundscale_round_ph): Likewise.
2832 (_mm_roundscale_sh): Likewise.
2833 (_mm_mask_roundscale_sh): Likewise.
2834 (_mm_maskz_roundscale_sh): Likewise.
2835 (_mm_roundscale_round_sh): Likewise.
2836 (_mm_mask_roundscale_round_sh): Likewise.
2837 (_mm_maskz_roundscale_round_sh): Likewise.
2838 * config/i386/avx512fp16vlintrin.h: (_mm_reduce_ph):
2840 (_mm_mask_reduce_ph): Likewise.
2841 (_mm_maskz_reduce_ph): Likewise.
2842 (_mm256_reduce_ph): Likewise.
2843 (_mm256_mask_reduce_ph): Likewise.
2844 (_mm256_maskz_reduce_ph): Likewise.
2845 (_mm_roundscale_ph): Likewise.
2846 (_mm_mask_roundscale_ph): Likewise.
2847 (_mm_maskz_roundscale_ph): Likewise.
2848 (_mm256_roundscale_ph): Likewise.
2849 (_mm256_mask_roundscale_ph): Likewise.
2850 (_mm256_maskz_roundscale_ph): Likewise.
2851 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2852 * config/i386/i386-builtin.def: Add corresponding new builtins.
2853 * config/i386/i386-expand.c
2854 (ix86_expand_args_builtin): Handle new builtin types.
2855 (ix86_expand_round_builtin): Ditto.
2856 * config/i386/sse.md (<mask_codefor>reducep<mode><mask_name>):
2858 (<mask_codefor>reducep<mode><mask_name><round_saeonly_name>):
2859 ... this, and adjust for round operands.
2860 (reduces<mode><mask_scalar_name>): Likewise, with ...
2861 (reduces<mode><mask_scalar_name><round_saeonly_scalar_name):
2863 (<avx512>_rndscale<mode><mask_name><round_saeonly_name>):
2864 Adjust for HF vector modes.
2865 (avx512f_rndscale<mode><mask_scalar_name><round_saeonly_scalar_name>):
2867 (*avx512f_rndscale<mode><round_saeonly_name>): Ditto.
2869 2021-09-14 liuhongt <hongtao.liu@intel.com>
2871 * config/i386/avx512fp16intrin.h: (_mm512_rcp_ph):
2873 (_mm512_mask_rcp_ph): Likewise.
2874 (_mm512_maskz_rcp_ph): Likewise.
2875 (_mm_rcp_sh): Likewise.
2876 (_mm_mask_rcp_sh): Likewise.
2877 (_mm_maskz_rcp_sh): Likewise.
2878 (_mm512_scalef_ph): Likewise.
2879 (_mm512_mask_scalef_ph): Likewise.
2880 (_mm512_maskz_scalef_ph): Likewise.
2881 (_mm512_scalef_round_ph): Likewise.
2882 (_mm512_mask_scalef_round_ph): Likewise.
2883 (_mm512_maskz_scalef_round_ph): Likewise.
2884 (_mm_scalef_sh): Likewise.
2885 (_mm_mask_scalef_sh): Likewise.
2886 (_mm_maskz_scalef_sh): Likewise.
2887 (_mm_scalef_round_sh): Likewise.
2888 (_mm_mask_scalef_round_sh): Likewise.
2889 (_mm_maskz_scalef_round_sh): Likewise.
2890 * config/i386/avx512fp16vlintrin.h (_mm_rcp_ph):
2892 (_mm256_rcp_ph): Likewise.
2893 (_mm_mask_rcp_ph): Likewise.
2894 (_mm256_mask_rcp_ph): Likewise.
2895 (_mm_maskz_rcp_ph): Likewise.
2896 (_mm256_maskz_rcp_ph): Likewise.
2897 (_mm_scalef_ph): Likewise.
2898 (_mm256_scalef_ph): Likewise.
2899 (_mm_mask_scalef_ph): Likewise.
2900 (_mm256_mask_scalef_ph): Likewise.
2901 (_mm_maskz_scalef_ph): Likewise.
2902 (_mm256_maskz_scalef_ph): Likewise.
2903 * config/i386/i386-builtin.def: Add new builtins.
2904 * config/i386/sse.md (VFH_AVX512VL): New.
2905 (avx512fp16_rcp<mode>2<mask_name>): Ditto.
2906 (avx512fp16_vmrcpv8hf2<mask_scalar_name>): Ditto.
2907 (avx512f_vmscalef<mode><mask_scalar_name><round_scalar_name>):
2908 Adjust to support HF vector modes.
2909 (<avx512>_scalef<mode><mask_name><round_name>): Ditto.
2911 2021-09-14 liuhongt <hongtao.liu@intel.com>
2913 * config/i386/avx512fp16intrin.h: (_mm512_sqrt_ph):
2915 (_mm512_mask_sqrt_ph): Likewise.
2916 (_mm512_maskz_sqrt_ph): Likewise.
2917 (_mm512_sqrt_round_ph): Likewise.
2918 (_mm512_mask_sqrt_round_ph): Likewise.
2919 (_mm512_maskz_sqrt_round_ph): Likewise.
2920 (_mm512_rsqrt_ph): Likewise.
2921 (_mm512_mask_rsqrt_ph): Likewise.
2922 (_mm512_maskz_rsqrt_ph): Likewise.
2923 (_mm_rsqrt_sh): Likewise.
2924 (_mm_mask_rsqrt_sh): Likewise.
2925 (_mm_maskz_rsqrt_sh): Likewise.
2926 (_mm_sqrt_sh): Likewise.
2927 (_mm_mask_sqrt_sh): Likewise.
2928 (_mm_maskz_sqrt_sh): Likewise.
2929 (_mm_sqrt_round_sh): Likewise.
2930 (_mm_mask_sqrt_round_sh): Likewise.
2931 (_mm_maskz_sqrt_round_sh): Likewise.
2932 * config/i386/avx512fp16vlintrin.h (_mm_sqrt_ph): New intrinsic.
2933 (_mm256_sqrt_ph): Likewise.
2934 (_mm_mask_sqrt_ph): Likewise.
2935 (_mm256_mask_sqrt_ph): Likewise.
2936 (_mm_maskz_sqrt_ph): Likewise.
2937 (_mm256_maskz_sqrt_ph): Likewise.
2938 (_mm_rsqrt_ph): Likewise.
2939 (_mm256_rsqrt_ph): Likewise.
2940 (_mm_mask_rsqrt_ph): Likewise.
2941 (_mm256_mask_rsqrt_ph): Likewise.
2942 (_mm_maskz_rsqrt_ph): Likewise.
2943 (_mm256_maskz_rsqrt_ph): Likewise.
2944 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2945 * config/i386/i386-builtin.def: Add corresponding new builtins.
2946 * config/i386/i386-expand.c
2947 (ix86_expand_args_builtin): Handle new builtins.
2948 (ix86_expand_round_builtin): Ditto.
2949 * config/i386/sse.md (VF_AVX512FP16VL): New.
2950 (sqrt<mode>2): Adjust for HF vector modes.
2951 (<sse>_sqrt<mode>2<mask_name><round_name>): Likewise.
2952 (<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
2954 (<sse>_rsqrt<mode>2<mask_name>): New.
2955 (avx512fp16_vmrsqrtv8hf2<mask_scalar_name>): Likewise.
2957 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
2960 * diagnostic-spec.c (warning_suppressed_at, copy_warning): Handle
2961 'RESERVED_LOCATION_P' locations.
2962 * warning-control.cc (get_nowarn_spec, suppress_warning)
2963 (copy_warning): Likewise.
2965 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
2967 * diagnostic-spec.h (typedef xint_hash_t): Use 'location_t' instead of...
2968 (typedef key_type_t): ... this. Remove.
2969 (nowarn_map): Document.
2970 * diagnostic-spec.c (nowarn_map): Likewise.
2971 * warning-control.cc (convert_to_key): Evolve functions into...
2972 (get_location): ... these. Adjust all users.
2974 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
2976 * warning-control.cc (copy_warning): Remove 'nowarn_map' setup.
2978 2021-09-13 Jason Merrill <jason@redhat.com>
2980 * params.opt: Add destructive-interference-size and
2981 constructive-interference-size.
2982 * doc/invoke.texi: Document them.
2983 * config/aarch64/aarch64.c (aarch64_override_options_internal):
2985 * config/arm/arm.c (arm_option_override): Set them.
2986 * config/i386/i386-options.c (ix86_option_override_internal):
2989 2021-09-13 Martin Liska <mliska@suse.cz>
2990 H.J. Lu <hjl.tools@gmail.com>
2993 * common/config/i386/cpuinfo.h (cpu_indicator_init): Add support
2994 for x86-64 micro levels for __builtin_cpu_supports.
2995 * common/config/i386/i386-cpuinfo.h (enum feature_priority):
2996 Add priorities for the micro-arch levels.
2997 (enum processor_features): Add new features.
2998 * common/config/i386/i386-isas.h: Add micro-arch features.
2999 * config/i386/i386-builtins.c (get_builtin_code_for_version):
3000 Support the micro-arch levels by callsing
3001 __builtin_cpu_supports.
3002 * doc/extend.texi: Document that the levels are support by
3003 __builtin_cpu_supports.
3005 2021-09-13 Andrew Pinski <apinski@marvell.com>
3008 * config/aarch64/aarch64-builtins.c (aarch64_fold_builtin_lane_check):
3010 (aarch64_general_fold_builtin): Handle AARCH64_SIMD_BUILTIN_LANE_CHECK.
3011 (aarch64_general_gimple_fold_builtin): Likewise.
3013 2021-09-13 Andrew Pinski <apinski@marvell.com>
3015 * config.gcc: Add m32r-*-linux* and m32rle-*-linux*
3016 to the Unsupported targets list.
3017 Remove support for m32r-*-linux* and m32rle-*-linux*.
3018 * config/m32r/linux.h: Removed.
3019 * config/m32r/t-linux: Removed.
3021 2021-09-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3024 * config/aarch64/aarch64.c (aarch64_classify_address): Don't allow
3025 register index for SVE predicate modes.
3027 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
3029 * tree-ssa-threadbackward.c
3030 (back_threader_profitability::profitable_path_p): Remove FSM
3032 (back_threader_registry::register_path): Same.
3033 * tree-ssa-threadedge.c
3034 (jump_threader::simplify_control_stmt_condition): Same.
3035 * tree-ssa-threadupdate.c (jt_path_registry::jt_path_registry):
3036 Add backedge_threads argument.
3037 (fwd_jt_path_registry::fwd_jt_path_registry): Pass
3038 backedge_threads argument.
3039 (back_jt_path_registry::back_jt_path_registry): Same.
3040 (dump_jump_thread_path): Adjust for FSM removal.
3041 (back_jt_path_registry::rewire_first_differing_edge): Same.
3042 (back_jt_path_registry::adjust_paths_after_duplication): Same.
3043 (back_jt_path_registry::update_cfg): Same.
3044 (jt_path_registry::register_jump_thread): Same.
3045 * tree-ssa-threadupdate.h (enum jump_thread_edge_type): Remove
3047 (class back_jt_path_registry): Add backedge_threads to
3050 2021-09-13 Martin Liska <mliska@suse.cz>
3053 * asan.h (sanitize_coverage_p): Handle when fn == NULL.
3055 2021-09-13 H.J. Lu <hjl.tools@gmail.com>
3058 * config/i386/i386.h (TARGET_AVX256_MOVE_BY_PIECES): New.
3059 (TARGET_AVX256_STORE_BY_PIECES): Likewise.
3060 (MOVE_MAX): Check TARGET_AVX256_MOVE_BY_PIECES and
3061 TARGET_AVX256_STORE_BY_PIECES instead of
3062 TARGET_AVX256_SPLIT_UNALIGNED_LOAD and
3063 TARGET_AVX256_SPLIT_UNALIGNED_STORE.
3064 (STORE_MAX_PIECES): Check TARGET_AVX256_STORE_BY_PIECES instead
3065 of TARGET_AVX256_SPLIT_UNALIGNED_STORE.
3066 * config/i386/x86-tune.def (X86_TUNE_AVX256_MOVE_BY_PIECES): New.
3067 (X86_TUNE_AVX256_STORE_BY_PIECES): Likewise.
3069 2021-09-13 liuhongt <hongtao.liu@intel.com>
3072 * expmed.c (extract_bit_field_using_extv): Use
3073 gen_lowpart_if_possible instead of gen_lowpart to avoid ICE.
3075 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
3077 * Makefile.in (OBJS): Add value-pointer-equiv.o.
3078 * gimple-ssa-evrp.c (class ssa_equiv_stack): Move to
3079 value-pointer-equiv.*.
3080 (ssa_equiv_stack::ssa_equiv_stack): Same.
3081 (ssa_equiv_stack::enter): Same.
3082 (ssa_equiv_stack::leave): Same.
3083 (ssa_equiv_stack::push_replacement): Same.
3084 (ssa_equiv_stack::get_replacement): Same.
3085 (is_pointer_ssa): Same.
3086 (class pointer_equiv_analyzer): Same.
3087 (pointer_equiv_analyzer::pointer_equiv_analyzer): Same.
3088 (pointer_equiv_analyzer::~pointer_equiv_analyzer): Same.
3089 (pointer_equiv_analyzer::set_global_equiv): Same.
3090 (pointer_equiv_analyzer::set_cond_equiv): Same.
3091 (pointer_equiv_analyzer::get_equiv): Same.
3092 (pointer_equiv_analyzer::enter): Same.
3093 (pointer_equiv_analyzer::leave): Same.
3094 (pointer_equiv_analyzer::get_equiv_expr): Same.
3095 (pta_valueize): Same.
3096 (pointer_equiv_analyzer::visit_stmt): Same.
3097 (pointer_equiv_analyzer::visit_edge): Same.
3098 (hybrid_folder::value_of_expr): Same.
3099 (hybrid_folder::value_on_edge): Same.
3100 * value-pointer-equiv.cc: New file.
3101 * value-pointer-equiv.h: New file.
3103 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
3106 * gimple-fold.c (gimple_fold_builtin_memory_op): Allow folding
3107 memcpy if the size is not more than MOVE_MAX * MOVE_RATIO.
3109 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
3112 * config/arm/arm.md (movmisaligndi): New define_expand.
3113 * config/arm/vec-common.md (movmisalign<mode>): Iterate over VDQ mode.
3115 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
3118 * emit-rtl.c (gen_highpart): Use adjust_address to handle
3119 MEM rather than calling simplify_gen_subreg.
3121 2021-09-13 Jan-Benedict Glaw <jbglaw@ług-owl.de>
3123 * config/alpha/vms.h (INIT_CUMULATIVE_ARGS): Wrap multi-statment
3124 define into a block.
3126 2021-09-13 Richard Biener <rguenther@suse.de>
3128 * config/darwin.h (DARWIN_PREFER_DWARF): Do not define.
3129 * config/i386/darwin.h (PREFERRED_DEBUGGING_TYPE): Do not
3130 change based on DARWIN_PREFER_DWARF not being defined.
3132 2021-09-13 Richard Biener <rguenther@suse.de>
3134 * config/i386/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
3135 to inherit from elfos.h
3137 2021-09-13 Richard Biener <rguenther@suse.de>
3139 * config.gcc: Add cr16-*-* to the list of obsoleted targets.
3141 2021-09-13 Richard Biener <rguenther@suse.de>
3143 * config/avr/elf.h (PREFERRED_DEBUGGING_TYPE): Remove
3144 override, pick up DWARF2_DEBUG define from elfos.h
3146 2021-09-13 Richard Biener <rguenther@suse.de>
3148 * config/rx/rx.h (PREFERRED_DEBUGGING_TYPE): Always define to
3151 2021-09-13 Richard Biener <rguenther@suse.de>
3153 * config/alpha/vms.h (PREFERRED_DEBUGGING_TYPE): Define to
3156 2021-09-13 Richard Biener <rguenther@suse.de>
3158 * config/i386/cygming.h: Always default to DWARF2 debugging.
3159 Do not define DBX_DEBUGGING_INFO, that's done via dbxcoff.h
3161 * doc/install.texi: Document binutils 2.16 as minimum
3162 requirement for mingw.
3164 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
3166 * config/rs6000/rs6000.c (struct rs6000_cost_data): New members
3167 nstmts, nloads and extra_ctor_cost.
3168 (rs6000_density_test): Add load density related heuristics. Do
3169 extra costing on vector construction statements if need.
3170 (rs6000_init_cost): Init new members.
3171 (rs6000_update_target_cost_per_stmt): New function.
3172 (rs6000_add_stmt_cost): Factor vect_nonmem hunk out to function
3173 rs6000_update_target_cost_per_stmt and call it.
3175 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
3177 * config/rs6000/rs6000.c (struct rs6000_cost_data): Remove typedef.
3178 (rs6000_init_cost): Adjust.
3180 2021-09-13 liuhongt <hongtao.liu@intel.com>
3182 * config/i386/i386.md: (UNSPEC_COPYSIGN): Remove.
3183 (UNSPEC_XORSIGN): Ditto.
3185 2021-09-12 Roger Sayle <roger@nextmovesoftware.com>
3187 * expr.c (convert_move): Preserve SUBREG_PROMOTED_VAR_P when
3188 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
3191 2021-09-11 Aldy Hernandez <aldyh@redhat.com>
3193 * tree-ssa-threadbackward.c (class back_threader_registry): Use
3194 back_jt_path_registry.
3195 * tree-ssa-threadedge.c (jump_threader::jump_threader): Use
3196 fwd_jt_path_registry.
3197 * tree-ssa-threadedge.h (class jump_threader): Same..
3198 * tree-ssa-threadupdate.c
3199 (jump_thread_path_registry::jump_thread_path_registry): Rename...
3200 (jt_path_registry::jt_path_registry): ...to this.
3201 (jump_thread_path_registry::~jump_thread_path_registry): Rename...
3202 (jt_path_registry::~jt_path_registry): ...this.
3203 (fwd_jt_path_registry::fwd_jt_path_registry): New.
3204 (fwd_jt_path_registry::~fwd_jt_path_registry): New.
3205 (jump_thread_path_registry::allocate_thread_edge): Rename...
3206 (jt_path_registry::allocate_thread_edge): ...to this.
3207 (jump_thread_path_registry::allocate_thread_path): Rename...
3208 (jt_path_registry::allocate_thread_path): ...to this.
3209 (jump_thread_path_registry::lookup_redirection_data): Rename...
3210 (fwd_jt_path_registry::lookup_redirection_data): ...to this.
3211 (jump_thread_path_registry::thread_block_1): Rename...
3212 (fwd_jt_path_registry::thread_block_1): ...to this.
3213 (jump_thread_path_registry::thread_block): Rename...
3214 (fwd_jt_path_registry::thread_block): ...to this.
3215 (jt_path_registry::thread_through_loop_header): Rename...
3216 (fwd_jt_path_registry::thread_through_loop_header): ...to this.
3217 (jump_thread_path_registry::mark_threaded_blocks): Rename...
3218 (fwd_jt_path_registry::mark_threaded_blocks): ...to this.
3219 (jump_thread_path_registry::debug_path): Rename...
3220 (jt_path_registry::debug_path): ...to this.
3221 (jump_thread_path_registry::dump): Rename...
3222 (jt_path_registry::debug): ...to this.
3223 (jump_thread_path_registry::rewire_first_differing_edge): Rename...
3224 (back_jt_path_registry::rewire_first_differing_edge): ...to this.
3225 (jump_thread_path_registry::adjust_paths_after_duplication): Rename...
3226 (back_jt_path_registry::adjust_paths_after_duplication): ...to this.
3227 (jump_thread_path_registry::duplicate_thread_path): Rename...
3228 (back_jt_path_registry::duplicate_thread_path): ...to this. Also,
3229 drop ill-formed candidates.
3230 (jump_thread_path_registry::remove_jump_threads_including): Rename...
3231 (fwd_jt_path_registry::remove_jump_threads_including): ...to this.
3232 (jt_path_registry::thread_through_all_blocks): New.
3233 (back_jt_path_registry::update_cfg): New.
3234 (fwd_jt_path_registry::update_cfg): New.
3235 (jump_thread_path_registry::register_jump_thread): Rename...
3236 (jt_path_registry::register_jump_thread): ...to this.
3237 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
3239 (class jt_path_registry): ...here.
3240 (class fwd_jt_path_registry): New.
3241 (class back_jt_path_registry): New.
3243 2021-09-10 liuhongt <hongtao.liu@intel.com>
3246 2021-09-01 liuhongt <hongtao.liu@intel.com>
3248 * emit-rtl.c (validate_subreg): Get rid of all float-int
3251 2021-09-10 Jakub Jelinek <jakub@redhat.com>
3253 * tree-core.h (enum omp_memory_order): Add OMP_MEMORY_ORDER_MASK,
3254 OMP_FAIL_MEMORY_ORDER_UNSPECIFIED, OMP_FAIL_MEMORY_ORDER_RELAXED,
3255 OMP_FAIL_MEMORY_ORDER_ACQUIRE, OMP_FAIL_MEMORY_ORDER_RELEASE,
3256 OMP_FAIL_MEMORY_ORDER_ACQ_REL, OMP_FAIL_MEMORY_ORDER_SEQ_CST and
3257 OMP_FAIL_MEMORY_ORDER_MASK enumerators.
3258 (OMP_FAIL_MEMORY_ORDER_SHIFT): Define.
3259 * gimple-pretty-print.c (dump_gimple_omp_atomic_load,
3260 dump_gimple_omp_atomic_store): Print [weak] for weak atomic
3262 * gimple.h (enum gf_mask): Change GF_OMP_ATOMIC_MEMORY_ORDER
3263 to 6-bit mask, adjust GF_OMP_ATOMIC_NEED_VALUE value and add
3265 (gimple_omp_atomic_weak_p, gimple_omp_atomic_set_weak): New inline
3267 * tree.h (OMP_ATOMIC_WEAK): Define.
3268 * tree-pretty-print.c (dump_omp_atomic_memory_order): Adjust for
3269 fail memory order being encoded in the same enum and also print
3270 fail clause if present.
3271 (dump_generic_node): Print weak clause if OMP_ATOMIC_WEAK.
3272 * gimplify.c (goa_stabilize_expr): Add target_expr and rhs arguments,
3273 handle pre_p == NULL case as a test mode that only returns value
3274 but doesn't change gimplify nor change anything otherwise, adjust
3275 recursive calls, add MODIFY_EXPR, ADDR_EXPR, COND_EXPR, TARGET_EXPR
3276 and CALL_EXPR handling, adjust COMPOUND_EXPR handling for
3277 __builtin_clear_padding calls, for !rhs gimplify as lvalue rather
3279 (gimplify_omp_atomic): Adjust goa_stabilize_expr caller. Handle
3280 COND_EXPR rhs. Set weak flag on gimple load/store for
3282 * omp-expand.c (omp_memory_order_to_fail_memmodel): New function.
3283 (omp_memory_order_to_memmodel): Adjust for fail clause encoded
3285 (expand_omp_atomic_cas): New function.
3286 (expand_omp_atomic_pipeline): Use omp_memory_order_to_fail_memmodel
3288 (expand_omp_atomic): Attempt to optimize atomic compare and exchange
3289 using expand_omp_atomic_cas.
3291 2021-09-10 Aldy Hernandez <aldyh@redhat.com>
3292 Michael Matz <matz@suse.de>
3294 * tree-pass.h (PROP_loop_opts_done): New.
3295 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
3296 Intersect with global range.
3297 * tree-ssa-loop.c (tree_ssa_loop_done): Set PROP_loop_opts_done.
3298 * tree-ssa-threadbackward.c
3299 (back_threader_profitability::profitable_path_p): Disable
3300 threading through latches until after loop optimizations have run.
3302 2021-09-10 David Faust <david.faust@oracle.com>
3304 * doc/invoke.texi: Document BPF -mcpu, -mjmpext, -mjmp32 and -malu32
3307 2021-09-10 David Faust <david.faust@oracle.com>
3309 * config/bpf/bpf-opts.h (bpf_isa_version): New enum.
3310 * config/bpf/bpf-protos.h (bpf_expand_cbranch): New.
3311 * config/bpf/bpf.c (bpf_option_override): Handle -mcpu option.
3312 (bpf_expand_cbranch): New function.
3313 * config/bpf/bpf.md (AM mode iterator): Conditionalize support for SI
3315 (zero_extendsidi2): Only use mov32 instruction if it is available.
3316 (SIM mode iterator): Conditionalize support for SI mode.
3317 (JM mode iterator): New.
3318 (cbranchdi4): Update name, use new JM iterator. Use bpf_expand_cbranch.
3319 (*branch_on_di): Update name, use new JM iterator.
3320 * config/bpf/bpf.opt: (mjmpext): New option.
3324 (bpf_isa): New enum.
3326 2021-09-10 David Faust <david.faust@oracle.com>
3328 * config/bpf/bpf.md (zero_extendhidi2): Add new output template
3329 for register-to-register extensions.
3330 (zero_extendqidi2): Likewise.
3332 2021-09-10 Richard Biener <rguenther@suse.de>
3334 PR middle-end/102273
3335 * internal-fn.c (expand_DEFERRED_INIT): Always expand non-SSA vars.
3337 2021-09-10 Richard Biener <rguenther@suse.de>
3339 PR middle-end/102269
3340 * gimplify.c (is_var_need_auto_init): Empty types do not need
3343 2021-09-10 Richard Biener <rguenther@suse.de>
3345 * configure.ac (--with-stabs): Remove.
3346 * configure: Regenerate.
3347 * doc/install.texi: Remove --with-stabs documentation.
3349 2021-09-10 liuhongt <hongtao.liu@intel.com>
3351 * config/i386/avx512fp16intrin.h: (_mm512_cmp_ph_mask):
3353 (_mm512_mask_cmp_ph_mask): Likewise.
3354 (_mm512_cmp_round_ph_mask): Likewise.
3355 (_mm512_mask_cmp_round_ph_mask): Likewise.
3356 (_mm_cmp_sh_mask): Likewise.
3357 (_mm_mask_cmp_sh_mask): Likewise.
3358 (_mm_cmp_round_sh_mask): Likewise.
3359 (_mm_mask_cmp_round_sh_mask): Likewise.
3360 (_mm_comieq_sh): Likewise.
3361 (_mm_comilt_sh): Likewise.
3362 (_mm_comile_sh): Likewise.
3363 (_mm_comigt_sh): Likewise.
3364 (_mm_comige_sh): Likewise.
3365 (_mm_comineq_sh): Likewise.
3366 (_mm_ucomieq_sh): Likewise.
3367 (_mm_ucomilt_sh): Likewise.
3368 (_mm_ucomile_sh): Likewise.
3369 (_mm_ucomigt_sh): Likewise.
3370 (_mm_ucomige_sh): Likewise.
3371 (_mm_ucomineq_sh): Likewise.
3372 (_mm_comi_round_sh): Likewise.
3373 (_mm_comi_sh): Likewise.
3374 * config/i386/avx512fp16vlintrin.h (_mm_cmp_ph_mask): New intrinsic.
3375 (_mm_mask_cmp_ph_mask): Likewise.
3376 (_mm256_cmp_ph_mask): Likewise.
3377 (_mm256_mask_cmp_ph_mask): Likewise.
3378 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3379 * config/i386/i386-builtin.def: Add corresponding new builtins.
3380 * config/i386/i386-expand.c
3381 (ix86_expand_args_builtin): Handle new builtin types.
3382 (ix86_expand_round_builtin): Ditto.
3383 * config/i386/i386.md (ssevecmode): Add HF mode.
3384 (MODEFH): New mode iterator.
3385 * config/i386/sse.md
3386 (V48H_AVX512VL): New mode iterator to support HF vector modes.
3387 Ajdust corresponding description.
3388 (ssecmpintprefix): New.
3389 (VI12_AVX512VL): Adjust to support HF vector modes.
3390 (cmp_imm_predicate): Likewise.
3391 (<avx512>_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
3393 (avx512f_vmcmp<mode>3<round_saeonly_name>): Likewise.
3394 (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Likewise.
3395 (<sse>_<unord>comi<round_saeonly_name>): Likewise.
3397 2021-09-10 liuhongt <hongtao.liu@intel.com>
3399 * config/i386/avx512fp16intrin.h: (_mm512_max_ph): New intrinsic.
3400 (_mm512_mask_max_ph): Likewise.
3401 (_mm512_maskz_max_ph): Likewise.
3402 (_mm512_min_ph): Likewise.
3403 (_mm512_mask_min_ph): Likewise.
3404 (_mm512_maskz_min_ph): Likewise.
3405 (_mm512_max_round_ph): Likewise.
3406 (_mm512_mask_max_round_ph): Likewise.
3407 (_mm512_maskz_max_round_ph): Likewise.
3408 (_mm512_min_round_ph): Likewise.
3409 (_mm512_mask_min_round_ph): Likewise.
3410 (_mm512_maskz_min_round_ph): Likewise.
3411 (_mm_max_sh): Likewise.
3412 (_mm_mask_max_sh): Likewise.
3413 (_mm_maskz_max_sh): Likewise.
3414 (_mm_min_sh): Likewise.
3415 (_mm_mask_min_sh): Likewise.
3416 (_mm_maskz_min_sh): Likewise.
3417 (_mm_max_round_sh): Likewise.
3418 (_mm_mask_max_round_sh): Likewise.
3419 (_mm_maskz_max_round_sh): Likewise.
3420 (_mm_min_round_sh): Likewise.
3421 (_mm_mask_min_round_sh): Likewise.
3422 (_mm_maskz_min_round_sh): Likewise.
3423 * config/i386/avx512fp16vlintrin.h (_mm_max_ph): New intrinsic.
3424 (_mm256_max_ph): Likewise.
3425 (_mm_mask_max_ph): Likewise.
3426 (_mm256_mask_max_ph): Likewise.
3427 (_mm_maskz_max_ph): Likewise.
3428 (_mm256_maskz_max_ph): Likewise.
3429 (_mm_min_ph): Likewise.
3430 (_mm256_min_ph): Likewise.
3431 (_mm_mask_min_ph): Likewise.
3432 (_mm256_mask_min_ph): Likewise.
3433 (_mm_maskz_min_ph): Likewise.
3434 (_mm256_maskz_min_ph): Likewise.
3435 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3436 * config/i386/i386-builtin.def: Add corresponding new builtins.
3437 * config/i386/i386-expand.c
3438 (ix86_expand_args_builtin): Handle new builtin types.
3439 * config/i386/sse.md
3440 (<code><mode>3<mask_name><round_saeonly_name>): Adjust to
3441 support HF vector modes.
3442 (*<code><mode>3<mask_name><round_saeonly_name>): Likewise.
3443 (ieee_<ieee_maxmin><mode>3<mask_name><round_saeonly_name>):
3445 (<sse>_vm<code><mode>3<mask_scalar_name><round_saeonly_scalar_name>):
3447 * config/i386/subst.md (round_saeonly_mode512bit_condition):
3448 Adjust for HF vector modes.
3450 2021-09-10 Liu, Hongtao <hongtao.liu@intel.com>
3452 * config/i386/avx512fp16intrin.h (_mm_add_sh): New intrinsic.
3453 (_mm_mask_add_sh): Likewise.
3454 (_mm_maskz_add_sh): Likewise.
3455 (_mm_sub_sh): Likewise.
3456 (_mm_mask_sub_sh): Likewise.
3457 (_mm_maskz_sub_sh): Likewise.
3458 (_mm_mul_sh): Likewise.
3459 (_mm_mask_mul_sh): Likewise.
3460 (_mm_maskz_mul_sh): Likewise.
3461 (_mm_div_sh): Likewise.
3462 (_mm_mask_div_sh): Likewise.
3463 (_mm_maskz_div_sh): Likewise.
3464 (_mm_add_round_sh): Likewise.
3465 (_mm_mask_add_round_sh): Likewise.
3466 (_mm_maskz_add_round_sh): Likewise.
3467 (_mm_sub_round_sh): Likewise.
3468 (_mm_mask_sub_round_sh): Likewise.
3469 (_mm_maskz_sub_round_sh): Likewise.
3470 (_mm_mul_round_sh): Likewise.
3471 (_mm_mask_mul_round_sh): Likewise.
3472 (_mm_maskz_mul_round_sh): Likewise.
3473 (_mm_div_round_sh): Likewise.
3474 (_mm_mask_div_round_sh): Likewise.
3475 (_mm_maskz_div_round_sh): Likewise.
3476 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3477 * config/i386/i386-builtin.def: Add corresponding new builtins.
3478 * config/i386/i386-expand.c
3479 (ix86_expand_round_builtin): Handle new builtins.
3480 * config/i386/sse.md (VF_128): Change description.
3481 (<sse>_vm<plusminus_insn><mode>3<mask_scalar_name><round_scalar_name>):
3482 Adjust to support HF vector modes.
3483 (<sse>_vm<multdiv_mnemonic><mode>3<mask_scalar_name><round_scalar_name>):
3486 2021-09-10 H.J. Lu <hjl.tools@gmail.com>
3488 * config/i386/i386-expand.c
3489 (ix86_avx256_split_vector_move_misalign): Handle V16HF mode.
3490 * config/i386/i386.c
3491 (ix86_preferred_simd_mode): Handle HF mode.
3492 * config/i386/sse.md (V_256H): New mode iterator.
3493 (avx_vextractf128<mode>): Use it.
3494 (VEC_INIT_MODE): Align vector HFmode condition to vector
3495 HImodes since there're no real HF instruction used.
3496 (VEC_INIT_HALF_MODE): Ditto.
3498 (VIHF_AVX512BW): Ditto.
3499 (*vec_extracthf): Ditto.
3500 (VEC_EXTRACT_MODE): Ditto.
3502 2021-09-10 Richard Biener <rguenther@suse.de>
3505 * config/dbx.h: Remove.
3506 * config/dbxcoff.h: Do not define PREFERRED_DEBUGGING_TYPE.
3507 * config/lynx.h: Likewise.
3509 2021-09-10 liuhongt <hongtao.liu@intel.com>
3511 * config/i386/i386-expand.c (ix86_expand_copysign): Expand
3512 right into ANDNOT + AND + IOR, using paradoxical subregs.
3513 (ix86_split_copysign_const): Remove.
3514 (ix86_split_copysign_var): Ditto.
3515 * config/i386/i386-protos.h (ix86_split_copysign_const): Dotto.
3516 (ix86_split_copysign_var): Ditto.
3517 * config/i386/i386.md (@copysign<mode>3_const): Ditto.
3518 (@copysign<mode>3_var): Ditto.
3520 2021-09-09 qing zhao <qing.zhao@oracle.com>
3522 * builtins.c (expand_builtin_memset): Make external visible.
3523 * builtins.h (expand_builtin_memset): Declare extern.
3524 * common.opt (ftrivial-auto-var-init=): New option.
3525 * doc/extend.texi: Document the uninitialized attribute.
3526 * doc/invoke.texi: Document -ftrivial-auto-var-init.
3527 * flag-types.h (enum auto_init_type): New enumerated type
3529 * gimple-fold.c (clear_padding_type): Add one new parameter.
3530 (clear_padding_union): Likewise.
3531 (clear_padding_emit_loop): Likewise.
3532 (clear_type_padding_in_mask): Likewise.
3533 (gimple_fold_builtin_clear_padding): Handle this new parameter.
3534 * gimplify.c (gimple_add_init_for_auto_var): New function.
3535 (gimple_add_padding_init_for_auto_var): New function.
3536 (is_var_need_auto_init): New function.
3537 (gimplify_decl_expr): Add initialization to automatic variables per
3539 (gimplify_call_expr): Add one new parameter for call to
3540 __builtin_clear_padding.
3541 (gimplify_init_constructor): Add padding initialization in the end.
3542 * internal-fn.c (INIT_PATTERN_VALUE): New macro.
3543 (expand_DEFERRED_INIT): New function.
3544 * internal-fn.def (DEFERRED_INIT): New internal function.
3545 * tree-cfg.c (verify_gimple_call): Verify calls to .DEFERRED_INIT.
3546 * tree-sra.c (generate_subtree_deferred_init): New function.
3547 (scan_function): Avoid setting cannot_scalarize_away_bitmap for
3548 calls to .DEFERRED_INIT.
3549 (sra_modify_deferred_init): New function.
3550 (sra_modify_function_body): Handle calls to DEFERRED_INIT specially.
3551 * tree-ssa-structalias.c (find_func_aliases_for_call): Likewise.
3552 * tree-ssa-uninit.c (warn_uninit): Handle calls to DEFERRED_INIT
3554 (check_defs): Likewise.
3555 (warn_uninitialized_vars): Likewise.
3556 * tree-ssa.c (ssa_undefined_value_p): Likewise.
3557 * tree.c (build_common_builtin_nodes): Build tree node for
3558 BUILT_IN_CLEAR_PADDING when needed.
3560 2021-09-09 Richard Biener <rguenther@suse.de>
3562 * tree-ssa-loop-im.c (fill_always_executed_in_1): Walk
3565 2021-09-09 Richard Biener <rguenther@suse.de>
3567 * tree-ssa-loop-im.c (fill_always_executed_in_1): Integrate
3568 DOM walk from get_loop_body_in_dom_order using a worklist
3571 2021-09-09 liuhongt <hongtao.liu@intel.com>
3573 * config.gcc: Add avx512fp16vlintrin.h.
3574 * config/i386/avx512fp16intrin.h: (_mm512_add_ph): New intrinsic.
3575 (_mm512_mask_add_ph): Likewise.
3576 (_mm512_maskz_add_ph): Likewise.
3577 (_mm512_sub_ph): Likewise.
3578 (_mm512_mask_sub_ph): Likewise.
3579 (_mm512_maskz_sub_ph): Likewise.
3580 (_mm512_mul_ph): Likewise.
3581 (_mm512_mask_mul_ph): Likewise.
3582 (_mm512_maskz_mul_ph): Likewise.
3583 (_mm512_div_ph): Likewise.
3584 (_mm512_mask_div_ph): Likewise.
3585 (_mm512_maskz_div_ph): Likewise.
3586 (_mm512_add_round_ph): Likewise.
3587 (_mm512_mask_add_round_ph): Likewise.
3588 (_mm512_maskz_add_round_ph): Likewise.
3589 (_mm512_sub_round_ph): Likewise.
3590 (_mm512_mask_sub_round_ph): Likewise.
3591 (_mm512_maskz_sub_round_ph): Likewise.
3592 (_mm512_mul_round_ph): Likewise.
3593 (_mm512_mask_mul_round_ph): Likewise.
3594 (_mm512_maskz_mul_round_ph): Likewise.
3595 (_mm512_div_round_ph): Likewise.
3596 (_mm512_mask_div_round_ph): Likewise.
3597 (_mm512_maskz_div_round_ph): Likewise.
3598 * config/i386/avx512fp16vlintrin.h: New header.
3599 * config/i386/i386-builtin-types.def (V16HF, V8HF, V32HF):
3600 Add new builtin types.
3601 * config/i386/i386-builtin.def: Add corresponding builtins.
3602 * config/i386/i386-expand.c
3603 (ix86_expand_args_builtin): Handle new builtin types.
3604 (ix86_expand_round_builtin): Likewise.
3605 * config/i386/immintrin.h: Include avx512fp16vlintrin.h
3606 * config/i386/sse.md (VFH): New mode_iterator.
3608 (avx512fmaskmode): Add HF vector modes.
3609 (avx512fmaskhalfmode): Likewise.
3610 (<plusminus_insn><mode>3<mask_name><round_name>): Adjust to for
3612 (*<plusminus_insn><mode>3<mask_name><round_name>): Likewise.
3613 (mul<mode>3<mask_name><round_name>): Likewise.
3614 (*mul<mode>3<mask_name><round_name>): Likewise.
3615 (div<mode>3): Likewise.
3616 (<sse>_div<mode>3<mask_name><round_name>): Likewise.
3617 * config/i386/subst.md (SUBST_V): Add HF vector modes.
3618 (SUBST_A): Likewise.
3619 (round_mode512bit_condition): Adjust for V32HFmode.
3621 2021-09-09 liuhongt <hongtao.liu@intel.com>
3624 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
3625 (reduc_plus_scal_v4sf): .. this, New define_expand.
3626 (reduc_plus_scal_v2df): .. and this, New define_expand.
3628 2021-09-09 liuhongt <hongtao.liu@intel.com>
3631 * config/i386/sse.md (*vec_extract<mode><ssescalarmodelower>_valign):
3634 2021-09-08 Jonathan Wakely <jwakely@redhat.com>
3637 * doc/trouble.texi (Copy Assignment): Fix description of
3638 behaviour and fix code in example.
3640 2021-09-08 Segher Boessenkool <segher@kernel.crashing.org>
3643 * config/rs6000/rs6000-logue.c (rs6000_emit_epilogue): For ELFv2 use
3644 r11 instead of r12 for restoring CR.
3646 2021-09-08 Jakub Jelinek <jakub@redhat.com>
3647 liuhongt <hongtao.liu@intel.com>
3650 * config/i386/i386.md (@xorsign<mode>3_1): Remove.
3651 * config/i386/i386-expand.c (ix86_expand_xorsign): Expand right away
3652 into AND with mask and XOR, using paradoxical subregs.
3653 (ix86_split_xorsign): Remove.
3654 * config/i386/i386-protos.h (ix86_split_xorsign): Remove.
3656 2021-09-08 Di Zhao <dizhao@os.amperecomputing.com>
3658 * tree-ssa-sccvn.c (vn_nary_op_insert_into): fix result compare
3660 2021-09-08 Jakub Jelinek <jakub@redhat.com>
3663 * config/i386/i386.md (xorsign<mode>3): If operands[1] is equal to
3664 operands[2], emit abs<mode>2 instead.
3665 (@xorsign<mode>3_1): Add early-clobbers for output operand, enable
3666 first alternative even for avx, add another alternative with
3667 =&Yv <- 0, Yv, Yvm constraints.
3668 * config/i386/i386-expand.c (ix86_split_xorsign): If op0 is equal
3669 to op1, emit vpandn instead.
3671 2021-09-08 liuhongt <hongtao.liu@intel.com>
3673 * config/i386/avx512fp16intrin.h (_mm_set_ph): New intrinsic.
3674 (_mm256_set_ph): Likewise.
3675 (_mm512_set_ph): Likewise.
3676 (_mm_setr_ph): Likewise.
3677 (_mm256_setr_ph): Likewise.
3678 (_mm512_setr_ph): Likewise.
3679 (_mm_set1_ph): Likewise.
3680 (_mm256_set1_ph): Likewise.
3681 (_mm512_set1_ph): Likewise.
3682 (_mm_setzero_ph): Likewise.
3683 (_mm256_setzero_ph): Likewise.
3684 (_mm512_setzero_ph): Likewise.
3685 (_mm_set_sh): Likewise.
3686 (_mm_load_sh): Likewise.
3687 (_mm_store_sh): Likewise.
3688 * config/i386/i386-builtin-types.def (V8HF): New type.
3689 (DEF_FUNCTION_TYPE (V8HF, V8HI)): New builtin function type
3690 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
3691 Support vector HFmodes.
3692 (ix86_expand_vector_init_one_nonzero): Likewise.
3693 (ix86_expand_vector_init_one_var): Likewise.
3694 (ix86_expand_vector_init_interleave): Likewise.
3695 (ix86_expand_vector_init_general): Likewise.
3696 (ix86_expand_vector_set): Likewise.
3697 (ix86_expand_vector_extract): Likewise.
3698 (ix86_expand_vector_init_concat): Likewise.
3699 (ix86_expand_sse_movcc): Handle vector HFmodes.
3700 (ix86_expand_vector_set_var): Ditto.
3701 * config/i386/i386-modes.def: Add HF vector modes in comment.
3702 * config/i386/i386.c (classify_argument): Add HF vector modes.
3703 (ix86_hard_regno_mode_ok): Allow HF vector modes for AVX512FP16.
3704 (ix86_vector_mode_supported_p): Likewise.
3705 (ix86_set_reg_reg_cost): Handle vector HFmode.
3706 (ix86_get_ssemov): Handle vector HFmode.
3707 (function_arg_advance_64): Pass unamed V16HFmode and V32HFmode
3709 (function_arg_advance_32): Pass V8HF/V16HF/V32HF by sse reg for 32bit
3711 (function_arg_advance_32): Ditto.
3712 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): New.
3713 (VALID_AVX256_REG_OR_OI_MODE): Rename to ..
3714 (VALID_AVX256_REG_OR_OI_VHF_MODE): .. this, and add V16HF.
3715 (VALID_SSE2_REG_VHF_MODE): New.
3716 (VALID_AVX512VL_128_REG_MODE): Add V8HF and TImode.
3717 (SSE_REG_MODE_P): Add vector HFmode.
3718 * config/i386/i386.md (mode): Add HF vector modes.
3719 (MODE_SIZE): Likewise.
3720 (ssemodesuffix): Add ph suffix for HF vector modes.
3721 * config/i386/sse.md (VFH_128): New mode iterator.
3722 (VMOVE): Adjust for HF vector modes.
3724 (V_256_512): Likewise.
3726 (avx512fmaskmode): Likewise.
3727 (shuffletype): Likewise.
3728 (sseinsnmode): Likewise.
3729 (ssedoublevecmode): Likewise.
3730 (ssehalfvecmode): Likewise.
3731 (ssehalfvecmodelower): Likewise.
3732 (ssePScmode): Likewise.
3733 (ssescalarmode): Likewise.
3734 (ssescalarmodelower): Likewise.
3735 (sseintprefix): Likewise.
3737 (bcstscalarsuff): Likewise.
3738 (xtg_mode): Likewise.
3739 (VI12HF_AVX512VL): New mode_iterator.
3740 (VF_AVX512FP16): Likewise.
3742 (VIHF_256): Likewise.
3743 (VIHF_AVX512BW): Likewise.
3744 (V16_256): Likewise.
3745 (V32_512): Likewise.
3746 (sseintmodesuffix): New mode_attr.
3747 (sse): Add scalar and vector HFmodes.
3748 (ssescalarmode): Add vector HFmode mapping.
3749 (ssescalarmodesuffix): Add sh suffix for HFmode.
3750 (*<sse>_vm<insn><mode>3): Use VFH_128.
3751 (*<sse>_vm<multdiv_mnemonic><mode>3): Likewise.
3752 (*ieee_<ieee_maxmin><mode>3): Likewise.
3753 (<avx512>_blendm<mode>): New define_insn.
3754 (vec_setv8hf): New define_expand.
3755 (vec_set<mode>_0): New define_insn for HF vector set.
3756 (*avx512fp16_movsh): Likewise.
3757 (avx512fp16_movsh): Likewise.
3758 (vec_extract_lo_v32hi): Rename to ...
3759 (vec_extract_lo_<mode>): ... this, and adjust to allow HF
3761 (vec_extract_hi_v32hi): Likewise.
3762 (vec_extract_hi_<mode>): Likewise.
3763 (vec_extract_lo_v16hi): Likewise.
3764 (vec_extract_lo_<mode>): Likewise.
3765 (vec_extract_hi_v16hi): Likewise.
3766 (vec_extract_hi_<mode>): Likewise.
3767 (vec_set_hi_v16hi): Likewise.
3768 (vec_set_hi_<mode>): Likewise.
3769 (vec_set_lo_v16hi): Likewise.
3770 (vec_set_lo_<mode>): Likewise.
3771 (*vec_extract<mode>_0): New define_insn_and_split for HF
3773 (*vec_extracthf): New define_insn.
3774 (VEC_EXTRACT_MODE): Add HF vector modes.
3775 (PINSR_MODE): Add V8HF.
3776 (sse2p4_1): Likewise.
3777 (pinsr_evex_isa): Likewise.
3778 (<sse2p4_1>_pinsr<ssemodesuffix>): Adjust to support
3779 insert for V8HFmode.
3780 (pbroadcast_evex_isa): Add HF vector modes.
3781 (AVX2_VEC_DUP_MODE): Likewise.
3782 (VEC_INIT_MODE): Likewise.
3783 (VEC_INIT_HALF_MODE): Likewise.
3784 (avx2_pbroadcast<mode>): Adjust to support HF vector mode
3786 (avx2_pbroadcast<mode>_1): Likewise.
3787 (<avx512>_vec_dup<mode>_1): Likewise.
3788 (<avx512>_vec_dup<mode><mask_name>): Likewise.
3789 (<mask_codefor><avx512>_vec_dup_gpr<mode><mask_name>):
3792 2021-09-08 Guo, Xuepeng <xuepeng.guo@intel.com>
3793 H.J. Lu <hongjiu.lu@intel.com>
3794 Liu Hongtao <hongtao.liu@intel.com>
3795 Wang Hongyu <hongyu.wang@intel.com>
3796 Xu Dianhong <dianhong.xu@intel.com>
3798 * common/config/i386/cpuinfo.h (get_available_features):
3799 Detect FEATURE_AVX512FP16.
3800 * common/config/i386/i386-common.c
3801 (OPTION_MASK_ISA_AVX512FP16_SET,
3802 OPTION_MASK_ISA_AVX512FP16_UNSET,
3803 OPTION_MASK_ISA2_AVX512FP16_SET,
3804 OPTION_MASK_ISA2_AVX512FP16_UNSET): New.
3805 (OPTION_MASK_ISA2_AVX512BW_UNSET,
3806 OPTION_MASK_ISA2_AVX512BF16_UNSET): Add AVX512FP16.
3807 (ix86_handle_option): Handle -mavx512fp16.
3808 * common/config/i386/i386-cpuinfo.h (enum processor_features):
3809 Add FEATURE_AVX512FP16.
3810 * common/config/i386/i386-isas.h: Add entry for AVX512FP16.
3811 * config.gcc: Add avx512fp16intrin.h.
3812 * config/i386/avx512fp16intrin.h: New intrinsic header.
3813 * config/i386/cpuid.h: Add bit_AVX512FP16.
3814 * config/i386/i386-builtin-types.def: (FLOAT16): New primitive type.
3815 * config/i386/i386-builtins.c: Support _Float16 type for i386
3817 (ix86_register_float16_builtin_type): New function.
3818 (ix86_float16_type_node): New.
3819 * config/i386/i386-c.c (ix86_target_macros_internal): Define
3821 * config/i386/i386-expand.c (ix86_expand_branch): Support
3823 (ix86_prepare_fp_compare_args): Adjust TARGET_SSE_MATH &&
3824 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
3825 (ix86_expand_fp_movcc): Ditto.
3826 * config/i386/i386-isa.def: Add PTA define for AVX512FP16.
3827 * config/i386/i386-options.c (isa2_opts): Add -mavx512fp16.
3828 (ix86_valid_target_attribute_inner_p): Add avx512fp16 attribute.
3829 * config/i386/i386.c (ix86_get_ssemov): Use
3830 vmovdqu16/vmovw/vmovsh for HFmode/HImode scalar or vector.
3831 (ix86_get_excess_precision): Use
3832 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when TARGET_AVX512FP16
3834 (sse_store_index): Use SFmode cost for HFmode cost.
3835 (inline_memory_move_cost): Add HFmode, and perfer SSE cost over
3836 GPR cost for HFmode.
3837 (ix86_hard_regno_mode_ok): Allow HImode in sse register.
3838 (ix86_mangle_type): Add manlging for _Float16 type.
3839 (inline_secondary_memory_needed): No memory is needed for
3840 16bit movement between gpr and sse reg under
3842 (ix86_multiplication_cost): Adjust TARGET_SSE_MATH &&
3843 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
3844 (ix86_division_cost): Ditto.
3845 (ix86_rtx_costs): Ditto.
3846 (ix86_add_stmt_cost): Ditto.
3847 (ix86_optab_supported_p): Ditto.
3848 * config/i386/i386.h (VALID_AVX512F_SCALAR_MODE): Add HFmode.
3849 (SSE_FLOAT_MODE_SSEMATH_OR_HF_P): Add HFmode.
3850 (PTA_SAPPHIRERAPIDS): Add PTA_AVX512FP16.
3851 * config/i386/i386.md (mode): Add HFmode.
3852 (MODE_SIZE): Add HFmode.
3853 (isa): Add avx512fp16.
3854 (enabled): Handle avx512fp16.
3855 (ssemodesuffix): Add sh suffix for HFmode.
3856 (comm): Add mult, div.
3857 (plusminusmultdiv): New code iterator.
3858 (insn): Add mult, div.
3859 (*movhf_internal): Adjust for avx512fp16 instruction.
3860 (*movhi_internal): Ditto.
3861 (*cmpi<unord>hf): New define_insn for HFmode.
3862 (*ieee_s<ieee_maxmin>hf3): Likewise.
3863 (extendhf<mode>2): Likewise.
3864 (trunc<mode>hf2): Likewise.
3865 (float<floatunssuffix><mode>hf2): Likewise.
3866 (*<insn>hf): Likewise.
3867 (cbranchhf4): New expander.
3868 (movhfcc): Likewise.
3869 (<insn>hf3): Likewise.
3872 * config/i386/i386.opt: Add mavx512fp16.
3873 * config/i386/immintrin.h: Include avx512fp16intrin.h.
3874 * doc/invoke.texi: Add mavx512fp16.
3875 * doc/extend.texi: Add avx512fp16 Usage Notes.
3877 2021-09-08 liuhongt <hongtao.liu@intel.com>
3879 * common.opt: Support -fexcess-precision=16.
3880 * config/aarch64/aarch64.c (aarch64_excess_precision): Return
3881 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when
3882 EXCESS_PRECISION_TYPE_FLOAT16.
3883 * config/arm/arm.c (arm_excess_precision): Ditto.
3884 * config/i386/i386.c (ix86_get_excess_precision): Ditto.
3885 * config/m68k/m68k.c (m68k_excess_precision): Issue an error
3886 when EXCESS_PRECISION_TYPE_FLOAT16.
3887 * config/s390/s390.c (s390_excess_precision): Ditto.
3888 * coretypes.h (enum excess_precision_type): Add
3889 EXCESS_PRECISION_TYPE_FLOAT16.
3890 * doc/tm.texi (TARGET_C_EXCESS_PRECISION): Update documents.
3891 * doc/tm.texi.in (TARGET_C_EXCESS_PRECISION): Ditto.
3892 * doc/extend.texi (Half-Precision): Document
3893 -fexcess-precision=16.
3894 * flag-types.h (enum excess_precision): Add
3895 EXCESS_PRECISION_FLOAT16.
3896 * target.def (excess_precision): Update document.
3897 * tree.c (excess_precision_type): Set excess_precision_type to
3898 EXCESS_PRECISION_FLOAT16 when -fexcess-precision=16.
3900 2021-09-08 liuhongt <hongtao.liu@intel.com>
3902 * doc/extend.texi: (@node Floating Types): Adjust the wording.
3903 (@node Half-Precision): Ditto.
3905 2021-09-07 Takayuki 'January June' Suwa <jjsuwa_sys3175@yahoo.co.jp>
3908 * config/xtensa/xtensa.c (xtensa_emit_move_sequence): Add
3909 'CONST_INT_P (src)' to the condition of the block that tries to
3910 eliminate literal when loading integer contant.
3912 2021-09-07 David Faust <david.faust@oracle.com>
3914 * doc/extend.texi (BPF Type Attributes) New node.
3915 Document new preserve_access_index attribute.
3916 Document new preserve_access_index builtin.
3917 * doc/invoke.texi: Document -mco-re and -mno-co-re options.
3919 2021-09-07 David Faust <david.faust@oracle.com>
3921 * config/bpf/bpf.c: Adjust includes.
3922 (bpf_handle_preserve_access_index_attribute): New function.
3923 (bpf_attribute_table): Use it here.
3924 (bpf_builtins): Add BPF_BUILTIN_PRESERVE_ACCESS_INDEX.
3925 (bpf_option_override): Handle "-mco-re" option.
3926 (bpf_asm_init_sections): New.
3927 (TARGET_ASM_INIT_SECTIONS): Redefine.
3928 (bpf_file_end): New.
3929 (TARGET_ASM_FILE_END): Redefine.
3930 (bpf_init_builtins): Add "__builtin_preserve_access_index".
3931 (bpf_core_compute, bpf_core_get_index): New.
3932 (is_attr_preserve_access): New.
3933 (bpf_expand_builtin): Handle new builtins.
3934 (bpf_core_newdecl, bpf_core_is_maybe_aggregate_access): New.
3935 (bpf_core_walk): New.
3936 (bpf_resolve_overloaded_builtin): New.
3937 (TARGET_RESOLVE_OVERLOADED_BUILTIN): Redefine.
3939 (pass_bpf_core_attr): New RTL pass.
3940 * config/bpf/bpf-passes.def: New file.
3941 * config/bpf/bpf-protos.h (make_pass_bpf_core_attr): New.
3942 * config/bpf/coreout.c: New file.
3943 * config/bpf/coreout.h: Likewise.
3944 * config/bpf/t-bpf (TM_H): Add $(srcdir)/config/bpf/coreout.h.
3945 (coreout.o): New rule.
3946 (PASSES_EXTRA): Add $(srcdir)/config/bpf/bpf-passes.def.
3947 * config.gcc (bpf): Add coreout.h to extra_headers.
3948 Add coreout.o to extra_objs.
3949 Add $(srcdir)/config/bpf/coreout.c to target_gtfiles.
3951 2021-09-07 David Faust <david.faust@oracle.com>
3953 * btfout.c (get_btf_id): Function is no longer static.
3954 * ctfc.h: Expose it here.
3956 2021-09-07 David Faust <david.faust@oracle.com>
3958 * ctfc.c (ctf_lookup_tree_type): New function.
3961 2021-09-07 David Faust <david.faust@oracle.com>
3963 * ctfc.c (ctf_dtd_lookup): Function is no longer static.
3964 * ctfc.h: Analogous change.
3966 2021-09-07 David Faust <david.faust@oracle.com>
3968 * dwarf2out.c (lookup_type_die): Function is no longer static.
3969 * dwarf2out.h: Expose it here.
3971 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
3973 * dwarf2ctf.c (ctf_debug_finalize): Make it static.
3974 (ctf_debug_early_finish): New definition.
3975 (ctf_debug_finish): Likewise.
3976 * dwarf2ctf.h (ctf_debug_finalize): Remove declaration.
3977 (ctf_debug_early_finish): New declaration.
3978 (ctf_debug_finish): Likewise.
3979 * dwarf2out.c (dwarf2out_finish): Invoke ctf_debug_finish.
3980 (dwarf2out_early_finish): Invoke ctf_debug_early_finish.
3982 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
3984 * config/bpf/bpf.c (bpf_option_override): For BPF backend, disable LTO
3985 support when compiling for CO-RE.
3986 * config/bpf/bpf.opt: Add new command line option -mco-re.
3988 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
3990 * flag-types.h (enum debug_info_type): Add new enum
3991 DINFO_TYPE_BTF_WITH_CORE.
3992 (BTF_WITH_CORE_DEBUG): New bitmask.
3993 * flags.h (btf_with_core_debuginfo_p): New declaration.
3994 * opts.c (btf_with_core_debuginfo_p): New definition.
3996 2021-09-07 Jason Merrill <jason@redhat.com>
3998 * tree.h (error_operand_p): Change to inline function.
4000 2021-09-07 Aldy Hernandez <aldyh@redhat.com>
4002 * tree-ssa-threadedge.c (forwarder_block_p): Rename to...
4003 (empty_block_with_phis_p): ...this.
4004 (potentially_threadable_block): Same.
4005 (jump_threader::thread_through_normal_block): Same.
4007 2021-09-07 Eric Botcazou <ebotcazou@adacore.com>
4010 * dwarf2out.c (mark_base_types): New overloaded function.
4011 (dwarf2out_early_finish): Invoke it on the COMDAT type list as well
4012 as the compilation unit, and call move_marked_base_types afterward.
4014 2021-09-07 H.J. Lu <hjl.tools@gmail.com>
4017 * config/i386/i386-expand.c (ix86_expand_convert_uns_sisf_sse):
4019 (ix86_expand_vector_convert_uns_vsivsf): Likewise.
4021 2021-09-07 Richard Biener <rguenther@suse.de>
4023 PR tree-optimization/102226
4024 * tree-vect-loop.c (vect_transform_cycle_phi): Record
4025 the converted value for the epilogue PHI use.
4027 2021-09-07 Martin Liska <mliska@suse.cz>
4029 PR gcov-profile/80223
4030 * ipa-inline.c (can_inline_edge_p): Similarly to sanitizer
4031 options, do not inline when no_profile_instrument_function
4032 attributes are different in early inliner. It's fine to inline
4033 it after PGO instrumentation.
4035 2021-09-07 Richard Biener <rguenther@suse.de>
4037 PR tree-optimization/101555
4038 * tree-ssa-pre.c (translate_vuse_through_block): Do not
4039 perform an alias walk to determine the validity of the
4040 mem at the start of the block which is already guaranteed
4041 by means of prune_clobbered_mems.
4042 (phi_translate_1): Pass edge to translate_vuse_through_block.
4044 2021-09-07 Xionghu Luo <luoxhu@linux.ibm.com>
4047 * config/rs6000/rs6000.md (fmod<mode>3): New define_expand.
4048 (remainder<mode>3): Likewise.
4050 2021-09-07 YunQiang Su <yunqiang.su@cipunited.com>
4052 * config/mips/mips.c (mips_file_start): add .module for
4055 2021-09-06 Roger Sayle <roger@nextmovesoftware.com>
4057 * wide-int.cc (wi::clz): Reorder tests to ensure the result
4058 is zero for all negative values.
4060 2021-09-06 Tobias Burnus <tobias@codesourcery.com>
4062 * doc/invoke.texi (-foffload-options): Fix @opindex.
4064 2021-09-06 H.J. Lu <hjl.tools@gmail.com>
4067 * config/i386/i386-expand.c (ix86_split_xorsign): Use operands[2].
4068 * config/i386/i386.md (@xorsign<mode>3_1): Add non-destructive
4069 source alternative for AVX.
4071 2021-09-06 liuhongt <hongtao.liu@intel.com>
4073 PR middle-end/102182
4074 * optabs.c (expand_fix): Add from1 to avoid from being
4077 2021-09-06 Eric Botcazou <ebotcazou@adacore.com>
4079 * dwarf2out.c (modified_type_die): Deal with all array types earlier
4080 and use local variable consistently throughout the function.
4082 2021-09-06 Jakub Jelinek <jakub@redhat.com>
4084 PR tree-optimization/102207
4085 * match.pd: Don't demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
4086 were promoted from signed to wider unsigned type.
4088 2021-09-06 Andrew Pinski <apinski@marvell.com>
4090 PR tree-optimization/63184
4091 * match.pd: Add simplification of pointer_diff of two pointer_plus
4092 with addr_expr in the first operand of each pointer_plus.
4093 Add simplificatoin of ne/eq of two pointer_plus with addr_expr
4094 in the first operand of each pointer_plus.
4096 2021-09-06 Richard Biener <rguenther@suse.de>
4098 PR tree-optimization/102176
4099 * tree-vect-slp.c (vect_slp_gather_vectorized_scalar_stmts):
4101 (vect_bb_slp_scalar_cost): Use the computed set of
4102 vectorized scalar stmts instead of relying on the out-of-date
4103 and not accurate PURE_SLP_STMT.
4104 (vect_bb_vectorization_profitable_p): Compute the set
4105 of vectorized scalar stmts.
4107 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
4109 * gimple-range-path.cc (path_range_query::range_of_stmt): Remove
4110 GIMPLE_COND special casing.
4111 (path_range_query::range_defined_in_block): Use range_of_stmt
4112 instead of calling fold_range directly.
4114 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
4116 * gimple-range-path.cc (path_range_query::range_of_expr): Set
4117 m_undefined_path when appropriate.
4118 (path_range_query::internal_range_of_expr): Copy from range_of_expr.
4119 (path_range_query::unreachable_path_p): New.
4120 (path_range_query::precompute_ranges): Set m_undefined_path.
4121 * gimple-range-path.h (path_range_query::unreachable_path_p): New.
4122 (path_range_query::internal_range_of_expr): New.
4123 * tree-ssa-threadbackward.c (back_threader::find_taken_edge_cond):
4124 Use unreachable_path_p.
4126 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
4128 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
4129 Remove argument and call find_taken_edge.
4130 (back_threader::resolve_phi): Do not calculate taken edge before
4131 calling maybe_register_path.
4132 (back_threader::find_paths_to_names): Same.
4134 2021-09-05 Jeff Law <jlaw@localhost.localdomain>
4136 * config/h8300/h8300.md (QHSI2 mode iterator): New mode iterator.
4137 * config/h8300/testcompare.md (store_c): Update name, use new
4139 (store_neg_c, store_shifted_c): New patterns.
4141 2021-09-03 Segher Boessenkool <segher@kernel.crashing.org>
4144 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): On ELFv2 use r11
4145 instead of r12 for CR save, in all cases.
4147 2021-09-03 Andrew Pinski <apinski@marvell.com>
4149 * config/aarch64/aarch64-sve-builtins.cc (register_vector_type):
4150 Handle error_mark_node as the type of the type_decl.
4152 2021-09-03 Andrew Pinski <apinski@marvell.com>
4154 * config/aarch64/aarch64-builtins.c (struct aarch64_simd_type_info):
4156 (aarch64_simd_types): Likewise.
4157 (aarch64_simd_intOI_type_node): Likewise.
4158 (aarch64_simd_intCI_type_node): Likewise.
4159 (aarch64_simd_intXI_type_node): Likewise.
4160 * config/aarch64/aarch64.h (aarch64_fp16_type_node): Likewise.
4161 (aarch64_fp16_ptr_type_node): Likewise.
4162 (aarch64_bf16_type_node): Likewise.
4163 (aarch64_bf16_ptr_type_node): Likewise.
4165 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4167 * range-op.cc (operator_minus::op1_op2_relation_effect): Abstract
4169 (minus_op1_op2_relation_effect): ...here.
4170 (class operator_pointer_diff): New.
4171 (operator_pointer_diff::op1_op2_relation_effect): Call
4172 minus_op1_op2_relation_effect.
4173 (integral_table::integral_table): Add entry for POINTER_DIFF_EXPR.
4175 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4177 * tree-ssa-threadbackward.c (back_threader::thread_through_all_blocks):
4178 Add may_peel_loop_headers.
4179 (back_threader_registry::thread_through_all_blocks): Same.
4180 (try_thread_blocks): Pass may_peel_loop_headers argument.
4181 (pass_early_thread_jumps::execute): Same.
4183 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4185 * tree-ssa-threadedge.c (has_phis_p): New.
4186 (forwarder_block_p): New.
4187 (potentially_threadable_block): Call forwarder_block_p.
4188 (jump_threader::thread_around_empty_blocks): Call has_phis_p.
4189 (jump_threader::thread_through_normal_block): Call
4192 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4194 * tree-ssa-threadbackward.c (back_threader::dump): New.
4195 (back_threader::debug): New.
4196 (back_threader_profitability::profitable_path_p): Dump blocks
4197 even if we are bailing early.
4199 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4201 * tree-ssa-threadupdate.c (cancel_thread): New.
4202 (jump_thread_path_registry::thread_block_1): Use cancel_thread.
4203 (jump_thread_path_registry::mark_threaded_blocks): Same.
4204 (jump_thread_path_registry::register_jump_thread): Same.
4206 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4208 * tree-ssa-threadedge.c (jt_state::push): Only call methods for
4209 which objects are available.
4210 (jt_state::pop): Same.
4211 (jt_state::register_equiv): Same.
4212 (jt_state::register_equivs_on_edge): Same.
4214 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4216 * tree-ssa-threadedge.c (jump_threader::thread_across_edge):
4217 Move pop until after a thread is registered.
4219 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4221 * tree-ssa-threadupdate.c (debug): New.
4223 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4225 * gimple-range-trace.cc (push_dump_file::push_dump_file): New.
4226 (push_dump_file::~push_dump_file): New.
4227 (dump_ranger): Change dump_file temporarily while dumping
4229 * gimple-range-trace.h (class push_dump_file): New.
4231 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4233 * gimple-range-trace.cc (debug_seed_ranger): Remove static.
4234 (dump_ranger): Dump function name.
4236 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4238 * gimple-range-path.cc (path_range_query::range_defined_in_block):
4239 Adjust for non-null.
4240 (path_range_query::adjust_for_non_null_uses): New.
4241 (path_range_query::precompute_ranges): Call
4242 adjust_for_non_null_uses.
4243 * gimple-range-path.h: Add m_non_null and
4244 adjust_for_non_null_uses.
4246 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4248 * gimple-range-path.cc (path_range_query::dump): Dump path
4250 (path_range_query::precompute_ranges): Dump entire path.
4252 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4254 * value-relation.cc (relation_oracle::debug): New.
4255 * value-relation.h (relation_oracle::debug): New.
4257 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4259 * tree-ssa-loop-ch.c: Remove unnecessary include file.
4261 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4263 * gimple-range-fold.cc (fold_using_range::postfold_gcond_edges):
4264 Skip statements with no defining BB.
4265 * gimple-range-path.cc (path_range_query::range_defined_in_block):
4266 Do not get confused by statements with no defining BB.
4268 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4270 * gimple-range-fold.cc (adjust_imagpart_expr): Move from
4271 gimple_range_adjustment. Add support for constants.
4272 (adjust_realpart_expr): New.
4273 (gimple_range_adjustment): Move IMAGPART_EXPR code to
4274 adjust_imagpart_expr.
4275 * range-op.cc (integral_table::integral_table): Add entry for
4278 2021-09-03 Jakub Jelinek <jakub@redhat.com>
4280 * omp-expand.c (expand_omp_atomic_pipeline): Use
4281 IFN_ATOMIC_COMPARE_EXCHANGE instead of
4282 BUILT_IN_SYNC_VAL_COMPARE_AND_SWAP_? so that memory order
4285 2021-09-03 Jakub Jelinek <jakub@redhat.com>
4288 * tree.h (DECL_FIELD_ABI_IGNORED): Changed into rvalue only macro
4289 that is false if DECL_BIT_FIELD.
4290 (SET_DECL_FIELD_ABI_IGNORED, DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD,
4291 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD): Define.
4292 * tree-streamer-out.c (pack_ts_decl_common_value_fields): For
4293 DECL_BIT_FIELD stream DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead
4294 of DECL_FIELD_ABI_IGNORED.
4295 * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Use
4296 SET_DECL_FIELD_ABI_IGNORED instead of writing to
4297 DECL_FIELD_ABI_IGNORED and for DECL_BIT_FIELD use
4298 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead.
4299 * lto-streamer-out.c (hash_tree): For DECL_BIT_FIELD hash
4300 DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead of DECL_FIELD_ABI_IGNORED.
4302 2021-09-03 liuhongt <hongtao.liu@intel.com>
4305 * config/i386/amxbf16intrin.h : Remove macro check for __AMX_BF16__.
4306 * config/i386/amxint8intrin.h : Remove macro check for __AMX_INT8__.
4307 * config/i386/amxtileintrin.h : Remove macro check for __AMX_TILE__.
4309 2021-09-02 Martin Sebor <msebor@redhat.com>
4311 PR tree-optimization/17506
4313 * tree-ssa-uninit.c (warn_uninit): Remove conditional guarding note.
4315 2021-09-02 Richard Biener <rguenther@suse.de>
4317 * tree-ssa-loop-im.c (fill_always_executed_in_1): Refine
4318 fix for PR78185 and continue processing when leaving
4321 2021-09-02 Jakub Jelinek <jakub@redhat.com>
4323 PR tree-optimization/99591
4324 * match.pd: Demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
4327 2021-09-02 Richard Biener <rguenther@suse.de>
4330 2021-09-02 Richard Biener <rguenther@suse.de>
4332 PR tree-optimization/102155
4333 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
4334 over a part of the RPO array and do not recurse here.
4335 Dump blocks marked as always executed.
4336 (fill_always_executed_in): Walk over the RPO array and
4337 process loops whose header we run into.
4338 (loop_invariant_motion_in_fun): Compute the first RPO
4339 using rev_post_order_and_mark_dfs_back_seme in iteration
4340 order and pass that to fill_always_executed_in.
4342 2021-09-02 liuhongt <hongtao.liu@intel.com>
4344 * config/i386/i386-modes.def (FLOAT_MODE): Define ieee HFmode.
4345 * config/i386/i386.c (enum x86_64_reg_class): Add
4347 (merge_classes): Handle X86_64_SSEHF_CLASS.
4348 (examine_argument): Ditto.
4349 (construct_container): Ditto.
4350 (classify_argument): Ditto, and set HFmode/HCmode to
4352 (function_value_32): Return _FLoat16/Complex Float16 by
4354 (function_value_64): Return _Float16/Complex Float16 by SSE
4356 (ix86_print_operand): Handle CONST_DOUBLE HFmode.
4357 (ix86_secondary_reload): Require gpr as intermediate register
4358 to store _Float16 from sse register when sse4 is not
4360 (ix86_libgcc_floating_mode_supported_p): Enable _FLoat16 under
4362 (ix86_scalar_mode_supported_p): Ditto.
4363 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Defined.
4364 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add HFmode.
4365 (VALID_INT_MODE_P): Add HFmode and HCmode.
4366 * config/i386/i386.md (*pushhf_rex64): New define_insn.
4368 (*movhf_internal): Ditto.
4369 * doc/extend.texi (Half-Precision Floating Point): Documemt
4372 2021-09-02 Richard Biener <rguenther@suse.de>
4374 PR tree-optimization/102155
4375 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
4376 over a part of the RPO array and do not recurse here.
4377 Dump blocks marked as always executed.
4378 (fill_always_executed_in): Walk over the RPO array and
4379 process loops whose header we run into.
4380 (loop_invariant_motion_in_fun): Compute the first RPO
4381 using rev_post_order_and_mark_dfs_back_seme in iteration
4382 order and pass that to fill_always_executed_in.
4384 2021-09-02 YunQiang Su <syq@debian.org>
4387 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
4389 * config/mips/mips.c (mips_module_isa_name): New.
4390 mips_file_start: add .module mipsREV to all asm output
4392 2021-09-01 Jeff Law <jlaw@localhost.localdomain>
4394 PR tree-optimization/102152
4395 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Reduce a vector
4396 comparison to a scalar comparison before calling
4397 update_stmt_if_modified.
4399 2021-09-01 Andrew Pinski <apinski@marvell.com>
4402 * config/aarch64/aarch64.c (aarch64_expand_setmem):
4403 Check STRICT_ALIGNMENT before creating an overlapping
4406 2021-09-01 Martin Sebor <msebor@redhat.com>
4408 * gimple-ssa-warn-access.cc (get_size_range): Add argument.
4409 (check_access): Pass additional argument.
4410 (check_memop_access): Remove template and make a member function.
4411 (maybe_check_dealloc_call): Make a pass_waccess member function.
4412 (class pass_waccess): Add, rename, and remove members.
4413 (pass_waccess::pass_waccess): Adjust to name change.
4414 (pass_waccess::~pass_waccess): Same.
4415 (check_alloca): Make a member function.
4416 (check_alloc_size_call): Same.
4417 (check_strcat): Same.
4418 (check_strncat): Same.
4419 (check_stxcpy): Same.
4420 (check_stxncpy): Same.
4421 (check_strncmp): Same.
4422 (maybe_warn_rdwr_sizes): Rename...
4423 (pass_waccess::maybe_check_access_sizes): ...to this.
4424 (pass_waccess::check_call): Adjust to name changes.
4425 (pass_waccess::maybe_check_dealloc_call): Make a pass_waccess member
4427 (pass_waccess::execute): Adjust to name changes.
4428 * gimple-ssa-warn-access.h (check_memop_access): Remove.
4429 * pointer-query.cc (access_ref::phi): Handle null pointer.
4430 (access_ref::inform_access): Same.
4431 (pointer_query::put_ref): Modify a cached value, not a copy of it.
4432 (pointer_query::dump): New function.
4433 (compute_objsize_r): Avoid overwriting access_ref::bndrng. Cache
4435 * pointer-query.h (pointer_query::dump): Declare.
4436 * tree-ssa-strlen.c (get_range): Simplify. Use function query.
4437 (dump_strlen_info): Use function query.
4438 (printf_strlen_execute): Factor code out into pointer_query::put_ref.
4440 2021-09-01 Thomas Schwinge <thomas@codesourcery.com>
4442 * tree.c (walk_tree_1) <OMP_CLAUSE>: Simplify.
4444 2021-09-01 Iain Sandoe <iain@sandoe.co.uk>
4446 * doc/extend.texi: Document unavailable attribute.
4447 * print-tree.c (print_node): Handle unavailable attribute.
4448 * tree-core.h (struct tree_base): Add a bit to carry unavailability.
4449 * tree.c (error_unavailable_use): New.
4450 * tree.h (TREE_UNAVAILABLE): New.
4451 (error_unavailable_use): New.
4453 2021-09-01 Jakub Jelinek <jakub@redhat.com>
4455 PR tree-optimization/102124
4456 * tree-vect-patterns.c (vect_recog_widen_op_pattern): For ORIG_CODE
4457 MINUS_EXPR, if itype is unsigned with smaller precision than type,
4458 add an extra cast to signed variant of itype to ensure sign-extension.
4460 2021-09-01 Martin Liska <mliska@suse.cz>
4462 * graph.c (draw_cfg_node_succ_edges): Do not color fallthru
4463 edges and rather use colors for TRUE and FALSE edges.
4465 2021-09-01 Richard Biener <rguenther@suse.de>
4467 PR tree-optimization/93491
4468 * tree-ssa-pre.c (compute_avail): Set BB_MAY_NOTRETURN
4469 after processing the stmt itself. Do not consider
4470 pure functions possibly not returning. Properly avoid
4471 adding possibly trapping calls to EXP_GEN when there's
4472 a preceeding possibly not returning call.
4473 * tree-ssa-sccvn.c (vn_reference_may_trap): Conservatively
4476 2021-09-01 Richard Biener <rguenther@suse.de>
4478 PR tree-optimization/102139
4479 * tree-vectorizer.h (vec_base_alignments): Adjust hash-map
4480 type to record a std::pair of the stmt-info and the innermost
4482 (dr_vec_info::group): New member.
4483 * tree-vect-data-refs.c (vect_record_base_alignment): Adjust.
4484 (vect_compute_data_ref_alignment): Verify the recorded
4485 base alignment can be used.
4486 (data_ref_pair): Remove.
4487 (dr_group_sort_cmp): Adjust.
4488 (vect_analyze_data_ref_accesses): Store the group-ID in the
4489 dr_vec_info and operate on a vector of dr_vec_infos.
4491 2021-09-01 YunQiang Su <yunqiang.su@cipunited.com>
4493 * read-md.c (md_reader::handle_enum): support value assignation.
4494 * doc/md.texi: record define_c_enum value assignation support.
4496 2021-09-01 Jakub Jelinek <jakub@redhat.com>
4498 PR tree-optimization/102141
4499 * gimple-ssa-store-merging.c (bswap_view_convert): Add BEFORE
4500 argument. If false, emit stmts after gsi instead of before, and
4502 (bswap_replace): Adjust callers. When converting output of bswap,
4503 emit VIEW_CONVERT prepratation stmts after a copy of gsi instead
4506 2021-09-01 liuhongt <hongtao.liu@intel.com>
4508 * emit-rtl.c (validate_subreg): Get rid of all float-int
4511 2021-09-01 liuhongt <hongtao.liu@intel.com>
4514 2021-08-30 liuhongt <hongtao.liu@intel.com>
4516 * expmed.c (extract_bit_field_1): Make sure we're playing with
4517 integral modes before call extract_integral_bit_field.
4518 (extract_integral_bit_field): Add a parameter of type
4519 scalar_int_mode which corresponds to of tmode.
4520 And call extract_and_convert_fixed_bit_field instead of
4521 extract_fixed_bit_field and convert_extracted_bit_field.
4522 (extract_and_convert_fixed_bit_field): New function, it's a
4523 combination of extract_fixed_bit_field and
4524 convert_extracted_bit_field.
4526 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
4528 * tree.c (walk_tree_1) <OMP_CLAUSE_TILE>: Handle three operands.
4530 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
4532 * omp-general.h (omp_is_reference): Rename to...
4533 (omp_privatize_by_reference): ... this. Adjust all users...
4534 * omp-general.c: ... here, ...
4535 * gimplify.c: ... here, ...
4536 * omp-expand.c: ... here, ...
4537 * omp-low.c: ... here.
4539 2021-08-31 Martin Sebor <msebor@redhat.com>
4541 * gimple-ssa-warn-access.cc (maybe_warn_alloc_args_overflow): Test
4542 pointer element for equality to zero, not that of the cotaining
4545 2021-08-31 Martin Sebor <msebor@redhat.com>
4547 * gcc-rich-location.h (gcc_rich_location): Make ctor explicit.
4549 2021-08-31 Martin Sebor <msebor@redhat.com>
4551 * function.h (function): Add comments.
4552 (get_range_query): Same. Add attribute returns nonnull.
4554 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
4556 * expr.c (convert_modes): Don't use subreg_promoted_mode on a
4557 SUBREG if it can't be guaranteed to a SUBREG_PROMOTED_VAR_P set.
4558 Instead use the standard (safer) is_a <scalar_int_mode> idiom.
4560 2021-08-31 Jeff Law <jlaw@localhost.localdomain>
4562 * config.gcc (cris-*-elf, cris-*-none): Remove dbxelf.h from
4564 (m32r-*-elf, m32rle-*-elf, m32r-*-linux): Likewise.
4565 (mn10300-*-*, am33_2.0-*-linux*): Likewise.
4566 (xtensa*-*-elf, xtensa*-*-linux, xtensa*-*-uclinux): Likewise.
4567 (m32c-*-elf*, m32c-*-rtems*): Likewise.
4568 * config/cris/cris.h (DBX_NO_XREFS): Remove.
4569 (DBX_CONTIN_LENGTH, DBX_CONTIN_CHAR): Likewise.
4570 * config/m32r/m32r.h (DBXOUT_SOURCE_LINE): Likewise.
4571 (DBX_DEBUGGING_INFO, DBX_CONTIN_LENGTH): Likewise.
4572 * config/mn10300/mn10300.h (DEFAULT_GDB_EXTENSIONS): Likewise.
4573 * config/mn10300/linux.h (DBX_REGISTER_NAMES): Likewise.
4575 2021-08-31 Marcel Vollweiler <marcel@codesourcery.com>
4577 * gimplify.c (gimplify_scan_omp_clauses): Error handling. 'ancestor' only
4578 allowed on target constructs and only with particular other clauses.
4579 * omp-expand.c (expand_omp_target): Output of 'sorry, not supported' if
4581 * omp-low.c (check_omp_nesting_restrictions): Error handling. No nested OpenMP
4582 structs when 'ancestor' is used.
4583 (scan_omp_1_stmt): No usage of OpenMP runtime routines in a target region when
4585 * tree-pretty-print.c (dump_omp_clause): Append 'ancestor'.
4586 * tree.h (OMP_CLAUSE_DEVICE_ANCESTOR): Define macro.
4588 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
4590 * expr.c (convert_modes): Preserve SUBREG_PROMOTED_VAR_P when
4591 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
4593 * simplify-rtx.c (simplify_unary_operation_1) [SIGN_EXTEND]:
4594 Likewise, preserve SUBREG_PROMOTED_VAR_P when creating a (wider)
4595 partial subreg from a SUBREG_PROMOTED_VAR_P subreg. Generate
4596 SIGN_EXTEND of the SUBREG_REG when a subreg would be paradoxical.
4597 [ZERO_EXTEND]: Likewise, preserve SUBREG_PROMOTED_VAR_P when
4598 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
4599 subreg. Generate ZERO_EXTEND of the SUBREG_REG when a subreg
4600 would be paradoxical.
4602 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
4604 * combine.c (combine_simplify_rtx): Avoid converting an explicit
4605 TRUNCATE into a lowpart SUBREG on !TRULY_NOOP_TRUNCATION targets.
4606 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
4608 2021-08-31 Richard Biener <rguenther@suse.de>
4610 PR tree-optimization/102142
4611 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Fix
4612 condition under which to unset the visited flag.
4614 2021-08-31 Richard Biener <rguenther@suse.de>
4616 PR middle-end/102129
4617 * tree-ssa-ter.c (find_replaceable_in_bb): Do not move
4618 possibly trapping expressions across calls.
4620 2021-08-31 Jakub Jelinek <jakub@redhat.com>
4622 PR tree-optimization/102134
4623 * tree-ssa-ccp.c (bit_value_binop) <case RSHIFT_EXPR>: If sgn is
4624 UNSIGNED and r1val | r1mask has MSB set, ensure lzcount doesn't
4627 2021-08-31 Andrew Pinski <apinski@marvell.com>
4630 * collect-utils.c (setup_signals): New declaration.
4631 * collect-utils.h (setup_signals): New function.
4632 * collect2.c (handler): Delete.
4633 (main): Instead of manually setting up the signals,
4634 just call setup_signals.
4635 * lto-wrapper.c (main): Likewise.
4637 2021-08-31 Andrew Pinski <apinski@marvell.com>
4640 * config/i386/i386-protos.h (x86_output_aligned_bss):
4641 Change align argument to unsigned type.
4642 (x86_elf_aligned_decl_common): Likewise.
4643 * config/i386/i386.c (x86_elf_aligned_decl_common): Likewise.
4644 (x86_output_aligned_bss): Likewise.
4646 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
4648 * config/mips/mips.c (mips_module_isa_name): New.
4649 mips_file_start: add .module mipsREV to all asm output
4651 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
4653 * config/mips/mips.h (struct mips_cpu_info): define enum mips_isa;
4654 use enum instead of int for 'isa' member.
4655 * config.gcc, config/mips/mips.c, config/mips/mips-cpus.def,
4656 config/mips/netbsd.h: replace hardcoded numbers with enum.
4658 2021-08-31 liuhongt <hongtao.liu@intel.com>
4660 * config/i386/sse.md (*<avx512>_ucmp<mode>3_1): Change from
4661 define_split to define_insn_and_split.
4662 (*avx2_eq<mode>3): Removed.
4663 (<avx512>_eq<mode>3<mask_scalar_merge_name>): Adjust pattern
4664 (<avx512>_eq<mode>3<mask_scalar_merge_name>_1): Rename to ..
4665 (*<avx512>_eq<mode>3<mask_scalar_merge_name>_1): .. this, and
4667 (*avx2_gt<mode>3): Removed.
4668 (<avx512>_gt<mode>3<mask_scalar_merge_name>): Change from
4669 define_insn to define_expand, and adjust pattern.
4670 (UNSPEC_MASKED_EQ, UNSPEC_MASKED_GT): Removed.
4672 2021-08-30 David Malcolm <dmalcolm@redhat.com>
4675 * Makefile.in (ANALYZER_OBJS): Add analyzer/call-info.o.
4677 2021-08-30 Jason Merrill <jason@redhat.com>
4679 * doc/invoke.texi: Document -Wmissing-requires.
4681 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
4683 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Remove
4684 TARGET_EXTRA_BUILTINS guard.
4686 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
4688 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Change
4689 initialization of V2DI_type_node and unsigned_V2DI_type_node.
4691 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
4693 * config/rs6000/darwin.h (SUBTARGET_INIT_BUILTINS): Use the new
4694 decl when new_builtins_are_live.
4695 * config/rs6000/rs6000-builtin-new.def (__builtin_cfstring): New
4698 2021-08-30 Pat Haugen <pthaugen@linux.ibm.com>
4700 * config/rs6000/rs6000-cpus.def (ISA_3_1_MASKS_SERVER): Add
4701 OPTION_MASK_P10_FUSION_2STORE.
4702 (POWERPC_MASKS): Likewise.
4703 * config/rs6000/rs6000.c (rs6000_option_override_internal): Enable
4704 store fusion for Power10.
4705 (is_fusable_store): New.
4706 (power10_sched_reorder): Likewise.
4707 (rs6000_sched_reorder): Do Power10 specific reordering.
4708 (rs6000_sched_reorder2): Likewise.
4709 * config/rs6000/rs6000.opt: Add new option.
4711 2021-08-30 Richard Biener <rguenther@suse.de>
4713 PR tree-optimization/102128
4714 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
4715 Move scanning for if-converted scalar code to the caller
4716 and instead delay clearing the visited flag for profitable
4718 (vect_slp_region): Cost all subgraphs before scheduling.
4719 For if-converted BB vectorization scan for scalar COND_EXPRs
4720 and do not vectorize if any found and the cost model is
4723 2021-08-30 Richard Biener <rguenther@suse.de>
4725 * common.opt (fexceptions): Mark
4726 EnabledBy(fnon-call-exceptions).
4727 * doc/invoke.texi (fnon-call-exceptions): Document this
4728 enables -fexceptions.
4730 2021-08-30 Sebastian Huber <sebastian.huber@embedded-brains.de>
4732 * tsystem.h (abort): Define abort() if inhibit_libc is defined and it
4733 is not already defined.
4735 2021-08-30 liuhongt <hongtao.liu@intel.com>
4737 * expmed.c (extract_bit_field_1): Make sure we're playing with
4738 integral modes before call extract_integral_bit_field.
4739 (extract_integral_bit_field): Add a parameter of type
4740 scalar_int_mode which corresponds to of tmode.
4741 And call extract_and_convert_fixed_bit_field instead of
4742 extract_fixed_bit_field and convert_extracted_bit_field.
4743 (extract_and_convert_fixed_bit_field): New function, it's a
4744 combination of extract_fixed_bit_field and
4745 convert_extracted_bit_field.
4747 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
4749 * config/darwin.c (darwin_libc_has_function): Do not run
4750 the checks for x86 or modern Darwin. Make sure that there
4751 is a value set for darwin_macosx_version_min before testing.
4753 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
4755 * config/i386/darwin.h (CLEAR_INSN_CACHE): New.
4757 2021-08-28 Jan Hubicka <hubicka@ucw.cz>
4759 * ipa-modref-tree.h (modref_access_node::merge): Break out
4760 logic combining offsets and logic merging ranges to ...
4761 (modref_access_node::combined_offsets): ... here
4762 (modref_access_node::update2): ... here
4763 (modref_access_node::closer_pair_p): New member function.
4764 (modref_access_node::forced_merge): New member function.
4765 (modre_ref_node::insert): Do merging when table is full.
4767 2021-08-28 YunQiang Su <yunqiang.su@cipunited.com>
4770 * config.gcc: MIPS: use N64 ABI by default if the triple end
4771 with -gnuabi64, which is used by Debian since 2013.
4773 2021-08-28 Alexandre Oliva <oliva@adacore.com>
4775 * ipa-modref.c (analyze_function): Skip debug stmts.
4776 * tree-inline.c (estimate_num_insn): Consider builtins even
4777 without a cgraph_node.
4779 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
4781 * config/h8300/bitfield.md (cstore<mode>4): Remove expander.
4782 * config/h8300/h8300.c (h8300_expand_branch): Remove function.
4783 * config/h8300/h8300-protos.h (h8300_expadn_branch): Remove prototype.
4784 * config/h8300/h8300.md (eqne): New code iterator.
4785 (geultu, geultu_to_c): Similarly.
4786 * config/h8300/testcompare.md (cstore<mode>4): Dummy expander.
4787 (store_c_<mode>, store_c_i_<mode>): New define_insn_and_splits
4788 (cmp<mode>_c): New pattern
4790 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
4792 * tree-ssa-dom.c (reduce_vector_comparison_to_scalar_comparison): New
4794 (dom_opt_dom_walker::optimize_stmt): Use it.
4796 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
4798 * config/darwin.c (finalize_ctors): Add a section-start linker-
4800 (finalize_dtors): Likewise.
4801 * config/darwin.h (MIN_LD64_INIT_TERM_START_LABELS): New.
4803 2021-08-27 Bill Schmidt <wschmidt@linux.ibm.com>
4805 * config/rs6000/rs6000-call.c (rs6000-builtins.h): New #include.
4806 (rs6000_init_builtins): Call rs6000_init_generated_builtins. Skip the
4807 old initialization logic when new builtins are enabled.
4808 * config/rs6000/rs6000-gen-builtins.c (write_decls): Rename
4809 rs6000_autoinit_builtins to rs6000_init_generated_builtins.
4810 (write_init_file): Likewise.
4812 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
4814 * configure.ac (darwin2[[0-9]]* | darwin19*): Alter use of
4815 gcc_GAS_CHECK_FEATURE to remove an extraneous parameter.
4816 (amdgcn-* | gcn-*) Likewise.
4818 2021-08-27 Anthony Sharp <anthonysharp15@gmail.com>
4820 * symbol-summary.h: Added missing template keyword.
4822 2021-08-27 Richard Biener <rguenther@suse.de>
4824 PR tree-optimization/45178
4825 * tree-ssa-dce.c (find_obviously_necessary_stmts): For
4826 infinite loops without exit do not mark control dependent
4827 edges of the latch necessary.
4829 2021-08-27 konglin1 <lingling.kong@intel.com>
4832 * config/i386/sse.md: (<avx512>scattersi<mode>): Add mask operand to
4834 (<avx512>scattersi<mode>): Likewise.
4835 (*avx512f_scattersi<VI48F:mode>): Merge mask operand to set_dest.
4836 (*avx512f_scatterdi<VI48F:mode>): Likewise
4838 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
4840 * config/rs6000/rs6000.c (rs6000_builtin_md_vectorized_function): Add
4841 support for built-in functions MISC_BUILTIN_DIVWE, MISC_BUILTIN_DIVWEU,
4842 MISC_BUILTIN_DIVDE, MISC_BUILTIN_DIVDEU, P10_BUILTIN_CFUGED,
4843 P10_BUILTIN_CNTLZDM, P10_BUILTIN_CNTTZDM, P10_BUILTIN_PDEPD and
4844 P10_BUILTIN_PEXTD on Power10.
4846 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
4848 * config/rs6000/rs6000-call.c (builtin_function_type): Add unsigned
4849 signedness for some Power10 bifs.
4851 2021-08-27 David Edelsohn <dje.gcc@gmail.com>
4854 * config/rs6000/rs6000.c (rs6000_adjust_field_align): Use
4855 computed alignment if the entire struct has attribute packed.
4857 2021-08-27 liuhongt <hongtao.liu@intel.com>
4861 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
4862 IX86_BUILTIN_SHUFPD512, IX86_BUILTIN_SHUFPS512,
4863 IX86_BUILTIN_SHUFPD256, IX86_BUILTIN_SHUFPS,
4864 IX86_BUILTIN_SHUFPS256.
4865 (ix86_masked_all_ones): New function.
4867 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
4869 * config/i386/i386.md (*btr<mode>_1): Call force_reg unconditionally.
4870 (conditional moves with memory inputs splitters): Ditto.
4871 * config/i386/sse.md (one_cmpl<mode>2): Simplify.
4873 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
4875 * ipa-modref-tree.h (modref_access_node::try_merge_with): Restart
4876 search after merging.
4878 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
4880 * config/rs6000/rs6000-overload.def: Add remaining overloads.
4882 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
4884 * config/rs6000/rs6000-builtin-new.def: Add cell stanza.
4886 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
4888 * config/rs6000/rs6000-builtin-new.def: Add ieee128-hw, dfp,
4889 crypto, and htm stanzas.
4891 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
4893 * config/rs6000/rs6000-builtin-new.def: Add mma stanza.
4895 2021-08-26 Martin Sebor <msebor@redhat.com>
4897 * tree-ssa-uninit.c (warn_uninit): Refactor and simplify.
4898 (warn_uninit_phi_uses): Remove argument from calls to warn_uninit.
4899 (warn_uninitialized_vars): Same. Reduce visibility of locals.
4900 (warn_uninitialized_phi): Same.
4902 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
4904 * tree-ssa-ccp.c (get_individual_bits): Helper function to
4905 extract the individual bits from a widest_int constant (mask).
4906 (gray_code_bit_flips): New read-only table for effiently
4907 enumerating permutations/combinations of bits.
4908 (bit_value_binop) [LROTATE_EXPR, RROTATE_EXPR]: Handle rotates
4909 by unknown counts that are guaranteed less than the target
4910 precision and four or fewer unknown bits by enumeration.
4911 [LSHIFT_EXPR, RSHIFT_EXPR]: Likewise, also handle shifts by
4912 enumeration under the same conditions. Handle remaining
4913 shifts as a mask based upon the minimum possible shift value.
4915 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
4916 Richard Biener <rguenther@suse.de>
4918 * match.pd (shift transformations): Remove a redundant
4919 !POINTER_TYPE_P check.
4921 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
4924 * config/i386/i386.md (cmove reg-reg move elimination peephole2s):
4925 Set all_regs to true in the call to replace_rtx.
4927 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
4929 * ipa-modref-tree.c (test_insert_search_collapse): Update test.
4930 * ipa-modref-tree.h (modref_base_node::insert): Be smarter when
4931 hiting --param modref-max-refs limit.
4932 (modref_tree:insert_base): Be smarter when hitting
4933 --param modref-max-bases limit. Add new parameter REF.
4934 (modref_tree:insert): Update.
4935 (modref_tree:merge): Update.
4936 * ipa-modref.c (read_modref_records): Update.
4938 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
4940 * params.opt: (modref-max-adjustments): Add full stop.
4942 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
4944 * ipa-modref-tree.h (modref_ref_node::verify): New member
4946 (modref_ref_node::insert): Use it.
4947 (modref_ref_node::try_mere_with): Fix off by one error.
4949 2021-08-26 Martin Liska <mliska@suse.cz>
4950 Stefan Kneifel <stefan.kneifel@bluewin.ch>
4952 * cgraph.h (create_version_clone_with_body): Add new parameter.
4953 * cgraphclones.c: Likewise.
4954 * multiple_target.c (create_dispatcher_calls): Do not use
4956 (create_target_clone): Likewise here.
4958 2021-08-26 Jonathan Yong <10walls@gmail.com>
4960 * doc/extend.texi: Add note about reserved priorities
4961 to the constructor attribute.
4963 2021-08-25 Martin Sebor <msebor@redhat.com>
4965 * gimple-range-cache.cc (ssa_global_cache::dump): Avoid printing
4966 range table header alone.
4967 * gimple-range.cc (gimple_ranger::export_global_ranges): Same.
4969 2021-08-25 Jan Hubicka <hubicka@ucw.cz>
4971 * doc/invoke.texi: Document --param modref-max-adjustments.
4972 * ipa-modref-tree.c (test_insert_search_collapse): Update.
4973 (test_merge): Update.
4974 * ipa-modref-tree.h (struct modref_access_node): Add adjustments;
4975 (modref_access_node::operator==): Fix handling of access ranges.
4976 (modref_access_node::contains): Constify parameter; handle also
4977 mismatched parm offsets.
4978 (modref_access_node::update): New function.
4979 (modref_access_node::merge): New function.
4980 (unspecified_modref_access_node): Update constructor.
4981 (modref_ref_node::insert_access): Add record_adjustments parameter;
4983 (modref_ref_node::try_merge_with): New private function.
4984 (modref_tree::insert): New record_adjustments parameter.
4985 (modref_tree::merge): New record_adjustments parameter.
4986 (modref_tree::copy_from): Update.
4987 * ipa-modref.c (dump_access): Dump adjustments field.
4988 (get_access): Update constructor.
4989 (record_access): Update call of insert.
4990 (record_access_lto): Update call of insert.
4991 (merge_call_side_effects): Add record_adjustments parameter.
4992 (get_access_for_fnspec): Update.
4993 (process_fnspec): Update.
4994 (analyze_call): Update.
4995 (analyze_function): Update.
4996 (read_modref_records): Update.
4997 (ipa_merge_modref_summary_after_inlining): Update.
4998 (propagate_unknown_call): Update.
4999 (modref_propagate_in_scc): Update.
5000 * params.opt (param-max-modref-adjustments=): New.
5002 2021-08-25 Michael Meissner <meissner@linux.ibm.com>
5004 * config/rs6000/vsx.md (UNSPEC_XXSPLTIDP): Rename from
5006 (xxspltiw_v4si): Use vecperm type attribute.
5007 (xxspltiw_v4si_inst): Use vecperm type attribute.
5008 (xxspltiw_v4sf_inst): Likewise.
5009 (xxspltidp_v2df): Use vecperm type attribute. Use
5010 UNSPEC_XXSPLTIDP instead of UNSPEC_XXSPLTID.
5011 (xxspltidp_v2df_inst): Likewise.
5012 (xxsplti32dx_v4si): Use vecperm type attribute.
5013 (xxsplti32dx_v4si_inst): Likewise.
5014 (xxsplti32dx_v4sf_inst): Likewise.
5015 (xxblend_<mode>): Likewise.
5016 (xxpermx): Likewise.
5017 (xxpermx_inst): Likewise.
5020 2021-08-25 Lewis Hyatt <lhyatt@gmail.com>
5023 * coretypes.h (typedef diagnostic_input_charset_callback): Declare.
5024 * diagnostic.c (diagnostic_initialize_input_context): New function.
5025 * diagnostic.h (diagnostic_initialize_input_context): Declare.
5026 * input.c (default_charset_callback): New function.
5027 (file_cache::initialize_input_context): New function.
5028 (file_cache_slot::create): Added ability to convert the input
5029 according to the input context.
5030 (file_cache::file_cache): Initialize the new input context.
5031 (class file_cache_slot): Added new m_alloc_offset member.
5032 (file_cache_slot::file_cache_slot): Initialize the new member.
5033 (file_cache_slot::~file_cache_slot): Handle potentially offset buffer.
5034 (file_cache_slot::maybe_grow): Likewise.
5035 (file_cache_slot::needs_read_p): Handle NULL fp, which is now possible.
5036 (file_cache_slot::get_next_line): Likewise.
5037 * input.h (class file_cache): Added input context member.
5039 2021-08-25 Richard Biener <rguenther@suse.de>
5041 PR tree-optimization/102046
5042 * tree-vect-slp.c (vect_build_slp_tree_2): Conservatively
5043 update ->any_pattern when swapping operands.
5045 2021-08-25 Hongyu Wang <hongyu.wang@intel.com>
5048 * config/i386/i386.c (ix86_live_on_entry): Adjust comment.
5049 (ix86_decompose_address): Remove retval check for ASHIFT,
5050 allow non-canonical zero extend if AND mask covers ASHIFT
5052 (ix86_legitimate_address_p): Adjust condition for decompose.
5053 (ix86_rtx_costs): Adjust cost for lea with non-canonical
5055 Co-Authored by: Uros Bizjak <ubizjak@gmail.com>
5057 2021-08-25 Jiufu Guo <guojiufu@linux.ibm.com>
5059 PR tree-optimization/101145
5060 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
5062 (number_of_iterations_lt): Invoke above function.
5063 (adjust_cond_for_loop_until_wrap):
5064 Merge to number_of_iterations_until_wrap.
5065 (number_of_iterations_cond): Update invokes for
5066 adjust_cond_for_loop_until_wrap and number_of_iterations_lt.
5068 2021-08-25 konglin1 <lingling.kong@intel.com>
5071 * config/i386/avx512dqintrin.h (_mm512_fpclass_ps_mask): Fix
5073 (_mm512_mask_fpclass_ps_mask): Ditto.
5075 2021-08-25 Kewen Lin <linkw@linux.ibm.com>
5077 * config/rs6000/altivec.md (vec_unpacku_hi_v16qi): Remove.
5078 (vec_unpacku_hi_v8hi): Likewise.
5079 (vec_unpacku_lo_v16qi): Likewise.
5080 (vec_unpacku_lo_v8hi): Likewise.
5081 (vec_unpacku_hi_<VP_small_lc>): New define_expand.
5082 (vec_unpacku_lo_<VP_small_lc>): Likewise.
5084 2021-08-24 David Edelsohn <dje.gcc@gmail.com>
5086 * config/rs6000/aix.h (SYSTEM_IMPLICIT_EXTERN_C): Delete.
5087 * config/rs6000/aix71.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
5088 * config/rs6000/aix72.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
5089 * config/rs6000/aix73.h (TARGET_AIX_VERSION): Increase to 73.
5091 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
5093 PR middle-end/102031
5094 * simplify-rtx.c (simplify_truncation): When comparing precisions
5095 use "subreg_prec" variable, not "subreg_mode".
5097 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
5099 * config/rs6000/rs6000-builtin-new.def: Add power10 and power10-64
5102 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
5104 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Initialize
5105 various pointer type nodes.
5106 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add enum
5107 values for various pointer types.
5108 (ptr_V16QI_type_node): New macro.
5109 (ptr_V1TI_type_node): New macro.
5110 (ptr_V2DI_type_node): New macro.
5111 (ptr_V2DF_type_node): New macro.
5112 (ptr_V4SI_type_node): New macro.
5113 (ptr_V4SF_type_node): New macro.
5114 (ptr_V8HI_type_node): New macro.
5115 (ptr_unsigned_V16QI_type_node): New macro.
5116 (ptr_unsigned_V1TI_type_node): New macro.
5117 (ptr_unsigned_V8HI_type_node): New macro.
5118 (ptr_unsigned_V4SI_type_node): New macro.
5119 (ptr_unsigned_V2DI_type_node): New macro.
5120 (ptr_bool_V16QI_type_node): New macro.
5121 (ptr_bool_V8HI_type_node): New macro.
5122 (ptr_bool_V4SI_type_node): New macro.
5123 (ptr_bool_V2DI_type_node): New macro.
5124 (ptr_bool_V1TI_type_node): New macro.
5125 (ptr_pixel_type_node): New macro.
5126 (ptr_intQI_type_node): New macro.
5127 (ptr_uintQI_type_node): New macro.
5128 (ptr_intHI_type_node): New macro.
5129 (ptr_uintHI_type_node): New macro.
5130 (ptr_intSI_type_node): New macro.
5131 (ptr_uintSI_type_node): New macro.
5132 (ptr_intDI_type_node): New macro.
5133 (ptr_uintDI_type_node): New macro.
5134 (ptr_intTI_type_node): New macro.
5135 (ptr_uintTI_type_node): New macro.
5136 (ptr_long_integer_type_node): New macro.
5137 (ptr_long_unsigned_type_node): New macro.
5138 (ptr_float_type_node): New macro.
5139 (ptr_double_type_node): New macro.
5140 (ptr_long_double_type_node): New macro.
5141 (ptr_dfloat64_type_node): New macro.
5142 (ptr_dfloat128_type_node): New macro.
5143 (ptr_ieee128_type_node): New macro.
5144 (ptr_ibm128_type_node): New macro.
5145 (ptr_vector_pair_type_node): New macro.
5146 (ptr_vector_quad_type_node): New macro.
5147 (ptr_long_long_integer_type_node): New macro.
5148 (ptr_long_long_unsigned_type_node): New macro.
5150 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
5152 * config/rs6000/rs6000-builtin-new.def: Add power9-vector, power9,
5153 and power9-64 stanzas.
5155 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
5156 Tom de Vries <tdevries@suse.de>
5158 * config.gcc (nvptx-*-*): Define {c,c++}_target_objs.
5159 * config/nvptx/nvptx-protos.h (nvptx_cpu_cpp_builtins): Prototype.
5160 * config/nvptx/nvptx.h (TARGET_CPU_CPP_BUILTINS): Implement with
5161 a call to the new nvptx_cpu_cpp_builtins function in nvptx-c.c.
5162 * config/nvptx/t-nvptx (nvptx-c.o): New rule.
5163 * config/nvptx/nvptx-c.c: New source file.
5164 (nvptx_cpu_cpp_builtins): Move implementation here.
5166 2021-08-24 Martin Sebor <msebor@redhat.com>
5168 PR middle-end/101600
5169 PR middle-end/101977
5170 * gimple-ssa-warn-access.cc (maybe_warn_for_bound): Tighten up
5171 the phrasing of a warning.
5172 (check_access): Use the remaining size after subtracting any offset
5173 rather than the whole object size.
5174 * pointer-query.cc (access_ref::get_ref): Clear BASE0 flag if it's
5175 clear for any nonnull PHI argument.
5176 (compute_objsize): Clear argument.
5178 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
5180 * config/rs6000/rs6000-builtin-new.def: Add power8-vector stanza.
5182 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
5184 * config/rs6000/rs6000-builtin-new.def: Add power7 and power7-64
5187 2021-08-24 Andrew MacLeod <amacleod@redhat.com>
5189 * value-relation.cc (rr_transitive_table): New.
5190 (relation_transitive): New.
5191 (value_relation::swap): Remove.
5192 (value_relation::apply_transitive): New.
5193 (relation_oracle::relation_oracle): Allocate a new tmp bitmap.
5194 (relation_oracle::register_relation): Call register_transitives.
5195 (relation_oracle::register_transitives): New.
5196 * value-relation.h (relation_oracle): Add new temporary bitmap and
5199 2021-08-24 H.J. Lu <hjl.tools@gmail.com>
5202 * config/i386/i386-expand.c (ix86_expand_vector_move): Broadcast
5203 from integer to a pseudo vector register.
5205 2021-08-24 Richard Biener <rguenther@suse.de>
5207 PR tree-optimization/100089
5208 * tree-vectorizer.h (vect_slp_bb): Rename to ...
5209 (vect_slp_if_converted_bb): ... this and get the original
5210 loop as new argument.
5211 * tree-vectorizer.c (try_vectorize_loop_1): Revert previous fix,
5212 pass original loop to vect_slp_if_converted_bb.
5213 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
5214 If orig_loop was passed scan the not vectorized stmts
5215 for COND_EXPRs and force not profitable if found.
5216 (vect_slp_region): Pass down all SLP instances to costing
5217 if orig_loop was specified.
5218 (vect_slp_bbs): Pass through orig_loop.
5219 (vect_slp_bb): Rename to ...
5220 (vect_slp_if_converted_bb): ... this and get the original
5221 loop as new argument.
5222 (vect_slp_function): Adjust.
5224 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
5227 * config/arm/arm.md (attribute arch): Add fix_vlldm.
5228 (arch_enabled): Use it.
5229 * config/arm/vfp.md (lazy_store_multiple_insn): Add alternative to
5230 use when erratum mitigation is needed.
5232 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
5235 * config/arm/arm.opt (mfix-cmse-cve-2021-35465): New option.
5236 * doc/invoke.texi (Arm Options): Document it.
5237 * config/arm/arm-cpus.in (quirk_vlldm): New feature bit.
5238 (ALL_QUIRKS): Add quirk_vlldm.
5239 (cortex-m33): Add quirk_vlldm.
5240 (cortex-m35p, cortex-m55): Likewise.
5241 * config/arm/arm.c (arm_option_override): Enable fix_vlldm if
5242 targetting an affected CPU and not explicitly controlled on
5245 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
5247 * config/arm/vfp.md (lazy_store_multiple_insn): Rewrite as valid RTL.
5248 (lazy_load_multiple_insn): Likewise.
5250 2021-08-24 liuhongt <hongtao.liu@intel.com>
5253 * config/i386/sse.md (<avx512>_vternlog<mode><sd_maskz_name>):
5254 Enable avx512 embedded broadcast.
5255 (*<avx512>_vternlog<mode>_all): Ditto.
5256 (<avx512>_vternlog<mode>_mask): Ditto.
5258 2021-08-24 liuhongt <hongtao.liu@intel.com>
5261 * config/i386/i386.c (ix86_rtx_costs): Define cost for
5263 * config/i386/i386.h (STRIP_UNARY): New macro.
5264 * config/i386/predicates.md (reg_or_notreg_operand): New
5266 * config/i386/sse.md (*<avx512>_vternlog<mode>_all): New define_insn.
5267 (*<avx512>_vternlog<mode>_1): New pre_reload
5268 define_insn_and_split.
5269 (*<avx512>_vternlog<mode>_2): Ditto.
5270 (*<avx512>_vternlog<mode>_3): Ditto.
5271 (any_logic1,any_logic2): New code iterator.
5272 (logic_op): New code attribute.
5273 (ternlogsuffix): Extend to VNxDF and VNxSF.
5275 2021-08-24 Richard Biener <rguenther@suse.de>
5277 * doc/invoke.texi (vect-inner-loop-cost-factor): Adjust.
5278 * params.opt (--param vect-inner-loop-cost-factor): Adjust
5280 * tree-vect-loop.c (vect_analyze_loop_form): Initialize
5281 inner_loop_cost_factor to the minimum of the estimated number
5282 of iterations of the inner loop and vect-inner-loop-cost-factor.
5284 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
5285 Richard Biener <rguenther@suse.de>
5287 * config/i386/i386-features.c (compute_convert_gain): Provide
5288 more accurate values for CONST_INT, when optimizing for size.
5289 * config/i386/i386.c (COSTS_N_BYTES): Move definition from here...
5290 * config/i386/i386.h (COSTS_N_BYTES): to here.
5292 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
5293 Jakub Jelinek <jakub@redhat.com>
5295 PR middle-end/102029
5296 * match.pd (shift transformations): Add an additional check for
5297 !POINTER_TYPE_P in the recently added left shift transformation.
5299 2021-08-24 liuhongt <hongtao.liu@intel.com>
5301 PR tree-optimization/100089
5302 * tree-vectorizer.c (try_vectorize_loop_1): Disable slp in
5303 loop vectorizer when cost model is very-cheap.
5305 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
5307 * config/rs6000/rs6000-gen-builtins.c (parse_bif_entry): Don't call
5308 asprintf, which is not available on AIX.
5310 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
5312 * config.gcc (target_gtfiles): Add ./rs6000-builtins.h.
5313 * config/rs6000/t-rs6000 (EXTRA_GTYPE_DEPS): Set.
5315 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
5317 * config.gcc (powerpc*-*-*): Add rs6000-builtins.o to extra_objs.
5318 * config/rs6000/rs6000-gen-builtins.c (main): Close init_file
5320 * config/rs6000/t-rs6000 (rs6000-gen-builtins.o): New target.
5321 (rbtree.o): Likewise.
5322 (rs6000-gen-builtins): Likewise.
5323 (rs6000-builtins.c): Likewise.
5324 (rs6000-builtins.h): Likewise.
5325 (rs6000.o): Add dependency.
5326 (EXTRA_HEADERS): Add rs6000-vecdefines.h.
5327 (rs6000-vecdefines.h): New target.
5328 (rs6000-builtins.o): Likewise.
5329 (rs6000-call.o): Add rs6000-builtins.h as a dependency.
5330 (rs6000-c.o): Likewise.
5332 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
5335 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace):
5336 Diagnose buffer overrun.
5337 (safe_inc_pos): Fix overrun detection.
5338 (match_identifier): Diagnose buffer overrun.
5339 (match_integer): Likewise.
5340 (match_to_right_bracket): Likewise.
5342 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
5344 * ipa-modref-tree.h (modref_access_node::range_info_useful_p):
5345 Improve range compare.
5346 (modref_access_node::contains): New member function.
5347 (modref_access_node::search): Remove.
5348 (modref_access_node::insert): Be smarter about subaccesses.
5350 2021-08-23 Thomas Schwinge <thomas@codesourcery.com>
5352 * config/i386/i386-options.c (ix86_omp_device_kind_arch_isa)
5353 <omp_device_arch> [ACCEL_COMPILER]: Match "intel_mic".
5354 * config/i386/t-omp-device (omp-device-properties-i386) <arch>:
5357 2021-08-23 Jeff Law <jlaw@localhost.localdomain>
5359 * config/h8300/h8300-protos.h (h8300_expand_epilogue): Add new
5361 * config/h8300/jumpcall.md (call, call_value): Restrict to
5362 !SIBLING_CALL_P cases.
5363 (subcall, sibcall_value): New patterns & expanders.
5364 * config/h8300/proepi.md (epilogue): Pass new argument to
5365 h8300_expand_epilogue.
5366 (sibcall_epilogue): New expander.
5367 * config/h8300/h8300.c (h8300_expand_epilogue): Handle sibcall
5369 (h8300_ok_for_sibcall_p): New function.
5370 (TARGET_FUNCTION_OK_FOR_SIBCALL): define.
5372 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
5374 * simplify-rtx.c (simplify_unary_operation_1): [TRUNCATE]:
5375 Handle case where the operand is already the desired mode.
5377 2021-08-23 Richard Biener <rguenther@suse.de>
5380 * tree-ssa-structalias.c (ipa_pta_execute): Check in_other_partition
5381 in addition to has_gimple_body.
5383 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
5385 PR middle-end/101949
5386 * ipa-modref.c (analyze_ssa_name_flags): Fix merging of
5389 2021-08-23 Martin Liska <mliska@suse.cz>
5391 * doc/invoke.texi: Put the option out of -mxl-mode-app-model
5394 2021-08-23 Richard Biener <rguenther@suse.de>
5396 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5397 Properly scale the inner loop cost only once.
5399 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
5401 * tree-ssa-ccp.c (bit_value_binop) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]:
5402 Provide bounds for unsigned (and signed with non-negative operands)
5403 division and modulus.
5405 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
5407 * simplify-rtx.c (simplify_truncation): Generalize simplification
5408 of (truncate:A (subreg:B X)).
5409 (simplify_unary_operation_1) [FLOAT_TRUNCATE, FLOAT_EXTEND,
5410 SIGN_EXTEND, ZERO_EXTEND]: Handle cases where the operand
5411 already has the desired machine mode.
5412 (test_scalar_int_ops): Add tests that useless extensions and
5413 truncations are optimized away.
5414 (test_scalar_int_ext_ops): New self-test function to confirm
5415 that truncations of extensions are correctly simplified.
5416 (test_scalar_int_ext_ops2): New self-test function to check
5417 truncations of truncations, extensions of extensions, and
5418 truncations of extensions.
5419 (test_scalar_ops): Call the above two functions with a
5420 representative sampling of integer machine modes.
5422 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
5424 * match.pd (shift transformations): Change the sign of an
5425 LSHIFT_EXPR if it reduces the number of explicit conversions.
5427 2021-08-23 Jakub Jelinek <jakub@redhat.com>
5429 PR tree-optimization/86723
5430 * gimple-ssa-store-merging.c (find_bswap_or_nop_finalize): Add
5431 cast64_to_32 argument, set *cast64_to_32 to false, unless n is
5432 non-memory permutation of 64-bit src which only has bytes of
5433 0 or [5..8] and n->range is 4.
5434 (find_bswap_or_nop): Add cast64_to_32 and mask arguments, adjust
5435 find_bswap_or_nop_finalize caller, support bswap with some bytes
5436 zeroed, as long as at least two bytes are not zeroed.
5437 (bswap_replace): Add mask argument and handle masking of bswap
5439 (maybe_optimize_vector_constructor): Adjust find_bswap_or_nop
5440 caller, punt if cast64_to_32 or mask is not all ones.
5441 (pass_optimize_bswap::execute): Adjust find_bswap_or_nop_finalize
5442 caller, for now punt if cast64_to_32.
5444 2021-08-23 Richard Biener <rguenther@suse.de>
5446 PR tree-optimization/79334
5447 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Record
5448 a type also for COMPONENT_REFs.
5449 (vn_reference_may_trap): Check ARRAY_REF with constant index
5450 against the array domain.
5452 2021-08-23 liuhongt <hongtao.liu@intel.com>
5455 * config/i386/sse.md (*avx512f_pshufb_truncv8hiv8qi_1): Add
5456 TARGET_AVX512BW to condition.
5458 2021-08-23 Jakub Jelinek <jakub@redhat.com>
5461 * dwarf2out.c (gen_variable_die): Add DW_AT_location for global
5462 register variables already during early_dwarf if possible.
5464 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
5466 * config/arm/arm_mve.h: Fix __arm_vctp16q return type.
5468 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
5471 * config/arm/arm.opt: Fix typo.
5472 * config/arm/t-rmprofile: Fix typo.
5474 2021-08-23 Jakub Jelinek <jakub@redhat.com>
5476 * tree.h (OMP_CLAUSE_GRAINSIZE_STRICT): Define.
5477 (OMP_CLAUSE_NUM_TASKS_STRICT): Define.
5478 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_GRAINSIZE,
5479 case OMP_CLAUSE_NUM_TASKS>: Print strict: modifier.
5480 * omp-expand.c (expand_task_call): Use GOMP_TASK_FLAG_STRICT in iflags
5481 if either grainsize or num_tasks clause has the strict modifier.
5483 2021-08-23 Martin Liska <mliska@suse.cz>
5485 * dbgcnt.def (DEBUG_COUNTER): New counter.
5486 * gimple.c (gimple_call_arg_flags): Use it in IPA PTA.
5488 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
5490 * ipa-modref.c (analyze_ssa_name_flags): Improve handling of return slot.
5492 2021-08-23 Xi Ruoyao <xry111@mengyan1223.wang>
5495 * config/mips/mips-protos.h (mips_msa_output_shift_immediate):
5497 * config/mips/mips.c (mips_msa_output_shift_immediate): New
5499 * config/mips/mips-msa.md (vashl<mode>3, vashr<mode>3,
5500 vlshr<mode>3): Call it.
5502 2021-08-22 Jan Hubicka <hubicka@ucw.cz>
5503 Martin Liska <mliska@suse.cz>
5505 PR middle-end/101949
5506 * ipa-modref.c (analyze_ssa_name_flags): Indirect call implies
5509 2021-08-21 Dragan Mladjenovic <OT_Dragan.Mladjenovic@mediatek.com>
5511 * config/mips/mips.c (mips_function_rodata_section,
5512 TARGET_ASM_FUNCTION_RODATA_SECTION): Removed.
5514 2021-08-21 John David Anglin <danglin@gcc.gnu.org>
5516 * config/pa/pa.c (pa_asm_output_aligned_common): Remove warning.
5518 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
5520 * configure.ac (thread-local storage support): Remove tls_first_major
5521 and tls_first_minor. Use "$conftest_s" to check support.
5522 * configure: Regenerate.
5524 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
5526 * configure.ac: Fixup formatting.
5528 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
5530 * acinclude.m4 (gcc_GAS_CHECK_FEATURE): Remove third argument and ...
5531 * configure.ac: ... update all callers.
5533 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
5536 * acinclude.m4 (_gcc_COMPUTE_GAS_VERSION, _gcc_GAS_VERSION_GTE_IFELSE)
5537 (gcc_GAS_VERSION_GTE_IFELSE): Remove.
5538 (gcc_GAS_CHECK_FEATURE): Do not handle in-tree case specially.
5539 * configure.ac: Remove gcc_cv_gas_major_version, gcc_cv_gas_minor_version.
5540 Remove remaining checks for in-tree assembler.
5541 * configure: Regenerate.
5543 2021-08-20 Jeff Law <jlaw@localhost.localdomain>
5545 * config/h8300/h8300.c (shift_alg_hi): Improve arithmetic shift right
5546 by 15 bits for H8/300H and H8/S. Improve logical shifts by 12
5548 (shift_alg_si): Improve arithmetic right shift by 28-30 bits for
5549 H8/300H. Improve arithmetic shift right by 15 bits for H8/S.
5550 Improve logical shifts by 27 bits for H8/S.
5551 (get_shift_alg): Corresponding changes.
5552 (h8300_option_override): Revert to loops for -Os when profitable.
5554 2021-08-20 Richard Biener <rguenther@suse.de>
5556 * tree-vect-data-refs.c (dr_group_sort_cmp): Do not compare
5558 (vect_analyze_data_ref_accesses): Likewise. Assign the BB
5559 index as group_id when dataref_groups were not computed.
5560 * tree-vect-slp.c (vect_slp_bbs): Bump current_group when
5561 we advace to the next BB.
5563 2021-08-20 Jakub Jelinek <jakub@redhat.com>
5565 * omp-builtins.def (BUILT_IN_GOMP_WARNING, BUILT_IN_GOMP_ERROR): New
5568 2021-08-20 Martin Liska <mliska@suse.cz>
5570 PR gcov-profile/89961
5571 * gcov.c (make_gcov_file_name): Rewrite using std::string.
5572 (mangle_name): Simplify, do not used the second argument.
5573 (strip_extention): New function.
5574 (get_md5sum): Likewise.
5575 (get_gcov_intermediate_filename): Handle properly -p and -x
5577 (output_gcov_file): Use string type.
5578 (generate_results): Likewise.
5579 (md5sum_to_hex): Remove.
5581 2021-08-20 Michael Meissner <meissner@linux.ibm.com>
5583 * config/rs6000/altivec.md (UNSPEC_XXEVAL): Move to vsx.md.
5584 (UNSPEC_XXSPLTIW): Move to vsx.md.
5585 (UNSPEC_XXSPLTID): Move to vsx.md.
5586 (UNSPEC_XXSPLTI32DX): Move to vsx.md.
5587 (UNSPEC_XXBLEND): Move to vsx.md.
5588 (UNSPEC_XXPERMX): Move to vsx.md.
5589 (VM3): Move to vsx.md.
5590 (VM3_char): Move to vsx.md.
5591 (xxspltiw_v4si): Move to vsx.md.
5592 (xxspltiw_v4sf): Move to vsx.md.
5593 (xxspltiw_v4sf_inst): Move to vsx.md.
5594 (xxspltidp_v2df): Move to vsx.md.
5595 (xxspltidp_v2df_inst): Move to vsx.md.
5596 (xxsplti32dx_v4si_inst): Move to vsx.md.
5597 (xxsplti32dx_v4sf): Move to vsx.md.
5598 (xxsplti32dx_v4sf_inst): Move to vsx.md.
5599 (xxblend_<mode>): Move to vsx.md.
5600 (xxpermx): Move to vsx.md.
5601 (xxpermx_inst): Move to vsx.md.
5602 * config/rs6000/vsx.md (UNSPEC_XXEVAL): Move from altivec.md.
5603 (UNSPEC_XXSPLTIW): Move from altivec.md.
5604 (UNSPEC_XXSPLTID): Move from altivec.md.
5605 (UNSPEC_XXSPLTI32DX): Move from altivec.md.
5606 (UNSPEC_XXBLEND): Move from altivec.md.
5607 (UNSPEC_XXPERMX): Move from altivec.md.
5608 (VM3): Move from altivec.md.
5609 (VM3_char): Move from altivec.md.
5610 (xxspltiw_v4si): Move from altivec.md.
5611 (xxspltiw_v4sf): Move from altivec.md.
5612 (xxspltiw_v4sf_inst): Move from altivec.md.
5613 (xxspltidp_v2df): Move from altivec.md.
5614 (xxspltidp_v2df_inst): Move from altivec.md.
5615 (xxsplti32dx_v4si_inst): Move from altivec.md.
5616 (xxsplti32dx_v4sf): Move from altivec.md.
5617 (xxsplti32dx_v4sf_inst): Move from altivec.md.
5618 (xxblend_<mode>): Move from altivec.md.
5619 (xxpermx): Move from altivec.md.
5620 (xxpermx_inst): Move from altivec.md.
5622 2021-08-19 Roger Sayle <roger@nextmovesoftware.com>
5624 * tree-vect-generic.c (expand_vector_operations_1): Use either
5625 gimplify_build1 or gimplify_build2 instead of gimple_build_assign
5626 when constructing scalar splat expressions.
5628 2021-08-19 Peter Bergner <bergner@linux.ibm.com>
5631 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Cast
5632 pointer to __vector_pair *.
5634 2021-08-19 Martin Sebor <msebor@redhat.com>
5636 * gimple-range.cc: Add comments.
5637 * gimple-range.h: Same.
5639 2021-08-19 Martin Sebor <msebor@redhat.com>
5641 PR middle-end/101984
5642 * gimple-ssa-warn-access.cc (pass_waccess::execute): Also call
5645 2021-08-19 Jeff Law <jlaw@localhost.localdomain>
5647 * config.gcc (h8300-*-elf*): Do not include dbxelf.h.
5648 (h8300-*-linux*, v850-*-rtems*, v850*-elf*): Likewise.
5649 * config/v850/v850.h (DEFAULT_GDB_EXTENSIONS): Remove.
5651 2021-08-19 Jakub Jelinek <jakub@redhat.com>
5653 PR middle-end/101950
5654 * optabs.c (expand_clrsb_using_clz): New function.
5655 (expand_unop): Use it as another clrsb expansion fallback.
5657 2021-08-19 liuhongt <hongtao.liu@intel.com>
5660 2021-07-28 liuhongt <hongtao.liu@intel.com>
5663 * config/i386/i386.h (processor_costs): Add new member
5665 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
5666 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
5667 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
5668 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
5669 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
5670 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
5671 generic_cost, core_cost): Initialize integer_to_sse same value
5673 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
5674 * config/i386/i386.c (ix86_builtin_vectorization_cost):
5675 Use integer_to_sse instead of sse_op to calculate the cost of
5678 2021-08-18 Iain Sandoe <iain@sandoe.co.uk>
5680 * config.gcc: Include rpath.opt for Darwin.
5681 * config/darwin.h (DRIVER_SELF_SPECS): Handle -rpath.
5683 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
5686 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
5689 2021-08-18 Jonathan Wright <jonathan.wright@arm.com>
5691 * config/aarch64/arm_neon.h (vld3_lane_f64): Use float RTL
5692 pattern and type cast.
5693 (vld4_lane_f32): Use float RTL pattern.
5694 (vld4q_lane_f64): Use float type cast.
5696 2021-08-18 Jan Hubicka <hubicka@ucw.cz>
5698 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Check also
5701 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
5703 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor): Extend.
5704 (test_map_of_type_with_ctor_and_dtor_expand): Add function.
5705 (hash_map_tests_c_tests): Call it.
5707 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
5709 * ggc.h (enum ggc_collect): New.
5710 (ggc_collect): Use it.
5711 * ggc-page.c: Adjust.
5712 * ggc-common.c: Likewise.
5713 * ggc-tests.c: Likewise.
5714 * read-rtl-function.c: Likewise.
5715 * selftest-run-tests.c: Likewise.
5716 * doc/gty.texi (Invoking the garbage collector): Likewise.
5718 2021-08-18 liuhongt <hongtao.liu@intel.com>
5721 * config/i386/i386.h (TARGET_V2DF_REDUCTION_PREFER_HADDPD):
5723 * config/i386/sse.md (*sse3_haddv2df3_low): Add
5724 TARGET_V2DF_REDUCTION_PREFER_HADDPD.
5725 (*sse3_hsubv2df3_low): Ditto.
5726 * config/i386/x86-tune.def
5727 (X86_TUNE_V2DF_REDUCTION_PREFER_HADDPD): New tune.
5729 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
5731 * gimple-range-gori.cc (gori_compute::gori_compute): Enable tracing.
5732 (gori_compute::compute_operand_range): Add tracing.
5733 (gori_compute::logical_combine): Ditto.
5734 (gori_compute::compute_logical_operands): Ditto.
5735 (gori_compute::compute_operand1_range): Ditto.
5736 (gori_compute::compute_operand2_range): Ditto.
5737 (gori_compute::outgoing_edge_range_p): Ditto.
5738 * gimple-range-gori.h (class gori_compute): Add range_tracer.
5740 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
5742 * flag-types.h (enum evrp_mode): Adjust evrp-mode values.
5743 * gimple-range-cache.cc (DEBUG_RANGE_CACHE): Relocate from.
5744 * gimple-range-trace.h (DEBUG_RANGE_CACHE): Here.
5745 * params.opt (--param=evrp-mode): Adjust options.
5747 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
5749 * Makefile.in (OBJS): Add gimple-range-trace.o.
5750 * gimple-range-cache.h (enable_new_values): Remove unused prototype.
5751 * gimple-range-fold.cc: Adjust headers.
5752 * gimple-range-trace.cc: New.
5753 * gimple-range-trace.h: New.
5754 * gimple-range.cc (gimple_ranger::gimple_ranger): Enable tracer.
5755 (gimple_ranger::range_of_expr): Add tracing.
5756 (gimple_ranger::range_on_entry): Ditto.
5757 (gimple_ranger::range_on_exit): Ditto.
5758 (gimple_ranger::range_on_edge): Ditto.
5759 (gimple_ranger::fold_range_internal): Ditto.
5760 (gimple_ranger::dump_bb): Do not calculate edge range twice.
5761 (trace_ranger::*): Remove.
5762 (enable_ranger): Never create a trace_ranger.
5763 (debug_seed_ranger): Move to gimple-range-trace.cc.
5764 (dump_ranger): Ditto.
5765 (debug_ranger): Ditto.
5766 * gimple-range.h: Include gimple-range-trace.h.
5767 (range_on_entry, range_on_exit): No longer virtual.
5768 (class trace_ranger): Remove.
5769 (DEBUG_RANGE_CACHE): Move to gimple-range-trace.h.
5771 2021-08-17 Martin Sebor <msebor@redhat.com>
5773 PR middle-end/101854
5774 * builtins.c (expand_builtin_alloca): Move warning code to check_alloca
5775 in gimple-ssa-warn-access.cc.
5776 * calls.c (alloc_max_size): Move code to check_alloca.
5777 (get_size_range): Move to pointer-query.cc.
5778 (maybe_warn_alloc_args_overflow): Move to gimple-ssa-warn-access.cc.
5779 (get_attr_nonstring_decl): Move to tree.c.
5780 (fntype_argno_type): Move to gimple-ssa-warn-access.cc.
5781 (append_attrname): Same.
5782 (maybe_warn_rdwr_sizes): Same.
5783 (initialize_argument_information): Move code to
5784 gimple-ssa-warn-access.cc.
5785 * calls.h (maybe_warn_alloc_args_overflow): Move to
5786 gimple-ssa-warn-access.h.
5787 (get_attr_nonstring_decl): Move to tree.h.
5788 (maybe_warn_nonstring_arg): Move to gimple-ssa-warn-access.h.
5789 (enum size_range_flags): Move to pointer-query.h.
5790 (get_size_range): Same.
5791 * gimple-ssa-warn-access.cc (has_location): Remove unused overload
5792 to avoid Clang -Wunused-function.
5793 (get_size_range): Declare static.
5794 (maybe_emit_free_warning): Rename...
5795 (maybe_check_dealloc_call): ...to this for consistency.
5796 (class pass_waccess): Add members.
5797 (pass_waccess::~pass_waccess): Defined.
5798 (alloc_max_size): Move here from calls.c.
5799 (maybe_warn_alloc_args_overflow): Same.
5800 (check_alloca): New function.
5801 (check_alloc_size_call): New function.
5802 (check_strncat): Handle another warning flag.
5803 (pass_waccess::check_builtin): Handle alloca.
5804 (fntype_argno_type): Move here from calls.c.
5805 (append_attrname): Same.
5806 (maybe_warn_rdwr_sizes): Same.
5807 (pass_waccess::check_call): Define.
5808 (check_nonstring_args): New function.
5809 (pass_waccess::check): Call new member functions.
5810 (pass_waccess::execute): Enable ranger.
5811 * gimple-ssa-warn-access.h (get_size_range): Move here from calls.h.
5812 (maybe_warn_nonstring_arg): Same.
5813 * gimple-ssa-warn-restrict.c: Remove #include.
5814 * pointer-query.cc (get_size_range): Move here from calls.c.
5815 * pointer-query.h (enum size_range_flags): Same.
5816 (get_size_range): Same.
5817 * tree.c (get_attr_nonstring_decl): Move here from calls.c.
5818 * tree.h (get_attr_nonstring_decl): Move here from calls.h.
5820 2021-08-17 Thomas Schwinge <thomas@codesourcery.com>
5822 * ggc.h (ggc_collect): Add 'force_collect' parameter.
5823 * ggc-page.c (ggc_collect): Use that one instead of global
5824 'ggc_force_collect'. Adjust all users.
5825 * doc/gty.texi (Invoking the garbage collector): Update.
5826 * ggc-internal.h (ggc_force_collect): Remove.
5827 * ggc-common.c (ggc_force_collect): Likewise.
5828 * selftest.h (forcibly_ggc_collect): Remove.
5829 * ggc-tests.c (selftest::forcibly_ggc_collect): Likewise.
5830 * read-rtl-function.c (test_loading_labels): Adjust.
5831 * selftest-run-tests.c (run_tests): Likewise.
5833 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
5835 * config/darwin.c (darwin_file_end): Reset and reclaim the
5836 section names table at the end of compile.
5838 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
5841 * config.in: Regenerate.
5842 * config/i386/darwin.h (EXTRA_ASM_OPTS): New
5843 (ASM_SPEC): Pass options to disable branch shortening where
5845 * configure: Regenerate.
5846 * configure.ac: Detect versions of 'as' that support the
5847 optimisation which has the bug.
5849 2021-08-17 Richard Biener <rguenther@suse.de>
5851 * optabs-query.c (supports_vec_gather_load_p): Also check
5853 (supports_vec_scatter_store_p): Likewise.
5854 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Fall
5855 back to masked variants if non-masked are not supported.
5856 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
5857 When we need to use masked gather/scatter but do not have
5858 a mask set up a constant true one.
5859 * tree-vect-stmts.c (vect_check_scalar_mask): Also allow
5862 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
5864 * tree-ssa-ccp.c (bit_value_binop) [MINUS_EXPR]: Use same
5865 algorithm as PLUS_EXPR to improve subtraction bit bounds.
5866 [POINTER_DIFF_EXPR]: Treat as synonymous with MINUS_EXPR.
5868 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
5870 * tree-ssa-ccp.c (bit_value_mult_const): New helper function to
5871 calculate the mask-value pair result of a multiplication by an
5873 (bit_value_binop) [MULT_EXPR]: Call it from here for
5874 multiplications by (sparse) non-negative constants.
5876 2021-08-17 Christophe Lyon <christophe.lyon@foss.st.com>
5879 * config.gcc (gcc_cv_initfini_array): Leave undefined for
5880 uclinuxfdpiceabi targets.
5882 2021-08-17 Alexandre Oliva <oliva@adacore.com>
5884 * tree-inline.c (maybe_move_debug_stmts_to_successors): Don't
5885 reverse debug stmts.
5887 2021-08-17 Alexandre Oliva <oliva@adacore.com>
5889 * tree-cfg.c (dump_function_to_file): Use fun, not cfun.
5891 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
5893 * config/aarch64/arm_neon.h (__LD4_LANE_FUNC): Delete.
5894 (__LD4Q_LANE_FUNC): Likewise.
5895 (vld4_lane_u8): Define without macro.
5896 (vld4_lane_u16): Likewise.
5897 (vld4_lane_u32): Likewise.
5898 (vld4_lane_u64): Likewise.
5899 (vld4_lane_s8): Likewise.
5900 (vld4_lane_s16): Likewise.
5901 (vld4_lane_s32): Likewise.
5902 (vld4_lane_s64): Likewise.
5903 (vld4_lane_f16): Likewise.
5904 (vld4_lane_f32): Likewise.
5905 (vld4_lane_f64): Likewise.
5906 (vld4_lane_p8): Likewise.
5907 (vld4_lane_p16): Likewise.
5908 (vld4_lane_p64): Likewise.
5909 (vld4q_lane_u8): Likewise.
5910 (vld4q_lane_u16): Likewise.
5911 (vld4q_lane_u32): Likewise.
5912 (vld4q_lane_u64): Likewise.
5913 (vld4q_lane_s8): Likewise.
5914 (vld4q_lane_s16): Likewise.
5915 (vld4q_lane_s32): Likewise.
5916 (vld4q_lane_s64): Likewise.
5917 (vld4q_lane_f16): Likewise.
5918 (vld4q_lane_f32): Likewise.
5919 (vld4q_lane_f64): Likewise.
5920 (vld4q_lane_p8): Likewise.
5921 (vld4q_lane_p16): Likewise.
5922 (vld4q_lane_p64): Likewise.
5923 (vld4_lane_bf16): Likewise.
5924 (vld4q_lane_bf16): Likewise.
5926 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
5928 * config/aarch64/arm_neon.h (__LD3_LANE_FUNC): Delete.
5929 (__LD3Q_LANE_FUNC): Delete.
5930 (vld3_lane_u8): Define without macro.
5931 (vld3_lane_u16): Likewise.
5932 (vld3_lane_u32): Likewise.
5933 (vld3_lane_u64): Likewise.
5934 (vld3_lane_s8): Likewise.
5935 (vld3_lane_s16): Likewise.
5936 (vld3_lane_s32): Likewise.
5937 (vld3_lane_s64): Likewise.
5938 (vld3_lane_f16): Likewise.
5939 (vld3_lane_f32): Likewise.
5940 (vld3_lane_f64): Likewise.
5941 (vld3_lane_p8): Likewise.
5942 (vld3_lane_p16): Likewise.
5943 (vld3_lane_p64): Likewise.
5944 (vld3q_lane_u8): Likewise.
5945 (vld3q_lane_u16): Likewise.
5946 (vld3q_lane_u32): Likewise.
5947 (vld3q_lane_u64): Likewise.
5948 (vld3q_lane_s8): Likewise.
5949 (vld3q_lane_s16): Likewise.
5950 (vld3q_lane_s32): Likewise.
5951 (vld3q_lane_s64): Likewise.
5952 (vld3q_lane_f16): Likewise.
5953 (vld3q_lane_f32): Likewise.
5954 (vld3q_lane_f64): Likewise.
5955 (vld3q_lane_p8): Likewise.
5956 (vld3q_lane_p16): Likewise.
5957 (vld3q_lane_p64): Likewise.
5958 (vld3_lane_bf16): Likewise.
5959 (vld3q_lane_bf16): Likewise.
5961 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
5963 * config/aarch64/arm_neon.h (__LD2_LANE_FUNC): Delete.
5964 (__LD2Q_LANE_FUNC): Likewise.
5965 (vld2_lane_u8): Define without macro.
5966 (vld2_lane_u16): Likewise.
5967 (vld2_lane_u32): Likewise.
5968 (vld2_lane_u64): Likewise.
5969 (vld2_lane_s8): Likewise.
5970 (vld2_lane_s16): Likewise.
5971 (vld2_lane_s32): Likewise.
5972 (vld2_lane_s64): Likewise.
5973 (vld2_lane_f16): Likewise.
5974 (vld2_lane_f32): Likewise.
5975 (vld2_lane_f64): Likewise.
5976 (vld2_lane_p8): Likewise.
5977 (vld2_lane_p16): Likewise.
5978 (vld2_lane_p64): Likewise.
5979 (vld2q_lane_u8): Likewise.
5980 (vld2q_lane_u16): Likewise.
5981 (vld2q_lane_u32): Likewise.
5982 (vld2q_lane_u64): Likewise.
5983 (vld2q_lane_s8): Likewise.
5984 (vld2q_lane_s16): Likewise.
5985 (vld2q_lane_s32): Likewise.
5986 (vld2q_lane_s64): Likewise.
5987 (vld2q_lane_f16): Likewise.
5988 (vld2q_lane_f32): Likewise.
5989 (vld2q_lane_f64): Likewise.
5990 (vld2q_lane_p8): Likewise.
5991 (vld2q_lane_p16): Likewise.
5992 (vld2q_lane_p64): Likewise.
5993 (vld2_lane_bf16): Likewise.
5994 (vld2q_lane_bf16): Likewise.
5996 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
5998 * haifa-sched.c (advance_one_cycle): Output more context-synchronization
6001 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
6003 * haifa-sched.c (enum rfs_decision, rfs_str): Add RFS_AUTOPREF.
6004 (rank_for_schedule): Use it.
6006 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
6008 PR rtl-optimization/91598
6009 * haifa-sched.c (autopref_rank_for_schedule): Prioritize "irrelevant"
6010 insns after memory reads and before memory writes.
6012 2021-08-17 Alistair_Lee <alistair.lee@arm.com>
6014 * rtl.h (CONST_VECTOR_P): New macro.
6015 * config/aarch64/aarch64.c (aarch64_get_sve_pred_bits): Use RTL
6016 code testing macros.
6017 (aarch64_ptrue_all_mode): Likewise.
6018 (aarch64_expand_mov_immediate): Likewise.
6019 (aarch64_const_vec_all_in_range_p): Likewise.
6020 (aarch64_rtx_costs): Likewise.
6021 (aarch64_legitimate_constant_p): Likewise.
6022 (aarch64_simd_valid_immediate): Likewise.
6023 (aarch64_simd_make_constant): Likewise.
6024 (aarch64_convert_mult_to_shift): Likewise.
6025 (aarch64_expand_sve_vec_perm): Likewise.
6026 (aarch64_vec_fpconst_pow_of_2): Likewise.
6028 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
6030 PR tree-optimization/101938
6031 * range-op.cc (operator_abs::op1_range): Special case
6032 -TYPE_MIN_VALUE for flag_wrapv.
6034 2021-08-17 Kewen Lin <linkw@linux.ibm.com>
6036 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Add the cost for
6039 2021-08-17 Jakub Jelinek <jakub@redhat.com>
6041 * tree.def (OMP_SCOPE): New tree code.
6042 * tree.h (OMP_SCOPE_BODY, OMP_SCOPE_CLAUSES): Define.
6043 * tree-nested.c (convert_nonlocal_reference_stmt,
6044 convert_local_reference_stmt, convert_gimple_call): Handle
6046 * tree-pretty-print.c (dump_generic_node): Handle OMP_SCOPE.
6047 * gimple.def (GIMPLE_OMP_SCOPE): New gimple code.
6048 * gimple.c (gimple_build_omp_scope): New function.
6049 (gimple_copy): Handle GIMPLE_OMP_SCOPE.
6050 * gimple.h (gimple_build_omp_scope): Declare.
6051 (gimple_has_substatements): Handle GIMPLE_OMP_SCOPE.
6052 (gimple_omp_scope_clauses, gimple_omp_scope_clauses_ptr,
6053 gimple_omp_scope_set_clauses): New inline functions.
6054 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_SCOPE.
6055 * gimple-pretty-print.c (dump_gimple_omp_scope): New function.
6056 (pp_gimple_stmt_1): Handle GIMPLE_OMP_SCOPE.
6057 * gimple-walk.c (walk_gimple_stmt): Likewise.
6058 * gimple-low.c (lower_stmt): Likewise.
6059 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
6060 (gimplify_scan_omp_clauses): For task reductions, handle OMP_SCOPE
6061 like ORT_WORKSHARE constructs. Adjust diagnostics for %<scope%>
6062 allowing task reductions. Reject inscan reductions on scope.
6063 (omp_find_stores_stmt): Handle GIMPLE_OMP_SCOPE.
6064 (gimplify_omp_workshare, gimplify_expr): Handle OMP_SCOPE.
6065 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_SCOPE.
6066 (estimate_num_insns): Likewise.
6067 * omp-low.c (build_outer_var_ref): Look through GIMPLE_OMP_SCOPE
6068 contexts if var isn't privatized there.
6069 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_SCOPE.
6070 (scan_omp_1_stmt): Likewise.
6071 (maybe_add_implicit_barrier_cancel): Look through outer
6073 (lower_omp_scope): New function.
6074 (lower_omp_task_reductions): Handle OMP_SCOPE.
6075 (lower_omp_1): Handle GIMPLE_OMP_SCOPE.
6076 (diagnose_sb_1, diagnose_sb_2): Likewise.
6077 * omp-expand.c (expand_omp_single): Support also GIMPLE_OMP_SCOPE.
6078 (expand_omp): Handle GIMPLE_OMP_SCOPE.
6079 (omp_make_gimple_edges): Likewise.
6080 * omp-builtins.def (BUILT_IN_GOMP_SCOPE_START): New built-in.
6082 2021-08-17 Richard Biener <rguenther@suse.de>
6084 PR tree-optimization/101925
6085 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Set
6086 reverse on COMPONENT_REF and ARRAY_REF according to
6087 what reverse_storage_order_for_component_p does.
6088 (vn_reference_eq): Compare reversed on reference ops.
6089 (reverse_storage_order_for_component_p): New overload.
6090 (vn_reference_lookup_3): Check reverse_storage_order_for_component_p
6091 on the reference looked up.
6093 2021-08-17 Jeff Law <jlaw@localhost.localdomain>
6095 * config/h8300/h8300.c (shift_alg_si): Avoid loops for most SImode
6097 (h8300_option_override): Use loops on H8/S more often when optimizing
6099 (get_shift_alg): Handle new "special" cases on H8/S. Simplify
6100 accordingly. Handle various arithmetic right shifts with special
6101 sequences that we couldn't handle before.
6103 2021-08-16 Jeff Law <jlaw@localhost.localdomain>
6105 * config.gcc (rl78-*-elf*): Do not include dbxelf.h.
6107 2021-08-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
6109 * config/sparc/rtemself.h (SPARC_GCOV_TYPE_SIZE): Define.
6110 * config/sparc/sparc.c (sparc_gcov_type_size): New.
6111 (TARGET_GCOV_TYPE_SIZE): Redefine if SPARC_GCOV_TYPE_SIZE is defined.
6112 * coverage.c (get_gcov_type): Use targetm.gcov_type_size().
6113 * doc/tm.texi (TARGET_GCOV_TYPE_SIZE): Add hook under "Misc".
6114 * doc/tm.texi.in: Regenerate.
6115 * target.def (gcov_type_size): New target hook.
6116 * targhooks.c (default_gcov_type_size): New.
6117 * targhooks.h (default_gcov_type_size): Declare.
6118 * tree-profile.c (gimple_gen_edge_profiler): Use precision of
6120 (gimple_gen_time_profiler): Likewise.
6122 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
6124 * dwarf2out.c (add_scalar_info): Deal with DW_AT_data_bit_offset.
6126 2021-08-16 Tobias Burnus <tobias@codesourcery.com>
6128 PR middle-end/101931
6129 * omp-low.c (omp_runtime_api_call): Update for routines
6130 added in the meanwhile.
6132 2021-08-16 Martin Liska <mliska@suse.cz>
6134 PR tree-optimization/100393
6135 * tree-switch-conversion.c (group_cluster::dump): Use
6136 get_comparison_count.
6137 (jump_table_cluster::find_jump_tables): Pre-compute number of
6138 comparisons and then decrement it. Cache also max_ratio.
6139 (jump_table_cluster::can_be_handled): Change signature.
6140 * tree-switch-conversion.h (get_comparison_count): New.
6142 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
6144 * dwarf2out.c (add_data_member_location_attribute): Use GNAT
6145 encodings only when -fgnat-encodings=all is specified.
6146 (add_bound_info): Likewise.
6147 (add_byte_size_attribute): Likewise.
6148 (gen_member_die): Likewise.
6150 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
6152 * omp-oacc-neuter-broadcast.cc
6153 (execute_omp_oacc_neuter_broadcast): Plug 'par' memory leak.
6155 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
6157 * omp-oacc-neuter-broadcast.cc
6158 (execute_omp_oacc_neuter_broadcast): Clarify memory management for
6161 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
6163 * omp-oacc-neuter-broadcast.cc (field_map): Move variable into...
6164 (execute_omp_oacc_neuter_broadcast): ... here.
6165 (install_var_field, build_receiver_ref, build_sender_ref): Take
6166 'field_map_t *' parameter. Adjust all users.
6167 (worker_single_copy, neuter_worker_single): Take a
6168 'record_field_map_t *' parameter. Adjust all users.
6170 2021-08-16 liuhongt <hongtao.liu@intel.com>
6173 * config/i386/i386.md (ldexp<mode>3): Force operands[1] to
6176 2021-08-16 Martin Liska <mliska@suse.cz>
6179 * multiple_target.c (create_dispatcher_calls): Make default
6180 function local only if it is a definition.
6182 2021-08-16 Martin Liska <mliska@suse.cz>
6185 * ipa-icf-gimple.c (func_checker::compare_ssa_name): Do not
6186 consider equal SSA_NAMEs when one is a param.
6188 2021-08-16 liuhongt <hongtao.liu@intel.com>
6191 * config/i386/i386-expand.c (ix86_expand_vec_perm_vpermt2):
6192 Support vpermi2b for V32QI/V16QImode.
6193 (ix86_extract_perm_from_pool_constant): New function.
6194 (ix86_expand_vec_one_operand_perm_avx512): Support
6195 vpermw/vpermb under TARGET_AVX512BW/TARGET_AVX512VBMI.
6196 (expand_vec_perm_1): Adjust comments for upper.
6197 * config/i386/i386-protos.h (ix86_extract_perm_from_pool_constant):
6199 * config/i386/predicates.md (permvar_truncate_operand): New predicate.
6200 (pshufb_truncv4siv4hi_operand): Ditto.
6201 (pshufb_truncv8hiv8qi_operand): Ditto.
6202 * config/i386/sse.md (*avx512bw_permvar_truncv16siv16hi_1):
6203 New pre_reload define_insn_and_split.
6204 (*avx512f_permvar_truncv8siv8hi_1): Ditto.
6205 (*avx512f_vpermvar_truncv8div8si_1): Ditto.
6206 (*avx512f_permvar_truncv32hiv32qi_1): Ditto.
6207 (*avx512f_permvar_truncv16hiv16qi_1): Ditto.
6208 (*avx512f_permvar_truncv4div4si_1): Ditto.
6209 (*avx512f_pshufb_truncv8hiv8qi_1): Ditto.
6210 (*avx512f_pshufb_truncv4siv4hi_1): Ditto.
6211 (*avx512f_pshufd_truncv2div2si_1): Ditto.
6213 2021-08-16 Kito Cheng <kito.cheng@sifive.com>
6215 * config/riscv/multilib-generator: Support code model option for
6217 * doc/install.texi: Add document of new option for
6218 --with-multilib-generator.
6220 2021-08-15 Clément Chigot <clement.chigot@atos.net>
6222 * config/rs6000/rs6000.c (xcoff_tls_exec_model_detected): New.
6223 (rs6000_legitimize_tls_address_aix): Use it.
6224 (rs6000_xcoff_file_end): Add ".ref __tls_get_addr" when
6225 xcoff_tls_exec_model_detected is true.
6227 2021-08-15 Jeff Law <jlaw@localhost.localdomain>
6229 * config/h8300/h8300.c (shift_alg_si): Retune H8/300H shifts
6230 to allow a bit more code growth, saving many dozens of cycles.
6231 (h8300_option_override): Adjus shift_alg_si if optimizing for
6233 (get_shift_alg): Use special + inline shifts for residuals
6236 2021-08-14 Stafford Horne <shorne@gmail.com>
6239 * config/or1k/or1k-opts.h: New file.
6240 * config/or1k/or1k.c (or1k_legitimize_address_1, print_reloc):
6241 Support generating gotha relocations if -mcmodel=large is
6243 * config/or1k/or1k.h (TARGET_CMODEL_SMALL, TARGET_CMODEL_LARGE):
6245 * config/or1k/or1k.opt (mcmodel=): New option.
6246 * doc/invoke.texi (OpenRISC Options): Document mcmodel.
6248 2021-08-14 Martin Sebor <msebor@redhat.com>
6250 PR middle-end/101791
6251 * gimple-ssa-warn-access.cc (new_delete_mismatch_p): Use new argument
6252 to valid_new_delete_pair_p.
6253 * tree.c (valid_new_delete_pair_p): Add argument.
6254 * tree.h (valid_new_delete_pair_p): Same.
6256 2021-08-14 Jakub Jelinek <jakub@redhat.com>
6259 * config/i386/i386-expand.c (expand_vec_perm_broadcast_1)
6260 <case E_V64QImode>: For this mode assert
6261 !TARGET_AVX512BW || d->perm[0] rather than !TARGET_AVX2 || d->perm[0].
6263 2021-08-13 Michael Meissner <meissner@linux.ibm.com>
6266 * config/rs6000/altivec.md (xxeval): Use register_predicate
6267 instead of altivec_register_predicate.
6269 2021-08-13 Martin Sebor <msebor@redhat.com>
6271 PR middle-end/101734
6272 * tree-ssa-uninit.c (maybe_warn_read_write_only): New function.
6273 (maybe_warn_operand): Call it.
6275 2021-08-13 Martin Liska <mliska@suse.cz>
6278 * attribs.c (decl_attributes): Make naked functions "noipa"
6281 2021-08-13 Martin Liska <mliska@suse.cz>
6284 * symtab.c (symtab_node::noninterposable_alias): Do not create
6285 local aliases for target_clone functions as the clonning pass
6288 2021-08-13 Martin Liska <mliska@suse.cz>
6290 * opts.c (LIVE_PATCHING_OPTION): Define.
6291 (control_options_for_live_patching): Use it in error messages.
6293 2021-08-13 Jan Hubicka <hubicka@ucw.cz>
6295 * ipa-modref.c (dump_eaf_flags): Dump EAF_NOREAD.
6296 (implicit_const_eaf_flags, implicit_pure_eaf_flags,
6297 ignore_stores_eaf_flags): New constants.
6298 (remove_useless_eaf_flags): New function.
6299 (eaf_flags_useful_p): Use it.
6300 (deref_flags): Add EAF_NOT_RETURNED if flag is unused;
6302 (modref_lattice::init): Add EAF_NOREAD.
6303 (modref_lattice::add_escape_point): Do not reacord escape point if
6305 (modref_lattice::merge): EAF_NOESCAPE implies EAF_NODIRECTESCAPE;
6306 use remove_useless_eaf_flags.
6307 (modref_lattice::merge_deref): Use ignore_stores_eaf_flags.
6308 (modref_lattice::merge_direct_load): Add EAF_NOREAD
6309 (analyze_ssa_name_flags): Fix handling EAF_NOT_RETURNED
6310 (analyze_parms): Use remove_useless_eaf_flags.
6311 (ipa_merge_modref_summary_after_inlining): Use ignore_stores_eaf_flags.
6312 (modref_merge_call_site_flags): Add caller and ecf_flags parameter;
6313 use remove_useless_eaf_flags.
6314 (modref_propagate_flags_in_scc): Update.
6315 * ipa-modref.h: Turn eaf_flags_t back to char.
6316 * tree-core.h (EAF_NOT_RETURNED): Fix.
6317 (EAF_NOREAD): New constant
6318 * tree-ssa-alias.c: (ref_maybe_used_by_call_p_1): Check for
6320 * tree-ssa-structalias.c (handle_rhs_call): Handle new flags.
6321 (handle_pure_call): Likewise.
6323 2021-08-12 Jakub Jelinek <jakub@redhat.com>
6325 * tree.def (OMP_MASKED): New tree code.
6326 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_FILTER.
6327 * tree.h (OMP_MASKED_BODY, OMP_MASKED_CLAUSES, OMP_MASKED_COMBINED,
6328 OMP_CLAUSE_FILTER_EXPR): Define.
6329 * tree.c (omp_clause_num_ops): Add OMP_CLAUSE_FILTER entry.
6330 (omp_clause_code_name): Likewise.
6331 (walk_tree_1): Handle OMP_CLAUSE_FILTER.
6332 * tree-nested.c (convert_nonlocal_omp_clauses,
6333 convert_local_omp_clauses): Handle OMP_CLAUSE_FILTER.
6334 (convert_nonlocal_reference_stmt, convert_local_reference_stmt,
6335 convert_gimple_call): Handle GIMPLE_OMP_MASTER.
6336 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_FILTER.
6337 (dump_generic_node): Handle OMP_MASTER.
6338 * gimple.def (GIMPLE_OMP_MASKED): New gimple code.
6339 * gimple.c (gimple_build_omp_masked): New function.
6340 (gimple_copy): Handle GIMPLE_OMP_MASKED.
6341 * gimple.h (gimple_build_omp_masked): Declare.
6342 (gimple_has_substatements): Handle GIMPLE_OMP_MASKED.
6343 (gimple_omp_masked_clauses, gimple_omp_masked_clauses_ptr,
6344 gimple_omp_masked_set_clauses): New inline functions.
6345 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_MASKED.
6346 * gimple-pretty-print.c (dump_gimple_omp_masked): New function.
6347 (pp_gimple_stmt_1): Handle GIMPLE_OMP_MASKED.
6348 * gimple-walk.c (walk_gimple_stmt): Likewise.
6349 * gimple-low.c (lower_stmt): Likewise.
6350 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
6351 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_FILTER. For clauses
6352 that take one expression rather than decl or constant, force
6353 gimplification of that into a SSA_NAME or temporary unless min
6355 (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_FILTER.
6356 (gimplify_expr): Handle OMP_MASKED.
6357 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_MASKED.
6358 (estimate_num_insns): Likewise.
6359 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_FILTER.
6360 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_MASKED. Adjust
6361 diagnostics for existence of masked construct.
6362 (scan_omp_1_stmt, lower_omp_master, lower_omp_1, diagnose_sb_1,
6363 diagnose_sb_2): Handle GIMPLE_OMP_MASKED.
6364 * omp-expand.c (expand_omp_synch, expand_omp, omp_make_gimple_edges):
6367 2021-08-12 Uroš Bizjak <ubizjak@gmail.com>
6370 * config/i386/i386.md (avx512f_scalef<mode>2): New insn pattern.
6371 (ldexp<mode>3): Use avx512f_scalef<mode>2.
6372 (UNSPEC_SCALEF): Move from ...
6373 * config/i386/sse.md (UNSPEC_SCALEF): ... here.
6375 2021-08-12 Jan Hubicka <hubicka@ucw.cz>
6377 * ipa-split.c (consider_split): Fix condition testing void functions.
6379 2021-08-12 Aldy Hernandez <aldyh@redhat.com>
6381 * doc/invoke.texi: Remove docs for threader-mode param.
6382 * flag-types.h (enum threader_mode): Remove.
6383 * params.opt: Remove threader-mode param.
6384 * tree-ssa-threadbackward.c (class back_threader): Remove
6385 path_is_unreachable_p.
6386 Make find_paths private.
6387 Add maybe_thread and thread_through_all_blocks.
6388 Remove reference marker for m_registry.
6389 Remove reference marker for m_profit.
6390 (back_threader::back_threader): Adjust for registry and profit not
6392 (dump_path): Move down.
6394 (class thread_jumps): Remove.
6395 (class back_threader_registry): Remove m_all_paths.
6397 (thread_jumps::thread_through_all_blocks): Move to back_threader
6399 (fsm_find_thread_path): Remove
6400 (back_threader::maybe_thread): New.
6401 (back_threader::thread_through_all_blocks): Move from
6403 (back_threader_registry::back_threader_registry): Remove
6405 (back_threader_registry::~back_threader_registry): Remove.
6406 (thread_jumps::find_taken_edge): Remove.
6407 (thread_jumps::check_subpath_and_update_thread_path): Remove.
6408 (thread_jumps::maybe_register_path): Remove.
6409 (thread_jumps::handle_phi): Remove.
6410 (handle_assignment_p): Remove.
6411 (thread_jumps::handle_assignment): Remove.
6412 (thread_jumps::fsm_find_control_statement_thread_paths): Remove.
6413 (thread_jumps::find_jump_threads_backwards): Remove.
6414 (thread_jumps::find_jump_threads_backwards_with_ranger): Remove.
6415 (try_thread_blocks): Rename find_jump_threads_backwards to
6417 (pass_early_thread_jumps::execute): Same.
6419 2021-08-12 Tobias Burnus <tobias@codesourcery.com>
6421 * tree-core.h (omp_clause_proc_bind_kind): Add
6422 OMP_CLAUSE_PROC_BIND_PRIMARY.
6423 * tree-pretty-print.c (dump_omp_clause): Add TODO comment to
6424 change 'master' to 'primary' in proc_bind for OpenMP 5.1.
6426 2021-08-12 Claudiu Zissulescu <claziss@synopsys.com>
6428 * common/config/arc/arc-common.c (arc_option_init_struct): Remove
6429 fno-common reference.
6430 * config/arc/arc.c (arc_override_options): Remove overriding of
6433 2021-08-12 Jakub Jelinek <jakub@redhat.com>
6436 * config/i386/i386-expand.c (ix86_expand_vec_one_operand_perm_avx512):
6437 If d->testing_p, return true after performing checks instead of
6438 actually expanding the insn.
6439 (expand_vec_perm_broadcast_1): Handle V32HImode - assert
6440 !TARGET_AVX512BW and return false.
6442 2021-08-12 Eric Botcazou <ebotcazou@gcc.gnu.org>
6444 * configure.ac (PE linker --disable-dynamicbase support): New check.
6445 * configure: Regenerate.
6446 * config.in: Likewise.
6447 * config/i386/mingw32.h (LINK_SPEC_DISABLE_DYNAMICBASE): New define.
6448 (LINK_SPEC): Use it.
6449 * config/i386/mingw-w64.h (LINK_SPEC_DISABLE_DYNAMICBASE): Likewise.
6450 (LINK_SPEC): Likewise.
6452 2021-08-12 liuhongt <hongtao.liu@intel.com>
6455 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_2): New
6456 post_reload define_insn_and_split.
6457 (*avx512bw_zero_extendv32qiv32hi2_2): Ditto.
6458 (*sse4_1_zero_extendv8qiv8hi2_4): Ditto.
6459 (*avx512f_zero_extendv16hiv16si2_2): Ditto.
6460 (*avx2_zero_extendv8hiv8si2_2): Ditto.
6461 (*sse4_1_zero_extendv4hiv4si2_4): Ditto.
6462 (*avx512f_zero_extendv8siv8di2_2): Ditto.
6463 (*avx2_zero_extendv4siv4di2_2): Ditto.
6464 (*sse4_1_zero_extendv2siv2di2_4): Ditto.
6465 (VI248_256, VI248_512, VI148_512, VI148_256, VI148_128): New
6468 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
6470 * config/rs6000/rs6000-builtin-new.def: Add always, power5, and
6473 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
6475 * config/rs6000/rs6000-builtin-new.def: Add vsx stanza.
6477 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
6479 * config/rs6000/rs6000-builtin-new.def: Finish altivec stanza.
6480 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Move
6481 initialization of pcvoid_type_node here...
6482 (altivec_init_builtins): ...from here.
6483 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add
6484 RS6000_BTI_const_ptr_void.
6485 (pcvoid_type_node): New macro.
6487 2021-08-11 Richard Biener <rguenther@suse.de>
6490 * tree-ssa-forwprop.c (pass_forwprop::execute): Do not decompose
6491 hard-register accesses.
6493 2021-08-11 Richard Biener <rguenther@suse.de>
6495 * tree-ssa-operands.c (operands_scanner::get_expr_operands):
6496 Do not look at COMPONENT_REF FIELD_DECLs TREE_THIS_VOLATILE
6497 to determine has_volatile_ops.
6499 2021-08-11 Eric Botcazou <ebotcazou@gcc.gnu.org>
6501 * cfgexpand.c (expand_used_vars): Reuse attribs local variable.
6503 2021-08-11 Jan Hubicka <hubicka@ucw.cz>
6504 Alexandre Oliva <oliva@adacore.com>
6506 * ipa-modref.c (modref_lattice::dump): Fix escape_point's min_flags
6508 (modref_lattice::merge_deref): Fix handling of indirect scape points.
6509 (update_escape_summary_1): Likewise.
6510 (update_escape_summary): Likewise.
6511 (ipa_merge_modref_summary_after_inlining): Likewise.
6513 2021-08-11 Richard Biener <rguenther@suse.de>
6515 PR middle-end/101858
6516 * fold-const.c (fold_binary_loc): Guard simplification
6517 of X < (cast) (1 << Y) to integer types.
6519 2021-08-11 Richard Biener <rguenther@suse.de>
6521 PR tree-optimization/101861
6522 * tree-vect-stmts.c (vectorizable_load): Fix error in
6523 previous change with regard to gather vectorization.
6525 2021-08-11 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
6528 * config/arm/arm_neon.h (vdup_n_s8): Replace call to builtin
6530 (vdup_n_s16): Likewise.
6531 (vdup_n_s32): Likewise.
6532 (vdup_n_s64): Likewise.
6533 (vdup_n_u8): Likewise.
6534 (vdup_n_u16): Likewise.
6535 (vdup_n_u32): Likewise.
6536 (vdup_n_u64): Likewise.
6537 (vdup_n_p8): Likewise.
6538 (vdup_n_p16): Likewise.
6539 (vdup_n_p64): Likewise.
6540 (vdup_n_f16): Likewise.
6541 (vdup_n_f32): Likewise.
6542 (vdupq_n_s8): Likewise.
6543 (vdupq_n_s16): Likewise.
6544 (vdupq_n_s32): Likewise.
6545 (vdupq_n_s64): Likewise.
6546 (vdupq_n_u8): Likewise.
6547 (vdupq_n_u16): Likewise.
6548 (vdupq_n_u32): Likewise.
6549 (vdupq_n_u64): Likewise.
6550 (vdupq_n_p8): Likewise.
6551 (vdupq_n_p16): Likewise.
6552 (vdupq_n_p64): Likewise.
6553 (vdupq_n_f16): Likewise.
6554 (vdupq_n_f32): Likewise.
6555 (vmov_n_s8): Replace call to builtin with call to corresponding
6557 (vmov_n_s16): Likewise.
6558 (vmov_n_s32): Likewise.
6559 (vmov_n_s64): Likewise.
6560 (vmov_n_u8): Likewise.
6561 (vmov_n_u16): Likewise.
6562 (vmov_n_u32): Likewise.
6563 (vmov_n_u64): Likewise.
6564 (vmov_n_p8): Likewise.
6565 (vmov_n_p16): Likewise.
6566 (vmov_n_f16): Likewise.
6567 (vmov_n_f32): Likewise.
6568 (vmovq_n_s8): Likewise.
6569 (vmovq_n_s16): Likewise.
6570 (vmovq_n_s32): Likewise.
6571 (vmovq_n_s64): Likewise.
6572 (vmovq_n_u8): Likewise.
6573 (vmovq_n_u16): Likewise.
6574 (vmovq_n_u32): Likewise.
6575 (vmovq_n_u64): Likewise.
6576 (vmovq_n_p8): Likewise.
6577 (vmovq_n_p16): Likewise.
6578 (vmovq_n_f16): Likewise.
6579 (vmovq_n_f32): Likewise.
6580 * config/arm/arm_neon_builtins.def: Remove entries for vdup_n.
6582 2021-08-11 liuhongt <hongtao.liu@intel.com>
6585 * config/i386/i386.md (ldexp<mode>3): Extend to vscalefs[sd]
6586 when TARGET_AVX512F and TARGET_SSE_MATH.
6588 2021-08-10 Jakub Jelinek <jakub@redhat.com>
6591 * config/i386/i386-expand.c (expand_vec_perm_even_odd): Return false
6592 for V32HImode if !TARGET_AVX512BW.
6593 (ix86_vectorize_vec_perm_const) <case E_V32HImode, case E_V64QImode>:
6594 If !TARGET_AVX512BW and TARGET_AVX512F and d.testing_p, don't fail
6595 early, but actually check the permutation.
6597 2021-08-10 Richard Biener <rguenther@suse.de>
6599 PR tree-optimization/101809
6600 * tree-vect-stmts.c (get_load_store_type): Allow emulated
6601 gathers with offset vector nunits being a constant multiple
6602 of the data vector nunits.
6603 (vect_get_gather_scatter_ops): Use the appropriate nunits
6604 for the offset vector defs.
6605 (vectorizable_store): Adjust call to
6606 vect_get_gather_scatter_ops.
6607 (vectorizable_load): Likewise. Handle the case of less
6608 offset vectors than data vectors.
6610 2021-08-10 Jakub Jelinek <jakub@redhat.com>
6613 * config/i386/sse.md (*avx512f_shuf_<shuffletype>64x2_1<mask_name>_1,
6614 *avx512f_shuf_<shuffletype>32x4_1<mask_name>_1): New define_insn
6617 2021-08-10 Richard Biener <rguenther@suse.de>
6619 PR tree-optimization/101801
6620 PR tree-optimization/101819
6621 * tree-vectorizer.h (vect_emulated_vector_p): Declare.
6622 * tree-vect-loop.c (vect_emulated_vector_p): New function.
6623 (vectorizable_reduction): Re-instantiate a check for emulated
6625 * tree-vect-stmts.c (vectorizable_shift): Likewise.
6626 (vectorizable_operation): Likewise. Cost emulated vector
6627 operations according to the scalar sequence synthesized by
6630 2021-08-10 Richard Biener <rguenther@suse.de>
6632 PR middle-end/101824
6633 * tree-nested.c (get_frame_field): Mark the COMPONENT_REF as
6634 volatile in case the variable was.
6636 2021-08-10 H.J. Lu <hjl.tools@gmail.com>
6639 * config/i386/constraints.md (BC): Document for integer SSE
6640 constant all bits set operand.
6641 (BF): New constraint for const floating-point all bits set
6643 * config/i386/i386.c (standard_sse_constant_p): Likewise.
6644 (standard_sse_constant_opcode): Likewise.
6645 * config/i386/sse.md (sseconstm1): New mode attribute.
6646 (mov<mode>_internal): Replace BC with <sseconstm1>.
6648 2021-08-10 liuhongt <hongtao.liu@intel.com>
6650 * config/i386/sse.md (cond_<insn><mode>): New expander.
6651 (VI248_AVX512VLBW): New mode iterator.
6652 * config/i386/predicates.md
6653 (nonimmediate_or_const_vec_dup_operand): New predicate.
6655 2021-08-09 Andrew MacLeod <amacleod@redhat.com>
6657 PR tree-optimization/101741
6658 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Check
6659 type of parameter for toupper/tolower.
6661 2021-08-09 Martin Jambor <mjambor@suse.cz>
6664 * ipa-prop.c (propagate_controlled_uses): Removed a spurious space.
6666 2021-08-09 Pat Haugen <pthaugen@linux.ibm.com>
6668 * config/rs6000/rs6000.c (is_load_insn1): Verify destination is a
6670 (is_store_insn1): Verify source is a register.
6672 2021-08-09 Uroš Bizjak <ubizjak@gmail.com>
6675 * config/i386/mmx.md (<any_logic:code>v2sf3):
6676 Rename from *mmx_<any_logic:code>v2sf3
6678 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6680 * config/nvptx/nvptx.c: Cross-reference parts adapted in
6681 'gcc/omp-oacc-neuter-broadcast.cc'.
6682 * omp-low.c: Likewise.
6683 * omp-oacc-neuter-broadcast.cc: Cross-reference parts adapted from
6686 2021-08-09 Julian Brown <julian@codesourcery.com>
6687 Kwok Cheung Yeung <kcy@codesourcery.com>
6688 Thomas Schwinge <thomas@codesourcery.com>
6690 * config/gcn/gcn.c (gcn_init_builtins): Override decls for
6691 BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START,
6692 BUILT_IN_GOACC_SINGLE_COPY_END and BUILT_IN_GOACC_BARRIER.
6693 (gcn_goacc_validate_dims): Turn on worker partitioning unconditionally.
6694 (gcn_fork_join): Update comment.
6695 * config/gcn/gcn.opt (flag_worker_partitioning): Remove.
6696 (macc_experimental_workers): Remove unused option.
6698 2021-08-09 Julian Brown <julian@codesourcery.com>
6699 Nathan Sidwell <nathan@codesourcery.com> (via 'gcc/config/nvptx/nvptx.c' master)
6700 Kwok Cheung Yeung <kcy@codesourcery.com>
6701 Thomas Schwinge <thomas@codesourcery.com>
6703 * Makefile.in (OBJS): Add omp-oacc-neuter-broadcast.o.
6704 * doc/tm.texi.in (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD):
6705 Add documentation hook.
6706 * doc/tm.texi: Regenerate.
6707 * omp-oacc-neuter-broadcast.cc: New file.
6708 * omp-builtins.def (BUILT_IN_GOACC_BARRIER)
6709 (BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START)
6710 (BUILT_IN_GOACC_SINGLE_COPY_END): New builtins.
6711 * passes.def (pass_omp_oacc_neuter_broadcast): Add pass.
6712 * target.def (goacc.create_worker_broadcast_record): Add target
6714 * tree-pass.h (make_pass_omp_oacc_neuter_broadcast): Add
6716 * config/gcn/gcn-protos.h (gcn_goacc_adjust_propagation_record):
6717 Rename prototype to...
6718 (gcn_goacc_create_worker_broadcast_record): ... this.
6719 * config/gcn/gcn-tree.c (gcn_goacc_adjust_propagation_record): Rename
6721 (gcn_goacc_create_worker_broadcast_record): ... this.
6722 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_PROPAGATION_RECORD):
6724 (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD): ... this.
6726 2021-08-09 Tejas Belagod <tejas.belagod@arm.com>
6729 * config/aarch64/aarch64-simd.md (vlshr<mode>3, vashr<mode>3): Use
6732 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6734 * Makefile.in (GTFILES): Remove '$(srcdir)/omp-offload.c'.
6736 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6738 * builtins.def (DEF_GOACC_BUILTIN, DEF_GOMP_BUILTIN): Don't
6739 consider '-foffload-abi'.
6740 * common.opt (-foffload-abi): Remove 'Var', 'Init'.
6741 * opts.c (common_handle_option) <-foffload-abi> [ACCEL_COMPILER]:
6744 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6746 * optc-gen.awk: Sanity check that 'Init' doesn't appear without
6749 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6751 * omp-builtins.def (BUILT_IN_ACC_GET_DEVICE_TYPE): Remove.
6753 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6755 * doc/gty.texi (Files): Update.
6757 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6759 * doc/gty.texi (Files): Fix GTY header file example.
6761 2021-08-09 Roger Sayle <roger@nextmovesoftware.com>
6763 * tree-ssa-ccp.c (value_mask_to_min_max): Helper function to
6764 determine the upper and lower bounds from a mask-value pair.
6765 (bit_value_unop) [ABS_EXPR, ABSU_EXPR]: Add support for
6766 absolute value and unsigned absolute value expressions.
6767 (bit_value_binop): Initialize *VAL's precision.
6768 [LT_EXPR, LE_EXPR]: Use value_mask_to_min_max to determine
6769 upper and lower bounds of operands. Add LE_EXPR/GE_EXPR
6770 support when the operands are unknown but potentially equal.
6771 [MIN_EXPR, MAX_EXPR]: Support minimum/maximum expressions.
6773 2021-08-09 Bin Cheng <bin.cheng@linux.alibaba.com>
6775 * config/aarch64/aarch64.md
6776 (*extend<SHORT:mode><GPI:mode>2_aarch64): Use %<GPI:w>0.
6778 2021-08-08 Sergei Trofimovich <siarheit@google.com>
6780 * lra-constraints.c: Fix s/otput/output/ typo.
6782 2021-08-06 Martin Sebor <msebor@redhat.com>
6784 * builtins.c (expand_builtin_memchr): Move to gimple-ssa-warn-access.cc.
6785 (expand_builtin_strcat): Same.
6786 (expand_builtin_stpncpy): Same.
6787 (expand_builtin_strncat): Same.
6788 (check_read_access): Same.
6789 (check_memop_access): Same.
6790 (expand_builtin_strlen): Move checks to gimple-ssa-warn-access.cc.
6791 (expand_builtin_strnlen): Same.
6792 (expand_builtin_memcpy): Same.
6793 (expand_builtin_memmove): Same.
6794 (expand_builtin_mempcpy): Same.
6795 (expand_builtin_strcpy): Same.
6796 (expand_builtin_strcpy_args): Same.
6797 (expand_builtin_stpcpy_1): Same.
6798 (expand_builtin_strncpy): Same.
6799 (expand_builtin_memset): Same.
6800 (expand_builtin_bzero): Same.
6801 (expand_builtin_strcmp): Same.
6802 (expand_builtin_strncmp): Same.
6803 (expand_builtin): Remove handlers.
6804 (fold_builtin_strlen): Add a comment.
6805 * builtins.h (check_access): Move to gimple-ssa-warn-access.cc.
6806 * calls.c (maybe_warn_nonstring_arg): Same.
6807 * diagnostic-spec.c (nowarn_spec_t::nowarn_spec_t): Add warning option.
6808 * gimple-fold.c (gimple_fold_builtin_strcpy): Pass argument to callee.
6809 (gimple_fold_builtin_stpcpy): Same.
6810 * gimple-ssa-warn-access.cc (has_location): New function.
6811 (get_location): Same.
6812 (get_callee_fndecl): Same.
6815 (warn_string_no_nul): Define.
6816 (unterminated_array): Same.
6817 (check_nul_terminated_array): Same.
6818 (maybe_warn_nonstring_arg): Same.
6819 (maybe_warn_for_bound): Same.
6820 (warn_for_access): Same.
6821 (check_access): Same.
6822 (check_memop_access): Same.
6823 (check_read_access): Same.
6824 (warn_dealloc_offset): Use helper functions.
6825 (maybe_emit_free_warning): Same.
6826 (class pass_waccess): Add members.
6827 (check_strcat): New function.
6828 (check_strncat): New function.
6829 (check_stxcpy): New function.
6830 (check_stxncpy): New function.
6831 (check_strncmp): New function.
6832 (pass_waccess::check_builtin): New function.
6833 (pass_waccess::check): Call it.
6834 * gimple-ssa-warn-access.h (warn_string_no_nul): Move here from
6836 (maybe_warn_for_bound): Same.
6837 (check_access): Same.
6838 (check_memop_access): Same.
6839 (check_read_access): Same.
6840 * pointer-query.h (struct access_data): Define a ctor overload.
6842 2021-08-06 Richard Biener <rguenther@suse.de>
6844 PR tree-optimization/101801
6845 * tree-vectorizer.h (vect_worthwhile_without_simd_p): Rename...
6846 (vect_can_vectorize_without_simd_p): ... to this.
6847 * tree-vect-loop.c (vect_worthwhile_without_simd_p): Rename...
6848 (vect_can_vectorize_without_simd_p): ... to this and fold
6849 in vect_min_worthwhile_factor.
6850 (vect_min_worthwhile_factor): Remove.
6851 (vectorizable_reduction): Adjust and remove the cost part.
6852 * tree-vect-stmts.c (vectorizable_shift): Likewise.
6853 (vectorizable_operation): Likewise.
6855 2021-08-06 Uroš Bizjak <ubizjak@gmail.com>
6858 * config/i386/i386.md (cmove reg-to-reg move elimination peephole2s):
6859 Add general_gr_operand predicate to operand 3.
6861 2021-08-06 Roger Sayle <roger@nextmovesoftware.com>
6863 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern): Use
6864 CFN_BUILT_IN_CLRSB* instead of BUILT_IN_CLRSB* for consistency.
6866 2021-08-06 Tamar Christina <tamar.christina@arm.com>
6868 * config/aarch64/aarch64-sve-builtins.cc (register_svpattern,
6869 register_svprfop): Pass vec<> by pointer.
6870 * langhooks-def.h (lhd_simulate_enum_decl): Likewise.
6871 * langhooks.c (lhd_simulate_enum_decl): Likewise.
6872 * langhooks.h (struct lang_hooks_for_types): Likewise.
6874 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
6876 * config/aarch64/arm_neon.h (vst1_bf16_x2): Use
6877 __builtin_memcpy instead of constructing an additional
6878 __builtin_aarch64_simd_oi one vector at a time.
6879 (vst1q_bf16_x2): Likewise.
6880 (vst1_bf16_x3): Use __builtin_memcpy instead of constructing
6881 an additional __builtin_aarch64_simd_ci one vector at a time.
6882 (vst1q_bf16_x3): Likewise.
6883 (vst1_bf16_x4): Use __builtin_memcpy instead of a union.
6884 (vst1q_bf16_x4): Likewise.
6885 (vst2_bf16): Use __builtin_memcpy instead of constructing an
6886 additional __builtin_aarch64_simd_oi one vector at a time.
6887 (vst2q_bf16): Likewise.
6888 (vst3_bf16): Use __builtin_memcpy instead of constructing an
6889 additional __builtin_aarch64_simd_ci mode one vector at a
6891 (vst3q_bf16): Likewise.
6892 (vst4_bf16): Use __builtin_memcpy instead of constructing an
6893 additional __builtin_aarch64_simd_xi one vector at a time.
6894 (vst4q_bf16): Likewise.
6896 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
6898 * config/aarch64/arm_neon.h (__ST2_LANE_FUNC): Delete.
6899 (__ST2Q_LANE_FUNC): Delete.
6900 (vst2_lane_f16): Use __builtin_memcpy to copy vector
6901 structure instead of constructing __builtin_aarch64_simd_oi
6902 one vector at a time.
6903 (vst2_lane_f32): Likewise.
6904 (vst2_lane_f64): Likewise.
6905 (vst2_lane_p8): Likewise.
6906 (vst2_lane_p16): Likewise.
6907 (vst2_lane_p64): Likewise.
6908 (vst2_lane_s8): Likewise.
6909 (vst2_lane_s16): Likewise.
6910 (vst2_lane_s32): Likewise.
6911 (vst2_lane_s64): Likewise.
6912 (vst2_lane_u8): Likewise.
6913 (vst2_lane_u16): Likewise.
6914 (vst2_lane_u32): Likewise.
6915 (vst2_lane_u64): Likewise.
6916 (vst2_lane_bf16): Likewise.
6917 (vst2q_lane_f16): Use __builtin_memcpy to copy vector
6918 structure instead of using a union.
6919 (vst2q_lane_f32): Likewise.
6920 (vst2q_lane_f64): Likewise.
6921 (vst2q_lane_p8): Likewise.
6922 (vst2q_lane_p16): Likewise.
6923 (vst2q_lane_p64): Likewise.
6924 (vst2q_lane_s8): Likewise.
6925 (vst2q_lane_s16): Likewise.
6926 (vst2q_lane_s32): Likewise.
6927 (vst2q_lane_s64): Likewise.
6928 (vst2q_lane_u8): Likewise.
6929 (vst2q_lane_u16): Likewise.
6930 (vst2q_lane_u32): Likewise.
6931 (vst2q_lane_u64): Likewise.
6932 (vst2q_lane_bf16): Likewise.
6934 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
6936 * config/aarch64/arm_neon.h (__ST3_LANE_FUNC): Delete.
6937 (__ST3Q_LANE_FUNC): Delete.
6938 (vst3_lane_f16): Use __builtin_memcpy to copy vector
6939 structure instead of constructing __builtin_aarch64_simd_ci
6940 one vector at a time.
6941 (vst3_lane_f32): Likewise.
6942 (vst3_lane_f64): Likewise.
6943 (vst3_lane_p8): Likewise.
6944 (vst3_lane_p16): Likewise.
6945 (vst3_lane_p64): Likewise.
6946 (vst3_lane_s8): Likewise.
6947 (vst3_lane_s16): Likewise.
6948 (vst3_lane_s32): Likewise.
6949 (vst3_lane_s64): Likewise.
6950 (vst3_lane_u8): Likewise.
6951 (vst3_lane_u16): Likewise.
6952 (vst3_lane_u32): Likewise.
6953 (vst3_lane_u64): Likewise.
6954 (vst3_lane_bf16): Likewise.
6955 (vst3q_lane_f16): Use __builtin_memcpy to copy vector
6956 structure instead of using a union.
6957 (vst3q_lane_f32): Likewise.
6958 (vst3q_lane_f64): Likewise.
6959 (vst3q_lane_p8): Likewise.
6960 (vst3q_lane_p16): Likewise.
6961 (vst3q_lane_p64): Likewise.
6962 (vst3q_lane_s8): Likewise.
6963 (vst3q_lane_s16): Likewise.
6964 (vst3q_lane_s32): Likewise.
6965 (vst3q_lane_s64): Likewise.
6966 (vst3q_lane_u8): Likewise.
6967 (vst3q_lane_u16): Likewise.
6968 (vst3q_lane_u32): Likewise.
6969 (vst3q_lane_u64): Likewise.
6970 (vst3q_lane_bf16): Likewise.
6972 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
6974 * config/aarch64/arm_neon.h (__ST4_LANE_FUNC): Delete.
6975 (__ST4Q_LANE_FUNC): Delete.
6976 (vst4_lane_f16): Use __builtin_memcpy to copy vector
6977 structure instead of constructing __builtin_aarch64_simd_xi
6978 one vector at a time.
6979 (vst4_lane_f32): Likewise.
6980 (vst4_lane_f64): Likewise.
6981 (vst4_lane_p8): Likewise.
6982 (vst4_lane_p16): Likewise.
6983 (vst4_lane_p64): Likewise.
6984 (vst4_lane_s8): Likewise.
6985 (vst4_lane_s16): Likewise.
6986 (vst4_lane_s32): Likewise.
6987 (vst4_lane_s64): Likewise.
6988 (vst4_lane_u8): Likewise.
6989 (vst4_lane_u16): Likewise.
6990 (vst4_lane_u32): Likewise.
6991 (vst4_lane_u64): Likewise.
6992 (vst4_lane_bf16): Likewise.
6993 (vst4q_lane_f16): Use __builtin_memcpy to copy vector
6994 structure instead of using a union.
6995 (vst4q_lane_f32): Likewise.
6996 (vst4q_lane_f64): Likewise.
6997 (vst4q_lane_p8): Likewise.
6998 (vst4q_lane_p16): Likewise.
6999 (vst4q_lane_p64): Likewise.
7000 (vst4q_lane_s8): Likewise.
7001 (vst4q_lane_s16): Likewise.
7002 (vst4q_lane_s32): Likewise.
7003 (vst4q_lane_s64): Likewise.
7004 (vst4q_lane_u8): Likewise.
7005 (vst4q_lane_u16): Likewise.
7006 (vst4q_lane_u32): Likewise.
7007 (vst4q_lane_u64): Likewise.
7008 (vst4q_lane_bf16): Likewise.
7010 2021-08-06 Martin Liska <mliska@suse.cz>
7012 * config/rs6000/rs6000.c (rs6000_option_override_internal): When
7013 a target option is restored, it can have
7014 rs6000_long_double_type_size set to FLOAT_PRECISION_TFmode
7015 and error should not be emitted.
7017 2021-08-06 Sebastian Huber <sebastian.huber@embedded-brains.de>
7019 * gcov-io.h (gcov_write): Declare.
7020 * gcov-io.c (gcov_write): New.
7021 (gcov_write_counter): Remove.
7022 (gcov_write_tag_length): Likewise.
7023 (gcov_write_summary): Replace gcov_write_tag_length() with calls to
7024 gcov_write_unsigned().
7025 * doc/invoke.texi (fprofile-info-section): Mention
7026 __gcov_info_to_gdca().
7028 2021-08-06 Martin Sebor <msebor@redhat.com>
7030 * dominance.c (prune_bbs_to_update_dominators): Adjust by-value vec
7031 arguments to by-reference.
7032 (iterate_fix_dominators): Same.
7033 * dominance.h (iterate_fix_dominators): Same.
7034 * ipa-prop.h: Call auto_vec::to_vec_legacy.
7035 * tree-data-ref.c (dump_data_dependence_relation): Adjust by-value vec
7036 arguments to by-reference.
7037 (debug_data_dependence_relation): Same.
7038 (dump_data_dependence_relations): Same.
7039 * tree-data-ref.h (debug_data_dependence_relation): Same.
7040 (dump_data_dependence_relations): Same.
7041 * tree-predcom.c (dump_chains): Same.
7042 (initialize_root_vars_lm): Same.
7043 (determine_unroll_factor): Same.
7044 (replace_phis_by_defined_names): Same.
7045 (insert_init_seqs): Same.
7046 (pcom_worker::tree_predictive_commoning_loop): Call
7047 auto_vec::to_vec_legacy.
7048 * tree-ssa-pre.c (insert_into_preds_of_block): Adjust by-value vec
7049 arguments to by-reference.
7050 * tree-ssa-threadbackward.c (populate_worklist): Same.
7051 (back_threader::resolve_def): Same.
7052 * tree-vect-data-refs.c (vect_check_nonzero_value): Same.
7053 (vect_enhance_data_refs_alignment): Same.
7054 (vect_check_lower_bound): Same.
7055 (vect_prune_runtime_alias_test_list): Same.
7056 (vect_permute_store_chain): Same.
7057 * tree-vect-slp-patterns.c (vect_normalize_conj_loc): Same.
7058 * tree-vect-stmts.c (vect_create_vectorized_demotion_stmts): Same.
7059 * tree-vectorizer.h (vect_permute_store_chain): Same.
7060 * vec.c (test_init): New function.
7061 (vec_c_tests): Call new function.
7062 * vec.h (vec): Declare ctors, dtor, and assignment.
7063 (auto_vec::vec_to_legacy): New function.
7064 (vec::copy): Adjust initialization.
7066 2021-08-05 H.J. Lu <hjl.tools@gmail.com>
7069 * config/i386/i386.c (ix86_can_inline_p): Ignore MASK_80387 if
7070 callee only uses GPRs.
7071 * config/i386/ia32intrin.h: Revert commit 5463cee2770.
7072 * config/i386/serializeintrin.h: Revert commit 71958f740f1.
7073 * config/i386/x86gprintrin.h: Add
7074 #pragma GCC target("general-regs-only") and #pragma GCC pop_options
7075 to disable non-GPR ISAs.
7077 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
7079 PR middle-end/101787
7080 * doc/md.texi (cond_ashl, cond_ashr, cond_lshr): Document.
7082 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
7084 * tree-vectorizer.h (vect_is_store_elt_extraction, vect_is_reduction)
7085 (vect_reduc_type, vect_embedded_comparison_type, vect_comparison_type)
7086 (vect_is_extending_load, vect_is_integer_truncation): New functions,
7087 moved from aarch64.c but given different names.
7088 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction)
7089 (aarch64_is_reduction, aarch64_reduc_type)
7090 (aarch64_embedded_comparison_type, aarch64_comparison_type)
7091 (aarch64_extending_load_p, aarch64_integer_truncation_p): Delete
7092 in favor of the above. Update callers accordingly.
7094 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
7097 * config/arm/arm-cpus.in (generic-armv7-a): Add quirk to suppress
7098 writing .cpu directive in asm output.
7099 * config/arm/arm.c (arm_identify_fpu_from_isa): New variable.
7100 (arm_last_printed_arch_string): Delete.
7101 (arm_last-printed_fpu_string): Delete.
7102 (arm_configure_build_target): If use of floating-point/SIMD is
7103 disabled, remove all fp/simd related features from the target ISA.
7104 (last_arm_targ_options): New variable.
7105 (arm_print_asm_arch_directives): Add new parameters. Change order
7106 of emitted directives and handle all cases here.
7107 (arm_file_start): Always call arm_print_asm_arch_directives, move
7108 all generation of .arch/.arch_extension here.
7109 (arm_file_end): Call arm_print_asm_arch.
7110 (arm_declare_function_name): Call arm_print_asm_arch_directives
7111 instead of printing .arch/.fpu directives directly.
7113 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
7115 * config/arm/arm.c (arm_configure_build_target): Don't call
7116 arm_option_reconfigure_globals.
7117 (arm_option_restore): Call arm_option_reconfigure_globals after
7118 reconfiguring the target.
7119 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
7121 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
7123 * config/arm/arm.c (arm_configure_build_target): Ensure the target's
7124 arch_name is always set.
7126 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
7128 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
7129 of vec_select high-half from being added into Neon subtract
7132 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
7134 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
7135 of vec_select high-half from being added into Neon add cost.
7137 2021-08-05 Kewen Lin <linkw@linux.ibm.com>
7139 * cfgloop.h (loops_list::loops_list): Add one optional argument
7140 root and adjust accordingly, update loop tree walking and factor
7142 * cfgloop.c (loops_list::walk_loop_tree): ... this. New function.
7144 2021-08-05 Eric Botcazou <ebotcazou@gcc.gnu.org>
7146 PR tree-optimization/101626
7147 * tree-sra.c (propagate_subaccesses_from_rhs): Do not set the
7148 reverse scalar storage order on a pointer or vector component.
7150 2021-08-05 liuhongt <hongtao.liu@intel.com>
7152 * config/i386/sse.md (cond_<code><mode>): New expander.
7154 2021-08-05 liuhongt <hongtao.liu@intel.com>
7156 * config/i386/sse.md (cond_<code><mode>): New expander.
7158 2021-08-05 liuhongt <hongtao.liu@intel.com>
7160 * config/i386/sse.md (cond_<code><mode>): New expander.
7162 2021-08-04 David Malcolm <dmalcolm@redhat.com>
7165 * Makefile.in (ANALYZER_OBJS): Add analyzer/region-model-asm.o.
7167 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
7170 * config/i386/i386.h (STORE_MAX_PIECES): Allow 16/32/64 bytes
7171 only if TARGET_INTER_UNIT_MOVES_TO_VEC is true.
7173 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
7176 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
7177 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
7178 data with SSE register from one memory location to another.
7180 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
7182 * config/s390/s390.c (expand_perm_with_vpdi): New function.
7183 (vectorize_vec_perm_const_1): Call expand_perm_with_vpdi.
7184 * config/s390/vector.md (*vpdi1<mode>, @vpdi1<mode>): Enable a
7185 parameterized expander.
7186 (*vpdi4<mode>, @vpdi4<mode>): Likewise.
7188 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
7190 * config/s390/s390.c (MAX_VECT_LEN): Define macro.
7191 (struct expand_vec_perm_d): Define struct.
7192 (expand_perm_with_merge): New function.
7193 (vectorize_vec_perm_const_1): New function.
7194 (s390_vectorize_vec_perm_const): New function.
7195 (TARGET_VECTORIZE_VEC_PERM_CONST): Define target macro.
7197 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
7199 * config/s390/vector.md (V_HW_64): Remove mode iterator.
7200 (*vec_load_pair<mode>): Use V_HW_2 instead of V_HW_64.
7201 * config/s390/vx-builtins.md
7202 (vec_scatter_element<V_HW_2:mode>_SI): Use V_HW_2 instead of
7205 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
7207 * config/s390/s390.md (UNSPEC_VEC_PERMI): Remove constant
7209 * config/s390/vector.md (*vpdi1<mode>, *vpdi4<mode>): New pattern
7211 * config/s390/vx-builtins.md (*vec_permi<mode>): Emit generic rtx
7212 instead of an unspec.
7214 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
7216 * config/s390/s390-modes.def: Add more vector modes to support
7217 concatenation of two vectors.
7218 * config/s390/s390-protos.h (s390_expand_merge_perm_const): Add
7220 (s390_expand_merge): Likewise.
7221 * config/s390/s390.c (s390_expand_merge_perm_const): New function.
7222 (s390_expand_merge): New function.
7223 * config/s390/s390.md (UNSPEC_VEC_MERGEH, UNSPEC_VEC_MERGEL):
7224 Remove constant definitions.
7225 * config/s390/vector.md (V_HW_2): Add mode iterators.
7226 (VI_HW_4, V_HW_4): Rename VI_HW_4 to V_HW_4.
7227 (vec_2x_nelts, vec_2x_wide): New mode attributes.
7228 (*vmrhb, *vmrlb, *vmrhh, *vmrlh, *vmrhf, *vmrlf, *vmrhg, *vmrlg):
7229 New pattern definitions.
7230 (vec_widen_umult_lo_<mode>, vec_widen_umult_hi_<mode>)
7231 (vec_widen_smult_lo_<mode>, vec_widen_smult_hi_<mode>)
7232 (vec_unpacks_lo_v4sf, vec_unpacks_hi_v4sf, vec_unpacks_lo_v2df)
7233 (vec_unpacks_hi_v2df): Adjust expanders to emit non-unspec RTX for
7235 * config/s390/vx-builtins.md (V_HW_4): Remove mode iterator. Now
7237 (vec_mergeh<mode>, vec_mergel<mode>): Use s390_expand_merge to
7238 emit vec merge pattern.
7240 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
7242 * config/aarch64/aarch64.c (aarch64_strip_extend_vec_half):
7244 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent cost of
7245 vec_select high-half from being added into Neon multiply
7247 * rtlanal.c (vec_series_highpart_p): Define.
7248 * rtlanal.h (vec_series_highpart_p): Declare.
7250 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
7252 * config/aarch64/aarch64.c (aarch64_strip_duplicate_vec_elt):
7254 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent
7255 vec_select cost from being added into Neon multiply cost.
7257 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
7259 * tree-vect-loop.c (vect_better_loop_vinfo_p): Detect cases in
7260 which old_loop_vinfo is an epilogue loop that handles a constant
7261 number of iterations.
7263 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
7265 * tree-vect-loop.c (vect_analyze_loop): Print a dump message
7266 when a reanalyzed loop fails to be cheaper than the current
7269 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
7271 * config/aarch64/aarch64.c: Fix a typo.
7273 2021-08-04 Vincent Lefèvre <vincent-gcc@vinc17.net>
7275 PR gcov-profile/101773
7276 * gcov-io.c (gcov_close): Check return code of a fclose.
7278 2021-08-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
7281 * dwarf2out.c (dwarf2out_assembly_start): Emit a dummy
7282 .file statement when needed.
7284 2021-08-04 Richard Biener <rguenther@suse.de>
7286 * tree-vect-data-refs.c (vect_check_gather_scatter):
7287 Include widening conversions only when the result is
7288 still handed by native gather or the current offset
7289 size not already matches the data size.
7290 Also succeed analysis in case there's no native support,
7291 noted by a IFN_LAST ifn and a NULL decl.
7292 (vect_analyze_data_refs): Always consider gathers.
7293 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
7294 Test for no IFN gather rather than decl gather.
7295 * tree-vect-stmts.c (vect_model_load_cost): Pass in the
7296 gather-scatter info and cost emulated gathers accordingly.
7297 (vect_truncate_gather_scatter_offset): Properly test for
7299 (vect_use_strided_gather_scatters_p): Likewise.
7300 (get_load_store_type): Handle emulated gathers and its
7302 (vectorizable_load): Likewise. Emulate them by extracting
7303 scalar offsets, doing scalar loads and a vector construct.
7305 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
7308 * expr.c (op_by_pieces_d::op_by_pieces_d): Add a max_pieces
7309 argument to set m_max_size.
7310 (move_by_pieces_d): Pass MOVE_MAX_PIECES to op_by_pieces_d.
7311 (store_by_pieces_d): Pass STORE_MAX_PIECES to op_by_pieces_d.
7312 (compare_by_pieces_d): Pass COMPARE_MAX_PIECES to op_by_pieces_d.
7314 2021-08-04 Roger Sayle <roger@nextmovesoftware.com>
7315 Marc Glisse <marc.glisse@inria.fr>
7317 * match.pd (bit_ior, bit_xor): Canonicalize (X*C1)|(X*C2) and
7318 (X*C1)^(X*C2) as X*(C1+C2), and related variants, using
7319 tree_nonzero_bits to ensure that operands are bit-wise disjoint.
7321 2021-08-04 Richard Biener <rguenther@suse.de>
7323 * tree-ssa-forwprop.c (pass_forwprop::execute): Split
7324 out code to decompose vector loads ...
7325 (optimize_vector_load): ... here. Generalize it to
7326 handle intermediate widening and TARGET_MEM_REF loads
7327 and apply it to loads with a supported vector mode as well.
7329 2021-08-04 Richard Biener <rguenther@suse.de>
7331 PR tree-optimization/101756
7332 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Make sure
7333 the result of the reduction epilogue is compatible to the original
7336 2021-08-04 liuhongt <hongtao.liu@intel.com>
7339 * config/i386/i386.md (peephole2): Refine predicate from
7340 register_operand to general_reg_operand.
7342 2021-08-04 Aldy Hernandez <aldyh@redhat.com>
7344 * gimple-range-path.h (path_range_query::dump): Mark override.
7346 2021-08-04 Richard Biener <rguenther@suse.de>
7348 PR tree-optimization/101769
7349 * tree-tailcall.c (eliminate_tail_call): Add the created loop
7350 for the first recursion and return it via the new output parameter.
7351 (optimize_tail_call): Pass through new output param.
7352 (tree_optimize_tail_calls_1): After creating all latches,
7353 add the created loop to the loop tree. Do not mark loops for fixup.
7355 2021-08-04 Martin Liska <mliska@suse.cz>
7357 * doc/invoke.texi: Document threader-mode param.
7359 2021-08-04 liuhongt <hongtao.liu@intel.com>
7361 * config/i386/sse.md (cond_fma<mode>): New expander.
7362 (cond_fms<mode>): Ditto.
7363 (cond_fnma<mode>): Ditto.
7364 (cond_fnms<mode>): Ditto.
7366 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
7368 * config/rs6000/vsx.md (*vsx_le_perm_store_<mode>): Use && instead of &.
7370 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
7372 * config/rs6000/constraints.md: Remove "e" from the list of available
7373 constraint characters.
7375 2021-08-03 Eugene Rozenfeld <erozen@microsoft.com>
7377 PR gcov-profile/71672
7378 * auto-profile.c (afdo_indirect_call): Fix setup of the historgram value for indirect calls.
7380 2021-08-03 Paul A. Clarke <pc@us.ibm.com>
7382 * config/rs6000/smmintrin.h (_mm_minpos_epu16): New.
7384 2021-08-03 H.J. Lu <hjl.tools@gmail.com>
7386 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): In 64-bit mode,
7387 try XMM31 to avoid vzeroupper.
7389 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7391 * doc/invoke.texi: Document -mtune=neoverse-512tvb and
7392 -mcpu=neoverse-512tvb.
7393 * config/aarch64/aarch64-cores.def (neoverse-512tvb): New entry.
7394 * config/aarch64/aarch64-tune.md: Regenerate.
7395 * config/aarch64/aarch64.c (neoverse512tvb_sve_vector_cost)
7396 (neoverse512tvb_sve_issue_info, neoverse512tvb_vec_issue_info)
7397 (neoverse512tvb_vector_cost, neoverse512tvb_tunings): New structures.
7398 (aarch64_adjust_body_cost_sve): Handle -mtune=neoverse-512tvb.
7399 (aarch64_adjust_body_cost): Likewise.
7401 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7403 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Only
7404 record issue information for operations that occur in the
7407 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7409 * config/aarch64/aarch64.c (aarch64_multiply_add_p): Add a vec_flags
7410 parameter. Detect cases in which an Advanced SIMD MLA would almost
7411 certainly require a MOV.
7412 (aarch64_count_ops): Update accordingly.
7414 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7416 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction): New
7417 function, split out from...
7418 (aarch64_detect_vector_stmt_subtype): ...here.
7419 (aarch64_add_stmt_cost): Treat extracting element 0 as free.
7421 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7423 * config/aarch64/aarch64-protos.h (sve_vec_cost):
7424 Add gather_load_x32_cost and gather_load_x64_cost.
7425 * config/aarch64/aarch64.c (generic_sve_vector_cost)
7426 (a64fx_sve_vector_cost, neoversev1_sve_vector_cost): Update
7427 accordingly, using the values given by the scalar_load * number
7428 of elements calculation that we used previously.
7429 (aarch64_detect_vector_stmt_subtype): Use the new fields.
7431 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7433 * config/aarch64/aarch64.c (aarch64_adjust_body_cost_sve): New
7434 function, split out from...
7435 (aarch64_adjust_body_cost): ...here.
7437 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7439 * config/aarch64/fractional-cost.h: New file.
7440 * config/aarch64/aarch64.c: Include <algorithm> (indirectly)
7441 and cost_fraction.h.
7442 (vec_cost_fraction): New typedef.
7443 (aarch64_detect_scalar_stmt_subtype): Use it for statement costs.
7444 (aarch64_detect_vector_stmt_subtype): Likewise.
7445 (aarch64_sve_adjust_stmt_cost, aarch64_adjust_stmt_cost): Likewise.
7446 (aarch64_estimate_min_cycles_per_iter): Use vec_cost_fraction
7448 (aarch64_adjust_body_cost): Likewise.
7449 (aarch64_test_cost_fraction): New function.
7450 (aarch64_run_selftests): Call it.
7452 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7454 * config/aarch64/aarch64-protos.h (tune_params::sve_width): Turn
7456 * config/aarch64/aarch64.c (aarch64_cmp_autovec_modes): Update
7458 (aarch64_estimated_poly_value): Likewise. Use the least significant
7459 set bit for the minimum and likely values. Use the most significant
7460 set bit for the maximum value.
7462 2021-08-03 liuhongt <hongtao.liu@intel.com>
7464 * config/i386/sse.md (cond_<insn><mode>): New expander.
7465 (cond_mul<mode>): Ditto.
7467 2021-08-03 Kewen Lin <linkw@linux.ibm.com>
7469 * tree-cfg.c (move_sese_region_to_fn): Fix typos on dloop.
7471 2021-08-03 liuhongt <hongtao.liu@intel.com>
7473 * config/i386/sse.md (cond_<insn><mode>):New expander.
7474 (cond_mul<mode>): Ditto.
7475 (cond_div<mode>): Ditto.
7477 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
7479 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Also
7480 check stack_realign_needed for stack realignment.
7481 (ix86_legitimate_constant_p): Always allow CONST_WIDE_INT smaller
7482 than the largest integer supported by vector register.
7483 * config/i386/i386.h (MAX_MOVE_MAX): New. Set to 64.
7484 (MOVE_MAX): Set to bytes of the largest integer supported by
7486 (STORE_MAX_PIECES): New.
7488 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
7490 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
7491 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
7492 data from one memory location to another.
7494 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
7497 * config/i386/i386.c (TARGET_GEN_MEMSET_SCRATCH_RTX): New.
7499 2021-08-02 Aldy Hernandez <aldyh@redhat.com>
7501 PR tree-optimization/101724
7502 * params.opt: Remove --param=threader-iterative.
7503 * tree-ssa-threadbackward.c (pass_thread_jumps::execute): Remove
7506 2021-08-02 Tom de Vries <tdevries@suse.de>
7508 PR middle-end/101665
7509 * doc/extend.texi (nonnull attribute): Improve documentation.
7511 2021-08-02 Andrew Pinski <apinski@marvell.com>
7513 PR rtl-optimization/101683
7514 * rtlanal.c (may_trap_p_1): Handle UNSIGNED_FIX.
7516 2021-08-02 Roger Sayle <roger@nextmovesoftware.com>
7518 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern):
7519 Renamed from cond_removal_in_popcount_clz_ctz_pattern.
7520 Add support for BSWAP, FFS, PARITY and CLRSB builtins.
7521 (tree_ssa_phiop_worker): Update call to function above.
7523 2021-08-01 H.J. Lu <hjl.tools@gmail.com>
7526 * config/i386/i386.md (bsr_rex64_1_zext): New.
7527 (combine splitter for constant - clzll): Replace gen_bsr_rex64_1
7528 with gen_bsr_rex64_1_zext.
7530 2021-07-31 Jakub Jelinek <jakub@redhat.com>
7533 * config/i386/i386.md (bsr_rex64_1, bsr_1, bsr_zext_1): New
7534 define_insn patterns.
7535 (*bsr_rex64_2, *bsr_2): New define_insn_and_split patterns.
7536 Add combine splitters for constant - clz.
7537 (clz<mode>2): Use a temporary pseudo for bsr result.
7539 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
7541 * config/rs6000/smmintrin.h (_mm_floor_pd, _mm_floor_ps,
7542 _mm_floor_sd, _mm_floor_ss): New.
7544 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
7546 * config/rs6000/smmintrin.h (_mm_ceil_pd, _mm_ceil_ps,
7547 _mm_ceil_sd, _mm_ceil_ss): New.
7549 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
7551 * config/rs6000/smmintrin.h (_mm_blend_pd, _mm_blendv_pd,
7552 _mm_blend_ps, _mm_blendv_ps): New.
7554 2021-07-30 Roger Sayle <roger@nextmovesoftware.com>
7555 Uroš Bizjak <ubizjak@gmail.com>
7557 * config/i386/i386.md (*dec_cmov<mode>): New define_insn_and_split
7558 to generate a conditional move using the carry flag after sub $1.
7559 (peephole2): Eliminate a register-to-register move by inverting
7560 the condition of a conditional move.
7562 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
7564 * config/mmix/mmix.md ("call", "call_value", "*call_real")
7565 ("*call_value_real"): Don't generate rtx mentioning the generic
7566 operands 1 and 2 to "call", and similarly for "call_value".
7567 * config/mmix/mmix.c (mmix_print_operand_punct_valid_p)
7568 (mmix_print_operand): Use '!' instead of 'p'.
7570 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
7572 * doc/md.texi (call): Correct information about operand 2.
7573 * config/mmix/mmix.md ("call", "call_value"): Remove fixed FIXMEs.
7575 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
7577 * range-op.cc (operator_trunc_mod::wi_fold): Fold constants.
7579 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
7581 * range-op.cc (operator_div::wi_fold): Return UNDEFINED for [0, 0] divisor.
7583 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
7585 * gimple-range-cache.cc (*::set_bb_range): Change const basic_block to
7587 (*::get_bb_range): Ditto.
7588 (*::bb_range_p): Ditto.
7589 * gimple-range-cache.h: Change prototypes.
7591 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
7594 * builtins.c (builtin_memcpy_read_str): Change the mode argument
7595 from scalar_int_mode to fixed_size_mode.
7596 (builtin_strncpy_read_str): Likewise.
7597 (gen_memset_value_from_prev): New function.
7598 (builtin_memset_read_str): Change the mode argument from
7599 scalar_int_mode to fixed_size_mode. Use gen_memset_value_from_prev
7600 and support CONST_VECTOR.
7601 (builtin_memset_gen_str): Likewise.
7602 (try_store_by_multiple_pieces): Use by_pieces_constfn to declare
7604 * builtins.h (builtin_strncpy_read_str): Replace scalar_int_mode
7605 with fixed_size_mode.
7606 (builtin_memset_read_str): Likewise.
7607 * expr.c (widest_int_mode_for_size): Renamed to ...
7608 (widest_fixed_size_mode_for_size): Add a bool argument to
7609 indicate if QI vector mode can be used.
7610 (by_pieces_ninsns): Call widest_fixed_size_mode_for_size
7611 instead of widest_int_mode_for_size.
7612 (pieces_addr::adjust): Change the mode argument from
7613 scalar_int_mode to fixed_size_mode.
7614 (op_by_pieces_d): Make m_len read-only. Add a bool member,
7615 m_qi_vector_mode, to indicate that QI vector mode can be used.
7616 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
7617 initialize m_qi_vector_mode. Call widest_fixed_size_mode_for_size
7618 instead of widest_int_mode_for_size.
7619 (op_by_pieces_d::get_usable_mode): Change the mode argument from
7620 scalar_int_mode to fixed_size_mode. Call
7621 widest_fixed_size_mode_for_size instead of
7622 widest_int_mode_for_size.
7623 (op_by_pieces_d::smallest_fixed_size_mode_for_size): New member
7624 function to return the smallest integer or QI vector mode.
7625 (op_by_pieces_d::run): Call widest_fixed_size_mode_for_size
7626 instead of widest_int_mode_for_size. Call
7627 smallest_fixed_size_mode_for_size instead of
7628 smallest_int_mode_for_size.
7629 (store_by_pieces_d::store_by_pieces_d): Add a bool argument to
7630 indicate that QI vector mode can be used and pass it to
7631 op_by_pieces_d::op_by_pieces_d.
7632 (can_store_by_pieces): Call widest_fixed_size_mode_for_size
7633 instead of widest_int_mode_for_size. Pass memsetp to
7634 widest_fixed_size_mode_for_size to support QI vector mode.
7635 Allow all CONST_VECTORs for memset if vec_duplicate is supported.
7636 (store_by_pieces): Pass memsetp to
7637 store_by_pieces_d::store_by_pieces_d.
7638 (clear_by_pieces_1): Removed.
7639 (clear_by_pieces): Replace clear_by_pieces_1 with
7640 builtin_memset_read_str and pass true to store_by_pieces_d to
7641 support vector mode broadcast.
7642 (string_cst_read_str): Change the mode argument from
7643 scalar_int_mode to fixed_size_mode.
7644 * expr.h (by_pieces_constfn): Change scalar_int_mode to
7646 (by_pieces_prev): Likewise.
7647 * rtl.h (lowpart_subreg_regno): New.
7648 * rtlanal.c (lowpart_subreg_regno): New. A wrapper around
7649 simplify_subreg_regno.
7650 * target.def (gen_memset_scratch_rtx): New hook.
7651 * doc/tm.texi.in: Add TARGET_GEN_MEMSET_SCRATCH_RTX.
7652 * doc/tm.texi: Regenerated.
7654 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
7657 * config/mips/mips.c (mips_atomic_assign_expand_fenv): Use
7658 TARGET_EXPR instead of MODIFY_EXPR.
7660 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
7663 * config/mips/mips-protos.h (mips_expand_vec_cmp_expr): Declare.
7664 * config/mips/mips.c (mips_expand_vec_cmp_expr): New function.
7665 * config/mips/mips-msa.md (vec_cmp<MSA:mode><mode_i>): New
7667 (vec_cmpu<IMSA:mode><mode_i>): New expander.
7669 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
7672 * config/i386/i386-options.c (ix86_option_override_internal):
7673 Don't enable LZCNT/POPCNT if they have been disabled explicitly.
7675 2021-07-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
7678 * config/arm/arm_neon.h (vld1_p64): Replace call to builtin by
7679 explicitly dereferencing __a.
7680 (vld1_s64): Likewise.
7681 (vld1_u64): Likewise.
7682 * config/arm/arm_neon_builtins.def (vld1): Remove entry for di
7683 and change to VAR13.
7685 2021-07-30 Aldy Hernandez <aldyh@redhat.com>
7687 * gimple-loop-versioning.cc (lv_dom_walker::lv_dom_walker): Remove
7688 use of m_range_analyzer.
7689 (loop_versioning::lv_dom_walker::before_dom_children): Same.
7690 (loop_versioning::lv_dom_walker::after_dom_children): Remove.
7691 (loop_versioning::prune_loop_conditions): Replace vr_values use
7692 with range_query interface.
7693 (pass_loop_versioning::execute): Use ranger.
7695 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
7698 * ipa-devirt.c (ipa_odr_read_section): Compare the precision of
7699 enum values, and emit a warning if they mismatch.
7701 2021-07-30 Kewen Lin <linkw@linux.ibm.com>
7703 * cfgloop.h (as_const): New function.
7704 (class loop_iterator): Rename to ...
7705 (class loops_list): ... this.
7706 (loop_iterator::next): Rename to ...
7707 (loops_list::Iter::fill_curr_loop): ... this and adjust.
7708 (loop_iterator::loop_iterator): Rename to ...
7709 (loops_list::loops_list): ... this and adjust.
7710 (loops_list::Iter): New class.
7711 (loops_list::iterator): New type.
7712 (loops_list::const_iterator): New type.
7713 (loops_list::begin): New function.
7714 (loops_list::end): Likewise.
7715 (loops_list::begin const): Likewise.
7716 (loops_list::end const): Likewise.
7717 (FOR_EACH_LOOP): Remove.
7718 (FOR_EACH_LOOP_FN): Remove.
7719 * cfgloop.c (flow_loops_dump): Adjust FOR_EACH_LOOP* with range-based
7720 for loop with loops_list instance.
7721 (sort_sibling_loops): Likewise.
7722 (disambiguate_loops_with_multiple_latches): Likewise.
7723 (verify_loop_structure): Likewise.
7724 * cfgloopmanip.c (create_preheaders): Likewise.
7725 (force_single_succ_latches): Likewise.
7726 * config/aarch64/falkor-tag-collision-avoidance.c
7727 (execute_tag_collision_avoidance): Likewise.
7728 * config/mn10300/mn10300.c (mn10300_scan_for_setlb_lcc): Likewise.
7729 * config/s390/s390.c (s390_adjust_loops): Likewise.
7730 * doc/loop.texi: Likewise.
7731 * gimple-loop-interchange.cc (pass_linterchange::execute): Likewise.
7732 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
7733 * gimple-loop-versioning.cc (loop_versioning::analyze_blocks): Likewise.
7734 (loop_versioning::make_versioning_decisions): Likewise.
7735 * gimple-ssa-split-paths.c (split_paths): Likewise.
7736 * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
7737 * graphite.c (canonicalize_loop_form): Likewise.
7738 (graphite_transform_loops): Likewise.
7739 * ipa-fnsummary.c (analyze_function_body): Likewise.
7740 * ipa-pure-const.c (analyze_function): Likewise.
7741 * loop-doloop.c (doloop_optimize_loops): Likewise.
7742 * loop-init.c (loop_optimizer_finalize): Likewise.
7743 (fix_loop_structure): Likewise.
7744 * loop-invariant.c (calculate_loop_reg_pressure): Likewise.
7745 (move_loop_invariants): Likewise.
7746 * loop-unroll.c (decide_unrolling): Likewise.
7747 (unroll_loops): Likewise.
7748 * modulo-sched.c (sms_schedule): Likewise.
7749 * predict.c (predict_loops): Likewise.
7750 (pass_profile::execute): Likewise.
7751 * profile.c (branch_prob): Likewise.
7752 * sel-sched-ir.c (sel_finish_pipelining): Likewise.
7753 (sel_find_rgns): Likewise.
7754 * tree-cfg.c (replace_loop_annotate): Likewise.
7755 (replace_uses_by): Likewise.
7756 (move_sese_region_to_fn): Likewise.
7757 * tree-if-conv.c (pass_if_conversion::execute): Likewise.
7758 * tree-loop-distribution.c (loop_distribution::execute): Likewise.
7759 * tree-parloops.c (parallelize_loops): Likewise.
7760 * tree-predcom.c (tree_predictive_commoning): Likewise.
7761 * tree-scalar-evolution.c (scev_initialize): Likewise.
7762 (scev_reset): Likewise.
7763 * tree-ssa-dce.c (find_obviously_necessary_stmts): Likewise.
7764 * tree-ssa-live.c (remove_unused_locals): Likewise.
7765 * tree-ssa-loop-ch.c (ch_base::copy_headers): Likewise.
7766 * tree-ssa-loop-im.c (analyze_memory_references): Likewise.
7767 (tree_ssa_lim_initialize): Likewise.
7768 * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables): Likewise.
7769 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize): Likewise.
7770 * tree-ssa-loop-manip.c (get_loops_exits): Likewise.
7771 * tree-ssa-loop-niter.c (estimate_numbers_of_iterations): Likewise.
7772 (free_numbers_of_iterations_estimates): Likewise.
7773 * tree-ssa-loop-prefetch.c (tree_ssa_prefetch_arrays): Likewise.
7774 * tree-ssa-loop-split.c (tree_ssa_split_loops): Likewise.
7775 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Likewise.
7776 * tree-ssa-loop.c (gate_oacc_kernels): Likewise.
7777 (pass_scev_cprop::execute): Likewise.
7778 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
7779 * tree-ssa-sccvn.c (do_rpo_vn): Likewise.
7780 * tree-ssa-threadupdate.c
7781 (jump_thread_path_registry::thread_through_all_blocks): Likewise.
7782 * tree-vectorizer.c (vectorize_loops): Likewise.
7783 * tree-vrp.c (vrp_asserts::find_assert_locations): Likewise.
7785 2021-07-29 Hans-Peter Nilsson <hp@bitrange.com>
7787 * config/mmix/mmix.c (mmix_function_arg_1): Avoid
7788 generating a VOIDmode register for e.g the
7789 function_arg_info::end_marker.
7791 2021-07-29 Jeff Law <jeffreyalaw@gmail.com>
7793 * config/h8300/h8300-modes.def: Add CCZ, CCV and CCC, drop CCZNV.
7794 * config/h8300/h8300.md (H8cc mode iterator): Add CCZ.
7795 (cc mode_attr): Similarly.
7796 (ccz subst_attr): Similarly.
7797 * config/h8300/jumpcall.md: Add new patterns for branch-on-bit.
7798 * config/h8300/testcompare.md: Remove various cc0 based patterns
7799 that had been commented out. Add pattern to set CCZ from a bit
7802 2021-07-29 Thomas Schwinge <thomas@codesourcery.com>
7803 Julian Brown <julian@codesourcery.com>
7804 Kwok Cheung Yeung <kcy@codesourcery.com>
7806 * omp-offload.c (oacc_loop_xform_head_tail, oacc_loop_process):
7807 'update_stmt' after modification.
7808 (pass_oacc_loop_designation): New function, extracted out of...
7809 (pass_oacc_device_lower): ... this.
7810 (pass_data_oacc_loop_designation, pass_oacc_loop_designation)
7811 (make_pass_oacc_loop_designation): New
7812 * passes.def: Add it.
7813 * tree-parloops.c (create_parallel_loop): Adjust.
7814 * tree-pass.h (make_pass_oacc_loop_designation): New.
7816 2021-07-29 Aldy Hernandez <aldyh@redhat.com>
7818 * flag-types.h (enum threader_mode): New.
7819 * params.opt: Add entry for --param=threader-mode.
7820 * tree-ssa-threadbackward.c (THREADER_ITERATIVE_MODE): New.
7821 (class back_threader): New.
7822 (back_threader::back_threader): New.
7823 (back_threader::~back_threader): New.
7824 (back_threader::maybe_register_path): New.
7825 (back_threader::find_taken_edge): New.
7826 (back_threader::find_taken_edge_switch): New.
7827 (back_threader::find_taken_edge_cond): New.
7828 (back_threader::resolve_def): New.
7829 (back_threader::resolve_phi): New.
7830 (back_threader::find_paths_to_names): New.
7831 (back_threader::find_paths): New.
7834 (thread_jumps::find_jump_threads_backwards): Call ranger threader.
7835 (thread_jumps::find_jump_threads_backwards_with_ranger): New.
7836 (pass_thread_jumps::execute): Abstract out code...
7837 (try_thread_blocks): ...here.
7838 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
7839 Abstract out threading candidate code to...
7840 (single_succ_to_potentially_threadable_block): ...here.
7841 * tree-ssa-threadedge.h (single_succ_to_potentially_threadable_block):
7843 * tree-ssa-threadupdate.c (register_jump_thread): Return boolean.
7844 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
7845 Return bool from register_jump_thread.
7847 2021-07-29 Andreas Krebbel <krebbel@linux.ibm.com>
7849 * target.def: in0 and in1 do not need to be registers.
7850 * doc/tm.texi: Regenerate.
7852 2021-07-29 liuhongt <hongtao.liu@intel.com>
7855 * config/i386/i386.c (ix86_widen_mult_cost): New function.
7856 (ix86_add_stmt_cost): Use ix86_widen_mult_cost for
7859 2021-07-29 Jiufu Guo <guojiufu@linux.ibm.com>
7862 * config/rs6000/rs6000.c (TARGET_PREFERRED_DOLOOP_MODE): New hook.
7863 (rs6000_preferred_doloop_mode): New hook.
7864 * doc/tm.texi: Regenerate.
7865 * doc/tm.texi.in: Add hook preferred_doloop_mode.
7866 * target.def (preferred_doloop_mode): New hook.
7867 * targhooks.c (default_preferred_doloop_mode): New hook.
7868 * targhooks.h (default_preferred_doloop_mode): New hook.
7869 * tree-ssa-loop-ivopts.c (compute_doloop_base_on_mode): New function.
7870 (add_iv_candidate_for_doloop): Call targetm.preferred_doloop_mode
7871 and compute_doloop_base_on_mode.
7873 2021-07-28 Martin Sebor <msebor@redhat.com>
7875 PR middle-end/101494
7876 * tree-ssa-uninit.c (maybe_warn_operand): Correct object offset
7877 and size computation.
7879 2021-07-28 Martin Sebor <msebor@redhat.com>
7881 PR middle-end/101601
7882 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Remove
7884 Handle pointers to functions.
7886 2021-07-28 Martin Sebor <msebor@redhat.com>
7888 * Makefile.in (OBJS): Add gimple-ssa-warn-access.o and pointer-query.o.
7889 * attribs.h (fndecl_dealloc_argno): Move fndecl_dealloc_argno to tree.h.
7890 * builtins.c (compute_objsize_r): Move to pointer-query.cc.
7891 (access_ref::access_ref): Same.
7892 (access_ref::phi): Same.
7893 (access_ref::get_ref): Same.
7894 (access_ref::size_remaining): Same.
7895 (access_ref::offset_in_range): Same.
7896 (access_ref::add_offset): Same.
7897 (access_ref::inform_access): Same.
7898 (ssa_name_limit_t::visit_phi): Same.
7899 (ssa_name_limit_t::leave_phi): Same.
7900 (ssa_name_limit_t::next): Same.
7901 (ssa_name_limit_t::next_phi): Same.
7902 (ssa_name_limit_t::~ssa_name_limit_t): Same.
7903 (pointer_query::pointer_query): Same.
7904 (pointer_query::get_ref): Same.
7905 (pointer_query::put_ref): Same.
7906 (pointer_query::flush_cache): Same.
7907 (warn_string_no_nul): Move to gimple-ssa-warn-access.cc.
7908 (check_nul_terminated_array): Same.
7909 (unterminated_array): Same.
7910 (maybe_warn_for_bound): Same.
7911 (check_read_access): Same.
7912 (warn_for_access): Same.
7913 (get_size_range): Same.
7914 (check_access): Same.
7915 (gimple_call_alloc_size): Move to tree.c.
7916 (gimple_parm_array_size): Move to pointer-query.cc.
7917 (get_offset_range): Same.
7918 (gimple_call_return_array): Same.
7919 (handle_min_max_size): Same.
7920 (handle_array_ref): Same.
7921 (handle_mem_ref): Same.
7922 (compute_objsize): Same.
7923 (gimple_call_alloc_p): Move to gimple-ssa-warn-access.cc.
7924 (call_dealloc_argno): Same.
7925 (fndecl_dealloc_argno): Same.
7926 (new_delete_mismatch_p): Same.
7927 (matching_alloc_calls_p): Same.
7928 (warn_dealloc_offset): Same.
7929 (maybe_emit_free_warning): Same.
7930 * builtins.h (check_nul_terminated_array): Move to
7931 gimple-ssa-warn-access.h.
7932 (check_nul_terminated_array): Same.
7933 (warn_string_no_nul): Same.
7934 (unterminated_array): Same.
7935 (class ssa_name_limit_t): Same.
7936 (class pointer_query): Same.
7937 (struct access_ref): Same.
7938 (class range_query): Same.
7939 (struct access_data): Same.
7940 (gimple_call_alloc_size): Same.
7941 (gimple_parm_array_size): Same.
7942 (compute_objsize): Same.
7943 (class access_data): Same.
7944 (maybe_emit_free_warning): Same.
7945 * calls.c (initialize_argument_information): Remove call to
7946 maybe_emit_free_warning.
7947 * gimple-array-bounds.cc: Include new header..
7948 * gimple-fold.c: Same.
7949 * gimple-ssa-sprintf.c: Same.
7950 * gimple-ssa-warn-restrict.c: Same.
7951 * passes.def: Add pass_warn_access.
7952 * tree-pass.h (make_pass_warn_access): Declare.
7953 * tree-ssa-strlen.c: Include new headers.
7954 * tree.c (fndecl_dealloc_argno): Move here from builtins.c.
7955 * tree.h (fndecl_dealloc_argno): Move here from attribs.h.
7956 * gimple-ssa-warn-access.cc: New file.
7957 * gimple-ssa-warn-access.h: New file.
7958 * pointer-query.cc: New file.
7959 * pointer-query.h: New file.
7961 2021-07-28 Jakub Jelinek <jakub@redhat.com>
7963 PR middle-end/101624
7964 * ubsan.c (maybe_instrument_pointer_overflow,
7965 instrument_object_size): Only test DECL_REGISTER on VAR_DECLs,
7966 PARM_DECLs or RESULT_DECLs.
7967 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Likewise.
7969 2021-07-28 Jakub Jelinek <jakub@redhat.com>
7971 PR middle-end/101642
7972 * match.pd (bswap16 (x) == bswap16 (y)): Cast both operands
7973 to type of bswap16 for comparison.
7974 (bswap16 (x) == cst): Cast bswap16 operand to type of cst.
7976 2021-07-28 Richard Biener <rguenther@suse.de>
7978 PR tree-optimization/101615
7979 * tree-vect-slp.c (vect_optimize_slp): Materialize permutes
7980 at CTOR SLP graph entries.
7982 2021-07-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7984 * config/aarch64/aarch64.md (*extend<SHORT:mode><GPI:mode>2_aarch64):
7985 Add "r,w" alternative.
7987 2021-07-28 H.J. Lu <hjl.tools@gmail.com>
7990 * config/i386/i386.c (ix86_avx_u128_mode_needed): Don't set
7991 AVX_U128_DIRTY when all bits are zero.
7993 2021-07-28 Richard Biener <rguenther@suse.de>
7995 PR tree-optimization/101615
7996 * tree-vect-slp.c (vect_optimize_slp): Pre-existing vector
7997 external nodes cannot be permuted so make them perm_out 0.
7999 2021-07-28 Andrew Stubbs <ams@codesourcery.com>
8002 * config.in: Regenerate.
8003 * config/gcn/gcn-hsa.h (A_FIJI): New define.
8004 (A_900): New define.
8005 (A_906): New define.
8006 (A_908): New define.
8007 (ASM_SPEC): Use A_FIJI, A_900, A_906 and A_908.
8008 * config/gcn/gcn.c (output_file_start): Adjust attributes according
8009 to the assembler capabilities.
8010 * config/gcn/mkoffload.c (main): Likewise.
8011 * configure: Regenerate.
8012 * configure.ac: Add tests for LLVM assembler attribute features.
8014 2021-07-28 Andrew MacLeod <amacleod@redhat.com>
8016 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Check for
8017 cond_false and cond_true on branches.
8019 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
8021 * config/aarch64/aarch64.c (aarch64_gen_adjusted_ldpstp): use
8024 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
8026 * alias.c (init_alias_analysis): Don't skip prologue/epilogue.
8028 2021-07-28 Jakub Jelinek <jakub@redhat.com>
8031 * config/i386/sse.md (vashr<mode>3): Split into vashrv8di3 expander
8032 and vashrv4di3 expander, where the latter requires just TARGET_AVX2
8033 and has special !TARGET_AVX512VL expansion.
8034 (vashrv2di3<mask_name>): Rename to ...
8035 (vashrv2di3): ... this. Change condition to TARGET_XOP || TARGET_AVX2
8036 and add special !TARGET_XOP && !TARGET_AVX512VL expansion.
8038 2021-07-28 Martin Uecker <muecker@gwdg.de>
8040 * calls.c (maybe_warn_rdwr_sizes): Correct argument
8041 numbers in warning that were switched.
8043 2021-07-28 Kewen Lin <linkw@linux.ibm.com>
8045 PR tree-optimization/101596
8046 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Fix wrong check
8047 by using new_type's precision instead.
8049 2021-07-28 liuhongt <hongtao.liu@intel.com>
8052 * config/i386/i386.h (processor_costs): Add new member
8054 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
8055 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
8056 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
8057 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
8058 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
8059 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
8060 generic_cost, core_cost): Initialize integer_to_sse same value
8062 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
8063 * config/i386/i386.c (ix86_builtin_vectorization_cost):
8064 Use integer_to_sse instead of sse_op to calculate the cost of
8067 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
8069 * config/rs6000/rs6000-gen-builtins.c (write_ovld_static_init): New
8071 (write_init_file): Call write_ovld_static_init.
8073 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
8075 * config/rs6000/rs6000-gen-builtins.c (write_bif_static_init): New
8077 (write_init_file): Call write_bif_static_init.
8079 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
8081 * config/rs6000/rs6000-gen-builtins.c (typemap): New struct.
8082 (TYPE_MAP_SIZE): New macro.
8083 (type_map): New initialized variable.
8084 (typemap_cmp): New function.
8085 (write_type_node): Likewise.
8086 (write_fntype_init): Implement.
8088 2021-07-27 Martin Sebor <msebor@redhat.com>
8090 PR tree-optimization/101584
8091 * tree-ssa-uninit.c (builtin_call_nomodifying_p): New function.
8092 (check_defs): Call it.
8094 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
8096 * tree-ssa-dom.c (dom_jump_threader_simplifier):
8097 Put avail_exprs_stack in the class, instead of passing it to
8098 jump_threader_simplifier.
8099 (dom_jump_threader_simplifier::simplify): Add state argument.
8100 (dom_opt_dom_walker): Add state.
8101 (pass_dominator::execute): Pass state to threader.
8102 (dom_opt_dom_walker::before_dom_children): Use state.
8103 * tree-ssa-threadedge.c (jump_threader::jump_threader): Replace
8105 (jump_threader::record_temporary_equivalences_from_phis):
8106 Register equivalences through the state variable.
8107 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
8108 Record ranges in a statement through the state variable.
8109 (jump_threader::simplify_control_stmt_condition): Pass state to
8111 (jump_threader::simplify_control_stmt_condition_1): Same.
8112 (jump_threader::thread_around_empty_blocks): Remove obsolete
8114 (jump_threader::thread_through_normal_block): Record equivalences
8115 on edge through the state variable.
8116 (jump_threader::thread_across_edge): Abstract state pushing.
8117 (jt_state::jt_state): New.
8118 (jt_state::push): New.
8119 (jt_state::pop): New.
8120 (jt_state::register_equiv): New.
8121 (jt_state::record_ranges_from_stmt): New.
8122 (jt_state::register_equivs_on_edge): New.
8123 (jump_threader_simplifier::jump_threader_simplifier): Move from
8125 (jump_threader_simplifier::simplify): Add state argument.
8126 * tree-ssa-threadedge.h (class jt_state): New.
8127 (class jump_threader): Add state to constructor.
8128 (class jump_threader_simplifier): Add state to simplify. Remove
8129 avail_exprs_stack from class.
8130 * tree-vrp.c (vrp_jump_threader_simplifier::simplify): Add state
8132 (vrp_jump_threader::vrp_jump_threader): Add state.
8133 (vrp_jump_threader::~vrp_jump_threader): Cleanup state.
8135 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
8137 * Makefile.in (OBJS): Add gimple-range-path.o.
8138 * gimple-range-path.cc: New file.
8139 * gimple-range-path.h: New file.
8141 2021-07-27 Jonathan Wright <jonathan.wright@arm.com>
8143 * config/aarch64/aarch64-simd.md: Push sign/zero-extension
8144 inside vec_duplicate for all patterns.
8145 * simplify-rtx.c (simplify_context::simplify_unary_operation_1):
8146 Push sign/zero-extension inside vec_duplicate.
8148 2021-07-27 Richard Biener <rguenther@suse.de>
8150 PR tree-optimization/101573
8151 * tree-ssa-uninit.c (warn_uninit_phi_uses): New function
8152 looking at uninitialized PHI arg defs in some constrained cases.
8153 (warn_uninitialized_vars): Call it.
8154 (execute_early_warn_uninitialized): Calculate dominators.
8156 2021-07-27 Richard Biener <rguenther@suse.de>
8158 PR tree-optimization/39821
8159 * tree-vect-stmts.c (vect_model_promotion_demotion_cost): Use
8160 vector_stmt for widening arithmetic.
8161 (vectorizable_conversion): Adjust.
8163 2021-07-27 Martin Jambor <mjambor@suse.cz>
8165 * cgraph.h (ipa_replace_map): New field force_load_ref.
8166 * ipa-prop.h (ipa_param_descriptor): Reduce precision of move_cost,
8167 aded new flag load_dereferenced, adjusted comments.
8168 (ipa_get_param_dereferenced): New function.
8169 (ipa_set_param_dereferenced): Likewise.
8170 * cgraphclones.c (cgraph_node::create_virtual_clone): Follow it.
8171 * ipa-cp.c: Include gimple.h.
8172 (ipcp_discover_new_direct_edges): Take into account dereferenced flag.
8173 (get_replacement_map): New parameter force_load_ref, set the
8174 appropriate flag in ipa_replace_map if set.
8175 (struct symbol_and_index_together): New type.
8176 (adjust_refs_in_act_callers): New function.
8177 (adjust_references_in_caller): Likewise.
8178 (create_specialized_node): When appropriate, call
8179 adjust_references_in_caller and force only load references.
8180 * ipa-prop.c (load_from_dereferenced_name): New function.
8181 (ipa_analyze_controlled_uses): Also detect loads from a
8182 dereference, harden testing of call statements.
8183 (ipa_write_node_info): Stream the dereferenced flag.
8184 (ipa_read_node_info): Likewise.
8185 (ipa_set_jf_constant): Also create refdesc when jump function
8186 references a variable.
8187 (cgraph_node_for_jfunc): Rename to symtab_node_for_jfunc, work
8188 also on references of variables and return a symtab_node. Adjust
8190 (propagate_controlled_uses): Also remove references to VAR_DECLs.
8192 2021-07-27 Jakub Jelinek <jakub@redhat.com>
8194 PR middle-end/101586
8195 * gimple-fold.c (clear_padding_type): Ignore FIELD_DECLs with byte
8196 positions above or equal to sz except for diagnostics of flexible
8199 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
8201 PR tree-optimization/78888
8202 * gimple-range-fold.cc (get_letter_range): New.
8203 (fold_using_range::range_of_builtin_call): Call get_letter_range.
8205 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
8207 PR tree-optimization/78888
8208 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Add cases
8209 for CFN_BUILT_IN_TOUPPER and CFN_BUILT_IN_TOLOWER.
8211 2021-07-26 Roger Sayle <roger@nextmovesoftware.com>
8212 Marc Glisse <marc.glisse@inria.fr>
8214 * match.pd (rotate): Simplify equality/inequality of rotations.
8215 (bswap): Simplify equality/inequality tests of byte swapping.
8217 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
8219 * range-op.cc (operator_bitwise_xor::op1_op2_relation_effect):
8222 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
8224 * range-op.cc (operator_lshift::fold_range): Pass rel to
8225 base class fold_range.
8226 (operator_rshift::fold_range): Same.
8228 2021-07-26 Ashimida <ashimida@linux.alibaba.com>
8231 * toplev.h (min_align_loops_log): Remove declaration.
8232 (min_align_jumps_log, min_align_labels_log): Likewise.
8233 (min_align_functions_log): Likewise.
8235 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
8237 * tree-vrp.c (vrp_simplify_cond_using_ranges): Rename vr_values
8239 (execute_vrp): Abstract out simplification of conditionals...
8240 (simplify_casted_conds): ...here.
8242 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
8244 * gimple-array-bounds.cc (array_bounds_checker::get_value_range):
8245 Add gimple argument.
8246 (array_bounds_checker::check_array_ref): Same.
8247 (array_bounds_checker::check_addr_expr): Same.
8248 (array_bounds_checker::check_array_bounds): Pass statement to
8249 check_array_bounds and check_addr_expr.
8250 * gimple-array-bounds.h (check_array_bounds): Add gimple argument.
8251 (check_addr_expr): Same.
8252 (get_value_range): Same.
8254 2021-07-26 Tamar Christina <tamar.christina@arm.com>
8256 * config/aarch64/aarch64-simd-builtins.def (sdot, udot): Rename to..
8257 (sdot_prod, udot_prod): ... This.
8258 * config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>): Merged
8260 (<sur>dot_prod<vsi2qi>): ... this.
8261 (aarch64_<sur>dot_lane<vsi2qi>, aarch64_<sur>dot_laneq<vsi2qi>):
8262 Change operands order.
8263 (<sur>sadv16qi): Use new operands order.
8264 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32,
8265 vdotq_s32): Use new RTL ordering.
8267 2021-07-26 Tamar Christina <tamar.christina@arm.com>
8269 * config/aarch64/aarch64-builtins.c (TYPES_TERNOP_SUSS,
8270 aarch64_types_ternop_suss_qualifiers): New.
8271 * config/aarch64/aarch64-simd-builtins.def (usdot_prod): Use it.
8272 * config/aarch64/aarch64-simd.md (usdot_prod<vsi2qi>): Re-organize RTL.
8273 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Use it.
8275 2021-07-23 Jakub Jelinek <jakub@redhat.com>
8277 PR rtl-optimization/101562
8278 * expmed.c (store_integral_bit_field): Only use movstrict_optab
8279 if the operand isn't paradoxical.
8281 2021-07-23 Aldy Hernandez <aldyh@redhat.com>
8283 * gimple-array-bounds.h (class array_bounds_checker): Change
8284 ranges type to range_query.
8286 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8288 * config/aarch64/arm_neon.h (vst1_s64_x2): Use
8289 __builtin_memcpy instead of constructing
8290 __builtin_aarch64_simd_oi one vector at a time.
8291 (vst1_u64_x2): Likewise.
8292 (vst1_f64_x2): Likewise.
8293 (vst1_s8_x2): Likewise.
8294 (vst1_p8_x2): Likewise.
8295 (vst1_s16_x2): Likewise.
8296 (vst1_p16_x2): Likewise.
8297 (vst1_s32_x2): Likewise.
8298 (vst1_u8_x2): Likewise.
8299 (vst1_u16_x2): Likewise.
8300 (vst1_u32_x2): Likewise.
8301 (vst1_f16_x2): Likewise.
8302 (vst1_f32_x2): Likewise.
8303 (vst1_p64_x2): Likewise.
8304 (vst1q_s8_x2): Likewise.
8305 (vst1q_p8_x2): Likewise.
8306 (vst1q_s16_x2): Likewise.
8307 (vst1q_p16_x2): Likewise.
8308 (vst1q_s32_x2): Likewise.
8309 (vst1q_s64_x2): Likewise.
8310 (vst1q_u8_x2): Likewise.
8311 (vst1q_u16_x2): Likewise.
8312 (vst1q_u32_x2): Likewise.
8313 (vst1q_u64_x2): Likewise.
8314 (vst1q_f16_x2): Likewise.
8315 (vst1q_f32_x2): Likewise.
8316 (vst1q_f64_x2): Likewise.
8317 (vst1q_p64_x2): Likewise.
8319 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8321 * config/aarch64/arm_neon.h (vst1_s64_x3): Use
8322 __builtin_memcpy instead of constructing
8323 __builtin_aarch64_simd_ci one vector at a time.
8324 (vst1_u64_x3): Likewise.
8325 (vst1_f64_x3): Likewise.
8326 (vst1_s8_x3): Likewise.
8327 (vst1_p8_x3): Likewise.
8328 (vst1_s16_x3): Likewise.
8329 (vst1_p16_x3): Likewise.
8330 (vst1_s32_x3): Likewise.
8331 (vst1_u8_x3): Likewise.
8332 (vst1_u16_x3): Likewise.
8333 (vst1_u32_x3): Likewise.
8334 (vst1_f16_x3): Likewise.
8335 (vst1_f32_x3): Likewise.
8336 (vst1_p64_x3): Likewise.
8337 (vst1q_s8_x3): Likewise.
8338 (vst1q_p8_x3): Likewise.
8339 (vst1q_s16_x3): Likewise.
8340 (vst1q_p16_x3): Likewise.
8341 (vst1q_s32_x3): Likewise.
8342 (vst1q_s64_x3): Likewise.
8343 (vst1q_u8_x3): Likewise.
8344 (vst1q_u16_x3): Likewise.
8345 (vst1q_u32_x3): Likewise.
8346 (vst1q_u64_x3): Likewise.
8347 (vst1q_f16_x3): Likewise.
8348 (vst1q_f32_x3): Likewise.
8349 (vst1q_f64_x3): Likewise.
8350 (vst1q_p64_x3): Likewise.
8352 2021-07-23 H.J. Lu <hjl.tools@gmail.com>
8355 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): Don't return
8356 hard register when LRA is in progress.
8358 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8360 * config/aarch64/arm_neon.h (vst1_s8_x4): Use
8361 __builtin_memcpy instead of using a union.
8362 (vst1q_s8_x4): Likewise.
8363 (vst1_s16_x4): Likewise.
8364 (vst1q_s16_x4): Likewise.
8365 (vst1_s32_x4): Likewise.
8366 (vst1q_s32_x4): Likewise.
8367 (vst1_u8_x4): Likewise.
8368 (vst1q_u8_x4): Likewise.
8369 (vst1_u16_x4): Likewise.
8370 (vst1q_u16_x4): Likewise.
8371 (vst1_u32_x4): Likewise.
8372 (vst1q_u32_x4): Likewise.
8373 (vst1_f16_x4): Likewise.
8374 (vst1q_f16_x4): Likewise.
8375 (vst1_f32_x4): Likewise.
8376 (vst1q_f32_x4): Likewise.
8377 (vst1_p8_x4): Likewise.
8378 (vst1q_p8_x4): Likewise.
8379 (vst1_p16_x4): Likewise.
8380 (vst1q_p16_x4): Likewise.
8381 (vst1_s64_x4): Likewise.
8382 (vst1_u64_x4): Likewise.
8383 (vst1_p64_x4): Likewise.
8384 (vst1q_s64_x4): Likewise.
8385 (vst1q_u64_x4): Likewise.
8386 (vst1q_p64_x4): Likewise.
8387 (vst1_f64_x4): Likewise.
8388 (vst1q_f64_x4): Likewise.
8390 2021-07-23 Jonathan Wrightt <jonathan.wright@arm.com>
8392 * config/aarch64/arm_neon.h (vst2_s64): Use __builtin_memcpy
8393 instead of constructing __builtin_aarch64_simd_oi one vector
8395 (vst2_u64): Likewise.
8396 (vst2_f64): Likewise.
8397 (vst2_s8): Likewise.
8398 (vst2_p8): Likewise.
8399 (vst2_s16): Likewise.
8400 (vst2_p16): Likewise.
8401 (vst2_s32): Likewise.
8402 (vst2_u8): Likewise.
8403 (vst2_u16): Likewise.
8404 (vst2_u32): Likewise.
8405 (vst2_f16): Likewise.
8406 (vst2_f32): Likewise.
8407 (vst2_p64): Likewise.
8408 (vst2q_s8): Likewise.
8409 (vst2q_p8): Likewise.
8410 (vst2q_s16): Likewise.
8411 (vst2q_p16): Likewise.
8412 (vst2q_s32): Likewise.
8413 (vst2q_s64): Likewise.
8414 (vst2q_u8): Likewise.
8415 (vst2q_u16): Likewise.
8416 (vst2q_u32): Likewise.
8417 (vst2q_u64): Likewise.
8418 (vst2q_f16): Likewise.
8419 (vst2q_f32): Likewise.
8420 (vst2q_f64): Likewise.
8421 (vst2q_p64): Likewise.
8423 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8425 * config/aarch64/arm_neon.h (vst3_s64): Use __builtin_memcpy
8426 instead of constructing __builtin_aarch64_simd_ci one vector
8428 (vst3_u64): Likewise.
8429 (vst3_f64): Likewise.
8430 (vst3_s8): Likewise.
8431 (vst3_p8): Likewise.
8432 (vst3_s16): Likewise.
8433 (vst3_p16): Likewise.
8434 (vst3_s32): Likewise.
8435 (vst3_u8): Likewise.
8436 (vst3_u16): Likewise.
8437 (vst3_u32): Likewise.
8438 (vst3_f16): Likewise.
8439 (vst3_f32): Likewise.
8440 (vst3_p64): Likewise.
8441 (vst3q_s8): Likewise.
8442 (vst3q_p8): Likewise.
8443 (vst3q_s16): Likewise.
8444 (vst3q_p16): Likewise.
8445 (vst3q_s32): Likewise.
8446 (vst3q_s64): Likewise.
8447 (vst3q_u8): Likewise.
8448 (vst3q_u16): Likewise.
8449 (vst3q_u32): Likewise.
8450 (vst3q_u64): Likewise.
8451 (vst3q_f16): Likewise.
8452 (vst3q_f32): Likewise.
8453 (vst3q_f64): Likewise.
8454 (vst3q_p64): Likewise.
8456 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8458 * config/aarch64/arm_neon.h (vst4_s64): Use __builtin_memcpy
8459 instead of constructing __builtin_aarch64_simd_xi one vector
8461 (vst4_u64): Likewise.
8462 (vst4_f64): Likewise.
8463 (vst4_s8): Likewise.
8464 (vst4_p8): Likewise.
8465 (vst4_s16): Likewise.
8466 (vst4_p16): Likewise.
8467 (vst4_s32): Likewise.
8468 (vst4_u8): Likewise.
8469 (vst4_u16): Likewise.
8470 (vst4_u32): Likewise.
8471 (vst4_f16): Likewise.
8472 (vst4_f32): Likewise.
8473 (vst4_p64): Likewise.
8474 (vst4q_s8): Likewise.
8475 (vst4q_p8): Likewise.
8476 (vst4q_s16): Likewise.
8477 (vst4q_p16): Likewise.
8478 (vst4q_s32): Likewise.
8479 (vst4q_s64): Likewise.
8480 (vst4q_u8): Likewise.
8481 (vst4q_u16): Likewise.
8482 (vst4q_u32): Likewise.
8483 (vst4q_u64): Likewise.
8484 (vst4q_f16): Likewise.
8485 (vst4q_f32): Likewise.
8486 (vst4q_f64): Likewise.
8487 (vst4q_p64): Likewise.
8489 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8491 * config/aarch64/arm_neon.h (vtbx4_s8): Use __builtin_memcpy
8492 instead of constructing __builtin_aarch64_simd_oi one vector
8494 (vtbx4_u8): Likewise.
8495 (vtbx4_p8): Likewise.
8497 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8499 * config/aarch64/arm_neon.h (vtbl3_s8): Use __builtin_memcpy
8500 instead of constructing __builtin_aarch64_simd_oi one vector
8502 (vtbl3_u8): Likewise.
8503 (vtbl3_p8): Likewise.
8504 (vtbl4_s8): Likewise.
8505 (vtbl4_u8): Likewise.
8506 (vtbl4_p8): Likewise.
8508 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8510 * config/aarch64/arm_neon.h (vqtbx2_s8): Use __builtin_memcpy
8511 instead of constructing __builtin_aarch64_simd_oi one vector
8513 (vqtbx2_u8): Likewise.
8514 (vqtbx2_p8): Likewise.
8515 (vqtbx2q_s8): Likewise.
8516 (vqtbx2q_u8): Likewise.
8517 (vqtbx2q_p8): Likewise.
8518 (vqtbx3_s8): Use __builtin_memcpy instead of constructing
8519 __builtin_aarch64_simd_ci one vector at a time.
8520 (vqtbx3_u8): Likewise.
8521 (vqtbx3_p8): Likewise.
8522 (vqtbx3q_s8): Likewise.
8523 (vqtbx3q_u8): Likewise.
8524 (vqtbx3q_p8): Likewise.
8525 (vqtbx4_s8): Use __builtin_memcpy instead of constructing
8526 __builtin_aarch64_simd_xi one vector at a time.
8527 (vqtbx4_u8): Likewise.
8528 (vqtbx4_p8): Likewise.
8529 (vqtbx4q_s8): Likewise.
8530 (vqtbx4q_u8): Likewise.
8531 (vqtbx4q_p8): Likewise.
8533 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8535 * config/aarch64/arm_neon.h (vqtbl2_s8): Use __builtin_memcpy
8536 instead of constructing __builtin_aarch64_simd_oi one vector
8538 (vqtbl2_u8): Likewise.
8539 (vqtbl2_p8): Likewise.
8540 (vqtbl2q_s8): Likewise.
8541 (vqtbl2q_u8): Likewise.
8542 (vqtbl2q_p8): Likewise.
8543 (vqtbl3_s8): Use __builtin_memcpy instead of constructing
8544 __builtin_aarch64_simd_ci one vector at a time.
8545 (vqtbl3_u8): Likewise.
8546 (vqtbl3_p8): Likewise.
8547 (vqtbl3q_s8): Likewise.
8548 (vqtbl3q_u8): Likewise.
8549 (vqtbl3q_p8): Likewise.
8550 (vqtbl4_s8): Use __builtin_memcpy instead of constructing
8551 __builtin_aarch64_simd_xi one vector at a time.
8552 (vqtbl4_u8): Likewise.
8553 (vqtbl4_p8): Likewise.
8554 (vqtbl4q_s8): Likewise.
8555 (vqtbl4q_u8): Likewise.
8556 (vqtbl4q_p8): Likewise.
8558 2021-07-23 Haochen Gui <guihaoc@gcc.gnu.org>
8561 * config/rs6000/rs6000.md (cstore<mode>4): Fix wrong fall through.
8563 2021-07-22 Andrew Pinski <apinski@marvell.com>
8565 PR tree-optimization/10153
8566 * tree-tailcall.c (create_tailcall_accumulator):
8567 Don't call fold_convert as the type should be correct already.
8568 (tree_optimize_tail_calls_1): Use build_{one,zero}_cst instead
8569 of integer_{one,zero}_node for the call of create_tailcall_accumulator.
8571 2021-07-22 Aldy Hernandez <aldyh@redhat.com>
8573 * gimple-range-cache.cc (non_null_ref::adjust_range): Replace
8574 varying_p check for null/non-null check.
8576 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
8578 PR tree-optimization/101511
8579 * value-relation.cc (relation_oracle::query_relation): Check if ssa1
8580 is in ssa2's equiv set, and don't trap if so.
8582 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
8584 PR tree-optimization/101497
8585 * gimple-range-fold.cc (fold_using_range::range_of_cond_expr): Check
8588 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
8590 PR tree-optimization/101496
8591 * vr-values.c (simplify_using_ranges::fold_cond): Call range_of_stmt
8592 first, then vrp_visit_cond_Stmt.
8594 2021-07-22 liuhongt <hongtao.liu@intel.com>
8596 * config/i386/i386-expand.c
8597 (ix86_broadcast_from_integer_constant): Rename to ..
8598 (ix86_broadcast_from_constant): .. this, and extend it to
8600 (ix86_expand_vector_move): Extend to float mode.
8601 * config/i386/i386-features.c
8602 (replace_constant_pool_with_broadcast): Remove.
8603 (remove_partial_avx_dependency_gate): Ditto.
8604 (constant_pool_broadcast): Ditto.
8605 (class pass_constant_pool_broadcast): Ditto.
8606 (make_pass_constant_pool_broadcast): Ditto.
8607 (remove_partial_avx_dependency): Adjust gate.
8608 * config/i386/i386-passes.def: Remove pass_constant_pool_broadcast.
8609 * config/i386/i386-protos.h
8610 (make_pass_constant_pool_broadcast): Remove.
8612 2021-07-22 liuhongt <hongtao.liu@intel.com>
8614 * config/i386/constraints.md (Wb): New constraint.
8616 * config/i386/i386.md (*ashlhi3_1): Extend to avx512 mask
8618 (*ashlqi3_1): Ditto.
8619 (*<insn><mode>3_1): Split to ..
8620 (*ashr<mode>3_1): this, ...
8621 (*lshr<mode>3_1): and this, also extend this pattern to avx512
8623 (*<insn><mode>3_1): Split to ..
8624 (*ashr<mode>3_1): this, ...
8625 (*lshrqi3_1): and this, also extend this pattern to avx512
8627 (*lshrhi3_1): And this, also extend this pattern to avx512
8629 * config/i386/sse.md (k<code><mode>): New define_split after
8630 it to convert generic shift pattern to mask shift ones.
8632 2021-07-21 Thomas Schwinge <thomas@codesourcery.com>
8633 Joseph Myers <joseph@codesourcery.com>
8634 Cesar Philippidis <cesar@codesourcery.com>
8636 * tree-core.h (omp_clause_code): Add 'OMP_CLAUSE_NOHOST'.
8637 * tree.c (omp_clause_num_ops, omp_clause_code_name, walk_tree_1):
8639 * tree-pretty-print.c (dump_omp_clause): Likewise.
8640 * omp-general.c (oacc_verify_routine_clauses): Likewise.
8641 * gimplify.c (gimplify_scan_omp_clauses)
8642 (gimplify_adjust_omp_clauses): Likewise.
8643 * tree-nested.c (convert_nonlocal_omp_clauses)
8644 (convert_local_omp_clauses): Likewise.
8645 * omp-low.c (scan_sharing_clauses): Likewise.
8646 * omp-offload.c (execute_oacc_device_lower): Update.
8648 2021-07-21 Martin Sebor <msebor@redhat.com>
8650 * tree-ssa-alias.c (walk_aliased_vdefs_1): Fix typos in a comment.
8652 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8654 * config/rs6000/rs6000-gen-builtins.c (write_init_bif_table):
8657 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8659 * config/rs6000/rs6000-gen-builtins.c (write_fntype): New
8661 (write_fntype_init): New stub function.
8662 (write_init_bif_table): Likewise.
8663 (write_init_ovld_table): New function.
8664 (write_init_file): Implement.
8666 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8668 * config/rs6000/rs6000-gen-builtins.c
8669 (write_autogenerated_header): New function.
8670 (write_decls): Likewise.
8671 (write_extern_fntype): New callback function.
8672 (write_header_file): Implement.
8674 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8676 * config/rs6000/rs6000-gen-builtins.c (write_defines_file):
8679 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8681 * config/rs6000/rs6000-gen-builtins.c (complete_vector_type): New
8683 (complete_base_type): Likewise.
8684 (construct_fntype_id): Likewise.
8685 (parse_bif_entry): Call contruct_fntype_id.
8686 (parse_ovld_entry): Likewise.
8688 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8690 * config/rs6000/rs6000-gen-builtins.c (ovld_stanza): New struct.
8691 (MAXOVLDSTANZAS): New macro.
8692 (ovld_stanzas): New variable.
8693 (curr_ovld_stanza): Likewise.
8694 (MAXOVLDS): New macro.
8695 (ovlddata): New struct.
8696 (ovlds): New variable.
8697 (curr_ovld): Likewise.
8698 (max_ovld_args): Likewise.
8699 (parse_ovld_entry): New function.
8700 (parse_ovld_stanza): Likewise.
8701 (parse_ovld): Implement.
8703 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8705 * config/rs6000/rs6000-gen-builtins.c (parse_bif_attrs):
8708 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8710 * config/rs6000/rs6000-gen-builtins.c (parse_args): New function.
8711 (parse_prototype): Implement.
8713 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8715 * config/rs6000/rs6000-gen-builtins.c (bif_stanza): New enum.
8716 (curr_bif_stanza): New variable.
8717 (stanza_entry): New struct.
8718 (stanza_map): New initialized variable.
8719 (enable_string): Likewise.
8720 (fnkinds): New enum.
8721 (typelist): New struct.
8722 (attrinfo): Likewise.
8723 (MAXRESTROPNDS): New macro.
8724 (prototype): New struct.
8725 (MAXBIFS): New macro.
8726 (bifdata): New struct.
8727 (bifs): New variable.
8728 (curr_bif): Likewise.
8729 (bif_order): Likewise.
8730 (bif_index): Likewise.
8731 (fatal): New function.
8732 (stanza_name_to_stanza): Likewise.
8733 (parse_bif_attrs): New stub function.
8734 (parse_prototype): Likewise.
8735 (parse_bif_entry): New function.
8736 (parse_bif_stanza): Likewise.
8737 (parse_bif): Implement.
8738 (set_bif_order): New function.
8739 (create_bif_order): Implement.
8741 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8743 * config/rs6000/rs6000-gen-builtins.c (rbtree.h): New #include.
8744 (num_bifs): New variable.
8745 (num_ovld_stanzas): Likewise.
8746 (num_ovlds): Likewise.
8747 (parse_codes): New enum.
8748 (bif_rbt): New variable.
8749 (ovld_rbt): Likewise.
8750 (fntype_rbt): Likewise.
8751 (bifo_rbt): Likewise.
8752 (parse_bif): New stub function.
8753 (create_bif_order): Likewise.
8754 (parse_ovld): Likewise.
8755 (write_header_file): Likewise.
8756 (write_init_file): Likewise.
8757 (write_defines_file): Likewise.
8758 (delete_output_files): New function.
8761 2021-07-21 H.J. Lu <hjl.tools@gmail.com>
8764 * config/i386/i386-builtin.def: Remove OPTION_MASK_ISA_SSE4_2
8765 from CRC32 _builtin functions.
8767 2021-07-21 Sebastian Huber <sebastian.huber@embedded-brains.de>
8769 * coverage.c (build_gcov_info_var_registration): Mark the object placed
8770 in the linker set as referenced so that it does not get optimized away.
8772 2021-07-21 Kito Cheng <kito.cheng@sifive.com>
8775 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
8777 * config.gcc (riscv*-*-*): Detect which python is available.
8779 2021-07-21 Jakub Jelinek <jakub@redhat.com>
8781 PR middle-end/101535
8782 * gimplify.c (omp_check_private): Properly skip ORT_TARGET_DATA
8783 contexts in which decl isn't privatized and for ORT_TARGET return
8784 false if decl is mapped.
8786 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
8788 * gimple-loop-jam.c: Include tree-ssa-sccvn.h.
8789 (tree_loop_unroll_and_jam): Run value-numbering on a loop that
8790 has been successfully unrolled.
8792 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
8794 * tree-ssa-loop-manip.c (determine_exit_conditions): Return a null
8795 exit condition if no tail loop is needed, and if the original exit
8796 condition should therefore be kept as-is.
8797 (tree_transform_and_unroll_loop): Handle that case here too.
8799 2021-07-21 Kewen Lin <linkw@linux.ibm.com>
8801 * tree-data-ref.c (free_dependence_relations): Adjust to pass vec
8803 (free_data_refs): Likewise.
8804 * tree-data-ref.h (free_dependence_relations): Likewise.
8805 (free_data_refs): Likewise.
8806 * tree-predcom.c (struct chain): Use auto_vec instead of vec for
8808 (struct component): Likewise.
8809 (pcom_worker::pcom_worker): Adjust for auto_vec and renaming changes.
8810 (pcom_worker::~pcom_worker): Likewise.
8811 (pcom_worker::release_chain): Adjust as auto_vec changes.
8812 (pcom_worker::loop): Rename to ...
8813 (pcom_worker::m_loop): ... this.
8814 (pcom_worker::datarefs): Rename to ...
8815 (pcom_worker::m_datarefs): ... this. Use auto_vec instead of vec.
8816 (pcom_worker::dependences): Rename to ...
8817 (pcom_worker::m_dependences): ... this. Use auto_vec instead of vec.
8818 (pcom_worker::chains): Rename to ...
8819 (pcom_worker::m_chains): ... this. Use auto_vec instead of vec.
8820 (pcom_worker::looparound_phis): Rename to ...
8821 (pcom_worker::m_looparound_phis): ... this. Use auto_vec instead of
8823 (pcom_worker::cache): Rename to ...
8824 (pcom_worker::m_cache): ... this. Use auto_vec instead of vec.
8825 (pcom_worker::release_chain): Adjust for auto_vec changes.
8826 (pcom_worker::release_chains): Adjust for auto_vec and renaming
8828 (release_component): Remove.
8829 (release_components): Adjust for release_component removal.
8830 (component_of): Adjust to use vec.
8831 (merge_comps): Likewise.
8832 (pcom_worker::aff_combination_dr_offset): Adjust for renaming changes.
8833 (pcom_worker::determine_offset): Likewise.
8834 (class comp_ptrs): Remove.
8835 (pcom_worker::split_data_refs_to_components): Adjust for renaming
8836 changes, for comp_ptrs removal with auto_vec.
8837 (pcom_worker::suitable_component_p): Adjust for renaming changes.
8838 (pcom_worker::filter_suitable_components): Adjust for release_component
8840 (pcom_worker::valid_initializer_p): Adjust for renaming changes.
8841 (pcom_worker::find_looparound_phi): Likewise.
8842 (pcom_worker::add_looparound_copies): Likewise.
8843 (pcom_worker::determine_roots_comp): Likewise.
8844 (pcom_worker::single_nonlooparound_use): Likewise.
8845 (pcom_worker::execute_pred_commoning_chain): Likewise.
8846 (pcom_worker::execute_pred_commoning): Likewise.
8847 (pcom_worker::try_combine_chains): Likewise.
8848 (pcom_worker::prepare_initializers_chain): Likewise.
8849 (pcom_worker::prepare_initializers): Likewise.
8850 (pcom_worker::prepare_finalizers_chain): Likewise.
8851 (pcom_worker::prepare_finalizers): Likewise.
8852 (pcom_worker::tree_predictive_commoning_loop): Likewise.
8854 2021-07-20 Martin Sebor <msebor@redhat.com>
8856 PR middle-end/101397
8857 * builtins.c (gimple_call_return_array): Add argument. Correct
8858 offsets for memchr, mempcpy, stpcpy, and stpncpy.
8859 (compute_objsize_r): Adjust offset computation for argument returning
8862 2021-07-20 Martin Sebor <msebor@redhat.com>
8864 PR middle-end/101300
8865 * tree-ssa-uninit.c (check_defs): Handle UBSAN built-ins.
8867 2021-07-20 Jeff Law <jlaw@localhost.localdomain>
8869 * function.c (assign_parm_setup_block): Use adjust_address instead
8870 of change_address to preserve MEM_EXPR and friends.
8872 2021-07-20 Martin Sebor <msebor@redhat.com>
8874 * cfgloop.h (single_likely_exit): Adjust by-value argument to
8876 * cfgloopanal.c (single_likely_exit): Same.
8877 * cgraph.h (struct cgraph_node): Same.
8878 * cgraphclones.c (cgraph_node::create_virtual_clone): Same.
8879 * genautomata.c (merge_states): Same.
8880 * genextract.c (VEC_char_to_string): Same.
8881 * genmatch.c (dt_node::gen_kids_1): Same.
8882 (walk_captures): Adjust by-value argument to by-reference.
8883 * gimple-ssa-store-merging.c (check_no_overlap): Adjust by-value argument
8884 to by-const-reference.
8885 * gimple.c (gimple_build_call_vec): Same.
8886 (gimple_build_call_internal_vec): Same.
8887 (gimple_build_switch): Same.
8888 (sort_case_labels): Same.
8889 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
8891 * gimple.h (gimple_build_call_vec): Adjust by-value argument to
8893 (gimple_build_call_internal_vec): Same.
8894 (gimple_build_switch): Same.
8895 (sort_case_labels): Same.
8896 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
8898 * haifa-sched.c (calc_priorities): Adjust by-value argument to
8900 (sched_init_luids): Same.
8901 (haifa_init_h_i_d): Same.
8902 * ipa-cp.c (ipa_get_indirect_edge_target_1): Same.
8903 (adjust_callers_for_value_intersection): Adjust by-value argument to
8905 (find_more_scalar_values_for_callers_subset): Adjust by-value argument to
8907 (find_more_contexts_for_caller_subset): Same.
8908 (find_aggregate_values_for_callers_subset): Same.
8909 (copy_useful_known_contexts): Same.
8910 * ipa-fnsummary.c (remap_edge_summaries): Same.
8911 (remap_freqcounting_predicate): Same.
8912 * ipa-inline.c (add_new_edges_to_heap): Adjust by-value argument to
8914 * ipa-predicate.c (predicate::remap_after_inlining): Adjust by-value argument
8915 to by-const-reference.
8916 * ipa-predicate.h (predicate::remap_after_inlining): Same.
8917 * ipa-prop.c (ipa_find_agg_cst_for_param): Same.
8918 * ipa-prop.h (ipa_find_agg_cst_for_param): Same.
8919 * ira-build.c (ira_loop_tree_body_rev_postorder): Same.
8920 * read-rtl.c (add_overload_instance): Same.
8921 * rtl.h (native_decode_rtx): Same.
8922 (native_decode_vector_rtx): Same.
8923 * sched-int.h (sched_init_luids): Same.
8924 (haifa_init_h_i_d): Same.
8925 * simplify-rtx.c (native_decode_vector_rtx): Same.
8926 (native_decode_rtx): Same.
8927 * tree-call-cdce.c (gen_shrink_wrap_conditions): Same.
8928 (shrink_wrap_one_built_in_call_with_conds): Same.
8929 (shrink_wrap_conditional_dead_built_in_calls): Same.
8930 * tree-data-ref.c (create_runtime_alias_checks): Same.
8931 (compute_all_dependences): Same.
8932 * tree-data-ref.h (compute_all_dependences): Same.
8933 (create_runtime_alias_checks): Same.
8934 (index_in_loop_nest): Same.
8935 * tree-if-conv.c (mask_exists): Same.
8936 * tree-loop-distribution.c (class loop_distribution): Same.
8937 (loop_distribution::create_rdg_vertices): Same.
8938 (dump_rdg_partitions): Same.
8939 (debug_rdg_partitions): Same.
8940 (partition_contains_all_rw): Same.
8941 (loop_distribution::distribute_loop): Same.
8942 * tree-parloops.c (oacc_entry_exit_ok_1): Same.
8943 (oacc_entry_exit_single_gang): Same.
8944 * tree-ssa-loop-im.c (hoist_memory_references): Same.
8945 (loop_suitable_for_sm): Same.
8946 * tree-ssa-loop-niter.c (bound_index): Same.
8947 * tree-ssa-reassoc.c (update_ops): Same.
8948 (swap_ops_for_binary_stmt): Same.
8949 (rewrite_expr_tree): Same.
8950 (rewrite_expr_tree_parallel): Same.
8951 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Same.
8952 * tree-ssa-sccvn.h (ao_ref_init_from_vn_reference): Same.
8953 * tree-ssa-structalias.c (process_all_all_constraints): Same.
8954 (make_constraints_to): Same.
8955 (handle_lhs_call): Same.
8956 (find_func_aliases_for_builtin_call): Same.
8957 (sort_fieldstack): Same.
8958 (check_for_overlaps): Same.
8959 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Same.
8960 (vect_create_cond_for_unequal_addrs): Same.
8961 (vect_create_cond_for_lower_bounds): Same.
8962 (vect_create_cond_for_alias_checks): Same.
8963 * tree-vect-slp-patterns.c (vect_validate_multiplication): Same.
8964 * tree-vect-slp.c (vect_analyze_slp_instance): Same.
8965 (vect_make_slp_decision): Same.
8966 (vect_slp_bbs): Same.
8967 (duplicate_and_interleave): Same.
8968 (vect_transform_slp_perm_load): Same.
8969 (vect_schedule_slp): Same.
8970 * tree-vectorizer.h (vect_transform_slp_perm_load): Same.
8971 (vect_schedule_slp): Same.
8972 (duplicate_and_interleave): Same.
8973 * tree.c (build_vector_from_ctor): Same.
8974 (build_vector): Same.
8975 (check_vector_cst): Same.
8976 (check_vector_cst_duplicate): Same.
8977 (check_vector_cst_fill): Same.
8978 (check_vector_cst_stepped): Same.
8979 * tree.h (build_vector_from_ctor): Same.
8981 2021-07-20 Jakub Jelinek <jakub@redhat.com>
8984 * config/rs6000/rs6000-protos.h (easy_altivec_constant): Change return
8985 type from bool to int.
8986 * config/rs6000/rs6000.c (vspltis_constant): Fix up handling the
8987 EASY_VECTOR_MSB case if either step or copies is not 1.
8988 (vspltis_shifted): Fix comment typo.
8989 (easy_altivec_constant): Change return type from bool to int, instead
8990 of returning true return byte size of the element mode that should be
8991 used to synthetize the constant.
8992 * config/rs6000/predicates.md (easy_vector_constant_msb): Require
8993 that vspltis_shifted is 0, handle the case where easy_altivec_constant
8994 assumes using different vector mode from CONST_VECTOR's mode.
8995 * config/rs6000/altivec.md (easy_vector_constant_msb splitter): Use
8996 easy_altivec_constant to determine mode in which -1 >> -1 should be
8997 performed, use rs6000_expand_vector_init instead of gen_vec_initv4sisi.
8999 2021-07-20 Richard Biener <rguenther@suse.de>
9002 * dwarf2out.h (dwarf_file_data): Add key member.
9003 * dwarf2out.c (dwarf_file_hasher::equal): Compare key.
9004 (dwarf_file_hasher::hash): Hash key.
9005 (lookup_filename): Remap the filename and store it in the
9006 filename member of dwarf_file_data when creating a new
9008 (file_name_acquire): Do not remap the filename again.
9009 (maybe_emit_file): Likewise.
9011 2021-07-20 Jonathan Wright <jonathan.wright@arm.com>
9013 * config/aarch64/aarch64-simd-builtins.def: Use two variant
9014 generators for all TBL/TBX intrinsics and rename to
9015 consistent forms: qtbl[1234] or qtbx[1234].
9016 * config/aarch64/aarch64-simd.md (aarch64_tbl1<mode>):
9018 (aarch64_qtbl1<mode>): This.
9019 (aarch64_tbx1<mode>): Rename to...
9020 (aarch64_qtbx1<mode>): This.
9021 (aarch64_tbl2v16qi): Delete.
9022 (aarch64_tbl3<mode>): Rename to...
9023 (aarch64_qtbl2<mode>): This.
9024 (aarch64_tbx4<mode>): Rename to...
9025 (aarch64_qtbx2<mode>): This.
9026 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_1): Use
9027 renamed qtbl1 and qtbl2 RTL patterns.
9028 * config/aarch64/arm_neon.h (vqtbl1_p8): Use renamed qtbl1
9030 (vqtbl1_s8): Likewise.
9031 (vqtbl1_u8): Likewise.
9032 (vqtbl1q_p8): Likewise.
9033 (vqtbl1q_s8): Likewise.
9034 (vqtbl1q_u8): Likewise.
9035 (vqtbx1_s8): Use renamed qtbx1 RTL pattern.
9036 (vqtbx1_u8): Likewise.
9037 (vqtbx1_p8): Likewise.
9038 (vqtbx1q_s8): Likewise.
9039 (vqtbx1q_u8): Likewise.
9040 (vqtbx1q_p8): Likewise.
9041 (vtbl1_s8): Use renamed qtbl1 RTL pattern.
9042 (vtbl1_u8): Likewise.
9043 (vtbl1_p8): Likewise.
9044 (vtbl2_s8): Likewise
9045 (vtbl2_u8): Likewise.
9046 (vtbl2_p8): Likewise.
9047 (vtbl3_s8): Use renamed qtbl2 RTL pattern.
9048 (vtbl3_u8): Likewise.
9049 (vtbl3_p8): Likewise.
9050 (vtbl4_s8): Likewise.
9051 (vtbl4_u8): Likewise.
9052 (vtbl4_p8): Likewise.
9053 (vtbx2_s8): Use renamed qtbx2 RTL pattern.
9054 (vtbx2_u8): Likewise.
9055 (vtbx2_p8): Likewise.
9056 (vqtbl2_s8): Use renamed qtbl2 RTL pattern.
9057 (vqtbl2_u8): Likewise.
9058 (vqtbl2_p8): Likewise.
9059 (vqtbl2q_s8): Likewise.
9060 (vqtbl2q_u8): Likewise.
9061 (vqtbl2q_p8): Likewise.
9062 (vqtbx2_s8): Use renamed qtbx2 RTL pattern.
9063 (vqtbx2_u8): Likewise.
9064 (vqtbx2_p8): Likewise.
9065 (vqtbx2q_s8): Likewise.
9066 (vqtbx2q_u8): Likewise.
9067 (vqtbx2q_p8): Likewise.
9068 (vtbx4_s8): Likewise.
9069 (vtbx4_u8): Likewise.
9070 (vtbx4_p8): Likewise.
9072 2021-07-20 Uroš Bizjak <ubizjak@gmail.com>
9075 * config/i386/sync.md (define_peephole2 atomic_storedi_fpu):
9077 (define_peephole2 atomic_loaddi_fpu): Ditto.
9079 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
9081 * config.gcc (riscv*-*-*): Detect which python is available.
9083 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
9085 * config/rs6000/vsx.md (mulhs_<mode>): Rename to...
9086 (smul<mode>3_highpart): ... this.
9087 (mulhu_<mode>): Rename to...
9088 (umul<mode>3_highpart): ... this.
9089 * config/rs6000/rs6000-builtin.def (MULHS_V2DI, MULHS_V4SI,
9090 MULHU_V2DI, MULHU_V4SI): Adjust.
9092 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
9094 PR tree-optimization/100696
9095 * internal-fn.c (first_commutative_argument): Add info for IFN_MULH.
9096 * internal-fn.def (IFN_MULH): New internal function.
9097 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Add support to
9098 recog normal multiply highpart as IFN_MULH.
9099 * config/i386/i386.c (ix86_add_stmt_cost): Adjust for combined
9102 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
9104 * config/elfos.h (CTF_DEBUGGING_INFO): New definition.
9105 (BTF_DEBUGGING_INFO): Likewise.
9106 * doc/tm.texi.in: Document the new macros.
9107 * doc/tm.texi: Regenerated.
9108 * toplev.c: Guard initialization of debug hooks.
9110 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
9112 * flags.h (ctf_debuginfo_p): New function declaration.
9113 * opts.c (ctf_debuginfo_p): New function definition.
9115 2021-07-19 Andrew Stubbs <ams@codesourcery.com>
9118 * config/gcn/gcn-hsa.h (DRIVER_SELF_SPECS): New.
9119 (ASM_SPEC): Set -mattr for xnack and sram-ecc.
9120 * config/gcn/gcn-opts.h (enum sram_ecc_type): New.
9121 * config/gcn/gcn-valu.md: Add a warning comment.
9122 * config/gcn/gcn.c (gcn_option_override): Add "sorry" for -mxnack.
9123 (output_file_start): Add xnack and sram-ecc state to ".amdgcn_target".
9124 * config/gcn/gcn.md: Add a warning comment.
9125 * config/gcn/gcn.opt: Add -mxnack and -msram-ecc.
9126 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): Remove
9128 (EF_AMDGPU_XNACK): New.
9129 (EF_AMDGPU_SRAM_ECC): New.
9131 (copy_early_debug_info): Use elf_flags.
9132 (main): Handle -mxnack and -msram-ecc options.
9133 * doc/invoke.texi: Document -mxnack and -msram-ecc.
9135 2021-07-19 Andrew Pinski <apinski@marvell.com>
9138 * config/aarch64/aarch64.md (csneg3_uxtw_insn): Rename to ...
9139 (*cs<neg_not_cs>3_uxtw_insn4): and extend to NEG_NOT.
9141 2021-07-19 Richard Biener <rguenther@suse.de>
9143 PR tree-optimization/101505
9144 * tree-vect-patterns.c (vect_determine_precisions): Walk
9145 PHIs also for loop vectorization.
9147 2021-07-19 Richard Biener <rguenther@suse.de>
9149 * gimple.h (gimple_expr_type): Remove.
9150 * doc/gimple.texi: Remove gimple_expr_type documentation.
9152 2021-07-19 Richard Biener <rguenther@suse.de>
9154 * tree-ssa-sccvn.c (vn_reference_eq): Handle NULL vr->type.
9155 (ao_ref_init_from_vn_reference): Likewise.
9156 (fully_constant_reference): Likewise.
9157 (vn_reference_lookup_call): Do not set vr->type to random
9159 * tree-ssa-pre.c (compute_avail): Do not try to PRE calls
9161 * tree-vect-generic.c (expand_vector_piecewise): Pass in
9162 whether we expanded parallel.
9163 (expand_vector_parallel): Adjust.
9164 (expand_vector_addition): Likewise.
9165 (expand_vector_comparison): Likewise.
9166 (expand_vector_operation): Likewise.
9167 (expand_vector_scalar_condition): Likewise.
9168 (expand_vector_conversion): Likewise.
9170 2021-07-19 Richard Biener <rguenther@suse.de>
9172 * tree-vrp.c (register_edge_assert_for_2): Use the
9174 (vrp_folder::fold_predicate_in): Likewise.
9175 * vr-values.c (gimple_assign_nonzero_p): Likewise.
9176 (vr_values::extract_range_from_comparison): Likewise.
9177 (vr_values::extract_range_from_ubsan_builtin): Use the
9178 type of the first operand.
9179 (vr_values::extract_range_basic): Push down type
9180 computation, use the appropriate LHS.
9181 (vr_values::extract_range_from_assignment): Use the
9184 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
9187 * common/config/i386/i386-common.c (ix86_handle_option): For
9188 -mgeneral-regs-only, enable the GPR only instructions which are
9189 enabled implicitly by SSE ISAs unless they have been disabled
9192 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
9195 * config/i386/i386.c (ix86_check_avx_upper_stores): Moved before
9196 ix86_avx_u128_mode_needed.
9197 (ix86_avx_u128_mode_needed): Return AVX_U128_DIRTY if callee
9198 returns AVX register.
9200 2021-07-17 Jan Hubicka <hubicka@ucw.cz>
9202 * tree-ssa-structalias.c (handle_rhs_call): Support EAF_NOT_RETURNED.
9203 (handle_const_call): Liekise
9204 (handle_pure_call): Liekise
9206 2021-07-17 Andrew MacLeod <amacleod@redhat.com>
9208 PR tree-optimization/96542
9209 * range-op.cc (range_operator::wi_fold_in_parts): New.
9210 (range_operator::fold_range): Call wi_fold_in_parts.
9211 (operator_lshift::wi_fold): Fix broken lshift by [0,0].
9212 * range-op.h (wi_fold_in_parts): Add prototype.
9214 2021-07-16 David Malcolm <dmalcolm@redhat.com>
9216 * doc/analyzer.texi: Add __analyzer_dump_state.
9218 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9220 * config/rs6000/rbtree.c: New file.
9221 * config/rs6000/rbtree.h: New file.
9223 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9225 * config/rs6000/rs6000-gen-builtins.c (restriction): New enum.
9226 (typeinfo): Add restr field.
9227 (match_bracketed_pair): New function.
9228 (match_const_restriction): Implement.
9230 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9232 * config/rs6000/rs6000-gen-builtins.c (match_basetype): Implement.
9234 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9236 * config/rs6000/rs6000-gen-builtins.c (void_status): New enum.
9237 (basetype): Likewise.
9238 (typeinfo): Likewise.
9239 (handle_pointer): New function.
9240 (match_basetype): New stub function.
9241 (match_const_restriction): Likewise.
9242 (match_type): New function.
9244 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9246 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace): New
9248 (advance_line): Likewise.
9249 (safe_inc_pos): Likewise.
9250 (match_identifier): Likewise.
9251 (match_integer): Likewise.
9252 (match_to_right_bracket): Likewise.
9254 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9256 * config/rs6000/rs6000-gen-builtins.c (bif_file): New variable.
9257 (ovld_file): Likewise.
9258 (header_file): Likewise.
9259 (init_file): Likewise.
9260 (defines_file): Likewise.
9261 (pgm_path): Likewise.
9262 (bif_path): Likewise.
9263 (ovld_path): Likewise.
9264 (header_path): Likewise.
9265 (init_path): Likewise.
9266 (defines_path): Likewise.
9267 (LINELEN): New macro.
9268 (linebuf): New variable.
9272 (bif_diag): New function.
9273 (ovld_diag): Likewise.
9275 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9277 * config/rs6000/rs6000-builtin-new.def: New.
9278 * config/rs6000/rs6000-overload.def: New.
9280 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9282 * config/rs6000/rs6000-gen-builtins.c: New.
9284 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9286 * Makefile.in (EXTRA_GTYPE_DEPS): New variable.
9287 (s-gtype): Depend on EXTRA_GTYPE_DEPS.
9288 * gengtype-state.c (state_writer::write_state_file_list): Add a
9289 parameter to the fileslist expression for the number of build
9291 (read_state_files_list): Detect build headers and strip the
9292 initial "./" or ".\" from their names.
9293 * gengtype.c (build_headers): New global variable.
9294 (num_build_headers): Likewise.
9295 (open_base_files): Emit #include for each build header.
9296 (main): Detect and count build headers.
9297 * gengtype.h (build_headers): New extern variable.
9298 (num_build_headers): Likewise.
9300 2021-07-16 Richard Biener <rguenther@suse.de>
9302 * gimple-ssa-store-merging.c (verify_symbolic_number_p): Use
9303 the type of the LHS.
9304 (find_bswap_or_nop_1): Likewise.
9305 (find_bswap_or_nop): Likewise.
9306 * tree-vectorizer.h (vect_get_smallest_scalar_type): Adjust
9308 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
9309 Remove unused parameters, pass in the scalar type. Fix
9310 internal store function handling.
9311 * tree-vect-stmts.c (vect_analyze_stmt): Remove assert.
9312 (vect_get_vector_types_for_stmt): Move down check for
9313 existing vector stmt after we've determined a scalar type.
9314 Pass down the used scalar type to vect_get_smallest_scalar_type.
9315 * tree-vect-generic.c (expand_vector_condition): Use
9316 the type of the LHS.
9317 (expand_vector_scalar_condition): Likewise.
9318 (expand_vector_operations_1): Likewise.
9319 * tree-vect-patterns.c (vect_widened_op_tree): Likewise.
9320 (vect_recog_dot_prod_pattern): Likewise.
9321 (vect_recog_sad_pattern): Likewise.
9322 (vect_recog_widen_op_pattern): Likewise.
9323 (vect_recog_widen_sum_pattern): Likewise.
9324 (vect_recog_mixed_size_cond_pattern): Likewise.
9326 2021-07-16 Jan Hubicka <hubicka@ucw.cz>
9328 * ipa-modref.c (struct escape_entry): Use eaf_fleags_t.
9329 (dump_eaf_flags): Dump EAF_NOT_RETURNED
9330 (eaf_flags_useful_p): Use eaf_fleags_t; handle const functions
9331 and EAF_NOT_RETURNED.
9332 (modref_summary::useful_p): Likewise.
9333 (modref_summary_lto::useful_p): Likewise.
9334 (struct) modref_summary_lto: Use eaf_fleags_t.
9335 (deref_flags): Handle EAF_NOT_RETURNED.
9336 (struct escape_point): Use min_flags.
9337 (modref_lattice::init): Add EAF_NOT_RETURNED.
9338 (merge_call_lhs_flags): Ignore EAF_NOT_RETURNED functions
9339 (analyze_ssa_name_flags): Clear EAF_NOT_RETURNED on return;
9341 (analyze_parms): Also analyze const functions; update conition on
9343 (modref_write): Update streaming.
9344 (read_section): Update streaming.
9345 (remap_arg_flags): Use eaf_flags_t.
9346 (modref_merge_call_site_flags): Hanlde EAF_NOT_RETURNED.
9347 * ipa-modref.h: (eaf_flags_t): New typedef.
9348 (struct modref_summary): Use eaf_flags_t.
9349 * tree-core.h (EAF_NOT_RETURNED): New constant.
9351 2021-07-16 Richard Biener <rguenther@suse.de>
9353 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Use
9354 the type of the LHS.
9355 (gimple_assign_nonnegative_warnv_p): Likewise.
9356 (gimple_call_nonnegative_warnv_p): Likewise. Return false
9357 if the call has no LHS.
9358 * gimple.c (gimple_could_trap_p_1): Use the type of the LHS.
9359 * tree-eh.c (stmt_could_throw_1_p): Likewise.
9360 * tree-inline.c (insert_init_stmt): Likewise.
9361 * tree-ssa-loop-niter.c (get_val_for): Likewise.
9362 * tree-outof-ssa.c (ssa_is_replaceable_p): Use the type of
9364 * tree-ssa-sccvn.c (init_vn_nary_op_from_stmt): Take a
9365 gassign *. Use the type of the lhs.
9366 (vn_nary_op_lookup_stmt): Adjust.
9367 (vn_nary_op_insert_stmt): Likewise.
9369 2021-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
9371 * config/s390/predicates.md (bras_sym_operand): Accept all
9372 functions in 64-bit mode, use UNSPEC_PLT31.
9373 (larl_operand): Use UNSPEC_PLT31.
9374 * config/s390/s390.c (s390_loadrelative_operand_p): Likewise.
9375 (legitimize_pic_address): Likewise.
9376 (s390_emit_tls_call_insn): Mark __tls_get_offset as function,
9378 (s390_delegitimize_address): Use UNSPEC_PLT31.
9379 (s390_output_addr_const_extra): Likewise.
9380 (print_operand): Add @PLT to TLS calls, handle %K.
9381 (s390_function_profiler): Mark __fentry__/_mcount as function,
9382 use %K, use UNSPEC_PLT31.
9383 (s390_output_mi_thunk): Use only UNSPEC_GOT, use %K.
9384 (s390_emit_call): Use UNSPEC_PLT31.
9385 (s390_emit_tpf_eh_return): Mark __tpf_eh_return as function.
9386 * config/s390/s390.md (UNSPEC_PLT31): Rename from UNSPEC_PLT.
9387 (*movdi_64): Use %K.
9388 (reload_base_64): Likewise.
9389 (*sibcall_brc): Likewise.
9390 (*sibcall_brcl): Likewise.
9391 (*sibcall_value_brc): Likewise.
9392 (*sibcall_value_brcl): Likewise.
9395 (*bras_r): Likewise.
9396 (*brasl_r): Likewise.
9397 (*bras_tls): Likewise.
9398 (*brasl_tls): Likewise.
9399 (main_base_64): Likewise.
9400 (reload_base_64): Likewise.
9401 (@split_stack_call<mode>): Likewise.
9403 2021-07-16 Richard Biener <rguenther@suse.de>
9405 PR tree-optimization/101467
9406 * tree-vect-stmts.c (vect_gen_while): Properly guard
9407 make_temp_ssa_name usage.
9409 2021-07-16 Cooper Qu <cooper.qu@linux.alibaba.com>
9411 * config.gcc: Don't use forked print-sysroot-suffix.sh and
9412 t-sysroot-suffix for C-SKY.
9413 * config/csky/print-sysroot-suffix.sh: Delete.
9414 * config/csky/t-csky-linux: Delete.
9415 * config/csky/t-sysroot-suffix: Define MULTILIB_DIRNAMES
9416 instead of CSKY_MULTILIB_DIRNAMES.
9418 2021-07-16 Richard Biener <rguenther@suse.de>
9420 * tree-vect-loop.c (vect_transform_cycle_phi): Correct sign
9421 conversion issues with the partial reduction of the reused
9424 2021-07-16 Richard Biener <rguenther@suse.de>
9426 * config/i386/i386-options.c (ix86_option_override_internal): Set
9427 param_vect_partial_vector_usage to zero if not set.
9429 2021-07-15 Uroš Bizjak <ubizjak@gmail.com>
9432 * config/i386/i386.h (VALID_SSE_REG_MODE): Add TDmode.
9433 (VALID_INT_MODE_P): Add SDmode and DDmode.
9434 Add TDmode for TARGET_64BIT.
9435 (VALID_DFP_MODE_P): Remove.
9436 * config/i386/i386.c (ix86_hard_regno_mode_ok):
9437 Do not use VALID_DFP_MODE_P.
9439 2021-07-15 Andrew MacLeod <amacleod@redhat.com>
9441 * gimple-range-fold.cc (adjust_pointer_diff_expr): Use
9443 (fold_using_range::fold_stmt): Ditto.
9444 (fold_using_range::range_of_range_op): Ditto.
9445 (fold_using_range::range_of_phi): Ditto.
9446 (fold_using_range::range_of_call): Ditto.
9447 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
9448 (fold_using_range::range_of_builtin_call): Ditto.
9449 (fold_using_range::range_of_cond_expr): Ditto.
9450 * gimple-range-fold.h (gimple_range_type): New.
9452 2021-07-15 Martin Sebor <msebor@redhat.com>
9455 * tree-ssa-strlen.c (handle_assign): New function.
9456 (maybe_warn_overflow): Add argument.
9457 (nonzero_bytes_for_type): New function.
9458 (count_nonzero_bytes): Handle more tree types. Call
9459 nonzero_bytes_for_tye.
9460 (count_nonzero_bytes): Handle types.
9461 (handle_store): Handle stores from function calls.
9462 (strlen_check_and_optimize_call): Move code to handle_assign. Call
9463 it for assignments from function calls.
9465 2021-07-15 David Malcolm <dmalcolm@redhat.com>
9470 * doc/invoke.texi: Add -Wanalyzer-use-of-uninitialized-value.
9472 2021-07-15 David Malcolm <dmalcolm@redhat.com>
9474 * doc/invoke.texi (-fdump-analyzer-exploded-paths): New.
9476 2021-07-15 Martin Sebor <msebor@redhat.com>
9480 * fold-const.c (operand_compare::operand_equal_p): Handle OEP_DECL_NAME.
9481 (operand_compare::verify_hash_value): Same.
9482 * tree-core.h (OEP_DECL_NAME): New.
9484 2021-07-15 Martin Jambor <mjambor@suse.cz>
9486 * profile-count.h (profile_count::value): Change the return type to
9488 * gimple-pretty-print.c (dump_gimple_bb_header): Adjust print
9490 * tree-cfg.c (dump_function_to_file): Likewise.
9492 2021-07-15 Bill Schmidt <wschmidt@linux.ibm.com>
9495 * config/rs6000/rs6000-p8swap.c (has_part_mult): New.
9496 (rs6000_analyze_swaps): Insns containing a subreg of a mult are
9499 2021-07-15 Richard Biener <rguenther@suse.de>
9501 * tree-vectorizer.h (vect_gen_while): Match up with
9503 * tree-vect-stmts.c (vect_gen_while): Adjust API to that
9504 of vect_gen_while_not.
9505 (vect_gen_while_not): Adjust.
9506 * tree-vect-loop-manip.c (vect_set_loop_controls_directly): Likewise.
9508 2021-07-15 Aldy Hernandez <aldyh@redhat.com>
9510 * gimple-range-cache.cc (non_null_ref::adjust_range): New.
9511 (ranger_cache::range_of_def): Call adjust_range.
9512 (ranger_cache::entry_range): Same.
9513 * gimple-range-cache.h (non_null_ref::adjust_range): New.
9514 * gimple-range.cc (gimple_ranger::range_of_expr): Call
9516 (gimple_ranger::range_on_entry): Same.
9518 2021-07-15 Tamar Christina <tamar.christina@arm.com>
9521 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9523 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
9525 2021-07-15 Tamar Christina <tamar.christina@arm.com>
9528 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9530 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
9531 (sdot_prod, udot_prod): ...These.
9532 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
9533 (aarch64_<sur>dot<vsi2qi>): Rename to...
9534 (<sur>dot_prod<vsi2qi>): ...This.
9535 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
9538 2021-07-15 Jakub Jelinek <jakub@redhat.com>
9540 PR middle-end/101437
9541 * gimplify.c (gimplify_expr): Throw away volatile reads from empty
9542 types even if they have non-BLKmode TYPE_MODE.
9544 2021-07-15 Richard Biener <rguenther@suse.de>
9547 * gcc.c (process_command): Process -gtoggle like process_options
9548 would after parsing options.
9550 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
9552 * cfgexpand.c (expand_asm_loc): Adjust.
9553 (expand_asm_stmt): Likewise.
9554 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Likewise.
9555 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
9556 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
9557 * config/avr/avr.c (avr_md_asm_adjust): Likewise.
9558 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
9559 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
9560 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
9561 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
9562 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
9563 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
9564 * config/s390/s390.c (s390_md_asm_adjust): Likewise.
9565 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
9566 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
9567 * doc/tm.texi: Regenerate.
9568 * target.def: Add location argument to md_asm_adjust.
9570 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
9572 * tree-diagnostic.c (diagnostic_report_current_function): Use the
9573 diagnostic's location, not input_location.
9575 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
9577 * cfgexpand.c (tree_conflicts_with_clobbers_p): Pass location to
9579 (expand_asm_stmt): Likewise.
9581 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
9583 * config/rs6000/rs6000.c (adjacent_mem_locations): Return the lower
9584 addressed memory rtx, if any.
9585 (rs6000_split_multireg_move): Fix code formatting.
9586 Handle MMA build built-ins with operands in adjacent memory locations.
9588 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
9590 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Move to later
9593 2021-07-14 Jason Merrill <jason@redhat.com>
9595 * sel-sched-ir.h (get_all_loop_exits): Use auto_vec.
9597 2021-07-14 Jason Merrill <jason@redhat.com>
9599 * doc/invoke.texi: -fdelete-dead-exceptions is on by default for
9602 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9604 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
9605 Remove erroneous line.
9607 2021-07-14 Andrew MacLeod <amacleod@redhat.com>
9609 * params.opt (param_evrp_mode): Change default.
9611 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9613 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
9614 (sdot_prod, udot_prod): ...These.
9615 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
9616 (aarch64_<sur>dot<vsi2qi>): Rename to...
9617 (<sur>dot_prod<vsi2qi>): ...This.
9618 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
9621 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9623 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
9625 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9627 * doc/sourcebuild.texi (arm_v8_2a_i8mm_neon_hw): Document.
9629 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9631 * config/arm/neon.md (usdot_prod<vsi2qi>): New.
9633 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9635 * config/aarch64/aarch64-simd.md (aarch64_usdot<vsi2qi>): Rename to...
9636 (usdot_prod<vsi2qi>): ... This.
9637 * config/aarch64/aarch64-simd-builtins.def (usdot): Rename to...
9638 (usdot_prod): ...This.
9639 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Likewise.
9640 * config/aarch64/aarch64-sve.md (@aarch64_<sur>dot_prod<vsi2qi>):
9642 (@<sur>dot_prod<vsi2qi>): ...This.
9643 * config/aarch64/aarch64-sve-builtins-base.cc
9644 (svusdot_impl::expand): Use it.
9646 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9648 * optabs.def (usdot_prod_optab): New.
9649 * doc/md.texi: Document it and clarify other dot prod optabs.
9650 * optabs-tree.h (enum optab_subtype): Add optab_vector_mixed_sign.
9651 * optabs-tree.c (optab_for_tree_code): Support usdot_prod_optab.
9652 * optabs.c (expand_widen_pattern_expr): Likewise.
9653 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
9654 * tree-vect-loop.c (vectorizable_reduction): Query dot-product kind.
9655 * tree-vect-patterns.c (vect_supportable_direct_optab_p): Take optional
9657 (vect_widened_op_tree): Optionally ignore
9659 (vect_recog_dot_prod_pattern): Support usdot_prod_optab.
9661 2021-07-14 H.J. Lu <hjl.tools@gmail.com>
9664 * config/i386/driver-i386.c (host_detect_local_cpu): Check
9665 "arch [32|64]" and "tune [32|64]" for 32-bit and 64-bit codegen.
9666 Enable UINTR only for 64-bit codegen.
9667 * config/i386/i386-options.c
9668 (ix86_option_override_internal::DEF_PTA): Skip PTA_UINTR if not
9670 * config/i386/i386.h (ARCH_ARG): New.
9671 (CC1_CPU_SPEC): Pass "[arch|tune] 32" for 32-bit codegen and
9672 "[arch|tune] 64" for 64-bit codegen.
9674 2021-07-14 Richard Biener <rguenther@suse.de>
9676 PR tree-optimization/101445
9677 * tree-vect-stmts.c (vectorizable_load): Do the gap adjustment
9678 of the IV in the correct direction for negative stride
9681 2021-07-14 Jakub Jelinek <jakub@redhat.com>
9684 * godump.c (godump_str_hash): New type.
9685 (godump_container::pot_dummy_types): Use string_hash instead of
9686 ptr_hash in the hash_set.
9688 2021-07-14 Richard Biener <rguenther@suse.de>
9690 * tree-vect-loop.c (vect_find_reusable_accumulator): Handle
9691 vector types where the old vector type has a multiple of
9692 the new vector type elements.
9693 (vect_create_partial_epilog): New function, split out from...
9694 (vect_create_epilog_for_reduction): ... here.
9695 (vect_transform_cycle_phi): Reduce the re-used accumulator
9696 to the new vector type.
9698 2021-07-14 Alexandre Oliva <oliva@adacore.com>
9700 * tree-ssa-alias.c (attr_fnspec::verify): Fix index in
9701 non-'t'-sized arg check.
9703 2021-07-14 Alexandre Oliva <oliva@adacore.com>
9705 * tree-cfg.c (cleanup_dead_labels_eh): Update
9706 post_landing_pad label upon change of landing pad block's
9708 (cleanup_dead_labels): Check that a removed label is not that
9711 2021-07-13 Jonathan Wright <jonathan.wright@arm.com>
9713 * combine.c (combine_simplify_rtx): Add vec_select -> subreg
9715 * config/aarch64/aarch64.md (*zero_extend<SHORT:mode><GPI:mode>2_aarch64):
9716 Add Neon to general purpose register case for zero-extend
9718 * config/arm/vfp.md (*arm_movsi_vfp): Remove "*" from *t -> r
9719 case to prevent some cases opting to go through memory.
9720 * cse.c (fold_rtx): Add vec_select -> subreg simplification.
9721 * rtl.c (rtvec_series_p): Define predicate to determine
9722 whether a vector contains a linear series of integers.
9723 * rtl.h (rtvec_series_p): Define.
9724 * rtlanal.c (vec_series_lowpart_p): Define predicate to
9725 determine if a vector selection is equivalent to the low part
9727 * rtlanal.h (vec_series_lowpart_p): Define.
9728 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
9729 Add vec_select -> subreg simplification.
9731 2021-07-13 Paul A. Clarke <pc@us.ibm.com>
9733 * config/rs6000/smmintrin.h (_mm_testz_si128, _mm_testc_si128,
9734 _mm_testnzc_si128, _mm_test_all_ones, _mm_test_all_zeros,
9735 _mm_test_mix_ones_zeros): New.
9737 2021-07-13 Roger Sayle <roger@nextmovesoftware.com>
9738 Richard Biener <rguenther@suse.de>
9740 * gimple.c (gimple_could_trap_p_1): Make S argument a
9741 "const gimple*". Preserve constness in call to
9742 gimple_asm_volatile_p.
9743 (gimple_could_trap_p): Make S argument a "const gimple*".
9744 * gimple.h (gimple_could_trap_p_1, gimple_could_trap_p):
9745 Update function prototypes.
9747 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9749 * tree-vectorizer.h (vect_reusable_accumulator): New structure.
9750 (_loop_vec_info::main_loop_edge): New field.
9751 (_loop_vec_info::skip_main_loop_edge): Likewise.
9752 (_loop_vec_info::skip_this_loop_edge): Likewise.
9753 (_loop_vec_info::reusable_accumulators): Likewise.
9754 (_stmt_vec_info::reduc_scalar_results): Likewise.
9755 (_stmt_vec_info::reused_accumulator): Likewise.
9756 (vect_get_main_loop_result): Declare.
9757 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize
9758 reduc_scalar_inputs.
9759 (vec_info::free_stmt_vec_info): Free reduc_scalar_inputs.
9760 * tree-vect-loop-manip.c (vect_get_main_loop_result): New function.
9761 (vect_do_peeling): Fill an epilogue loop's main_loop_edge,
9762 skip_main_loop_edge and skip_this_loop_edge fields.
9763 * tree-vect-loop.c (INCLUDE_ALGORITHM): Define.
9764 (vect_emit_reduction_init_stmts): New function.
9765 (get_initial_def_for_reduction): Use it.
9766 (get_initial_defs_for_reduction): Likewise. Change the vinfo
9767 parameter to a loop_vec_info.
9768 (vect_create_epilog_for_reduction): Store the scalar results
9769 in the reduc_info. If an epilogue loop is reusing an accumulator
9770 from the main loop, and if the epilogue loop can also be skipped,
9771 try to place the reduction code in the join block. Record
9772 accumulators that could potentially be reused by epilogue loops.
9773 (vect_transform_cycle_phi): When vectorizing epilogue loops,
9774 try to reuse accumulators from the main loop. Record the initial
9775 value in reduc_info for non-SLP reductions too.
9777 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9779 * tree-vect-loop.c (get_initial_def_for_reduction): Remove
9780 adjustment handling. Take the neutral value as an argument,
9781 in place of the code argument.
9782 (vect_transform_cycle_phi): Update accordingly. Handle the
9783 initial values of cond reductions separately from code reductions.
9784 Choose the adjustment here rather than in
9785 get_initial_def_for_reduction. Sink the splat of vec_initial_def.
9787 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9789 * tree-vect-loop.c (neutral_op_for_slp_reduction): Replace with...
9790 (neutral_op_for_reduction): ...this, providing a more general
9792 (vect_create_epilog_for_reduction): Update accordingly.
9793 (vectorizable_reduction): Likewise.
9794 (vect_transform_cycle_phi): Likewise.
9796 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9798 * tree-vect-loop.c (get_initial_def_for_reduction): Take the
9799 reduc_info instead of the original stmt_vec_info.
9800 (vect_transform_cycle_phi): Update accordingly.
9802 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9804 * tree-vect-loop.c (get_initial_defs_for_reduction): Take the
9805 reduc_info as an additional parameter.
9806 (vect_transform_cycle_phi): Update accordingly.
9808 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9810 * tree-vectorizer.h: Include tree-ssa-operands.h.
9811 (vect_phi_initial_value): New function.
9812 * tree-vect-loop.c (neutral_op_for_slp_reduction): Use it.
9813 (get_initial_defs_for_reduction, info_for_reduction): Likewise.
9814 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
9815 (vect_transform_cycle_phi, vectorizable_induction): Likewise.
9817 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9819 * tree-vect-loop.c (vect_create_epilog_for_reduction): Convert
9820 the phi results to vectype after creating them. Remove later
9821 conversion code that thus becomes redundant.
9823 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9825 * tree-vect-loop.c (vect_create_epilog_for_reduction): Replace
9826 the new_phis vector with a reduc_inputs vector. Combine handling
9827 of reduction chains and ncopies > 1.
9829 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9831 * tree-vect-loop.c (vect_create_epilog_for_reduction): Truncate
9832 scalar_results to group_size elements after reducing down from
9833 N*group_size elements. Construct an array_slice of the live-out
9834 stmts and assert that there is one stmt per scalar result.
9836 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9838 * tree-vect-loop.c (vect_create_epilog_for_reduction): Remove
9839 nested_in_vect_loop and use double_reduc everywhere. Remove dead
9840 assignment to "loop".
9842 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9844 * internal-fn.c (vectorized_internal_fn_supported_p): Handle
9845 vector types first. For scalar types, consider both the preferred
9846 vector mode and the alternative vector modes.
9847 * optabs-query.c (can_vec_mask_load_store_p): Use the same
9848 structure as above, in particular using related_vector_mode
9849 for modes provided by autovectorize_vector_modes.
9851 2021-07-13 Jakub Jelinek <jakub@redhat.com>
9852 Richard Biener <rguenther@suse.de>
9854 PR tree-optimization/101419
9855 * tree-pass.h (PROP_objsz): Define.
9856 (make_pass_early_object_sizes): Declare.
9857 * passes.def (pass_all_early_optimizations): Rename pass_object_sizes
9858 there to pass_early_object_sizes, drop parameter.
9859 (pass_all_optimizations): Move pass_object_sizes right after pass_ccp,
9860 drop parameter, move pass_post_ipa_warn right after that.
9861 * tree-object-size.c (pass_object_sizes::execute): Rename to...
9862 (object_sizes_execute): ... this. Add insert_min_max_p argument.
9863 (pass_data_object_sizes): Move after object_sizes_execute.
9864 (pass_object_sizes): Likewise. In execute method call
9865 object_sizes_execute, drop set_pass_param method and insert_min_max_p
9866 non-static data member and its initializer in the ctor.
9867 (pass_data_early_object_sizes, pass_early_object_sizes,
9868 make_pass_early_object_sizes): New.
9869 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Use
9870 (cfun->curr_properties & PROP_objsz) instead of cfun->after_inlining.
9872 2021-07-13 Kito Cheng <kito.cheng@sifive.com>
9875 * config/riscv/constraints.md ("S"): Update description and remove
9877 * doc/md.texi (Machine Constraints): Document the 'S' constraints
9880 2021-07-13 Richard Biener <rguenther@suse.de>
9883 2021-07-12 Richard Biener <rguenther@suse.de>
9885 * tree-vect-slp.c (vect_slp_region): Show the number of
9886 SLP graph entries in the optimization message.
9888 2021-07-13 Michael Meissner <meissner@linux.ibm.com>
9890 * config/rs6000/altivec.md (xxspltiw_v4sf): Change local variable
9892 * config/rs6000/rs6000-protos.h (rs6000_const_f32_to_i32): Change
9893 return type to long.
9894 * config/rs6000/rs6000.c (rs6000_const_f32_to_i32): Change return
9897 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
9899 * gimple-range-fold.cc (fold_using_range::range_of_builtin_ubsan_call):
9900 Query relation between the 2 operands and use it.
9902 2021-07-12 Sergei Trofimovich <siarheit@google.com>
9904 * doc/cfg.texi: Fix s/ei_safe_safe/ei_safe_edge/ typo.
9906 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
9909 * config/i386/predicates.md (vec_setm_sse41_operand):
9910 Rename from vec_setm_operand.
9911 (vec_setm_avx2_operand): New predicate.
9912 * config/i386/sse.md (vec_set<V_128:mode>): Use V_128 mode iterator.
9913 Use vec_setm_sse41_operand as operand 2 predicate.
9914 (vec_set<V_256_512:mode): New expander.
9915 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_sse41_operand
9916 as operand 2 predicate.
9918 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
9920 PR tree-optimization/101335
9921 * range-op.cc (operator_cast::lhs_op1_relation): Delete.
9923 2021-07-12 Andrew Pinski <apinski@marvell.com>
9925 * tree-ssa-phiopt.c (match_simplify_replacement): Move
9926 insert of the sequence before the movement of the
9927 statement. Check if to see if the statement is used
9928 outside of the original phi to see if we should move it.
9930 2021-07-12 Richard Biener <rguenther@suse.de>
9932 * dump-context.h (debug_dump_context::debug_dump_context):
9933 Add FILE * parameter defaulted to stderr.
9934 * dumpfile.c (debug_dump_context::debug_dump_context): Adjust.
9935 * tree-vect-slp.c (dot_slp_tree): New functions.
9937 2021-07-12 Richard Biener <rguenther@suse.de>
9939 PR tree-optimization/101373
9940 * tree-ssa-pre.c (prune_clobbered_mems): Also prune trapping
9941 references when the BB may not return.
9942 (compute_avail): Pass in the function we're working on and
9943 replace cfun references with it. Externally throwing
9944 const calls also possibly terminate the function.
9945 (pass_pre::execute): Pass down the function we're working on.
9946 * gcse.c (compute_hash_table_work): Externally throwing
9947 const/pure calls also need record_last_mem_set_info.
9948 * postreload-gcse.c (record_opr_changes): Looping or externally
9949 throwing const/pure calls also need record_last_mem_set_info.
9951 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
9953 * recog.c (memory_address_addr_space_p): Change the type to bool.
9954 Return true/false instead of 1/0.
9955 (offsettable_memref_p): Ditto.
9956 (offsettable_nonstrict_memref_p): Ditto.
9957 (offsettable_address_addr_space_p): Ditto.
9958 Change the type of addressp indirect function to bool.
9959 * recog.h (memory_address_addr_space_p): Change the type to bool.
9960 (strict_memory_address_addr_space_p): Ditto.
9961 (offsettable_memref_p): Ditto.
9962 (offsettable_nonstrict_memref_p): Ditto.
9963 (offsettable_address_addr_space_p): Ditto.
9964 * reload.c (maybe_memory_address_addr_space_p): Ditto.
9965 (strict_memory_address_addr_space_p): Change the type to bool.
9966 Return true/false instead of 1/0.
9967 (maybe_memory_address_addr_space_p): Change the type to bool.
9969 2021-07-12 Richard Biener <rguenther@suse.de>
9971 * tree-vect-slp.c (vect_slp_region): Show the number of
9972 SLP graph entries in the optimization message.
9974 2021-07-12 Richard Biener <rguenther@suse.de>
9976 PR tree-optimization/101394
9977 * tree-ssa-pre.c (do_pre_regular_insertion): Avoid inserting
9978 copies from abnormals for a full redundancy.
9980 2021-07-12 Richard Biener <rguenther@suse.de>
9982 PR middle-end/101423
9983 * gimple.c (gimple_could_trap_p_1): Internal function calls
9985 * tree-eh.c (tree_could_trap_p): Likewise.
9987 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
9990 * config/arm/arm_neon.h (vmul_n_u32): Replace call to builtin with
9992 (vmulq_n_u32): Likewise.
9993 (vmul_n_f32): Gate __a * __b on __FAST_MATH__.
9994 (vmulq_n_f32): Likewise.
9995 (vmul_n_f16): Likewise.
9996 (vmulq_n_f16): Likewise.
9998 2021-07-12 Martin Liska <mliska@suse.cz>
10000 PR sanitizer/101425
10001 * gcc.c (check_offload_target_name): Call
10002 candidates_list_and_hint only if we have a candidate.
10004 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
10007 * config/arm/neon.md (vec_init): Move to ...
10008 * config/arm/vec-common.md (vec_init): ... here.
10009 Change the pattern's mode to VDQX and gate it on VALID_MVE_MODE.
10011 2021-07-12 Roger Sayle <roger@nextmovesoftware.com>
10013 PR tree-optimization/101403
10014 * match.pd ((T)bswap(X)>>C): Correctly handle cases where
10015 signedness of the shift is not the same as the signedness of
10016 the type extension.
10018 2021-07-09 Roger Sayle <roger@nextmovesoftware.com>
10019 Uroš Bizjak <ubizjak@gmail.com>
10021 * config/i386/i386.md (*divmodsi4_const): Optimize SImode
10022 divmod of a constant numerator with new define_insn_and_split.
10024 2021-07-09 Iain Sandoe <iain@sandoe.co.uk>
10027 * config/i386/i386-expand.c (ix86_expand_call): If a call is
10028 to a non-local-binding, or local but to a public symbol, then
10029 assume that it might be indirected via the lazy symbol binder.
10030 Mark R10 and R10 as clobbered in that case.
10032 2021-07-09 Eric Botcazou <ebotcazou@adacore.com>
10035 * gcc.c (ASM_DEBUG_DWARF_OPTION): Set again to --gdwarf2 in
10036 the case where HAVE_AS_WORKING_DWARF_N_FLAG is not defined
10037 and HAVE_LD_BROKEN_PE_DWARF5 is defined.
10039 2021-07-09 Uroš Bizjak <ubizjak@gmail.com>
10041 * config/i386/i386.md (*udivmodsi4_pow2_zext_1): Limit the
10042 log2 range of operands[3] to [1,31].
10043 (*udivmodsi4_pow2_zext_2): Ditto. Correct insn RTX pattern.
10045 2021-07-09 Sergei Trofimovich <siarheit@google.com>
10047 * doc/md.texi: Don't split @smallexample in multiple @groups.
10049 2021-07-09 Sergei Trofimovich <siarheit@google.com>
10051 * doc/md.texi: Add missing 'see' word.
10053 2021-07-09 Andrew Pinski <apinski@marvell.com>
10055 * tree-ssa-phiopt.c (phiopt_early_allow): Change arguments
10056 to take sequence and gimple_match_op. Accept the case where
10057 op is a SSA_NAME and one statement in the sequence.
10058 Also allow constants.
10059 (gimple_simplify_phiopt): Always pass a sequence to resimplify.
10060 Update call to phiopt_early_allow. Discard the sequence if not
10063 2021-07-09 Xi Ruoyao <xry111@mengyan1223.wang>
10068 * config/mips/mips.c (mips_const_insns): Use MSA_SUPPORTED_MODE_P
10069 instead of ISA_HAS_MSA.
10070 (mips_expand_vec_unpack): Likewise.
10071 (mips_expand_vector_init): Likewise.
10073 2021-07-09 Kewen Lin <linkw@linux.ibm.com>
10075 * config/rs6000/vsx.md (mods_<mode>): Rename to...
10076 (mod<mode>3): ... this.
10077 (modu_<mode>): Rename to...
10078 (umod<mode>3): ... this.
10079 * config/rs6000/rs6000-builtin.def (MODS_V2DI, MODS_V4SI, MODU_V2DI,
10080 MODU_V4SI): Adjust.
10082 2021-07-08 Jeff Law <jeffreyalaw@gmail.com>
10084 * config/h8300/shiftrotate.md (variable shifts): Expose condition
10085 code handling for the test before the loop.
10087 2021-07-08 Martin Jambor <mjambor@suse.cz>
10090 * ipa-sra.c (class isra_call_summary): New member
10091 m_before_any_store, initialize it in the constructor.
10092 (isra_call_summary::dump): Dump the new field.
10093 (ipa_sra_call_summaries::duplicate): Copy it.
10094 (process_scan_results): Set it.
10095 (isra_write_edge_summary): Stream it.
10096 (isra_read_edge_summary): Likewise.
10097 (param_splitting_across_edge): Only override
10098 safe_to_import_accesses if m_before_any_store is set.
10100 2021-07-08 Martin Sebor <msebor@redhat.com>
10102 PR bootstrap/101374
10103 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref):
10104 Use Object Size Type 0 instead of 1.
10106 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
10108 * tree-vect-loop.c (vectorizable_reduction): Remove always-true
10111 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
10113 * match.pd: Simplify an extend-operate-truncate sequence involving
10116 2021-07-08 Roger Sayle <roger@nextmovesoftware.com>
10117 Richard Biener <rguenther@suse.de>
10119 PR tree-optimization/40210
10120 * match.pd (bswap optimizations): Simplify (bswap(x)>>C1)&C2 as
10121 (x>>C3)&C2 when possible. Simplify bswap(x)>>C1 as ((T)x)>>C2
10122 when possible. Simplify bswap(x)&C1 as (x>>C2)&C1 when 0<=C1<=255.
10124 2021-07-08 Uroš Bizjak <ubizjak@gmail.com>
10127 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
10129 * config/i386/mmx.md (V_32): New mode iterator.
10130 (mov<V_32:mode>): Use V_32 mode iterator.
10131 (*mov<V_32:mode>_internal): Ditto.
10132 (*push<V_32:mode>2_rex64): Ditto.
10133 (*push<V_32:mode>2): Ditto.
10134 (movmisalign<V_32:mode>): Ditto.
10135 (mmx_<any_shiftrt:insn>v1si3): New insn pattern.
10136 (sse4_1_<any_extend:code>v2qiv2hi2): Ditto.
10137 (vec_unpacks_lo_v4qi): New expander.
10138 (vec_unpacks_hi_v4qi): Ditto.
10139 (vec_unpacku_lo_v4qi): Ditto.
10140 (vec_unpacku_hi_v4qi): Ditto.
10141 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add V1SImode.
10142 (VALID_INT_MODE_P): Ditto.
10144 2021-07-08 Michael Meissner <meissner@linux.ibm.com>
10147 * config/rs6000/rs6000.md (udivti3): New insn.
10148 (divti3): New insn.
10149 (umodti3): New insn.
10150 (modti3): New insn.
10152 2021-07-07 Martin Sebor <msebor@redhat.com>
10154 PR tree-optimization/100137
10155 PR tree-optimization/99121
10156 PR tree-optimization/97027
10157 * builtins.c (access_ref::access_ref): Also set offmax.
10158 (access_ref::offset_in_range): Define new function.
10159 (access_ref::add_offset): Set offmax.
10160 (access_ref::inform_access): Handle access_none.
10161 (handle_mem_ref): Clear ostype.
10162 (compute_objsize_r): Handle ASSERT_EXPR.
10163 * builtins.h (struct access_ref): Add offmax member.
10164 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Use
10165 compute_objsize() and simplify.
10167 2021-07-07 Peter Bergner <bergner@linux.ibm.com>
10169 * config/rs6000/rs6000-call.c (mma_init_builtins): Use VSX_BUILTIN_LXVP
10170 and VSX_BUILTIN_STXVP.
10172 2021-07-07 Martin Sebor <msebor@redhat.com>
10175 * config/aarch64/aarch64.c (aarch64_simd_lane_bounds): Remove
10176 a stray %K from error_at() missed in r12-2088.
10178 2021-07-07 Richard Biener <rguenther@suse.de>
10180 PR tree-optimization/99728
10181 * tree-ssa-loop-im.c (gather_mem_refs_stmt): Record
10183 (mem_refs_may_alias_p): Add assert we handled aggregate
10185 (sm_seq_valid_bb): Give up when running into aggregate copies.
10186 (ref_indep_loop_p): Handle aggregate copies as never
10187 being invariant themselves but allow other refs to be
10188 disambiguated against them.
10189 (can_sm_ref_p): Do not try to apply store-motion to aggregate
10192 2021-07-06 Indu Bhagat <indu.bhagat@oracle.com>
10195 * dwarf2ctf.c (ctf_get_AT_data_member_location): Multiply by 8 to get
10198 2021-07-06 Martin Sebor <msebor@redhat.com>
10200 * gimple-pretty-print.c (percent_G_format): Remove.
10201 * tree-diagnostic.c (default_tree_printer): Remove calls.
10202 * tree-pretty-print.c (percent_K_format): Remove.
10203 * tree-pretty-print.h (percent_K_format): Remove.
10205 2021-07-06 Martin Sebor <msebor@redhat.com>
10207 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_builtin):
10208 Remove %K and use error_at.
10209 (aarch64_expand_fcmla_builtin): Same.
10210 (aarch64_expand_builtin_tme): Same.
10211 (aarch64_expand_builtin_memtag): Same.
10212 * config/arm/arm-builtins.c (arm_expand_acle_builtin): Same.
10213 (arm_expand_builtin): Same.
10214 * config/arm/arm.c (bounds_check): Same.
10216 2021-07-06 Martin Sebor <msebor@redhat.com>
10218 * builtins.c (warn_string_no_nul): Remove %G.
10219 (maybe_warn_for_bound): Same.
10220 (warn_for_access): Same.
10221 (check_access): Same.
10222 (check_strncat_sizes): Same.
10223 (expand_builtin_strncat): Same.
10224 (expand_builtin_strncmp): Same.
10225 (expand_builtin): Same.
10226 (expand_builtin_object_size): Same.
10227 (warn_dealloc_offset): Same.
10228 (maybe_emit_free_warning): Same.
10229 * calls.c (maybe_warn_alloc_args_overflow): Same.
10230 (maybe_warn_nonstring_arg): Same.
10231 (maybe_warn_rdwr_sizes): Same.
10232 * expr.c (expand_expr_real_1): Remove %K.
10233 * gimple-fold.c (gimple_fold_builtin_strncpy): Remove %G.
10234 (gimple_fold_builtin_strncat): Same.
10235 * gimple-ssa-sprintf.c (format_directive): Same.
10236 (handle_printf_call): Same.
10237 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Same.
10238 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
10239 (maybe_diag_access_bounds): Same. Call gimple_location.
10240 (check_bounds_or_overlap): Same.
10241 * trans-mem.c (ipa_tm_scan_irr_block): Remove %K. Simplify.
10242 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Remove %G.
10243 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
10244 (maybe_diag_stxncpy_trunc): Same.
10245 (handle_builtin_stxncpy_strncat): Same.
10246 (maybe_warn_pointless_strcmp): Same.
10247 * tree-ssa-uninit.c (maybe_warn_operand): Same.
10249 2021-07-06 Uroš Bizjak <ubizjak@gmail.com>
10252 * config/i386/predicates.md (vec_setm_operand): Enable
10253 register_operand for TARGET_SSE4_1.
10254 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_operand
10255 as operand 2 predicate. Call ix86_expand_vector_set_var
10256 for non-constant index operand.
10257 (vec_setv4qi): Use vec_setm_mmx_operand as operand 2 predicate.
10258 Call ix86_expand_vector_set_var for non-constant index operand.
10260 2021-07-06 Jeff Law <jeffreyalaw@gmail.com>
10262 * config/h8300/jumpcall.md (*branch): When possible, generate
10263 the comparison in CCZN mode.
10264 * config/h8300/predicates.md (simple_memory_operand): Reject all
10265 auto-increment addressing modes.
10267 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
10269 PR bootstrap/100246
10270 * config/i386/i386.h (struct stringop_algs): Define a CTOR for
10273 2021-07-06 Richard Biener <rguenther@suse.de>
10275 * doc/md.texi (vec_fmaddsub<mode>4): Document.
10276 (vec_fmsubadd<mode>4): Likewise.
10277 * optabs.def (vec_fmaddsub$a4): Add.
10278 (vec_fmsubadd$a4): Likewise.
10279 * internal-fn.def (IFN_VEC_FMADDSUB): Add.
10280 (IFN_VEC_FMSUBADD): Likewise.
10281 * tree-vect-slp-patterns.c (addsub_pattern::recognize):
10282 Refactor to handle IFN_VEC_FMADDSUB and IFN_VEC_FMSUBADD.
10283 (addsub_pattern::build): Likewise.
10284 * tree-vect-slp.c (vect_optimize_slp): CFN_VEC_FMADDSUB
10285 and CFN_VEC_FMSUBADD are not transparent for permutes.
10286 * config/i386/sse.md (vec_fmaddsub<mode>4): New expander.
10287 (vec_fmsubadd<mode>4): Likewise.
10289 2021-07-06 Richard Biener <rguenther@suse.de>
10291 * doc/invoke.texi (fmove-loop-stores): Document.
10292 * common.opt (fmove-loop-stores): New option.
10293 * opts.c (default_options_table): Enable -fmove-loop-stores
10294 at -O1 but not -Og.
10295 * tree-ssa-loop-im.c (pass_lim::execute): Pass
10296 flag_move_loop_stores instead of true to
10297 loop_invariant_motion_in_fun.
10299 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
10301 * doc/install.texi: Document --with-dsymutil.
10303 2021-07-06 Andrew Pinski <apinski@marvell.com>
10305 PR tree-optimization/101256
10306 * dbgcnt.def (phiopt_edge_range): New counter.
10307 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
10308 Check to make sure the new name is defined in the same
10309 bb as the conditional before duplicating range info.
10310 Also add debug counter.
10312 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
10314 PR rtl-optimization/100328
10315 * config/i386/i386-options.c (ix86_option_override_internal):
10316 Set param_ira_consider_dup_in_all_alts to 0.
10318 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
10320 PR rtl-optimization/100328
10321 * doc/invoke.texi (ira-consider-dup-in-all-alts): Document new
10323 * ira.c (ira_get_dup_out_num): Adjust as parameter
10324 param_ira_consider_dup_in_all_alts.
10325 * params.opt (ira-consider-dup-in-all-alts): New.
10326 * ira-conflicts.c (process_regs_for_copy): Add one parameter
10327 single_input_op_has_cstr_p.
10328 (get_freq_for_shuffle_copy): New function.
10329 (add_insn_allocno_copies): Adjust as single_input_op_has_cstr_p.
10330 * ira-int.h (ira_get_dup_out_num): Add one bool parameter.
10332 2021-07-05 Jeff Law <jeffreyalaw@gmail.com>
10334 * config/h8300/shiftrotate.md (shift-by-variable patterns): Update to
10335 generate condition code aware RTL directly.
10337 2021-07-05 Andrew Pinski <apinski@marvell.com>
10339 PR tree-optimization/101039
10340 * match.pd (A CMP 0 ? A : -A): New patterns.
10341 * tree-ssa-phiopt.c (abs_replacement): Delete function.
10342 (tree_ssa_phiopt_worker): Don't call abs_replacement.
10343 Update comment about abs_replacement.
10345 2021-07-05 Andrew Pinski <apinski@marvell.com>
10347 * tree-ssa-phiopt.c (gimple_simplify_phiopt):
10348 If "A ? B : C" fails to simplify, try "(!A) ? C : B".
10350 2021-07-05 Andrew Pinski <apinski@marvell.com>
10352 * tree-ssa-phiopt.c (match_simplify_replacement):
10353 Add early_p argument. Call gimple_simplify_phiopt
10354 instead of gimple_simplify.
10355 (tree_ssa_phiopt_worker): Update call to
10356 match_simplify_replacement and allow unconditionally.
10357 (phiopt_early_allow): New function.
10358 (gimple_simplify_phiopt): New function.
10360 2021-07-05 Andrew Pinski <apinski@marvell.com>
10362 PR middle-end/101237
10363 * fold-const.c (negate_expr_p): Remove call to element_mode
10364 and TREE_MODE/TREE_TYPE when calling HONOR_SIGNED_ZEROS,
10365 HONOR_SIGN_DEPENDENT_ROUNDING, and HONOR_SNANS.
10366 (fold_negate_expr_1): Likewise.
10367 (const_unop): Likewise.
10368 (fold_cond_expr_with_comparison): Likewise.
10369 (fold_binary_loc): Likewise.
10370 (fold_ternary_loc): Likewise.
10371 (tree_call_nonnegative_warnv_p): Likewise.
10372 * match.pd (-(A + B) -> (-B) - A): Likewise.
10374 2021-07-05 Iain Sandoe <iain@sandoe.co.uk>
10376 * configure.ac: Handle --with-dsymutil in the same way as we
10377 do for the assembler and linker. (DEFAULT_DSYMUTIL): New.
10378 Extract the type and version for the dsymutil configured or
10379 found by the default searches.
10380 * config.in: Regenerated.
10381 * configure: Regenerated.
10382 * collect2.c (do_dsymutil): Handle locating dsymutil in the
10383 same way as for the assembler and linker.
10384 * config/darwin.h (DSYMUTIL): Delete.
10385 * gcc.c: Report a configured dsymutil correctly.
10386 * exec-tool.in: Allow for dsymutil.
10388 2021-07-05 Uroš Bizjak <ubizjak@gmail.com>
10390 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
10391 Handle V4QI and V2HI modes.
10392 (expand_vec_perm_blend): Allow 4-byte vector modes with TARGET_SSE4_1.
10393 Handle V4QI mode. Emit mmx_pblendvb32 for 4-byte modes.
10394 (expand_vec_perm_pshufb): Rewrite to use switch statemets.
10395 Handle 4-byte dual operands with TARGET_XOP and single operands
10396 with TARGET_SSSE3. Emit mmx_ppermv32 for TARGET_XOP and
10397 mmx_pshufbv4qi3 for TARGET_SSSE3.
10398 (expand_vec_perm_pblendv): Allow 4-byte vector modes with TARGET_SSE4_1.
10399 (expand_vec_perm_interleave2): Allow 4-byte vector modes.
10400 (expand_vec_perm_pshufb2): Allow 4-byte vector modes with TARGET_SSSE3.
10401 (expand_vec_perm_even_odd_1): Handle V4QI mode.
10402 (expand_vec_perm_broadcast_1): Handle V4QI mode.
10403 (ix86_vectorize_vec_perm_const): Handle V4QI mode.
10404 * config/i386/mmx.md (mmx_ppermv32): New insn pattern.
10405 (mmx_pshufbv4qi3): Ditto.
10406 (*mmx_pblendw32): Ditto.
10407 (*mmx_pblendw64): Rename from *mmx_pblendw.
10408 (mmx_punpckhbw_low): New insn_and_split pattern.
10409 (mmx_punpcklbw_low): Ditto.
10411 2021-07-05 Richard Biener <rguenther@suse.de>
10413 * tree-vect-loop-manip.c (vect_loop_versioning): Do not
10414 set LOOP_C_INFINITE on the vectorized loop.
10416 2021-07-05 Richard Biener <rguenther@suse.de>
10418 PR middle-end/101291
10419 * cfgloopmanip.c (loop_version): Set the loop copy of the
10420 versioned loop to the new loop.
10422 2021-07-04 Iain Sandoe <iain@sandoe.co.uk>
10425 * config.gcc: Ensure that Darwin biarch definitions are
10426 added before i386.h.
10427 * config/i386/darwin.h (TARGET_64BIT): Remove.
10428 (PR80556_WORKAROUND): New.
10429 (REAL_LIBGCC_SPEC): Amend to use PR80556_WORKAROUND.
10430 (DARWIN_SUBARCH_SPEC): New.
10431 * config/i386/darwin32-biarch.h (TARGET_64BIT_DEFAULT,
10432 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
10433 (REAL_LIBGCC_SPEC): Remove.
10434 * config/i386/darwin64-biarch.h (TARGET_64BIT_DEFAULT,
10435 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
10436 (REAL_LIBGCC_SPEC): Remove.
10438 2021-07-03 H.J. Lu <hjl.tools@gmail.com>
10440 PR middle-end/101294
10441 * expr.c (store_constructor): Don't use vec_duplicate on vector.
10443 2021-07-02 Martin Sebor <msebor@redhat.com>
10445 PR middle-end/98871
10446 PR middle-end/98512
10447 * diagnostic.c (get_any_inlining_info): New.
10448 (update_effective_level_from_pragmas): Handle inlining context.
10449 (diagnostic_enabled): Same.
10450 (diagnostic_report_diagnostic): Same.
10451 * diagnostic.h (struct diagnostic_info): Add ctor.
10452 (struct diagnostic_context): Add new member.
10453 * tree-diagnostic.c (set_inlining_locations): New.
10454 (tree_diagnostics_defaults): Set new callback pointer.
10456 2021-07-02 Peter Bergner <bergner@linux.ibm.com>
10458 * config/rs6000/rs6000-builtin.def (BU_MMA_PAIR_LD, BU_MMA_PAIR_ST):
10460 (__builtin_vsx_lxvp, __builtin_vsx_stxvp): New built-ins.
10461 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Expand
10462 lxvp and stxvp built-ins.
10463 (mma_init_builtins): Handle lxvp and stxvp built-ins.
10464 (builtin_function_type): Likewise.
10465 * doc/extend.texi (__builtin_vsx_lxvp, __builtin_mma_stxvp): Document.
10467 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
10469 * config/h8300/h8300-protos.h (compute_a_shift_cc): Accept
10470 additional argument for the code.
10471 * config/h8300/h8300.c (compute_a_shift_cc): Accept additional
10472 argument for the code. Just return if the ZN bits are useful or
10473 not rather than the old style CC_* enums.
10474 * config/h8300/shiftrotate.md (shiftqi_noscratch): Move before
10475 more generic shiftqi patterns.
10476 (shifthi_noscratch, shiftsi_noscratch): Similarly.
10477 (shiftqi_noscratch_set_flags): New pattern.
10478 (shifthi_noscratch_set_flags, shiftsi_noscratch_set_flags): Likewise.
10480 2021-07-02 Andrew MacLeod <amacleod@redhat.com>
10482 PR tree-optimization/101223
10483 * range-op.cc (build_lt): Add -1 for signed values.
10484 (built_gt): Subtract -1 for signed values.
10486 2021-07-02 David Faust <david.faust@oracle.com>
10488 * btfout.c (get_btf_kind): Support BTF_KIND_FLOAT.
10489 (btf_asm_type): Likewise.
10491 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
10493 * config/h8300/h8300-protos.h (output_a_shift): Make first argument
10494 an array of rtx rather than a pointer to rtx. Add code argument.
10495 (compute_a_shift_length): Similarly.
10496 * config/h8300/h8300.c (h8300_shift_costs): Adjust now that the
10497 shift itself isn't an operand. Create dummy operand[0] to carry
10498 a mode and pass a suitable rtx code to compute_a_shift_length.
10499 (get_shift_alg): Adjust operand number of clobber in output templates.
10500 (output_a_shift): Make first argument an array of rtx rather than
10501 a pointer to rtx. Add code argument for the type of shift.
10502 Adjust now that the shift itself is no longer an operand.
10503 (compute_a_shift_length): Similarly.
10504 * config/h8300/shiftrotate.md (shiftqi, shifthi, shiftsi): Use an
10505 iterator rather than nshift_operator.
10506 (shiftqi_noscratch, shifthi_noscratch, shiftsi_noscratch): Likewise.
10507 (shiftqi_clobber_flags): Adjust to API changes in output_a_shift
10508 and compute_a_shift_length.
10509 (shiftqi_noscratch_clobber_flags): Likewise.
10510 (shifthi_noscratch_clobber_flags): Likewise.
10511 (shiftsi_noscratch_clobber_flags): Likewise.
10513 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
10516 * config/darwin.h (DSYMUTIL_SPEC): Do not try to run
10517 dsymutil for BTF/CTF.
10519 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
10522 * config/darwin.h (CTF_INFO_SECTION_NAME): Update the
10523 segment to include BTF.
10524 (BTF_INFO_SECTION_NAME): New.
10526 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
10528 * config/m32r/m32r-protos.h (call_operand): Adjust return type.
10529 (small_data_operand, memreg_operand, small_insn_p): Likewise.
10530 * config/m32r/m32r.c (call_operand): Adjust return type.
10531 (small_data_operand, memreg_operand): Likewise.
10533 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
10535 * config/frv/frv-protos.h (integer_register_operand): Adjust return
10537 (frv_load_operand, gpr_or_fpr_operand, gpr_no_subreg_operand): Likewise.
10538 (fpr_or_int6_operand, gpr_or_int_operand); Likewise.
10539 (gpr_or_int12_operand, gpr_or_int10_operand); Likewise.
10540 (move_source_operand, move_destination_operand): Likewise.
10541 (condexec_source_operand, condexec_dest_operand): Likewise.
10542 (lr_operand, gpr_or_memory_operand, fpr_or_memory_operand): Likewise.
10543 (reg_or_0_operand, fcc_operand, icc_operand, cc_operand): Likewise.
10544 (fcr_operand, icr_operand, cr_operand, call_operand): Likewise.
10545 (fpr_operand, even_reg_operand, odd_reg_operand): Likewise.
10546 (even_gpr_operand, odd_gpr_operand, quad_fpr_operand): Likewise.
10547 (even_fpr_operand, odd_fpr_operand): Likewise.
10548 (dbl_memory_one_insn_operand, dbl_memory_two_insn_operand): Likewise.
10549 (int12_operand, int6_operand, int5_operand, uint5_operand): Likewise.
10550 (uint4_operand, uint1_operand, int_2word_operand): Likewise
10551 (upper_int16_operand, uint16_operand, symbolic_operand): Likewise.
10552 (relational_operator, float_relational_operator): Likewise.
10553 (ccr_eqne_operator, minmax_operator): Likewise.
10554 (condexec_si_binary_operator, condexec_si_media_operator): Likewise.
10555 (condexec_si_divide_operator, condexec_si_unary_operator): Likewise.
10556 (condexec_sf_conv_operator, condexec_sf_add_operator): Likewise.
10557 (intop_compare_operator, acc_operand, even_acc_operand): Likewise.
10558 (quad_acc_operand, accg_operand): Likewise.
10560 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
10562 * config/stormy16/stormy16-protos.h (xstormy16_below_100_symbol): Change
10563 return type to a bool.
10564 (nonimmediate_nonstack_operand): Likewise.
10565 (xstormy16_splittable_below100_operand): Likewise.
10566 * config/stormy16/stormy16.c (xstormy16_below_100_symbol): Fix
10568 (xstormy16_splittable_below100_operand): Likewise.
10570 2021-07-02 Richard Biener <rguenther@suse.de>
10572 PR tree-optimization/101293
10573 * tree-ssa-loop-im.c (mem_ref_hasher::equal): Compare MEM_REF bases
10574 with combined offsets.
10575 (gather_mem_refs_stmt): Hash MEM_REFs as if their offset were
10576 combined with the rest of the offset.
10578 2021-07-02 Eric Botcazou <ebotcazou@adacore.com>
10580 * config/i386/i386.c (asm_preferred_eh_data_format): Always use the
10581 PIC encodings for PE-COFF targets.
10583 2021-07-02 Jakub Jelinek <jakub@redhat.com>
10586 * config/i386/i386-expand.c (ix86_broadcast_from_integer_constant):
10587 Return nullptr for TImode inner mode.
10589 2021-07-02 Richard Biener <rguenther@suse.de>
10591 PR tree-optimization/101280
10592 PR tree-optimization/101173
10593 * gimple-loop-interchange.cc
10594 (tree_loop_interchange::valid_data_dependences): Properly
10595 guard all dependence checks with DDR_REVERSED_P or its
10598 2021-07-02 Hongyu Wang <hongyu.wang@intel.com>
10600 * config/i386/i386-expand.c (ix86_expand_builtin):
10601 Add branch to clear odata when ZF is set for asedecenc_expand
10602 and wideaesdecenc_expand.
10604 2021-07-02 Eugene Rozenfeld <erozen@microsoft.com>
10606 * config/i386/gcc-auto-profile: regenerate
10608 2021-07-02 liuhongt <hongtao.liu@intel.com>
10610 * config/i386/sse.md (trunc<mode><pmov_dst_4>2): Refined to ..
10611 (trunc<mode><pmov_dst_4_lower>2): this.
10613 2021-07-01 David Malcolm <dmalcolm@redhat.com>
10615 * diagnostic.h (diagnostic_context::m_file_cache): New field.
10616 * input.c (class fcache): Rename to...
10617 (class file_cache_slot): ...this, making most members private and
10618 prefixing fields with "m_".
10619 (file_cache_slot::get_file_path): New accessor.
10620 (file_cache_slot::get_use_count): New accessor.
10621 (file_cache_slot::missing_trailing_newline_p): New accessor.
10622 (file_cache_slot::inc_use_count): New.
10623 (fcache_buffer_size): Move to...
10624 (file_cache_slot::buffer_size): ...here.
10625 (fcache_line_record_size): Move to...
10626 (file_cache_slot::line_record_size): ...here.
10627 (fcache_tab): Delete, in favor of global_dc->m_file_cache.
10628 (fcache_tab_size): Move to file_cache::num_file_slots.
10629 (diagnostic_file_cache_init): Update for move of fcache_tab
10630 to global_dc->m_file_cache.
10631 (diagnostic_file_cache_fini): Likewise.
10632 (lookup_file_in_cache_tab): Convert to...
10633 (file_cache::lookup_file): ...this.
10634 (diagnostics_file_cache_forcibly_evict_file): Update for move of
10635 fcache_tab to global_dc->m_file_cache, moving most of
10636 implementation to...
10637 (file_cache::forcibly_evict_file): ...this new function and...
10638 (file_cache_slot::evict): ...this new function.
10639 (evicted_cache_tab_entry): Convert to...
10640 (file_cache::evicted_cache_tab_entry): ...this.
10641 (add_file_to_cache_tab): Convert to...
10642 (file_cache::add_file): ...this, moving bulk of implementation
10644 (file_cache_slot::create): ..this new function.
10645 (file_cache::file_cache): New.
10646 (file_cache::~file_cache): New.
10647 (lookup_or_add_file_to_cache_tab): Convert to...
10648 (file_cache::lookup_or_add_file): ..this new function.
10649 (fcache::fcache): Rename to...
10650 (file_cache_slot::file_cache_slot): ...this, adding "m_" prefixes
10652 (fcache::~fcache): Rename to...
10653 (file_cache_slot::~file_cache_slot): ...this, adding "m_" prefixes
10655 (needs_read): Convert to...
10656 (file_cache_slot::needs_read_p): ...this.
10657 (needs_grow): Convert to...
10658 (file_cache_slot::needs_grow_p): ...this.
10659 (maybe_grow): Convert to...
10660 (file_cache_slot::maybe_grow): ...this.
10661 (read_data): Convert to...
10662 (file_cache_slot::read_data): ...this.
10663 (maybe_read_data): Convert to...
10664 (file_cache_slot::maybe_read_data): ...this.
10665 (get_next_line): Convert to...
10666 (file_cache_slot::get_next_line): ...this.
10667 (goto_next_line): Convert to...
10668 (file_cache_slot::goto_next_line): ...this.
10669 (read_line_num): Convert to...
10670 (file_cache_slot::read_line_num): ...this.
10671 (location_get_source_line): Update for moving of globals to
10672 global_dc->m_file_cache.
10673 (location_missing_trailing_newline): Likewise.
10674 * input.h (class file_cache_slot): New forward decl.
10675 (class file_cache): New.
10677 2021-07-01 Michael Meissner <meissner@linux.ibm.com>
10679 * config/rs6000/rs6000.c (rs6000_maybe_emit_fp_cmove): Add IEEE
10680 128-bit floating point conditional move support.
10681 (have_compare_and_set_mask): Add IEEE 128-bit floating point
10683 * config/rs6000/rs6000.md (mov<mode>cc, IEEE128 iterator): New insn.
10684 (mov<mode>cc_p10, IEEE128 iterator): New insn.
10685 (mov<mode>cc_invert_p10, IEEE128 iterator): New insn.
10686 (fpmask<mode>, IEEE128 iterator): New insn.
10687 (xxsel<mode>, IEEE128 iterator): New insn.
10689 2021-07-01 Iain Sandoe <iain@sandoe.co.uk>
10692 * config/darwin.h (CTF_INFO_SECTION_NAME): New.
10694 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
10696 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
10698 * config/i386/i386-protos.h (ix86_expand_vector_init_duplicate):
10700 * config/i386/sse.md (INT_BROADCAST_MODE): New mode iterator.
10701 (vec_duplicate<mode>): New expander.
10703 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
10706 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
10708 (ix86_byte_broadcast): New function.
10709 (ix86_convert_const_wide_int_to_broadcast): Likewise.
10710 (ix86_expand_move): Convert CONST_WIDE_INT to broadcast if mode
10711 size is 16 bytes or bigger.
10712 (ix86_broadcast_from_integer_constant): New function.
10713 (ix86_expand_vector_move): Convert CONST_WIDE_INT and CONST_VECTOR
10714 to broadcast if mode size is 16 bytes or bigger.
10715 * config/i386/i386-protos.h (ix86_gen_scratch_sse_rtx): New
10717 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): New function.
10719 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
10721 * config/i386/predicates.md (ix86_endbr_immediate_operand):
10722 Return true/false instead of 1/0.
10723 (movq_parallel): Ditto.
10725 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
10727 * recog.c (general_operand): Return true/false instead of 1/0.
10728 (register_operand): Ditto.
10729 (immediate_operand): Ditto.
10730 (const_int_operand): Ditto.
10731 (const_scalar_int_operand): Ditto.
10732 (const_double_operand): Ditto.
10733 (push_operand): Ditto.
10734 (pop_operand): Ditto.
10735 (memory_operand): Ditto.
10736 (indirect_operand): Ditto.
10738 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
10740 * genpreds.c (write_predicate_subfunction):
10741 Change the type of written subfunction to bool.
10742 (write_one_predicate_function):
10743 Change the type of written function to bool.
10744 (write_tm_preds_h): Ditto.
10745 * recog.h (*insn_operand_predicate_fn): Change the type to bool.
10746 * recog.c (general_operand): Change the type to bool.
10747 (address_operand): Ditto.
10748 (register_operand): Ditto.
10749 (pmode_register_operand): Ditto.
10750 (scratch_operand): Ditto.
10751 (immediate_operand): Ditto.
10752 (const_int_operand): Ditto.
10753 (const_scalar_int_operand): Ditto.
10754 (const_double_operand): Ditto.
10755 (nonimmediate_operand): Ditto.
10756 (nonmemory_operand): Ditto.
10757 (push_operand): Ditto.
10758 (pop_operand): Ditto.
10759 (memory_operand): Ditto.
10760 (indirect_operand): Ditto.
10761 (ordered_comparison_operator): Ditto.
10762 (comparison_operator): Ditto.
10763 * config/i386/i386-expand.c (ix86_expand_sse_cmp):
10764 Change the type of indirect predicate function to bool.
10765 * config/rs6000/rs6000.c (easy_vector_constant):
10766 Change the type to bool.
10767 * config/mips/mips-protos.h (m16_based_address_p):
10768 Change the type of operand 3 to bool.
10770 2021-07-01 Richard Biener <rguenther@suse.de>
10772 PR tree-optimization/101280
10773 PR tree-optimization/101173
10774 * gimple-loop-interchange.cc
10775 (tree_loop_interchange::valid_data_dependences): Revert
10776 previous change and instead correctly handle DDR_REVERSED_P
10779 2021-07-01 Richard Biener <rguenther@suse.de>
10781 PR tree-optimization/101278
10782 * tree-ssa-dse.c (dse_classify_store): First check for
10783 uses, then ignore stmt for chaining purposes.
10785 2021-07-01 Richard Biener <rguenther@suse.de>
10787 PR tree-optimization/100778
10788 * tree-vect-slp.c (vect_schedule_slp_node): Do not place trapping
10789 vectorized ops ahead of their scalar BB.
10791 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
10794 * config/i386/i386.md (*nabs<dwi>2_doubleword):
10795 New insn_and_split pattern.
10796 (*nabs<dwi>2_1): Ditto.
10797 * config/i386/i386-features.c
10798 (general_scalar_chain::compute_convert_gain):
10799 Handle (NEG (ABS (...))) RTX. Rewrite src code
10800 scanner as switch statement.
10801 (general_scalar_chain::convert_insn):
10802 Handle (NEG (ABS (...))) RTX.
10803 (general_scalar_to_vector_candidate_p):
10804 Detect (NEG (ABS (...))) RTX. Reorder case statements
10805 for (AND (NOT (...) ...)) fallthrough.
10807 2021-07-01 Richard Biener <rguenther@suse.de>
10809 PR tree-optimization/101178
10810 * tree-vect-slp.c (slpg_vertex::materialize): Remove.
10811 (slpg::perm_in): Add.
10812 (slpg::get_perm_in): Remove.
10813 (slpg::get_perm_materialized): Add.
10814 (vect_optimize_slp): Handle VEC_PERM nodes more optimally
10815 during permute propagation and materialization.
10817 2021-07-01 Jakub Jelinek <jakub@redhat.com>
10820 * dwarf2out.c (loc_list_from_tree_1): Handle COMPOUND_LITERAL_EXPR.
10822 2021-07-01 Jakub Jelinek <jakub@redhat.com>
10824 PR middle-end/94366
10825 * omp-low.c (lower_rec_input_clauses): Rename is_fp_and_or to
10826 is_truth_op, set it for TRUTH_*IF_EXPR regardless of new_var's type,
10827 use boolean_type_node instead of integer_type_node as NE_EXPR type.
10828 (lower_reduction_clauses): Likewise.
10830 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
10832 * config/gcn/gcn.c: Include dwarf2.h.
10833 (gcn_addr_space_debug): New function.
10834 (TARGET_ADDR_SPACE_DEBUG): New hook.
10836 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
10838 * common/config/gcn/gcn-common.c
10839 (gcn_option_optimization_table): Change OPT_fomit_frame_pointer to -O3.
10840 * config/gcn/gcn.c (gcn_expand_prologue): Prefer the frame pointer
10842 (gcn_expand_prologue): Prefer the frame pointer when emitting CFI.
10843 (gcn_frame_pointer_rqd): New function.
10844 (TARGET_FRAME_POINTER_REQUIRED): New hook.
10846 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
10848 * config/gcn/gcn.c (move_callee_saved_registers): Emit CFI notes for
10849 prologue register saves.
10850 (gcn_debug_unwind_info): Use UI_DWARF2.
10851 (gcn_dwarf_register_number): Map DWARF_LINK_REGISTER to DWARF PC.
10852 (gcn_dwarf_register_span): DWARF_LINK_REGISTER doesn't span.
10853 * config/gcn/gcn.h: (DWARF_FRAME_RETURN_COLUMN): New define.
10854 (DWARF_LINK_REGISTER): New define.
10855 (FIRST_PSEUDO_REGISTER): Increment.
10856 (FIXED_REGISTERS): Add entry for DWARF_LINK_REGISTER.
10857 (CALL_USED_REGISTERS): Likewise.
10858 (REGISTER_NAMES): Likewise.
10860 2021-06-30 Richard Biener <rguenther@suse.de>
10862 PR tree-optimization/101267
10863 * tree-vect-stmts.c (vect_check_scalar_mask): Adjust
10864 API and use SLP compatible interface of vect_is_simple_use.
10865 Reject not vectorized SLP defs for callers that do not support
10867 (vect_check_store_rhs): Handle masked stores and pass down
10868 the appropriate operator index.
10869 (vectorizable_call): Adjust.
10870 (vectorizable_store): Likewise.
10871 (vectorizable_load): Likewise. Handle SLP pecularity of
10873 (vect_is_simple_use): Remove special-casing of masked stores.
10875 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
10877 * common.opt (foffload): Remove help as Driver only.
10878 * gcc.c (display_help): Add -foffload.
10880 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
10882 * gcc.c (close_at_file, execute): Replace alloca by XALLOCAVEC.
10883 (check_offload_target_name): Fix splitting OFFLOAD_TARGETS into
10884 a candidate list; better inform no offload target is configured
10885 and fix hint extraction when passed target is not '\0' at [len].
10886 * common.opt (foffload): Add tailing '.'.
10887 (foffload-options): Likewise; fix flag name in the help string.
10889 2021-06-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
10892 * config/arm/arm_neon.h: Move vabs intrinsics before vcage_f32.
10893 (vcage_f32): Gate comparison on __FAST_MATH__.
10894 (vcageq_f32): Likewise.
10895 (vcale_f32): Likewise.
10896 (vcaleq_f32): Likewise.
10897 (vcagt_f32): Likewise.
10898 (vcagtq_f32): Likewise.
10899 (vcalt_f32): Likewise.
10900 (vcaltq_f32): Likewise.
10901 (vcage_f16): Likewise.
10902 (vcageq_f16): Likewise.
10903 (vcale_f16): Likewise.
10904 (vcaleq_f16): Likewise.
10905 (vcagt_f16): Likewise.
10906 (vcagtq_f16): Likewise.
10907 (vcalt_f16): Likewise.
10908 (vcaltq_f16): Likewise.
10910 2021-06-30 Richard Biener <rguenther@suse.de>
10912 PR tree-optimization/101264
10913 * tree-vect-slp.c (vect_optimize_slp): Propagate the
10914 computed perm_in to all "any" permute successors
10915 we cannot de-duplicate immediately.
10917 2021-06-30 liuhongt <hongtao.liu@intel.com>
10920 * config/i386/sse.md
10921 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>):
10923 (avx512f_sfixupimm<mode><maskz_scalar_name><round_saeonly_name>):
10925 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>"): Refined.
10926 * config/i386/subst.md (maskz_scalar): New define_subst.
10927 (maskz_scalar_name): New subst_attr.
10928 (maskz_scalar_op5): Ditto.
10929 (round_saeonly_maskz_scalar_op5): Ditto.
10930 (round_saeonly_maskz_scalar_operand5): Ditto.
10932 2021-06-30 David Edelsohn <dje.gcc@gmail.com>
10934 * config/rs6000/rs6000.c (rs6000_xcoff_section_type_flags):
10935 Increase code CSECT alignment to at least 32 bytes.
10936 * config/rs6000/xcoff.h (TEXT_SECTION_ASM_OP): Add 32 byte
10937 alignment designation.
10939 2021-06-29 Sergei Trofimovich <siarheit@google.com>
10941 * doc/generic.texi: Fix s/net yet/not yet/ typo.
10943 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
10945 PR tree-optimization/101254
10946 * range-op.cc (operator_minus::op1_op2_relation_effect): Check for
10947 wrapping/non-wrapping when setting the result range.
10949 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
10951 * value-query.cc (gimple_range_global): Allow phis.
10953 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
10955 * vr-values.c (vr_values::vrp_stmt_computes_nonzero): Use stmt.
10956 (simplify_using_ranges::op_with_boolean_value_range_p): Add a
10957 statement for location context.
10958 (check_for_binary_op_overflow): Ditto.
10959 (simplify_using_ranges::get_vr_for_comparison): Ditto.
10960 (simplify_using_ranges::compare_name_with_value): Ditto.
10961 (simplify_using_ranges::compare_names): Ditto.
10962 (vrp_evaluate_conditional_warnv_with_ops_using_ranges): Ditto.
10963 (simplify_using_ranges::simplify_truth_ops_using_ranges): Ditto.
10964 (simplify_using_ranges::simplify_min_or_max_using_ranges): Ditto.
10965 (simplify_using_ranges::simplify_internal_call_using_ranges): Ditto.
10966 (simplify_using_ranges::two_valued_val_range_p): Ditto.
10967 (simplify_using_ranges::simplify): Ditto.
10968 * vr-values.h: Adjust prototypes.
10970 2021-06-29 Uroš Bizjak <ubizjak@gmail.com>
10973 * config/i386/mmx.md (vec_addsubv2sf3): New insn pattern.
10975 2021-06-29 Julian Brown <julian@codesourcery.com>
10977 * config/gcn/gcn.c (gcn_init_libfuncs): New function.
10978 (TARGET_INIT_LIBFUNCS): Define target hook using above function.
10979 * config/gcn/gcn.h (UNITS_PER_WORD): Define to 8 for IN_LIBGCC2, 4
10981 (LIBGCC2_UNITS_PER_WORD, BITS_PER_WORD): Remove definitions.
10982 (MAX_FIXED_MODE_SIZE): Change to 128.
10984 2021-06-29 Julian Brown <julian@codesourcery.com>
10986 * config/gcn/gcn.md (UNSPEC_FLBIT_INT): New unspec constant.
10987 (s_mnemonic): Add clrsb.
10988 (gcn_flbit<mode>_int): Add insn pattern for SImode/DImode.
10989 (clrsb<mode>2): Add expander for SImode/DImode.
10991 2021-06-29 Julian Brown <julian@codesourcery.com>
10993 * config/gcn/gcn.md (<su>mulsidi3, <su>mulsidi3_reg, <su>mulsidi3_imm,
10994 muldi3): Add patterns.
10996 2021-06-29 Julian Brown <julian@codesourcery.com>
10998 * config/gcn/gcn.md (<su>mulsi3_highpart): Change to expander.
10999 (<su>mulsi3_highpart_reg, <su>mulsi3_highpart_imm): New patterns.
11001 2021-06-29 Julian Brown <julian@codesourcery.com>
11003 * config/gcn/gcn.md (mulsi3): Make s_mulk_i32 variant clobber SCC.
11005 2021-06-29 Joseph Myers <joseph@codesourcery.com>
11007 * btfout.c, ctfout.c: Include "memmodel.h".
11009 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
11011 * gcc.c (check_offload_target_name): Cast len argument to
11012 %q.*s to 'int'; avoid -Wstringop-truncation warning.
11014 2021-06-29 Richard Biener <rguenther@suse.de>
11016 * tree-vect-slp.c (vect_optimize_slp): Forward propagate
11017 to "any" permute nodes and relax "any" permute proapgation
11018 during iterative backward propagation.
11020 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
11023 * common.opt (-foffload=): Update description.
11024 (-foffload-options=): New.
11025 * doc/invoke.texi (C Language Options): Document
11026 -foffload and -foffload-options.
11027 * gcc.c (check_offload_target_name): New, split off from
11028 handle_foffload_option.
11029 (check_foffload_target_names): New.
11030 (handle_foffload_option): Handle -foffload=default.
11031 (driver_handle_option): Update for -foffload-options.
11032 * lto-opts.c (lto_write_options): Use -foffload-options
11033 instead of -foffload.
11034 * lto-wrapper.c (merge_and_complain, append_offload_options):
11036 * opts.c (common_handle_option): Likewise.
11038 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
11040 * doc/invoke.texi (C Language Options): Sort options
11041 alphabetically in optlist and also the description itself.
11042 Remove leftover -fallow-single-precision from and add missing
11043 -fgnu-tm to the optlist.
11045 2021-06-29 Richard Biener <rguenther@suse.de>
11047 * tree-vect-slp.c (slpg_vertex::visited): Remove.
11048 (vect_slp_perms_eq): Handle -1 permutes.
11049 (vect_optimize_slp): Rewrite permute propagation.
11051 2021-06-29 Jakub Jelinek <jakub@redhat.com>
11054 * match.pd ((intptr_t)x eq/ne CST to x eq/ne (typeof x) CST): Don't
11055 perform the optimization in GENERIC when sanitizing and x has a
11058 2021-06-29 Richard Biener <rguenther@suse.de>
11060 PR tree-optimization/101242
11061 * tree-vect-slp.c (vect_slp_build_vertices): Force-add
11062 PHIs with not represented initial values as leafs.
11064 2021-06-29 Jan-Benedict Glaw <jbglaw@getslash.de>
11066 * config/pdp11/pdp11.h (ASM_OUTPUT_SKIP): Fix signedness warning.
11067 * config/pdp11/pdp11.c (pdp11_asm_print_operand_punct_valid_p): Remove
11068 "register" keyword.
11069 (pdp11_initial_elimination_offset) Remove unused variable.
11070 (pdp11_cmp_length) Ditto.
11071 (pdp11_insn_cost): Ditto, and fix signedness warning.
11073 2021-06-29 David Edelsohn <dje.gcc@gmail.com>
11075 * btfout.c: Include tm_p.h.
11078 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
11080 * config/bpf/bpf.c (bpf_expand_prologue): Do not mark insns as
11082 (bpf_expand_epilogue): Likewise.
11083 * config/bpf/bpf.h (DWARF2_FRAME_INFO): Define to 0.
11084 Do not define DBX_DEBUGGING_INFO.
11086 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
11088 * doc/invoke.texi: Document the CTF and BTF debug info options.
11090 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
11091 David Faust <david.faust@oracle.com>
11092 Jose E. Marchesi <jose.marchesi@oracle.com>
11093 Weimin Pan <weimin.pan@oracle.com>
11095 * Makefile.in: Add ctfc.*, ctfout.c and btfout.c files to
11096 GTFILES. Add new object files.
11097 * common.opt: Add CTF and BTF debug info options.
11098 * btfout.c: New file.
11099 * ctfc.c: Likewise.
11100 * ctfc.h: Likewise.
11101 * ctfout.c: Likewise.
11102 * dwarf2ctf.c: Likewise.
11103 * dwarf2ctf.h: Likewise.
11104 * dwarf2cfi.c (dwarf2out_do_frame): Acknowledge CTF_DEBUG and
11106 * dwarf2out.c (dwarf2out_source_line): Likewise.
11107 (dwarf2out_finish): Skip emitting DWARF if CTF or BTF are to
11109 (debug_format_do_cu): New function.
11110 (dwarf2out_early_finish): Traverse DIEs and emit CTF/BTF for
11112 Include dwarf2ctf.c.
11113 * final.c (dwarf2_debug_info_emitted_p): Acknowledge DWARF-based debug
11115 * flag-types.h (enum debug_info_type): Add CTF_DEBUG and BTF_DEBUG.
11116 (CTF_DEBUG): New bitmask.
11117 (BTF_DEBUG): Likewise.
11118 (enum ctf_debug_info_levels): New enum.
11119 * gengtype.c (open_base_files): Handle ctfc.h.
11120 (main): Handle uint32_t type.
11121 * flags.h (btf_debuginfo_p): New definition.
11122 (dwarf_based_debuginfo_p): Likewise.
11123 * opts.c (debug_type_names): Add entries for CTF and BTF.
11124 (btf_debuginfo_p): New function.
11125 (dwarf_based_debuginfo_p): Likewise.
11126 (common_handle_option): Handle -gctfN and -gbtf options.
11127 (set_debug_level): Set CTF_DEBUG, BTF_DEBUG whenever appropriate.
11128 * toplev.c (process_options): Inform the user and ignore -gctfLEVEL if
11131 2021-06-28 Jose E. Marchesi <jose.marchesi@oracle.com>
11133 * dwarf2out.c (AT_class): Function is no longer static.
11134 (AT_int): Likewise.
11135 (AT_unsigned): Likewise.
11136 (AT_loc): Likewise.
11137 (get_AT): Likewise.
11138 (get_AT_string): Likewise.
11139 (get_AT_flag): Likewise.
11140 (get_AT_unsigned): Likewise.
11141 (get_AT_ref): Likewise.
11142 (new_die_raw): Likewise.
11143 (lookup_decl_die): Likewise.
11144 (base_type_die): Likewise.
11145 (add_name_attribute): Likewise.
11146 (add_AT_int): Likewise.
11147 (add_AT_unsigned): Likewise.
11148 (add_AT_loc): Likewise.
11149 (dw_get_die_tag): New function.
11150 (dw_get_die_child): Likewise.
11151 (dw_get_die_sib): Likewise.
11152 (struct dwarf_file_data): Move from here to dwarf2out.h
11153 (struct dw_attr_struct): Likewise.
11154 * dwarf2out.h: Analogous changes.
11156 2021-06-28 Martin Jambor <mjambor@suse.cz>
11159 * ipa-param-manipulation.h (class ipa_param_body_adjustments): New
11160 members m_dead_stmts and m_dead_ssas.
11161 * ipa-param-manipulation.c
11162 (ipa_param_body_adjustments::mark_dead_statements): New function.
11163 (ipa_param_body_adjustments::common_initialization): Call it on
11164 all removed but not split parameters.
11165 (ipa_param_body_adjustments::ipa_param_body_adjustments): Initialize
11167 (ipa_param_body_adjustments::modify_call_stmt): Remove arguments that
11169 * tree-inline.c (remap_gimple_stmt): Do not copy dead statements, reset
11170 dead debug statements.
11171 (copy_phis_for_bb): Do not copy dead PHI nodes.
11173 2021-06-28 Martin Jambor <mjambor@suse.cz>
11176 * symtab-clones.h (clone_info): Removed member param_adjustments.
11177 * ipa-param-manipulation.h: Adjust initial comment to reflect how we
11178 deal with pass-through splits now.
11179 (ipa_param_performed_split): Removed.
11180 (ipa_param_adjustments::modify_call): Adjusted parameters.
11181 (class ipa_param_body_adjustments): Adjusted parameters of
11182 register_replacement, modify_gimple_stmt and modify_call_stmt.
11183 (ipa_verify_edge_has_no_modifications): Declare.
11184 (ipa_edge_modifications_finalize): Declare.
11185 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Remove
11186 performed_splits processing, pas only edge to padjs->modify_call,
11187 check that call arguments were not modified if they should not have
11189 * cgraphclones.c (cgraph_node::create_clone): Do not copy performed
11191 * ipa-param-manipulation.c (struct pass_through_split_map): New type.
11192 (ipa_edge_modification_info): Likewise.
11193 (ipa_edge_modification_sum): Likewise.
11194 (ipa_edge_modifications): New edge summary.
11195 (ipa_verify_edge_has_no_modifications): New function.
11196 (transitive_split_p): Removed.
11197 (transitive_split_map): Likewise.
11198 (init_transitive_splits): Likewise.
11199 (ipa_param_adjustments::modify_call): Adjusted to use the new edge
11200 summary instead of performed_splits.
11201 (ipa_param_body_adjustments::register_replacement): Drop dummy
11202 parameter, set base_index of the created ipa_param_body_replacement.
11203 (phi_arg_will_live_p): New function.
11204 (ipa_param_body_adjustments::common_initialization): Do not create
11205 IPA_SRA dummy decls.
11206 (simple_tree_swap_info): Removed.
11207 (remap_split_decl_to_dummy): Likewise.
11208 (record_argument_state_1): New function.
11209 (record_argument_state): Likewise.
11210 (ipa_param_body_adjustments::modify_call_stmt): New parameter
11211 orig_stmt. Do not work with dummy decls, save necessary info about
11212 changes to ipa_edge_modifications.
11213 (ipa_param_body_adjustments::modify_gimple_stmt): New parameter
11214 orig_stmt, pass it to modify_call_stmt.
11215 (ipa_param_body_adjustments::modify_cfun_body): Adjust call to
11216 modify_gimple_stmt.
11217 (ipa_edge_modifications_finalize): New function.
11218 * tree-inline.c (remap_gimple_stmt): Pass original statement to
11219 modify_gimple_stmt.
11220 (copy_phis_for_bb): Do not copy dead PHI nodes.
11221 (expand_call_inline): Do not remap performed_splits.
11222 (update_clone_info): Likewise.
11223 * toplev.c: Include ipa-param-manipulation.h.
11224 (toplev::finalize): Call ipa_edge_modifications_finalize.
11226 2021-06-28 Andrew Pinski <apinski@marvell.com>
11228 * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Duplicate range
11229 info if we're the only things setting the target PHI.
11230 (value_replacement): Don't duplicate range here.
11231 (minmax_replacement): Likewise.
11233 2021-06-28 Richard Biener <rguenther@suse.de>
11235 PR tree-optimization/101229
11236 * gimple-walk.c (gimple_walk_op): Handle PHIs.
11238 2021-06-28 Martin Liska <mliska@suse.cz>
11240 * config/v850/v850.c (construct_dispose_instruction): Allocate
11242 (construct_prepare_instruction): Likewise.
11244 2021-06-28 Martin Liska <mliska@suse.cz>
11246 * config/v850/v850.c (v850_option_override): Build default
11248 (v850_can_inline_p): New. Allow MASK_PROLOG_FUNCTION to be
11249 ignored for inlining.
11250 (TARGET_CAN_INLINE_P): New.
11252 2021-06-28 Richard Biener <rguenther@suse.de>
11254 PR tree-optimization/101207
11255 * tree-vect-slp.c (vect_optimize_slp): Do BB reduction
11256 permute eliding for load permutations properly.
11258 2021-06-28 Richard Biener <rguenther@suse.de>
11260 PR tree-optimization/101173
11261 * gimple-loop-interchange.cc
11262 (tree_loop_interchange::valid_data_dependences): Disallow outer
11263 loop dependence distance of zero.
11265 2021-06-28 liuhongt <hongtao.liu@intel.com>
11268 * config/i386/sse.md (*avx_cmp<mode>3_lt): New
11269 define_insn_and_split.
11270 (*avx_cmp<mode>3_ltint): Ditto.
11271 (*avx2_pcmp<mode>3_3): Ditto.
11272 (*avx2_pcmp<mode>3_4): Ditto.
11273 (*avx2_pcmp<mode>3_5): Ditto.
11275 2021-06-28 liuhongt <hongtao.liu@intel.com>
11277 * config/i386/i386-builtin.def (IX86_BUILTIN_BLENDVPD256,
11278 IX86_BUILTIN_BLENDVPS256, IX86_BUILTIN_PBLENDVB256,
11279 IX86_BUILTIN_BLENDVPD, IX86_BUILTIN_BLENDVPS,
11280 IX86_BUILTIN_PBLENDVB128): Replace icode with
11282 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold blendv
11284 * config/i386/sse.md (*<sse4_1_avx2>_pblendvb_lt_subreg_not):
11285 New pre_reload splitter.
11287 2021-06-27 Andrew Pinski <apinski@marvell.com>
11289 PR middle-end/101230
11290 * fold-const.c (fold_ternary_loc): Check
11291 the return value of invert_tree_comparison.
11293 2021-06-27 David Edelsohn <dje.gcc@gmail.com>
11295 * config.gcc: Add SPDX License Identifier.
11296 (powerpc-ibm-aix789): Default to aix73.h.
11297 (powerpc-ibm-aix7.2.*.*): New stanza.
11298 * config/rs6000/aix72.h: Add SPDX License Identifier.
11299 * config/rs6000/aix73.h: New file.
11301 2021-06-26 Jason Merrill <jason@redhat.com>
11303 * except.c: #include "dwarf2.h" instead of "dwarf2out.h".
11305 2021-06-26 Andrew Pinski <apinski@marvell.com>
11307 * genmatch.c (lower_cond): Copy for_subst_vec
11308 for the simplify also.
11309 (lower): Swap the order for lower_for and lower_cond.
11311 2021-06-26 Andrew Pinski <apinski@marvell.com>
11313 * tree-ssa-phiopt.c (match_simplify_replacement): Reset
11314 flow senatitive info on the moved ssa set.
11316 2021-06-26 Andrew Pinski <apinski@marvell.com>
11318 * fold-const.c (fold_cond_expr_with_comparison):
11319 Exand arg0 into comp_code, arg00, and arg01.
11320 (fold_ternary_loc): Use invert_tree_comparison
11321 instead of fold_invert_truthvalue for the case
11322 where we have A CMP B ? C : A.
11324 2021-06-25 Martin Sebor <msebor@redhat.com>
11326 PR middle-end/101216
11327 * calls.c (maybe_warn_rdwr_sizes): Use the no_warning constant.
11329 2021-06-25 Jeff Law <jeffreyalaw@gmail.com>
11331 * config/h8300/h8300.c (select_cc_mode): Handle ASHIFTRT and LSHIFTRT.
11333 2021-06-25 Richard Biener <rguenther@suse.de>
11335 PR tree-optimization/101202
11336 * tree-vect-slp.c (vect_optimize_slp): Explicitely handle
11339 2021-06-25 Richard Biener <rguenther@suse.de>
11341 * tree-vect-slp-patterns.c (addsub_pattern::build): Copy
11342 STMT_VINFO_REDUC_DEF from the original representative.
11344 2021-06-25 Martin Sebor <msebor@redhat.com>
11346 * builtins.c (warn_string_no_nul): Replace uses of TREE_NO_WARNING,
11347 gimple_no_warning_p and gimple_set_no_warning with
11348 warning_suppressed_p, and suppress_warning.
11350 (maybe_warn_for_bound): Same.
11351 (warn_for_access): Same.
11352 (check_access): Same.
11353 (expand_builtin_strncmp): Same.
11354 (fold_builtin_varargs): Same.
11355 * calls.c (maybe_warn_nonstring_arg): Same.
11356 (maybe_warn_rdwr_sizes): Same.
11357 * cfgexpand.c (expand_call_stmt): Same.
11358 * cgraphunit.c (check_global_declaration): Same.
11359 * fold-const.c (fold_undefer_overflow_warnings): Same.
11360 (fold_truth_not_expr): Same.
11361 (fold_unary_loc): Same.
11362 (fold_checksum_tree): Same.
11363 * gimple-array-bounds.cc (array_bounds_checker::check_array_ref): Same.
11364 (array_bounds_checker::check_mem_ref): Same.
11365 (array_bounds_checker::check_addr_expr): Same.
11366 (array_bounds_checker::check_array_bounds): Same.
11367 * gimple-expr.c (copy_var_decl): Same.
11368 * gimple-fold.c (gimple_fold_builtin_strcpy): Same.
11369 (gimple_fold_builtin_strncat): Same.
11370 (gimple_fold_builtin_stxcpy_chk): Same.
11371 (gimple_fold_builtin_stpcpy): Same.
11372 (gimple_fold_builtin_sprintf): Same.
11373 (fold_stmt_1): Same.
11374 * gimple-ssa-isolate-paths.c (diag_returned_locals): Same.
11375 * gimple-ssa-nonnull-compare.c (do_warn_nonnull_compare): Same.
11376 * gimple-ssa-sprintf.c (handle_printf_call): Same.
11377 * gimple-ssa-store-merging.c (imm_store_chain_info::output_merged_store): Same.
11378 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
11379 * gimple-ssa-warn-restrict.h: Adjust declarations.
11380 (maybe_diag_access_bounds): Replace uses of TREE_NO_WARNING,
11381 gimple_no_warning_p and gimple_set_no_warning with
11382 warning_suppressed_p, and suppress_warning.
11383 (check_call): Same.
11384 (check_bounds_or_overlap): Same.
11385 * gimple.c (gimple_build_call_from_tree): Same.
11386 * gimplify.c (gimplify_return_expr): Same.
11387 (gimplify_cond_expr): Same.
11388 (gimplify_modify_expr_complex_part): Same.
11389 (gimplify_modify_expr): Same.
11390 (gimple_push_cleanup): Same.
11391 (gimplify_expr): Same.
11392 * omp-expand.c (expand_omp_for_generic): Same.
11393 (expand_omp_taskloop_for_outer): Same.
11394 * omp-low.c (lower_rec_input_clauses): Same.
11395 (lower_lastprivate_clauses): Same.
11396 (lower_send_clauses): Same.
11397 (lower_omp_target): Same.
11398 * tree-cfg.c (pass_warn_function_return::execute): Same.
11399 * tree-complex.c (create_one_component_var): Same.
11400 * tree-inline.c (remap_gimple_op_r): Same.
11401 (copy_tree_body_r): Same.
11402 (declare_return_variable): Same.
11403 (expand_call_inline): Same.
11404 * tree-nested.c (lookup_field_for_decl): Same.
11405 * tree-sra.c (create_access_replacement): Same.
11406 (generate_subtree_copies): Same.
11407 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Same.
11408 * tree-ssa-forwprop.c (combine_cond_expr_cond): Same.
11409 * tree-ssa-loop-ch.c (ch_base::copy_headers): Same.
11410 * tree-ssa-loop-im.c (execute_sm): Same.
11411 * tree-ssa-phiopt.c (cond_store_replacement): Same.
11412 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
11413 (handle_builtin_strcpy): Same.
11414 (maybe_diag_stxncpy_trunc): Same.
11415 (handle_builtin_stxncpy_strncat): Same.
11416 (handle_builtin_strcat): Same.
11417 * tree-ssa-uninit.c (get_no_uninit_warning): Same.
11418 (set_no_uninit_warning): Same.
11419 (uninit_undefined_value_p): Same.
11420 (warn_uninit): Same.
11421 (maybe_warn_operand): Same.
11422 * tree-vrp.c (compare_values_warnv): Same.
11423 * vr-values.c (vr_values::extract_range_for_var_from_comparison_expr): Same.
11424 (test_for_singularity): Same.
11425 * gimple.h (warning_suppressed_p): New function.
11426 (suppress_warning): Same.
11427 (copy_no_warning): Same.
11428 (gimple_set_block): Call gimple_set_location.
11429 (gimple_set_location): Call copy_warning.
11431 2021-06-25 Martin Sebor <msebor@redhat.com>
11433 * tree.h (warning_suppressed_at, copy_warning,
11434 warning_suppressed_p, suppress_warning): New functions.
11436 2021-06-25 Martin Sebor <msebor@redhat.com>
11438 * Makefile.in (OBJS-libcommon): Add diagnostic-spec.o.
11439 * gengtype.c (open_base_files): Add diagnostic-spec.h.
11440 * diagnostic-spec.c: New file.
11441 * diagnostic-spec.h: New file.
11442 * tree.h (no_warning, all_warnings, suppress_warning_at): New
11444 * warning-control.cc: New file.
11446 2021-06-25 liuhongt <hongtao.liu@intel.com>
11449 * config/i386/i386.c (x86_order_regs_for_local_alloc):
11452 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
11454 PR tree-optimization/101189
11455 * gimple-range-fold.cc (fold_using_range::range_of_range_op): Pass
11456 LHS range of condition to postfold routine.
11457 (fold_using_range::postfold_gcond_edges): Only process the TRUE or
11458 FALSE edge if the LHS range supports it being taken.
11459 * gimple-range-fold.h (postfold_gcond_edges): Add range parameter.
11461 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
11463 * value-relation.cc (equiv_oracle::dump): Do not dump NULL blocks.
11464 (relation_oracle::find_relation_block): Check correct bitmap.
11465 (relation_oracle::dump): Do not dump NULL blocks.
11467 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
11469 * gimple-range-cache.cc (ranger_cache::propagate_cache): Call
11470 range_on_edge instead of manually calculating.
11472 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
11474 * range-op.cc: Fix comment.
11476 2021-06-24 Uroš Bizjak <ubizjak@gmail.com>
11479 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
11480 Handle V8QI and V4HI modes.
11481 * config/i386/mmx.md (sse4_1_<any_extend:code>v4qiv4hi2):
11483 (sse4_1_<any_extend:code>v4qiv4hi2): Ditto.
11484 (mmxpackmode): New mode attribute.
11485 (vec_pack_trunc_<mmxpackmode:mode>): New expander.
11486 (mmxunpackmode): New mode attribute.
11487 (vec_unpacks_lo_<mmxunpackmode:mode>): New expander.
11488 (vec_unpacks_hi_<mmxunpackmode:mode>): Ditto.
11489 (vec_unpacku_lo_<mmxunpackmode:mode>): Ditto.
11490 (vec_unpacku_hi_<mmxunpackmode:mode>): Ditto.
11491 * config/i386/i386.md (extsuffix): Move from ...
11492 * config/i386/sse.md: ... here.
11494 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
11496 * dwarf2out.c (dwarf2out_assembly_start): Emit .file 0 marker here..
11497 (dwarf2out_finish): ...instead of here.
11499 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
11501 * configure.ac (--gdwarf-5 option): Use objdump instead of readelf.
11502 (working --gdwarf-4/--gdwarf-5 for all sources): Likewise.
11503 (--gdwarf-4 not refusing generated .debug_line): Adjust for Windows.
11504 * configure: Regenerate.
11506 2021-06-24 Richard Biener <rguenther@suse.de>
11508 * config/i386/sse.md (vec_addsubv4df3, vec_addsubv2df3,
11509 vec_addsubv8sf3, vec_addsubv4sf3): Merge into ...
11510 (vec_addsub<mode>3): ... using a new addsub_cst mode attribute.
11512 2021-06-24 Richard Biener <rguenther@suse.de>
11514 * config/i386/sse.md (avx_addsubv4df3): Rename to
11516 (avx_addsubv8sf3): Rename to vec_addsubv8sf3.
11517 (sse3_addsubv2df3): Rename to vec_addsubv2df3.
11518 (sse3_addsubv4sf3): Rename to vec_addsubv4sf3.
11519 * config/i386/i386-builtin.def: Adjust.
11520 * internal-fn.def (VEC_ADDSUB): New internal optab fn.
11521 * optabs.def (vec_addsub_optab): New optab.
11522 * tree-vect-slp-patterns.c (class addsub_pattern): New.
11523 (slp_patterns): Add addsub_pattern.
11524 * tree-vect-slp.c (vect_optimize_slp): Disable propagation
11525 across CFN_VEC_ADDSUB.
11526 * tree-vectorizer.h (vect_pattern::vect_pattern): Make
11528 * doc/md.texi (vec_addsub<mode>3): Document.
11530 2021-06-24 Jakub Jelinek <jakub@redhat.com>
11532 PR middle-end/101170
11533 * df-scan.c (df_ref_record): For paradoxical big-endian SUBREGs
11534 where regno + subreg_regno_offset wraps around use 0 as starting
11537 2021-06-24 Jakub Jelinek <jakub@redhat.com>
11539 PR middle-end/101172
11540 * stor-layout.c (finish_bitfield_representative): If nextf has
11541 error_mark_node type, set repr type to error_mark_node too.
11543 2021-06-24 Ilya Leoshkevich <iii@linux.ibm.com>
11545 * config/s390/s390.c (s390_function_profiler): Ignore labelno
11547 * config/s390/s390.h (NO_PROFILE_COUNTERS): Define.
11549 2021-06-24 Richard Biener <rguenther@suse.de>
11551 * tree-vect-slp.c (vect_optimize_slp): Do not propagate
11552 across operations that have different semantics on different
11555 2021-06-24 Jakub Jelinek <jakub@redhat.com>
11557 * tree.h (OMP_CLAUSE_MAP_IN_REDUCTION): Document meaning for OpenMP.
11558 * gimplify.c (gimplify_scan_omp_clauses): For OpenMP map clauses
11559 with OMP_CLAUSE_MAP_IN_REDUCTION flag partially defer gimplification
11560 of non-decl OMP_CLAUSE_DECL. For OMP_CLAUSE_IN_REDUCTION on
11561 OMP_TARGET user outer_ctx instead of ctx for placeholders and
11562 initializer/combiner gimplification.
11563 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_MAP_IN_REDUCTION
11564 on target constructs.
11565 (lower_rec_input_clauses): Likewise.
11566 (lower_omp_target): Likewise.
11567 * omp-expand.c (expand_omp_target): Temporarily ignore nowait clause
11568 on target if in_reduction is present.
11570 2021-06-24 Kewen Lin <linkw@linux.ibm.com>
11572 * tree-predcom.c (class pcom_worker): New class.
11573 (release_chain): Renamed to...
11574 (pcom_worker::release_chain): ...this.
11575 (release_chains): Renamed to...
11576 (pcom_worker::release_chains): ...this.
11577 (aff_combination_dr_offset): Renamed to...
11578 (pcom_worker::aff_combination_dr_offset): ...this.
11579 (determine_offset): Renamed to...
11580 (pcom_worker::determine_offset): ...this.
11581 (class comp_ptrs): New class.
11582 (split_data_refs_to_components): Renamed to...
11583 (pcom_worker::split_data_refs_to_components): ...this,
11584 and update with class comp_ptrs.
11585 (suitable_component_p): Renamed to...
11586 (pcom_worker::suitable_component_p): ...this.
11587 (filter_suitable_components): Renamed to...
11588 (pcom_worker::filter_suitable_components): ...this.
11589 (valid_initializer_p): Renamed to...
11590 (pcom_worker::valid_initializer_p): ...this.
11591 (find_looparound_phi): Renamed to...
11592 (pcom_worker::find_looparound_phi): ...this.
11593 (add_looparound_copies): Renamed to...
11594 (pcom_worker::add_looparound_copies): ...this.
11595 (determine_roots_comp): Renamed to...
11596 (pcom_worker::determine_roots_comp): ...this.
11597 (determine_roots): Renamed to...
11598 (pcom_worker::determine_roots): ...this.
11599 (single_nonlooparound_use): Renamed to...
11600 (pcom_worker::single_nonlooparound_use): ...this.
11601 (remove_stmt): Renamed to...
11602 (pcom_worker::remove_stmt): ...this.
11603 (execute_pred_commoning_chain): Renamed to...
11604 (pcom_worker::execute_pred_commoning_chain): ...this.
11605 (execute_pred_commoning): Renamed to...
11606 (pcom_worker::execute_pred_commoning): ...this.
11607 (struct epcc_data): New member worker.
11608 (execute_pred_commoning_cbck): Call execute_pred_commoning
11609 with pcom_worker pointer.
11610 (find_use_stmt): Renamed to...
11611 (pcom_worker::find_use_stmt): ...this.
11612 (find_associative_operation_root): Renamed to...
11613 (pcom_worker::find_associative_operation_root): ...this.
11614 (find_common_use_stmt): Renamed to...
11615 (pcom_worker::find_common_use_stmt): ...this.
11616 (combinable_refs_p): Renamed to...
11617 (pcom_worker::combinable_refs_p): ...this.
11618 (reassociate_to_the_same_stmt): Renamed to...
11619 (pcom_worker::reassociate_to_the_same_stmt): ...this.
11620 (stmt_combining_refs): Renamed to...
11621 (pcom_worker::stmt_combining_refs): ...this.
11622 (combine_chains): Renamed to...
11623 (pcom_worker::combine_chains): ...this.
11624 (try_combine_chains): Renamed to...
11625 (pcom_worker::try_combine_chains): ...this.
11626 (prepare_initializers_chain): Renamed to...
11627 (pcom_worker::prepare_initializers_chain): ...this.
11628 (prepare_initializers): Renamed to...
11629 (pcom_worker::prepare_initializers): ...this.
11630 (prepare_finalizers_chain): Renamed to...
11631 (pcom_worker::prepare_finalizers_chain): ...this.
11632 (prepare_finalizers): Renamed to...
11633 (pcom_worker::prepare_finalizers): ...this.
11634 (tree_predictive_commoning_loop): Renamed to...
11635 (pcom_worker::tree_predictive_commoning_loop): ...this, adjust
11636 some calls and remove some cleanup code.
11637 (tree_predictive_commoning): Adjusted to use pcom_worker instance.
11638 (static variable looparound_phis): Remove.
11639 (static variable name_expansions): Remove.
11641 2021-06-24 Richard Biener <rguenther@suse.de>
11643 * tree-vect-slp.c (slpg_vertex): New struct.
11644 (vect_slp_build_vertices): Adjust.
11645 (vect_optimize_slp): Likewise. Maintain an outgoing permute
11646 and a materialized one.
11648 2021-06-24 Richard Biener <rguenther@suse.de>
11650 PR tree-optimization/101105
11651 * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
11652 Only ignore steps when they are equal or scalar order is preserved.
11654 2021-06-24 liuhongt <hongtao.liu@intel.com>
11657 * config/i386/i386-expand.c (ix86_expand_vec_interleave):
11658 Adjust comments for ix86_expand_vecop_qihi2.
11659 (ix86_expand_vecmul_qihi): Renamed to ..
11660 (ix86_expand_vecop_qihi2): Adjust function prototype to
11661 support shift operation, add static to definition.
11662 (ix86_expand_vec_shift_qihi_constant): Add static to definition.
11663 (ix86_expand_vecop_qihi): Call ix86_expand_vecop_qihi2 and
11664 ix86_expand_vec_shift_qihi_constant.
11665 * config/i386/i386-protos.h (ix86_expand_vecmul_qihi): Deleted.
11666 (ix86_expand_vec_shift_qihi_constant): Deleted.
11667 * config/i386/sse.md (VI12_256_512_AVX512VL): New mode
11669 (mulv8qi3): Call ix86_expand_vecop_qihi directly, add
11670 condition TARGET_64BIT.
11671 (mul<mode>3): Ditto.
11672 (<insn><mode>3): Ditto.
11673 (vlshr<mode>3): Extend to support avx512 vlshr.
11674 (v<insn><mode>3): New expander for
11676 (v<insn>v8qi3): Ditto.
11677 (vashrv8hi3<mask_name>): Renamed to ..
11678 (vashr<mode>3): And extend to support V16QImode for avx512.
11679 (vashrv16qi3): Deleted.
11680 (vashrv2di3<mask_name>): Extend expander to support avx512
11683 2021-06-23 Dimitar Dimitrov <dimitar@dinux.eu>
11685 * doc/lto.texi (Design Overview): Update that slim objects are
11688 2021-06-23 Aaron Sawdey <acsawdey@linux.ibm.com>
11690 * config/rs6000/rs6000-cpus.def: Take OPTION_MASK_PCREL_OPT out
11691 of OTHER_POWER10_MASKS so it will not be enabled by default.
11693 2021-06-23 Richard Biener <rguenther@suse.de>
11694 Martin Jambor <mjambor@suse.cz>
11696 * tree-inline.c (setup_one_parameter): Set TREE_READONLY of the
11697 param replacement unconditionally. Adjust comment.
11699 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
11701 * Makefile.in (OBJS): Add gimple-range-fold.o
11702 * gimple-range-fold.cc: New.
11703 * gimple-range-fold.h: New.
11704 * gimple-range-gori.cc (gimple_range_calc_op1): Move to here.
11705 (gimple_range_calc_op2): Ditto.
11706 * gimple-range-gori.h: Move prototypes to here.
11707 * gimple-range.cc: Adjust include files.
11708 (fur_source:fur_source): Relocate to gimple-range-fold.cc.
11709 (fur_source::get_operand): Ditto.
11710 (fur_source::get_phi_operand): Ditto.
11711 (fur_source::query_relation): Ditto.
11712 (fur_source::register_relation): Ditto.
11713 (class fur_edge): Ditto.
11714 (fur_edge::fur_edge): Ditto.
11715 (fur_edge::get_operand): Ditto.
11716 (fur_edge::get_phi_operand): Ditto.
11717 (fur_stmt::fur_stmt): Ditto.
11718 (fur_stmt::get_operand): Ditto.
11719 (fur_stmt::get_phi_operand): Ditto.
11720 (fur_stmt::query_relation): Ditto.
11721 (class fur_depend): Relocate to gimple-range-fold.h.
11722 (fur_depend::fur_depend): Relocate to gimple-range-fold.cc.
11723 (fur_depend::register_relation): Ditto.
11724 (fur_depend::register_relation): Ditto.
11725 (class fur_list): Ditto.
11726 (fur_list::fur_list): Ditto.
11727 (fur_list::get_operand): Ditto.
11728 (fur_list::get_phi_operand): Ditto.
11729 (fold_range): Ditto.
11730 (adjust_pointer_diff_expr): Ditto.
11731 (gimple_range_adjustment): Ditto.
11732 (gimple_range_base_of_assignment): Ditto.
11733 (gimple_range_operand1): Ditto.
11734 (gimple_range_operand2): Ditto.
11735 (gimple_range_calc_op1): Relocate to gimple-range-gori.cc.
11736 (gimple_range_calc_op2): Ditto.
11737 (fold_using_range::fold_stmt): Relocate to gimple-range-fold.cc.
11738 (fold_using_range::range_of_range_op): Ditto.
11739 (fold_using_range::range_of_address): Ditto.
11740 (fold_using_range::range_of_phi): Ditto.
11741 (fold_using_range::range_of_call): Ditto.
11742 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
11743 (fold_using_range::range_of_builtin_call): Ditto.
11744 (fold_using_range::range_of_cond_expr): Ditto.
11745 (fold_using_range::range_of_ssa_name_with_loop_info): Ditto.
11746 (fold_using_range::relation_fold_and_or): Ditto.
11747 (fold_using_range::postfold_gcond_edges): Ditto.
11748 * gimple-range.h: Add gimple-range-fold.h to include files. Change
11749 GIMPLE_RANGE_STMT_H to GIMPLE_RANGE_H.
11750 (gimple_range_handler): Relocate to gimple-range-fold.h.
11751 (gimple_range_ssa_p): Ditto.
11752 (range_compatible_p): Ditto.
11753 (class fur_source): Ditto.
11754 (class fur_stmt): Ditto.
11755 (class fold_using_range): Ditto.
11756 (gimple_range_calc_op1): Relocate to gimple-range-gori.h
11757 (gimple_range_calc_op2): Ditto.
11759 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
11761 PR tree-optimization/101148
11762 PR tree-optimization/101014
11763 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
11764 (ranger_cache::~ranger_cache): Adjust.
11765 (ranger_cache::block_range): Check if propagation disallowed.
11766 (ranger_cache::propagate_cache): Disallow propagation if new value
11767 can't be stored properly.
11768 * gimple-range-cache.h (ranger_cache::m_propfail): New member.
11770 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
11772 * gimple-range-cache.cc (class ssa_block_ranges): Adjust prototype.
11773 (sbr_vector::set_bb_range): Return true.
11774 (class sbr_sparse_bitmap): Adjust.
11775 (sbr_sparse_bitmap::set_bb_range): Return value.
11776 (block_range_cache::set_bb_range): Return value.
11777 (ranger_cache::propagate_cache): Use return value to print msg.
11778 * gimple-range-cache.h (class block_range_cache): Adjust.
11780 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
11782 * gimple-range.cc (dump_bb): Use range_on_edge from the cache.
11784 2021-06-23 Jeff Law <jeffreyalaw@gmail.com>
11786 * config/h8300/logical.md (<code><mode>3<ccnz>): Use <cczn>
11787 so this pattern can be used for test/compare removal. Pass
11788 current insn to compute_logical_op_length and output_logical_op.
11789 * config/h8300/h8300.c (compute_logical_op_cc): Remove.
11790 (h8300_and_costs): Add argument to compute_logical_op_length.
11791 (output_logical_op): Add new argument. Use it to determine if the
11792 condition codes are used and adjust the output accordingly.
11793 (compute_logical_op_length): Add new argument and update length
11794 computations when condition codes are used.
11795 * config/h8300/h8300-protos.h (compute_logical_op_length): Update
11797 (output_logical_op): Likewise.
11799 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
11802 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
11803 Handle 64bit modes for TARGET_XOP. Use indirect gen_* functions.
11804 * config/i386/mmx.md (mmx_ppermv64): New insn pattern.
11805 * config/i386/i386.md (unspec): Move UNSPEC_XOP_PERMUTE from ...
11806 * config/i386/sse.md (unspec): ... here.
11808 2021-06-23 Martin Liska <mliska@suse.cz>
11811 * optc-save-gen.awk: Put back arm_fp16_format to
11814 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
11817 * config/i386/i386.md (bsr_rex64): Add zero-flag setting RTX.
11820 (clz<mode>2): Update RTX pattern for additions.
11822 2021-06-23 Jakub Jelinek <jakub@redhat.com>
11824 PR middle-end/101167
11825 * omp-low.c (lower_omp_regimplify_p): Regimplify also PARM_DECLs
11826 and RESULT_DECLs that have DECL_HAS_VALUE_EXPR_P set.
11828 2021-06-22 Sergei Trofimovich <siarheit@google.com>
11830 * doc/rtl.texi: drop unbalanced parenthesis.
11832 2021-06-22 Richard Biener <rguenther@suse.de>
11834 PR middle-end/101156
11835 * gimplify.c (gimplify_expr): Remove premature incorrect
11838 2021-06-22 Jakub Jelinek <jakub@redhat.com>
11840 PR tree-optimization/101159
11841 * tree-vect-patterns.c (vect_recog_popcount_pattern): Fix some
11844 2021-06-22 Jakub Jelinek <jakub@redhat.com>
11846 PR middle-end/101160
11847 * function.c (assign_parms): For decl_result with TYPE_EMPTY_P type
11848 clear crtl->return_rtx instead of keeping it referencing a pseudo.
11850 2021-06-22 Jakub Jelinek <jakub@redhat.com>
11851 Andrew Pinski <apinski@marvell.com>
11853 PR tree-optimization/101162
11854 * fold-const.c (range_check_type): Handle OFFSET_TYPE like pointer
11857 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11859 * range-op.cc (range_relational_tests): New.
11860 (range_op_tests): Call range_relational_tests.
11862 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11864 * range-op.cc (operator_cast::lhs_op1_relation): New.
11865 (operator_identity::lhs_op1_relation): Mew.
11867 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11869 * range-op.cc (operator_minus::op1_op2_relation_effect): New.
11871 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11873 * range-op.cc (operator_plus::lhs_op1_relation): New.
11874 (operator_plus::lhs_op2_relation): New.
11876 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11878 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a
11879 relation_oracle if dominators exist.
11880 (ranger_cache::~ranger_cache): Dispose of oracle.
11881 (ranger_cache::dump_bb): Dump oracle.
11882 * gimple-range.cc (fur_source::fur_source): New.
11883 (fur_source::get_operand): Use mmeber query.
11884 (fur_source::get_phi_operand): Use member_query.
11885 (fur_source::query_relation): New.
11886 (fur_source::register_dependency): Delete.
11887 (fur_source::register_relation): New.
11888 (fur_edge::fur_edge): Adjust.
11889 (fur_edge::get_phi_operand): Fix comment.
11890 (fur_edge::query): Delete.
11891 (fur_stmt::fur_stmt): Adjust.
11892 (fur_stmt::query): Delete.
11893 (fur_depend::fur_depend): Adjust.
11894 (fur_depend::register_relation): New.
11895 (fur_depend::register_relation): New.
11896 (fur_list::fur_list): Adjust.
11897 (fur_list::get_operand): Use member query.
11898 (fold_using_range::range_of_range_op): Process and query relations.
11899 (fold_using_range::range_of_address): Adjust dependency call.
11900 (fold_using_range::range_of_phi): Ditto.
11901 (gimple_ranger::gimple_ranger): New. Use ranger_ache oracle.
11902 (fold_using_range::relation_fold_and_or): New.
11903 (fold_using_range::postfold_gcond_edges): New.
11904 * gimple-range.h (class gimple_ranger): Adjust.
11905 (class fur_source): Adjust members.
11906 (class fur_stmt): Ditto.
11907 (class fold_using_range): Ditto.
11909 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11911 * range-op.cc (range_operator::wi_fold): Apply relation effect.
11912 (range_operator::fold_range): Adjust and apply relation effect.
11913 (*::fold_range): Add relation parameters.
11914 (*::op1_range): Ditto.
11915 (*::op2_range): Ditto.
11916 (range_operator::lhs_op1_relation): New.
11917 (range_operator::lhs_op2_relation): New.
11918 (range_operator::op1_op2_relation): New.
11919 (range_operator::op1_op2_relation_effect): New.
11920 (relop_early_resolve): New.
11921 (operator_equal::op1_op2_relation): New.
11922 (operator_equal::fold_range): Call relop_early_resolve.
11923 (operator_not_equal::op1_op2_relation): New.
11924 (operator_not_equal::fold_range): Call relop_early_resolve.
11925 (operator_lt::op1_op2_relation): New.
11926 (operator_lt::fold_range): Call relop_early_resolve.
11927 (operator_le::op1_op2_relation): New.
11928 (operator_le::fold_range): Call relop_early_resolve.
11929 (operator_gt::op1_op2_relation): New.
11930 (operator_gt::fold_range): Call relop_early_resolve.
11931 (operator_ge::op1_op2_relation): New.
11932 (operator_ge::fold_range): Call relop_early_resolve.
11933 * range-op.h (class range_operator): Adjust parameters and methods.
11935 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11937 * Makefile.in (OBJS): Add value-relation.o.
11938 * gimple-range.h: Adjust include files.
11939 * tree-data-ref.c: Adjust include file order.
11940 * value-query.cc (range_query::get_value_range): Default to no oracle.
11941 (range_query::query_relation): New.
11942 (range_query::query_relation): New.
11943 * value-query.h (class range_query): Adjust.
11944 * value-relation.cc: New.
11945 * value-relation.h: New.
11947 2021-06-22 Richard Biener <rguenther@suse.de>
11949 PR tree-optimization/101151
11950 * tree-ssa-sink.c (statement_sink_location): Expand irreducible
11953 2021-06-22 Jojo R <rjiejie@linux.alibaba.com>
11955 * config/riscv/riscv.c (thead_c906_tune_info): New.
11956 (riscv_tune_info_table): Use new tune.
11958 2021-06-22 Richard Biener <rguenther@suse.de>
11960 PR tree-optimization/101158
11961 * tree-vect-slp.c (vect_build_slp_tree_1): Move same operand
11962 checking after checking for matching operation.
11964 2021-06-22 Richard Biener <rguenther@suse.de>
11966 PR tree-optimization/101159
11967 * tree-vect-patterns.c (vect_recog_popcount_pattern): Add
11968 missing NULL vectype check.
11970 2021-06-22 Richard Biener <rguenther@suse.de>
11972 PR tree-optimization/101154
11973 * tree-vect-slp.c (vect_build_slp_tree_2): Fix out-of-bound access.
11975 2021-06-22 Jakub Jelinek <jakub@redhat.com>
11978 * config/i386/i386-protos.h (ix86_last_zero_store_uid): Declare.
11979 * config/i386/i386-expand.c (ix86_last_zero_store_uid): New variable.
11980 * config/i386/i386.c (ix86_expand_prologue): Clear it.
11981 * config/i386/i386.md (peephole2s for 1/2/4 stores of const0_rtx):
11982 Remove "" from match_operand. Emit new insns using emit_move_insn and
11983 set ix86_last_zero_store_uid to INSN_UID of the last store.
11984 Add peephole2s for 1/2/4 stores of const0_rtx following previous
11987 2021-06-22 Martin Liska <mliska@suse.cz>
11989 * auto-profile.c (AUTO_PROFILE_VERSION): Bump as string format
11992 2021-06-22 Martin Liska <mliska@suse.cz>
11994 * gcov-io.h: Remove padding entries.
11996 2021-06-22 liuhongt <hongtao.liu@intel.com>
11998 PR tree-optimization/97770
11999 * tree-vect-patterns.c (vect_recog_popcount_pattern):
12001 (vect_recog_func vect_vect_recog_func_ptrs): Add new pattern.
12003 2021-06-22 liuhongt <hongtao.liu@intel.com>
12006 * config/i386/i386-builtin.def (BDESC): Adjust builtin name.
12007 * config/i386/sse.md (<avx512>_expand<mode>_mask): Rename to ..
12008 (expand<mode>_mask): this ..
12009 (*expand<mode>_mask): New pre_reload splitter to transform
12010 v{,p}expand* to vmov* when mask is zero, all ones, or has all
12011 ones in it's lower part, otherwise still generate
12014 2021-06-22 liuhongt <hongtao.liu@intel.com>
12017 * config/i386/i386-expand.c
12018 (ix86_expand_special_args_builtin): Keep constm1_operand only
12019 if it satisfies insn's operand predicate.
12021 2021-06-21 Jason Merrill <jason@redhat.com>
12024 * df-scan.c (df_ref_record): Check that regno < endregno.
12025 * function.c (assign_parms, expand_function_end): Do nothing with a
12026 TYPE_EMPTY_P result.
12028 2021-06-21 Richard Biener <rguenther@suse.de>
12030 PR tree-optimization/101120
12031 * tree-vect-data-refs.c (bump_vector_ptr): Fold the
12033 * tree-vect-slp.c (vect_transform_slp_perm_load): Add
12034 DR chain DCE capability.
12035 * tree-vectorizer.h (vect_transform_slp_perm_load): Adjust.
12036 * tree-vect-stmts.c (vectorizable_load): Remove unused
12037 loads in the DR chain for SLP.
12039 2021-06-21 Jakub Jelinek <jakub@redhat.com>
12041 PR inline-asm/100785
12042 * gimplify.c (gimplify_asm_expr): Don't diagnose errors if
12043 output or input operands were already error_mark_node.
12044 * cfgexpand.c (expand_asm_stmt): If errors are emitted,
12045 remove all inputs, outputs and clobbers from the asm and
12046 set template to "".
12048 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
12050 * config/arm/arm_neon.h (vceq_s8): Replace builtin with __a == __b.
12051 (vceq_s16): Likewise.
12052 (vceq_s32): Likewise.
12053 (vceq_u8): Likewise.
12054 (vceq_u16): Likewise.
12055 (vceq_u32): Likewise.
12056 (vceq_p8): Likewise.
12057 (vceqq_s8): Likewise.
12058 (vceqq_s16): Likewise.
12059 (vceqq_s32): Likewise.
12060 (vceqq_u8): Likewise.
12061 (vceqq_u16): Likewise.
12062 (vceqq_u32): Likewise.
12063 (vceqq_p8): Likewise.
12064 (vceq_f32): Gate __a == __b on __FAST_MATH__.
12065 (vceqq_f32): Likewise.
12066 (vceq_f16): Likewise.
12067 (vceqq_f16): Likewise.
12069 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
12072 * config/arm/iterators.md (NEON_VACMP): Remove.
12073 * config/arm/neon.md (neon_vca<cmp_op><mode>): Use GLTE instead of GTGE
12075 (neon_vca<cmp_op><mode>_insn): Likewise.
12076 (neon_vca<cmp_op_unsp><mode>_insn_unspec): Use NEON_VAGLTE instead of
12079 2021-06-21 Richard Biener <rguenther@suse.de>
12081 PR tree-optimization/101121
12082 * tree-vect-slp.c (vect_build_slp_tree_2): To not fail fatally
12083 when we just lack a stmt with the desired op when doing permutation.
12084 (vect_build_slp_tree): When caching a failed SLP build attempt
12085 assert that at least one lane is marked as not matching.
12087 2021-06-21 liuhongt <hongtao.liu@intel.com>
12090 * config/i386/i386.md: (*anddi_1): Disparage slightly the mask
12091 register alternative.
12092 (*and<mode>_1): Ditto.
12094 (*andn<mode>_1): Ditto.
12095 (*<code><mode>_1): Ditto.
12096 (*<code>qi_1): Ditto.
12097 (*one_cmpl<mode>2_1): Ditto.
12098 (*one_cmplsi2_1_zext): Ditto.
12099 (*one_cmplqi2_1): Ditto.
12100 * config/i386/i386.c (x86_order_regs_for_local_alloc): Change
12101 the order of mask registers to be before general registers.
12103 2021-06-21 Roger Sayle <roger@nextmovesoftware.com>
12106 * config/i386/i386.md: New define_peephole2s to shrink writing
12107 1, 2 or 4 consecutive zeros to memory when optimizing for size.
12109 2021-06-18 Jeff Law <jeffreyalaw@gmail.com>
12111 * config/h8300/h8300.c (h8300_select_cc_mode): Handle SYMBOL_REF.
12112 * config/h8300/logical.md (<code><mode>3 logcial expander): Generate
12113 more efficient code when the source can be trivially simplified.
12115 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
12117 * gimple-range-cache.cc (ranger_cache::range_of_def): Calculate
12118 a range if global is not available.
12119 (ranger_cache::entry_range): Fallback to range_of_def.
12120 * gimple-range-cache.h (range_of_def): Adjust prototype.
12122 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
12124 PR tree-optimization/101014
12125 * gimple-range-cache.cc (ranger_cache::ranger_cache): Remove poor
12127 (ranger_cache::~ranger_cache): Ditto.
12128 (ranger_cache::enable_new_values): Delete.
12129 (ranger_cache::push_poor_value): Delete.
12130 (ranger_cache::range_of_def): Remove poor value processing.
12131 (ranger_cache::entry_range): Ditto.
12132 (ranger_cache::fill_block_cache): Ditto.
12133 * gimple-range-cache.h (class ranger_cache): Remove poor value members.
12134 * gimple-range.cc (gimple_ranger::range_of_expr): Remove call.
12135 * gimple-range.h (class gimple_ranger): Adjust.
12137 2021-06-18 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
12140 * common/config/arm/arm-common.c (arm_canon_arch_option_1): New function
12141 derived from arm_canon_arch.
12142 (arm_canon_arch_option): Call it.
12143 (arm_canon_arch_multilib_option): New function.
12144 * config/arm/arm-cpus.in (IGNORE_FOR_MULTILIB): New fgroup.
12145 * config/arm/arm.h (arm_canon_arch_multilib_option): New prototype.
12146 (CANON_ARCH_MULTILIB_SPEC_FUNCTION): New macro.
12147 (MULTILIB_ARCH_CANONICAL_SPECS): New macro.
12148 (DRIVER_SELF_SPECS): Add MULTILIB_ARCH_CANONICAL_SPECS.
12149 * config/arm/arm.opt (mlibarch): New option.
12150 * config/arm/t-rmprofile (MULTILIB_MATCHES): For armv8*-m, replace use
12151 of march on RHS with mlibarch.
12153 2021-06-18 Marcel Vollweiler <marcel@codesourcery.com>
12155 * config.in: Regenerate.
12156 * config/gcn/gcn.c (print_operand_address): Fix for global_load assembler
12158 * configure: Regenerate.
12159 * configure.ac: Fix for global_load assembler functions.
12161 2021-06-18 Richard Biener <rguenther@suse.de>
12163 PR tree-optimization/101112
12164 * tree-vect-slp.c (vect_slp_linearize_chain): Fix condition
12165 to lookup a pattern stmt def.
12167 2021-06-18 Jakub Jelinek <jakub@redhat.com>
12169 PR middle-end/101062
12170 * stor-layout.c (finish_bitfield_layout): Don't add bitfield
12171 representatives in QUAL_UNION_TYPE.
12173 2021-06-18 Andrew Pinski <apinski@marvell.com>
12175 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
12176 Add counting of how many times it is done.
12177 (factor_out_conditional_conversion): Likewise.
12178 (match_simplify_replacement): Likewise.
12179 (value_replacement): Likewise.
12180 (spaceship_replacement): Likewise.
12181 (cond_store_replacement): Likewise.
12182 (cond_if_else_store_replacement_1): Likewise.
12183 (hoist_adjacent_loads): Likewise.
12185 2021-06-18 Andrew Pinski <apinski@marvell.com>
12187 * tree-cfg.c (verify_gimple_assign_unary): Reject point and offset
12188 types on NEGATE_EXPR, ABS_EXPR, BIT_NOT_EXPR, PAREN_EXPR and CNONJ_EXPR.
12189 (verify_gimple_assign_binary): Reject point and offset types on
12190 MULT_EXPR, MULT_HIGHPART_EXPR, TRUNC_DIV_EXPR, CEIL_DIV_EXPR,
12191 FLOOR_DIV_EXPR, ROUND_DIV_EXPR, TRUNC_MOD_EXPR, CEIL_MOD_EXPR,
12192 FLOOR_MOD_EXPR, ROUND_MOD_EXPR, RDIV_EXPR, and EXACT_DIV_EXPR.
12194 2021-06-18 Michael Meissner <meissner@linux.ibm.com>
12196 * config/rs6000/rs6000.c (rs6000_emit_minmax): Add support for ISA
12197 3.1 IEEE 128-bit floating point xsmaxcqp/xsmincqp instructions.
12198 * config/rs6000/rs6000.md (s<minmax><mode>3, IEEE128 iterator):
12201 2021-06-17 Aaron Sawdey <acsawdey@linux.ibm.com>
12203 * config/rs6000/genfusion.pl (gen_logical_addsubf): Add
12204 earlyclobber to alts 0/1.
12205 (gen_addadd): Add earlyclobber to alts 0/1.
12206 * config/rs6000/fusion.md: Regenerate file.
12208 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12210 * cfgloopanal.c (get_loop_hot_path): Make path an auto_vec.
12212 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
12214 * gimple-range-cache.cc: Comment cleanups.
12215 * gimple-range-gori.cc: Comment cleanups.
12216 * gimple-range.cc: Comment/spacing cleanups
12217 * value-range.h: Comment cleanups.
12219 2021-06-17 H.J. Lu <hjl.tools@gmail.com>
12222 * calls.c (expand_call): Replace PUSH_ARGS with
12223 targetm.calls.push_argument (0).
12224 (emit_library_call_value_1): Likewise.
12225 * defaults.h (PUSH_ARGS): Removed.
12226 (PUSH_ARGS_REVERSED): Replace PUSH_ARGS with
12227 targetm.calls.push_argument (0).
12228 * expr.c (block_move_libcall_safe_for_call_parm): Likewise.
12229 (emit_push_insn): Pass the number bytes to push to
12230 targetm.calls.push_argument and pass 0 if ARGS_ADDR is 0.
12231 * hooks.c (hook_bool_uint_true): New.
12232 * hooks.h (hook_bool_uint_true): Likewise.
12233 * rtlanal.c (nonzero_bits1): Replace PUSH_ARGS with
12234 targetm.calls.push_argument (0).
12235 * target.def (push_argument): Add a targetm.calls hook.
12236 * targhooks.c (default_push_argument): New.
12237 * targhooks.h (default_push_argument): Likewise.
12238 * config/bpf/bpf.h (PUSH_ARGS): Removed.
12239 * config/cr16/cr16.c (TARGET_PUSH_ARGUMENT): New.
12240 * config/cr16/cr16.h (PUSH_ARGS): Removed.
12241 * config/i386/i386.c (ix86_push_argument): New.
12242 (TARGET_PUSH_ARGUMENT): Likewise.
12243 * config/i386/i386.h (PUSH_ARGS): Removed.
12244 * config/m32c/m32c.c (TARGET_PUSH_ARGUMENT): New.
12245 * config/m32c/m32c.h (PUSH_ARGS): Removed.
12246 * config/nios2/nios2.h (PUSH_ARGS): Likewise.
12247 * config/pru/pru.h (PUSH_ARGS): Likewise.
12248 * doc/tm.texi.in: Remove PUSH_ARGS documentation. Add
12249 TARGET_PUSH_ARGUMENT hook.
12250 * doc/tm.texi: Regenerated.
12252 2021-06-17 Uroš Bizjak <ubizjak@gmail.com>
12255 * config/i386/i386-expand.c (expand_vector_set_var):
12256 Handle V2FS mode remapping. Pass TARGET_MMX_WITH_SSE to
12257 ix86_expand_vector_init_duplicate.
12258 (ix86_expand_vector_init_duplicate): Emit insv_1 for
12259 QImode for !TARGET_PARTIAL_REG_STALL.
12260 * config/i386/predicates.md (vec_setm_mmx_operand): New predicate.
12261 * config/i386/mmx.md (vec_setv2sf): Use vec_setm_mmx_operand
12262 as operand 2 predicate. Call ix86_expand_vector_set_var
12263 for non-constant index operand.
12264 (vec_setv2si): Ditto.
12265 (vec_setv4hi): Ditto.
12266 (vec_setv8qi): ditto.
12268 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
12270 PR tree-optimization/100790
12271 * gimple-range.cc (range_of_builtin_call): Cleanup clz and ctz
12274 2021-06-17 Martin Liska <mliska@suse.cz>
12276 * doc/invoke.texi: Use consistently -O1 instead of -O.
12278 2021-06-17 Martin Liska <mliska@suse.cz>
12280 * gcov-io.h: Update documentation entry about string format.
12282 2021-06-17 Marius Hillenbrand <mhillen@linux.ibm.com>
12285 * config/s390/vecintrin.h (vec_doublee): Fix to use
12286 __builtin_s390_vflls.
12287 (vec_floate): Fix to use __builtin_s390_vflrd.
12289 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12291 * dominance.c (get_dominated_to_depth): Return auto_vec<basic_block>.
12292 * dominance.h (get_dominated_to_depth): Likewise.
12293 (get_all_dominated_blocks): Likewise.
12294 * cfgcleanup.c (delete_unreachable_blocks): Adjust.
12295 * gcse.c (hoist_code): Likewise.
12296 * tree-cfg.c (remove_edge_and_dominated_blocks): Likewise.
12297 * tree-parloops.c (oacc_entry_exit_ok): Likewise.
12298 * tree-ssa-dce.c (eliminate_unnecessary_stmts): Likewise.
12299 * tree-ssa-phiprop.c (pass_phiprop::execute): Likewise.
12301 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12303 * dominance.c (get_dominated_by_region): Return auto_vec<basic_block>.
12304 * dominance.h (get_dominated_by_region): Likewise.
12305 * tree-cfg.c (gimple_duplicate_sese_region): Adjust.
12306 (gimple_duplicate_sese_tail): Likewise.
12307 (move_sese_region_to_fn): Likewise.
12309 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12311 * dominance.c (get_dominated_by): Return auto_vec<basic_block>.
12312 * dominance.h (get_dominated_by): Likewise.
12313 * auto-profile.c (afdo_find_equiv_class): Adjust.
12314 * cfgloopmanip.c (duplicate_loop_to_header_edge): Likewise.
12315 * loop-unroll.c (unroll_loop_runtime_iterations): Likewise.
12316 * tree-cfg.c (test_linear_chain): Likewise.
12317 (test_diamond): Likewise.
12319 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12321 * cfgloop.h (get_loop_hot_path): Return auto_vec<basic_block>.
12322 * cfgloopanal.c (get_loop_hot_path): Likewise.
12323 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Likewise.
12325 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12327 * cgraph.c (cgraph_node::collect_callers): Return
12328 auto_vec<cgraph_edge *>.
12329 * cgraph.h (cgraph_node::collect_callers): Likewise.
12330 * ipa-cp.c (create_specialized_node): Adjust.
12331 (decide_about_value): Likewise.
12332 (decide_whether_version_node): Likewise.
12333 * ipa-sra.c (process_isra_node_results): Likewise.
12335 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12337 * vec.h (vl_ptr>::using_auto_storage): Handle null m_vec.
12338 (auto_vec<T, 0>::auto_vec): Define move constructor, and delete copy
12340 (auto_vec<T, 0>::operator=): Define move assignment and delete copy
12343 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
12345 * gimple-range.cc (debug_seed_ranger): New.
12346 (dump_ranger): New.
12347 (debug_ranger): New.
12349 2021-06-17 Richard Biener <rguenther@suse.de>
12351 PR tree-optimization/54400
12352 * tree-vectorizer.h (enum slp_instance_kind): Add
12353 slp_inst_kind_bb_reduc.
12354 (reduction_fn_for_scalar_code): Declare.
12355 * tree-vect-data-refs.c (vect_slp_analyze_instance_dependence):
12356 Check SLP_INSTANCE_KIND instead of looking at the
12358 (vect_slp_analyze_instance_alignment): Likewise.
12359 * tree-vect-loop.c (reduction_fn_for_scalar_code): Export.
12360 * tree-vect-slp.c (vect_slp_linearize_chain): Split out
12361 chain linearization from vect_build_slp_tree_2 and generalize
12362 for the use of BB reduction vectorization.
12363 (vect_build_slp_tree_2): Adjust accordingly.
12364 (vect_optimize_slp): Elide permutes at the root of BB reduction
12366 (vectorizable_bb_reduc_epilogue): New function.
12367 (vect_slp_prune_covered_roots): Likewise.
12368 (vect_slp_analyze_operations): Use them.
12369 (vect_slp_check_for_constructors): Recognize associatable
12370 chains for BB reduction vectorization.
12371 (vectorize_slp_instance_root_stmt): Generate code for the
12372 BB reduction epilogue.
12374 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
12376 * gimple-range-gori.cc (gori_compute::has_edge_range_p): Check with
12378 (gori_compute::may_recompute_p): New.
12379 (gori_compute::outgoing_edge_range_p): Perform recomputations.
12380 * gimple-range-gori.h (class gori_compute): Add prototype.
12382 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
12384 * gimple-range-cache.cc (ranger_cache::range_on_edge): Always return
12385 true when a range can be calculated.
12386 * gimple-range.cc (gimple_ranger::dump_bb): Check has_edge_range_p.
12388 2021-06-16 Martin Sebor <msebor@redhat.com>
12390 * doc/invoke.texi (-Wmismatched-dealloc, -Wmismatched-new-delete):
12391 Correct documented defaults.
12393 2021-06-16 Andrew MacLeod <amacleod@redhat.com>
12395 * gimple-range-cache.cc (ranger_cache::ranger_cache): Initialize
12396 m_new_value_p directly.
12398 2021-06-16 Uroš Bizjak <ubizjak@gmail.com>
12401 * config/i386/i386-expand.c (expand_vec_perm_2perm_pblendv):
12402 Handle 64bit modes for TARGET_SSE4_1.
12403 (expand_vec_perm_pshufb2): Handle 64bit modes for TARGET_SSSE3.
12404 (expand_vec_perm_even_odd_pack): Handle V4HI mode.
12405 (expand_vec_perm_even_odd_1) <case E_V4HImode>: Expand via
12406 expand_vec_perm_pshufb2 for TARGET_SSSE3 and via
12407 expand_vec_perm_even_odd_pack for TARGET_SSE4_1.
12408 * config/i386/mmx.md (mmx_packusdw): New insn pattern.
12410 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
12412 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn<mode>):
12413 Change to an expander that emits the correct instruction
12414 depending on endianness.
12415 (aarch64_<sur><addsub>hn<mode>_insn_le): Define.
12416 (aarch64_<sur><addsub>hn<mode>_insn_be): Define.
12418 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
12420 * config/aarch64/aarch64-simd-builtins.def: Split generator
12421 for aarch64_<su>qmovn builtins into scalar and vector
12423 * config/aarch64/aarch64-simd.md (aarch64_<su>qmovn<mode>_insn_le):
12425 (aarch64_<su>qmovn<mode>_insn_be): Define.
12426 (aarch64_<su>qmovn<mode>): Split into scalar and vector
12427 variants. Change vector variant to an expander that emits the
12428 correct instruction depending on endianness.
12430 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
12432 * config/aarch64/aarch64-simd-builtins.def: Split generator
12433 for aarch64_sqmovun builtins into scalar and vector variants.
12434 * config/aarch64/aarch64-simd.md (aarch64_sqmovun<mode>):
12435 Split into scalar and vector variants. Change vector variant
12436 to an expander that emits the correct instruction depending
12438 (aarch64_sqmovun<mode>_insn_le): Define.
12439 (aarch64_sqmovun<mode>_insn_be): Define.
12441 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
12443 * config/aarch64/aarch64-simd.md (aarch64_xtn<mode>_insn_le):
12444 Define - modeling zero-high-half semantics.
12445 (aarch64_xtn<mode>): Change to an expander that emits the
12446 appropriate instruction depending on endianness.
12447 (aarch64_xtn<mode>_insn_be): Define - modeling zero-high-half
12449 (aarch64_xtn2<mode>_le): Rename to...
12450 (aarch64_xtn2<mode>_insn_le): This.
12451 (aarch64_xtn2<mode>_be): Rename to...
12452 (aarch64_xtn2<mode>_insn_be): This.
12453 (vec_pack_trunc_<mode>): Emit truncation instruction instead
12455 * config/aarch64/iterators.md (Vnarrowd): Add Vnarrowd mode
12456 attribute iterator.
12458 2021-06-16 Martin Jambor <mjambor@suse.cz>
12460 PR tree-optimization/100453
12461 * tree-sra.c (create_access): Disqualify any const candidates
12462 which are written to.
12463 (sra_modify_expr): Do not store sub-replacements back to a const base.
12464 (handle_unscalarized_data_in_subtree): Likewise.
12465 (sra_modify_assign): Likewise. Earlier, use TREE_READONLy test
12466 instead of constant_decl_p.
12468 2021-06-16 Jakub Jelinek <jakub@redhat.com>
12470 PR middle-end/101062
12471 * stor-layout.c (finish_bitfield_representative): For fields in unions
12472 assume nextf is always NULL.
12473 (finish_bitfield_layout): Compute bit field representatives also in
12474 unions, but handle it as if each bitfield was the only field in the
12477 2021-06-16 Richard Biener <rguenther@suse.de>
12479 PR tree-optimization/101088
12480 * tree-ssa-loop-im.c (sm_seq_valid_bb): Only look for
12481 supported refs on edges. Do not assert same ref but
12482 different kind stores are unsuported but mark them so.
12483 (hoist_memory_references): Only look for supported refs
12486 2021-06-16 Roger Sayle <roger@nextmovesoftware.com>
12488 PR rtl-optimization/46235
12489 * config/i386/i386.md: New define_split for bt followed by cmov.
12490 (*bt<mode>_setcqi): New define_insn_and_split for bt followed by setc.
12491 (*bt<mode>_setncqi): New define_insn_and_split for bt then setnc.
12492 (*bt<mode>_setnc<mode>): New define_insn_and_split for bt followed
12493 by setnc with zero extension.
12495 2021-06-16 Richard Biener <rguenther@suse.de>
12497 PR tree-optimization/101083
12498 * tree-vect-slp.c (vect_slp_build_two_operator_nodes): Get
12499 vectype as argument.
12500 (vect_build_slp_tree_2): Adjust.
12502 2021-06-15 Martin Sebor <msebor@redhat.com>
12504 PR middle-end/100876
12505 * builtins.c: (gimple_call_return_array): Account for size_t
12506 mangling as either unsigned int or unsigned long
12508 2021-06-15 Jeff Law <jeffreyalaw@gmail.com>
12510 * compare-elim.c (try_eliminate_compare): Run DCE to clean things
12511 up before eliminating comparisons.
12513 2021-06-15 Aldy Hernandez <aldyh@redhat.com>
12515 * range-op.cc (operator_bitwise_or::wi_fold): Make sure
12516 nonzero|X is nonzero.
12517 (range_op_bitwise_and_tests): Add tests for above.
12519 2021-06-15 Carl Love <cel@us.ibm.com>
12522 * config/rs6000/rs6000-builtin.def (VCMPEQUT): Fix the ICODE for the
12524 (VRLQ, VSLQ, VSRQ, VSRAQ): Remove unused BU_P10_OVERLOAD_2
12527 2021-06-15 Tobias Burnus <tobias@codesourcery.com>
12530 * gimplify.c (enum gimplify_defaultmap_kind): Add GDMK_SCALAR_TARGET.
12531 (struct gimplify_omp_ctx): Extend defaultmap array by one.
12532 (new_omp_context): Init defaultmap[GDMK_SCALAR_TARGET].
12533 (omp_notice_variable): Update type classification for Fortran.
12534 (gimplify_scan_omp_clauses): Update calls for new argument; handle
12535 GDMK_SCALAR_TARGET; for Fortran, GDMK_POINTER avoid GOVD_MAP_0LEN_ARRAY.
12536 * langhooks-def.h (lhd_omp_scalar_p): Add 'ptr_ok' argument.
12537 * langhooks.c (lhd_omp_scalar_p): Likewise.
12538 (LANG_HOOKS_OMP_ALLOCATABLE_P, LANG_HOOKS_OMP_SCALAR_TARGET_P): New.
12539 (LANG_HOOKS_DECLS): Add them.
12540 * langhooks.h (struct lang_hooks_for_decls): Add new hooks, update
12541 omp_scalar_p pointer type to include the new bool argument.
12543 2021-06-15 David Malcolm <dmalcolm@redhat.com>
12545 * doc/analyzer.texi
12546 (Special Functions for Debugging the Analyzer): Add
12547 __analyzer_dump_capacity.
12549 2021-06-15 Jakub Jelinek <jakub@redhat.com>
12552 * expr.c (expand_expr_real_2) <case VEC_PACK_FIX_TRUNC_EXPR,
12553 case VEC_PACK_TRUNC_EXPR>: Clear subtarget when changing mode.
12555 2021-06-15 Richard Biener <rguenther@suse.de>
12557 * cfgloopanal.c (mark_irreducible_loops): Use a dominance
12558 check to identify loop latches.
12559 * cfgloop.c (verify_loop_structure): Likewise.
12560 * loop-init.c (apply_loop_flags): Allow marked irreducible
12561 regions even with multiple latches.
12562 * predict.c (rebuild_frequencies): Simplify.
12564 2021-06-15 Richard Biener <rguenther@suse.de>
12566 * tree-ssa-threadupdate.c
12567 (jump_thread_path_registry::mark_threaded_blocks): Assert we
12568 have marked irreducible regions.
12570 2021-06-14 Martin Sebor <msebor@redhat.com>
12573 * builtins.c (gimple_call_return_array): Check for attribute fn spec.
12574 Handle calls to placement new.
12575 (ndecl_dealloc_argno): Avoid placement delete.
12577 2021-06-14 Peter Bergner <bergner@linux.ibm.com>
12580 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Use
12581 create_tmp_reg_or_ssa_name().
12583 2021-06-14 Andrew MacLeod <amacleod@redhat.com>
12585 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
12586 (ranger_cache::enable_new_values): Set to specified value and
12587 return the old value.
12588 (ranger_cache::disable_new_values): Delete.
12589 (ranger_cache::fill_block_cache): Disable non 1st order derived
12591 * gimple-range-cache.h (ranger_cache): Adjust prototypes.
12592 * gimple-range.cc (gimple_ranger::range_of_expr): Adjust.
12594 2021-06-14 Uroš Bizjak <ubizjak@gmail.com>
12597 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const):
12598 Return true early when testing with V2HImode.
12599 * config/i386/mmx.md (*punpckwd): Split to sse2_pshuflw_1.
12601 2021-06-14 Christophe Lyon <christophe.lyon@linaro.org>
12603 * config/arm/mve.md (mve_vec_unpack<US>_lo_<mode>): New pattern.
12604 (mve_vec_unpack<US>_hi_<mode>): New pattern.
12605 (@mve_vec_pack_trunc_lo_<mode>): New pattern.
12606 (mve_vmovntq_<supf><mode>): Prefix with '@'.
12607 * config/arm/neon.md (vec_unpack<US>_hi_<mode>): Move to
12609 (vec_unpack<US>_lo_<mode>): Likewise.
12610 (vec_pack_trunc_<mode>): Rename to
12611 neon_quad_vec_pack_trunc_<mode>.
12612 * config/arm/vec-common.md (vec_unpack<US>_hi_<mode>): New
12614 (vec_unpack<US>_lo_<mode>): New.
12615 (vec_pack_trunc_<mode>): New.
12617 2021-06-14 Richard Biener <rguenther@suse.de>
12619 PR tree-optimization/100934
12620 * tree-ssa-dom.c (pass_dominator::execute): Properly
12621 mark irreducible regions.
12623 2021-06-14 Martin Liska <mliska@suse.cz>
12625 * doc/invoke.texi: Put r{...} on the same line as @item.
12627 2021-06-14 Martin Liska <mliska@suse.cz>
12629 * doc/invoke.texi: Add missing newline.
12631 2021-06-14 Martin Liska <mliska@suse.cz>
12633 * doc/invoke.texi: Remove '+' charasters.
12635 2021-06-14 Claudiu Zissulescu <claziss@synopsys.com>
12637 * config.gcc (arc): Add support for with_cpu option.
12638 * config/arc/arc.h (OPTION_DEFAULT_SPECS): Add fpu.
12640 2021-06-14 Richard Biener <rguenther@suse.de>
12642 PR tree-optimization/101031
12643 * tree-ssa-strlen.c (maybe_invalidate): Increment max_size
12644 instead of size when accounting for a possibly string
12647 2021-06-14 Martin Liska <mliska@suse.cz>
12649 * gimple-ssa-evrp.c (pointer_equiv_analyzer::~pointer_equiv_analyzer): Use delete[].
12651 2021-06-14 Aldy Hernandez <aldyh@redhat.com>
12653 * value-query.cc (gimple_range_global): Call get_range_global
12654 if called after inlining.
12656 2021-06-13 Uroš Bizjak <ubizjak@gmail.com>
12659 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
12660 Emit constant permutation insn directly from here.
12662 2021-06-13 Trevor Saunders <tbsaunde@tbsaunde.org>
12664 * attribs.c (find_attribute_namespace): Iterate over vec<> with
12666 * auto-profile.c (afdo_find_equiv_class): Likewise.
12667 * gcc.c (do_specs_vec): Likewise.
12668 (do_spec_1): Likewise.
12669 (driver::set_up_specs): Likewise.
12670 * gimple-loop-jam.c (any_access_function_variant_p): Likewise.
12671 * gimple-ssa-store-merging.c (compatible_load_p): Likewise.
12672 (imm_store_chain_info::try_coalesce_bswap): Likewise.
12673 (imm_store_chain_info::coalesce_immediate_stores): Likewise.
12674 (get_location_for_stmts): Likewise.
12675 * graphite-poly.c (print_iteration_domains): Likewise.
12676 (free_poly_bb): Likewise.
12677 (remove_gbbs_in_scop): Likewise.
12678 (free_scop): Likewise.
12679 (dump_gbb_cases): Likewise.
12680 (dump_gbb_conditions): Likewise.
12681 (print_pdrs): Likewise.
12682 (print_scop): Likewise.
12683 * ifcvt.c (cond_move_process_if_block): Likewise.
12684 * lower-subreg.c (decompose_multiword_subregs): Likewise.
12685 * regcprop.c (pass_cprop_hardreg::execute): Likewise.
12686 * sanopt.c (sanitize_rewrite_addressable_params): Likewise.
12687 * sel-sched-dump.c (dump_insn_vector): Likewise.
12688 * store-motion.c (store_ops_ok): Likewise.
12689 (store_killed_in_insn): Likewise.
12690 * timevar.c (timer::named_items::print): Likewise.
12691 * tree-cfgcleanup.c (cleanup_control_flow_pre): Likewise.
12692 (cleanup_tree_cfg_noloop): Likewise.
12693 * tree-data-ref.c (dump_data_references): Likewise.
12694 (print_dir_vectors): Likewise.
12695 (print_dist_vectors): Likewise.
12696 (dump_data_dependence_relations): Likewise.
12697 (dump_dist_dir_vectors): Likewise.
12698 (dump_ddrs): Likewise.
12699 (create_runtime_alias_checks): Likewise.
12700 (free_subscripts): Likewise.
12701 (save_dist_v): Likewise.
12702 (save_dir_v): Likewise.
12703 (invariant_access_functions): Likewise.
12704 (same_access_functions): Likewise.
12705 (access_functions_are_affine_or_constant_p): Likewise.
12706 (find_data_references_in_stmt): Likewise.
12707 (graphite_find_data_references_in_stmt): Likewise.
12708 (free_dependence_relations): Likewise.
12709 (free_data_refs): Likewise.
12710 * tree-inline.c (copy_debug_stmts): Likewise.
12711 * tree-into-ssa.c (dump_currdefs): Likewise.
12712 (rewrite_update_phi_arguments): Likewise.
12713 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
12714 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
12716 (vect_slp_analyze_node_dependences): Likewise.
12717 (vect_slp_analyze_instance_dependence): Likewise.
12718 (vect_record_base_alignments): Likewise.
12719 (vect_get_peeling_costs_all_drs): Likewise.
12720 (vect_peeling_supportable): Likewise.
12721 * tree-vectorizer.c (vec_info::~vec_info): Likewise.
12722 (vec_info::free_stmt_vec_infos): Likewise.
12724 2021-06-13 Jeff Law <jeffreyalaw@gmail.com>
12726 * config/h8300/logical.md (<code>qi3_1<cczn>): New pattern.
12727 (andqi3_1<cczn>): Removed.
12728 (<ors>qi3_1): Do not split for IOR/XOR a single bit.
12729 (H8/SX bit logicals): Split out from other patterns.
12730 * config/h8300/multiply.md (mulqihi3_const<cczn>): Renamed from
12731 mulqihi3_const_clobber_flags.
12732 (mulqihi3<cczn>, mulhisi3_const<cczn>, mulhisi3<cczn>): Similarly
12734 2021-06-13 H.J. Lu <hjl.tools@gmail.com>
12737 * config/i386/i386.c (ix86_expand_prologue): Set red_zone_used
12738 to true if red zone is used.
12739 (ix86_output_indirect_jmp): Replace ix86_red_zone_size with
12740 ix86_red_zone_used.
12741 * config/i386/i386.h (machine_function): Add red_zone_used.
12742 (ix86_red_zone_size): Removed.
12743 (ix86_red_zone_used): New.
12744 * config/i386/i386.md (peephole2 patterns): Replace
12745 ix86_red_zone_size with ix86_red_zone_used.
12747 2021-06-12 Jason Merrill <jason@redhat.com>
12749 * doc/extend.texi (unused variable attribute): Applies to
12750 structure fields as well.
12752 2021-06-12 Eugene Rozenfeld <erozen@microsoft.com>
12754 * auto-profile.c (read_profile): fix a typo in an error string
12756 2021-06-11 Thomas Schwinge <thomas@codesourcery.com>
12758 * tree-pretty-print.h (dump_omp_clauses): Add 'bool = true'
12760 * tree-pretty-print.c (dump_omp_clauses): Update.
12761 (dump_generic_node) <OMP_CLAUSE>: Use it.
12763 2021-06-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
12766 * config/arm/arm_mve.h (__arm_vld1q): Change __ARM_mve_coerce(p0,
12767 int8_t const *) to __ARM_mve_coerce1(p0, int8_t *) in the argument for
12768 the polymorphic variants matching code.
12769 (__arm_vld1q_z): Likewise.
12770 (__arm_vld2q): Likewise.
12771 (__arm_vld4q): Likewise.
12772 (__arm_vldrbq_gather_offset): Likewise.
12773 (__arm_vldrbq_gather_offset_z): Likewise.
12775 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
12777 PR tree-optimization/96392
12778 * fold-const.h (tree_expr_maybe_real_minus_zero_p): Fix prototype.
12780 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
12782 PR tree-optimization/96392
12783 * fold-const.c (fold_real_zero_addition_p): Take both arguments
12784 of the addition or subtraction, not just the zero. Use this
12785 other argument in tests for signaling NaNs and signed zeros.
12786 (tree_expr_maybe_real_minus_zero_p): New predicate.
12787 * fold-const.h (fold_real_zero_addition_p): Update prototype.
12788 (tree_expr_maybe_real_minus_zero_p): New function prototype.
12789 * match.pd: Update calls to fold_real_zero_addition_p.
12790 Replace HONOR_NANS with tree_expr_maybe_nan_p.
12791 Replace HONOR_SIGNED_ZEROS with tree_expr_maybe_real_minus_zero_p.
12792 Replace HONOR_SNANS with tree_expr_maybe_signaling_nan_p.
12793 * tree-ssa-reassoc.c (eliminate_using_constants): Update
12794 call to fold_real_zero_addition_p.
12796 2021-06-11 Richard Biener <rguenther@suse.de>
12798 PR tree-optimization/101025
12799 * tree-ssa-loop-im.c (sm_seq_valid_bb): Make sure to process
12800 all refs that require dependence checking.
12802 2021-06-11 Richard Biener <rguenther@suse.de>
12804 PR tree-optimization/101028
12805 * tree-vect-slp.c (vect_build_slp_tree_2): When SLP
12806 reassoc discovery fails fatally, mark appropriate lanes
12809 2021-06-11 Richard Biener <rguenther@suse.de>
12811 PR tree-optimization/101026
12812 * tree-vect-slp.c (vect_build_slp_tree_2): Make sure we
12813 have a representative for the associated chain nodes.
12815 2021-06-11 Jakub Jelinek <jakub@redhat.com>
12817 PR rtl-optimization/101008
12818 * simplify-rtx.c (relational_result): New function.
12819 (simplify_logical_relational_operation,
12820 simplify_relational_operation): Use it.
12822 2021-06-11 Jakub Jelinek <jakub@redhat.com>
12825 * config/i386/sse.md (*vec_concat<mode>_0_1): Require TARGET_SSE2.
12827 2021-06-11 Uroš Bizjak <ubizjak@gmail.com>
12830 * config/i386/i386-expand.c (expand_vec_perm_pshufb): Return
12831 false if the permutation can be implemented with constant
12832 permutation instruction in wider mode.
12833 (canonicalize_vector_int_perm): Move above expand_vec_perm_pshufb.
12834 Handle V8QImode and V4HImode.
12836 2021-06-11 Martin Liska <mliska@suse.cz>
12838 PR gcov-profile/100788
12839 * common.opt: Add new option.
12840 * coverage.c (coverage_begin_function): Emit warning instead on
12841 the internal compiler error.
12842 * doc/invoke.texi: Document the option.
12843 * toplev.c (process_options): Enable it by default.
12845 2021-06-11 Richard Biener <rguenther@suse.de>
12847 PR middle-end/101009
12848 * tree-data-ref.c (build_classic_dist_vector_1): Make sure
12849 to set *init_b to true when we encounter a constant equal
12851 (compute_affine_dependence): Also dump the actual DR_REF.
12853 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
12855 PR tree-optimization/100984
12856 * gimple-ssa-evrp.c (ssa_equiv_stack): Use auto_vec for
12857 replacements table.
12858 (ssa_equiv_stack::~ssa_equiv_stack): Remove.
12860 2021-06-11 Kewen Lin <linkw@linux.ibm.com>
12862 * config/rs6000/rs6000.md
12863 (floatsi<SFDF:mode>2_lfiwax_<QHI:mode>_mem_zext): New
12864 define_insn_and_split.
12866 2021-06-11 Richard Biener <rguenther@suse.de>
12868 * tree-vect-slp.c (vect_build_slp_tree_2): Use stablesort
12869 to sort operands of the associative chain.
12871 2021-06-11 Richard Biener <rguenther@suse.de>
12873 * system.h (gcc_stablesort_r): Declare.
12874 * sort.cc (gcc_sort_r): Support stable sort.
12875 (gcc_stablesort_r): Define.
12876 * vec.h (vec<>::stablesort): Add.
12878 2021-06-10 Uroš Bizjak <ubizjak@gmail.com>
12881 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
12882 Handle V2SF mode. Emit SHUFPS to fixup unpack-high for V2SF mode.
12883 (expand_vec_perm_blend): Handle 64bit modes for TARGET_SSE4_1.
12884 (expand_vec_perm_pshufb): Handle 64bit modes for TARGET_SSSE3.
12885 (expand_vec_perm_pblendv): Handle 64bit modes for TARGET_SSE4_1.
12886 (expand_vec_perm_interleave2): Handle 64bit modes.
12887 (expand_vec_perm_even_odd_pack): Handle V8QI mode.
12888 (expand_vec_perm_even_odd_1): Ditto.
12889 (ix86_vectorize_vec_perm_const): Ditto.
12890 * config/i386/i386.md (UNSPEC_PSHUFB): Move from ...
12891 * config/i386/sse.md: ... here.
12892 * config/i386/mmx.md (*vec_interleave_lowv2sf):
12893 New insn_and_split pattern.
12894 (*vec_interleave_highv2sf): Ditto.
12895 (mmx_pshufbv8qi3): New insn pattern.
12896 (*mmx_pblendw): Ditto.
12898 2021-06-10 Peter Bergner <bergner@linux.ibm.com>
12900 * config/rs6000/rs6000-builtin.def (build_pair): New built-in.
12901 (build_acc): Likewise.
12902 * config/rs6000/rs6000-call.c (mma_expand_builtin): Swap assemble
12903 source operands in little-endian mode.
12904 (rs6000_gimple_fold_mma_builtin): Handle VSX_BUILTIN_BUILD_PAIR.
12905 (mma_init_builtins): Likewise.
12906 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Handle endianness
12907 ordering for the MMA assemble and build source operands.
12908 * doc/extend.texi (__builtin_vsx_build_acc, __builtin_mma_build_pair):
12910 (__builtin_mma_assemble_acc, __builtin_mma_assemble_pair): Remove
12913 2021-06-10 Jeff Law <jeffreyalaw@gmail.com>
12915 * config/h8300/h8300.c (select_cc_mode): Handle MEM. Use
12917 * config/h8300/extensions.md: Replace _clobber_flags patterns
12920 2021-06-10 Robin Dapp <rdapp@linux.ibm.com>
12922 * config/s390/vector.md (vcond_mask_<mode><mode>): Change to
12923 (vcond_mask_<mode><tointvec>): this.
12925 2021-06-10 Andrew Stubbs <ams@codesourcery.com>
12926 Thomas Schwinge <thomas@codesourcery.com>
12928 * omp-builtins.def (BUILT_IN_GOACC_ENTER_EXIT_DATA): Split into...
12929 (BUILT_IN_GOACC_ENTER_DATA, BUILT_IN_GOACC_EXIT_DATA): ... these.
12930 * gimple.h (enum gf_mask): Split
12931 'GF_OMP_TARGET_KIND_OACC_ENTER_EXIT_DATA' into
12932 'GF_OMP_TARGET_KIND_OACC_ENTER_DATA' and
12933 'GF_OMP_TARGET_KIND_OACC_EXIT_DATA'.
12934 (is_gimple_omp_oacc): Update.
12935 * gimple-pretty-print.c (dump_gimple_omp_target): Likewise.
12936 * gimplify.c (gimplify_omp_target_update): Likewise.
12937 * omp-expand.c (expand_omp_target, build_omp_regions_1)
12938 (omp_make_gimple_edges): Likewise.
12939 * omp-low.c (check_omp_nesting_restrictions, lower_omp_target):
12942 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
12944 * value-query.cc (value_query::value_on_edge): Rename name to
12946 (range_query::range_on_edge): Same.
12947 (range_query::value_of_expr): Same.
12948 (range_query::value_on_edge): Same.
12949 * value-query.h (class value_query): Same.
12950 (class range_query): Same.
12952 2021-06-10 Richard Biener <rguenther@suse.de>
12954 PR tree-optimization/101003
12955 * tree-vect-slp.c (vect_build_slp_tree_2): Appropriately
12956 use the pattern stmt defs when linearizing a chain.
12958 2021-06-10 Jakub Jelinek <jakub@redhat.com>
12961 * ifcvt.c (noce_get_alt_condition, noce_try_abs): Use
12962 prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
12964 2021-06-10 Clement Chigot <clement.chigot@atos.net>
12966 * config/rs6000/aix71.h (ASM_CPU_SPEC): Add Power10 directive.
12967 * config/rs6000/aix72.h (ASM_CPU_SPEC): Likewise.
12969 2021-06-09 Andrew Pinski <apinski@marvell.com>
12971 PR tree-optimization/100925
12972 * match.pd (a ? CST1 : CST2): Limit transformations
12973 that would produce a negative to integeral types only.
12974 Change !POINTER_TYPE_P to INTEGRAL_TYPE_P also.
12976 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
12979 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
12981 * doc/tm.texi: Correctly update.
12983 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
12985 * doc/tm.texi: Correctly update.
12987 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
12990 * doc/tm.texi.in (Trampolines): Add a missing blank line.
12992 2021-06-09 Paul Eggert <eggert@cs.ucla.edu>
12995 * doc/invoke.texi (Code Gen Options); Document that -fno-trampolines
12996 and -ftrampolines work only with Ada.
12997 * doc/tm.texi.in (Trampolines): Likewise.
12998 * doc/tm.texi: Regenerated.
13000 2021-06-09 Carl Love <cel@us.ibm.com>
13002 * config/rs6000/altivec.h (vec_signextll, vec_signexti, vec_signextq):
13003 Add define for new builtins.
13004 * config/rs6000/altivec.md(altivec_vreveti2): Add define_expand.
13005 * config/rs6000/rs6000-builtin.def (VSIGNEXTI, VSIGNEXTLL): Add
13006 overloaded builtin definitions.
13007 (VSIGNEXTSB2W, VSIGNEXTSH2W, VSIGNEXTSB2D, VSIGNEXTSH2D,VSIGNEXTSW2D,
13008 VSIGNEXTSD2Q): Add builtin expansions.
13009 (SIGNEXT): Add P10 overload definition.
13010 * config/rs6000/rs6000-call.c (P9V_BUILTIN_VEC_VSIGNEXTI, P9V_BUILTIN_VEC_VSIGNEXTLL,
13011 P10_BUILTIN_VEC_SIGNEXT): Add overloaded argument definitions.
13012 * config/rs6000/vsx.md (vsx_sign_extend_v2di_v1ti): Add define_insn.
13013 (vsignextend_v2di_v1ti, vsignextend_qi_<mode>, vsignextend_hi_<mode>,
13014 vsignextend_si_v2di)[VIlong]: Add define_expand.
13015 Make define_insn vsx_sign_extend_si_v2di visible.
13016 * doc/extend.texi: Add documentation for the vec_signexti,
13017 vec_signextll builtins and vec_signextq.
13019 2021-06-09 Carl Love <cel@us.ibm.com>
13021 * config/rs6000/rs6000.c (__fixkfti, __fixunskfti, __floattikf,
13022 __floatuntikf): Names changed to __fixkfti_sw, __fixunskfti_sw,
13023 __floattikf_sw, __floatuntikf_sw respectively.
13024 * config/rs6000/rs6000.md (floatti<mode>2, floatunsti<mode>2,
13025 fix_trunc<mode>ti2, fixuns_trunc<mode>ti2): Add
13026 define_insn for mode IEEE 128.
13028 2021-06-09 Carl Love <cel@us.ibm.com>
13030 * config/rs6000/altivec.md (altivec_vslq, altivec_vsrq):
13031 Rename to altivec_vslq_<mode>, altivec_vsrq_<mode>, mode VEC_TI.
13032 * config/rs6000/vector.md (VEC_TI): Was named VSX_TI in vsx.md.
13033 (vashlv1ti3): Change to vashl<mode>3, mode VEC_TI.
13034 (vlshrv1ti3): Change to vlshr<mode>3, mode VEC_TI.
13035 * config/rs6000/vsx.md (VSX_TI): Remove define_mode_iterator. Update
13036 uses of VSX_TI to VEC_TI.
13038 2021-06-09 Carl Love <cel@us.ibm.com>
13040 * config/rs6000/dfp.md (floattitd2, fixtdti2): New define_insns.
13042 2021-06-09 Carl Love <cel@us.ibm.com>
13044 * config/rs6000/altivec.h (vec_dive, vec_mod): Add define for new
13046 * config/rs6000/altivec.md (UNSPEC_VMULEUD, UNSPEC_VMULESD,
13047 UNSPEC_VMULOUD, UNSPEC_VMULOSD): New unspecs.
13048 (altivec_eqv1ti, altivec_gtv1ti, altivec_gtuv1ti, altivec_vmuleud,
13049 altivec_vmuloud, altivec_vmulesd, altivec_vmulosd, altivec_vrlq,
13050 altivec_vrlqmi, altivec_vrlqmi_inst, altivec_vrlqnm,
13051 altivec_vrlqnm_inst, altivec_vslq, altivec_vsrq, altivec_vsraq,
13052 altivec_vcmpequt_p, altivec_vcmpgtst_p, altivec_vcmpgtut_p): New
13054 (vec_widen_umult_even_v2di, vec_widen_smult_even_v2di,
13055 vec_widen_umult_odd_v2di, vec_widen_smult_odd_v2di, altivec_vrlqmi,
13056 altivec_vrlqnm): New define_expands.
13057 * config/rs6000/rs6000-builtin.def (VCMPEQUT_P, VCMPGTST_P,
13058 VCMPGTUT_P): Add macro expansions.
13059 (BU_P10V_AV_P): Add builtin predicate definition.
13060 (VCMPGTUT, VCMPGTST, VCMPEQUT, CMPNET, CMPGE_1TI,
13061 CMPGE_U1TI, CMPLE_1TI, CMPLE_U1TI, VNOR_V1TI_UNS, VNOR_V1TI, VCMPNET_P,
13062 VCMPAET_P, VMULEUD, VMULESD, VMULOUD, VMULOSD, VRLQ,
13063 VSLQ, VSRQ, VSRAQ, VRLQNM, DIV_V1TI, UDIV_V1TI, DIVES_V1TI, DIVEU_V1TI,
13064 MODS_V1TI, MODU_V1TI, VRLQMI): New macro expansions.
13065 (VRLQ, VSLQ, VSRQ, VSRAQ, DIVE, MOD): New overload expansions.
13066 * config/rs6000/rs6000-call.c (P10_BUILTIN_VCMPEQUT,
13067 P10V_BUILTIN_CMPGE_1TI, P10V_BUILTIN_CMPGE_U1TI,
13068 P10V_BUILTIN_VCMPGTUT, P10V_BUILTIN_VCMPGTST,
13069 P10V_BUILTIN_CMPLE_1TI, P10V_BUILTIN_VCMPLE_U1TI,
13070 P10V_BUILTIN_DIV_V1TI, P10V_BUILTIN_UDIV_V1TI,
13071 P10V_BUILTIN_VMULESD, P10V_BUILTIN_VMULEUD,
13072 P10V_BUILTIN_VMULOSD, P10V_BUILTIN_VMULOUD,
13073 P10V_BUILTIN_VNOR_V1TI, P10V_BUILTIN_VNOR_V1TI_UNS,
13074 P10V_BUILTIN_VRLQ, P10V_BUILTIN_VRLQMI,
13075 P10V_BUILTIN_VRLQNM, P10V_BUILTIN_VSLQ,
13076 P10V_BUILTIN_VSRQ, P10V_BUILTIN_VSRAQ,
13077 P10V_BUILTIN_VCMPGTUT_P, P10V_BUILTIN_VCMPGTST_P,
13078 P10V_BUILTIN_VCMPEQUT_P, P10V_BUILTIN_VCMPGTUT_P,
13079 P10V_BUILTIN_VCMPGTST_P, P10V_BUILTIN_CMPNET,
13080 P10V_BUILTIN_VCMPNET_P, P10V_BUILTIN_VCMPAET_P,
13081 P10V_BUILTIN_DIVES_V1TI, P10V_BUILTIN_MODS_V1TI,
13082 P10V_BUILTIN_MODU_V1TI):
13083 New overloaded definitions.
13084 (rs6000_gimple_fold_builtin) [P10V_BUILTIN_VCMPEQUT,
13085 P10V_BUILTIN_CMPNET, P10V_BUILTIN_CMPGE_1TI,
13086 P10V_BUILTIN_CMPGE_U1TI, P10V_BUILTIN_VCMPGTUT,
13087 P10V_BUILTIN_VCMPGTST, P10V_BUILTIN_CMPLE_1TI,
13088 P10V_BUILTIN_CMPLE_U1TI]: New case statements.
13089 (rs6000_init_builtins) [bool_V1TI_type_node, int_ftype_int_v1ti_v1ti]:
13091 (altivec_init_builtins): New E_V1TImode case statement.
13092 (builtin_function_type)[P10_BUILTIN_128BIT_VMULEUD,
13093 P10_BUILTIN_128BIT_VMULOUD, P10_BUILTIN_128BIT_DIVEU_V1TI,
13094 P10_BUILTIN_128BIT_MODU_V1TI, P10_BUILTIN_CMPGE_U1TI,
13095 P10_BUILTIN_VCMPGTUT, P10_BUILTIN_VCMPEQUT]: New case statements.
13096 * config/rs6000/rs6000.c (rs6000_handle_altivec_attribute) [E_TImode,
13097 E_V1TImode]: New case statements.
13098 * config/rs6000/rs6000.h (rs6000_builtin_type_index): New enum
13099 value RS6000_BTI_bool_V1TI.
13100 * config/rs6000/vector.md (vector_gtv1ti,vector_nltv1ti,
13101 vector_gtuv1ti, vector_nltuv1ti, vector_ngtv1ti, vector_ngtuv1ti,
13102 vector_eq_v1ti_p, vector_ne_v1ti_p, vector_ae_v1ti_p,
13103 vector_gt_v1ti_p, vector_gtu_v1ti_p, vrotlv1ti3, vashlv1ti3,
13104 vlshrv1ti3, vashrv1ti3): New define_expands.
13105 * config/rs6000/vsx.md (UNSPEC_VSX_DIVSQ, UNSPEC_VSX_DIVUQ,
13106 UNSPEC_VSX_DIVESQ, UNSPEC_VSX_DIVEUQ, UNSPEC_VSX_MODSQ,
13107 UNSPEC_VSX_MODUQ): New unspecs.
13108 (mulv2di3, vsx_div_v1ti, vsx_udiv_v1ti, vsx_dives_v1ti,
13109 vsx_diveu_v1ti, vsx_mods_v1ti, vsx_modu_v1ti, xxswapd_v1ti): New
13111 (vcmpnet): New define_expand.
13112 * doc/extend.texi: Add documentation for the new builtins vec_rl,
13113 vec_rlmi, vec_rlnm, vec_sl, vec_sr, vec_sra, vec_mule, vec_mulo,
13114 vec_div, vec_dive, vec_mod, vec_cmpeq, vec_cmpne, vec_cmpgt, vec_cmplt,
13115 vec_cmpge, vec_cmple, vec_all_eq, vec_all_ne, vec_all_gt, vec_all_lt,
13116 vec_all_ge, vec_all_le, vec_any_eq, vec_any_ne, vec_any_gt, vec_any_lt,
13117 vec_any_ge, vec_any_le.
13119 2021-06-09 Carl Love <cel@us.ibm.com>
13121 * config/rs6000/altivec.md (altivec_vrl<VI_char>mi): Fix
13122 bug in argument generation.
13124 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
13126 * config/arm/iterators.md (<supf>): Remove VCLZQ_U, VCLZQ_S.
13128 * config/arm/mve.md (mve_vclzq_<supf><mode>): Add '@' prefix,
13129 remove <supf> iterator.
13130 (mve_vclzq_u<mode>): New.
13131 * config/arm/neon.md (clz<mode>2): Rename to neon_vclz<mode>.
13132 (neon_vclz<mode): Move to ...
13133 * config/arm/unspecs.md (VCLZQ_U, VCLZQ_S): Remove.
13134 * config/arm/vec-common.md: ... here. Add support for MVE.
13136 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
13138 * config/arm/mve.md (mve_vhaddq_<supf><mode>): Prefix with '@'.
13139 (@mve_vrhaddq_<supf><mode): Likewise.
13140 * config/arm/neon.md (neon_v<r>hadd<sup><mode>): Likewise.
13141 * config/arm/vec-common.md (avg<mode>3_floor, uavg<mode>3_floor)
13142 (avg<mode>3_ceil", uavg<mode>3_ceil): New patterns.
13144 2021-06-09 imba-tjd <109224573@qq.com>
13146 * doc/invoke.texi: Fix typo.
13148 2021-06-09 Roger Sayle <roger@nextmovesoftware.com>
13150 PR middle-end/53267
13151 * fold-const-call.c (fold_const_call_sss) [CASE_CFN_FMOD]:
13152 Support evaluation of fmod/fmodf/fmodl at compile-time.
13154 2021-06-09 Richard Biener <rguenther@suse.de>
13156 PR tree-optimization/100981
13157 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
13158 gimple_get_lhs to also handle calls.
13159 * tree-vect-slp-patterns.c (complex_pattern::build): Transfer
13162 2021-06-09 Richard Biener <rguenther@suse.de>
13164 PR tree-optimization/97832
13165 * tree-vectorizer.h (_slp_tree::failed): New.
13166 * tree-vect-slp.c (_slp_tree::_slp_tree): Initialize
13168 (_slp_tree::~_slp_tree): Free failed.
13169 (vect_build_slp_tree): Retain failed nodes and record
13170 matches in them, copying that back out when running
13171 into a cached fail. Dump start and end of discovery.
13172 (dt_sort_cmp): New.
13173 (vect_build_slp_tree_2): Handle associatable chains
13174 together doing more aggressive operand swapping.
13176 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
13179 * config.gcc (gcc_cv_initfini_array): Set to yes for Linux and
13181 * doc/install.texi: Require glibc 2.1 and binutils 2.12 for
13182 Linux and GNU targets.
13184 2021-06-09 Richard Biener <rguenther@suse.de>
13186 * tree-vect-stmts.c (vect_is_simple_use): Always get dt
13189 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
13191 * config/arc/arc.md (loop_end): Change it to
13192 define_insn_and_split.
13194 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
13196 * config/arc/arc.md (maddhisi4): Use VMAC2H instruction.
13197 (machi): New pattern.
13198 (umaddhisi4): Use VMAC2HU instruction.
13199 (umachi): New pattern.
13201 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
13203 * config/arc/arc-protos.h (arc_split_move_p): New prototype.
13204 * config/arc/arc.c (arc_split_move_p): New function.
13205 (arc_split_move): Clean up.
13206 * config/arc/arc.md (movdi_insn): Clean up, use arc_split_move_p.
13207 (movdf_insn): Likewise.
13208 * config/arc/simdext.md (mov<VWH>_insn): Likewise.
13210 2021-06-09 Uroš Bizjak <ubizjak@gmail.com>
13213 * config/i386/i386.c (print_operand_address_as): Rename "no_rip"
13214 argument to "raw". Do not emit segment overrides when "raw" is true.
13216 2021-06-09 Martin Liska <mliska@suse.cz>
13218 * doc/gcov.texi: Create a proper JSON files.
13219 * doc/invoke.texi: Remove dots in order to make it a valid
13222 2021-06-09 Xionghu Luo <luoxhu@linux.ibm.com>
13224 * config/rs6000/rs6000-p8swap.c (pattern_is_rotate64): New.
13225 (insn_is_load_p): Use pattern_is_rotate64.
13226 (insn_is_swap_p): Likewise.
13227 (quad_aligned_load_p): Likewise.
13228 (const_load_sequence_p): Likewise.
13229 (replace_swapped_aligned_load): Likewise.
13230 (recombine_lvx_pattern): Likewise.
13231 (recombine_stvx_pattern): Likewise.
13233 2021-06-09 Andrew MacLeod <amacleod@redhat.com>
13235 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Use a
13236 fur_stmt source record.
13237 * gimple-range.cc (fur_source::get_operand): Generic range query.
13238 (fur_source::get_phi_operand): New.
13239 (fur_source::register_dependency): New.
13240 (fur_source::query): New.
13241 (class fur_edge): New. Edge source for operands.
13242 (fur_edge::fur_edge): New.
13243 (fur_edge::get_operand): New.
13244 (fur_edge::get_phi_operand): New.
13245 (fur_edge::query): New.
13246 (fur_stmt::fur_stmt): New.
13247 (fur_stmt::get_operand): New.
13248 (fur_stmt::get_phi_operand): New.
13249 (fur_stmt::query): New.
13250 (class fur_depend): New. Statement source and process dependencies.
13251 (fur_depend::fur_depend): New.
13252 (fur_depend::register_dependency): New.
13253 (class fur_list): New. List source for operands.
13254 (fur_list::fur_list): New.
13255 (fur_list::get_operand): New.
13256 (fur_list::get_phi_operand): New.
13257 (fold_range): New. Instantiate appropriate fur_source class and fold.
13258 (fold_using_range::range_of_range_op): Use new API.
13259 (fold_using_range::range_of_address): Ditto.
13260 (fold_using_range::range_of_phi): Ditto.
13261 (imple_ranger::fold_range_internal): Use fur_depend class.
13262 (fold_using_range::range_of_ssa_name_with_loop_info): Use new API.
13263 * gimple-range.h (class fur_source): Now a base class.
13264 (class fur_stmt): New.
13265 (fold_range): New prototypes.
13266 (fur_source::fur_source): Delete.
13268 2021-06-08 Andrew Pinski <apinski@marvell.com>
13270 PR tree-optimization/25290
13271 * tree-ssa-phiopt.c (xor_replacement): Delete.
13272 (tree_ssa_phiopt_worker): Delete use of xor_replacement.
13273 (match_simplify_replacement): Allow one cheap preparation
13274 statement that can be moved to before the if.
13276 2021-06-08 Pat Haugen <pthaugen@linux.ibm.com>
13278 * config/rs6000/power10.md (power10-fused-load, power10-fused-store,
13279 power10-fused_alu, power10-fused-vec, power10-fused-branch): New.
13281 2021-06-08 Jeff Law <jeffreyalaw@gmail.com>
13283 * config/h8300/logical.md (andqi3_1): Move BCLR case into define_insn_and_split.
13284 Create length attribute on define_insn_and_split. Only split for cases which we
13286 (andqi3_1<cczn>): Renamed from andqi3_1_clobber_flags. Only handle AND here and
13287 fix length computation.
13288 (b<code><mode>msx): Combine QImode and HImode H8/SX patterns using iterator.
13290 2021-06-08 Richard Biener <rguenther@suse.de>
13292 PR tree-optimization/100923
13293 * tree-ssa-sccvn.c (valueize_refs_1): Take a pointer to
13294 the operand vector to be valueized.
13295 (valueize_refs): Likewise.
13296 (valueize_shared_reference_ops_from_ref): Adjust.
13297 (valueize_shared_reference_ops_from_call): Likewise.
13298 (vn_reference_lookup_3): Likewise.
13299 (vn_reference_lookup_pieces): Likewise. Re-valueize
13300 with honoring availability when we are about to create
13301 the ao_ref and valueized before.
13302 (vn_reference_lookup): Likewise.
13303 (vn_reference_insert_pieces): Adjust.
13305 2021-06-08 Richard Biener <rguenther@suse.de>
13307 * tree-vectorizer.h (_slp_instance::root_stmt): Change to...
13308 (_slp_instance::root_stmts): ... a vector.
13309 (SLP_INSTANCE_ROOT_STMT): Rename to ...
13310 (SLP_INSTANCE_ROOT_STMTS): ... this.
13311 (slp_root::root): Change to...
13312 (slp_root::roots): ... a vector.
13313 (slp_root::slp_root): Adjust.
13314 * tree-vect-slp.c (_slp_instance::location): Adjust.
13315 (vect_free_slp_instance): Release the root stmt vector.
13316 (vect_build_slp_instance): Adjust.
13317 (vect_analyze_slp): Likewise.
13318 (_bb_vec_info::~_bb_vec_info): Likewise.
13319 (vect_slp_analyze_operations): Likewise.
13320 (vect_bb_vectorization_profitable_p): Likewise. Adjust
13321 costs for the root stmt.
13322 (vect_slp_check_for_constructors): Gather all BIT_INSERT_EXPRs
13324 (vect_slp_analyze_bb_1): Simplify by marking all root stmts
13326 (vectorize_slp_instance_root_stmt): Adjust.
13327 (vect_schedule_slp): Likewise.
13329 2021-06-08 Aldy Hernandez <aldyh@redhat.com>
13331 * gimple-ssa-evrp.c (class ssa_equiv_stack): New.
13332 (ssa_equiv_stack::ssa_equiv_stack): New.
13333 (ssa_equiv_stack::~ssa_equiv_stack): New.
13334 (ssa_equiv_stack::enter): New.
13335 (ssa_equiv_stack::leave): New.
13336 (ssa_equiv_stack::push_replacement): New.
13337 (ssa_equiv_stack::get_replacement): New.
13338 (is_pointer_ssa): New.
13339 (class pointer_equiv_analyzer): New.
13340 (pointer_equiv_analyzer::pointer_equiv_analyzer): New.
13341 (pointer_equiv_analyzer::~pointer_equiv_analyzer): New.
13342 (pointer_equiv_analyzer::set_global_equiv): New.
13343 (pointer_equiv_analyzer::set_cond_equiv): New.
13344 (pointer_equiv_analyzer::get_equiv): New.
13345 (pointer_equiv_analyzer::enter): New.
13346 (pointer_equiv_analyzer::leave): New.
13347 (pointer_equiv_analyzer::get_equiv_expr): New.
13348 (pta_valueize): New.
13349 (pointer_equiv_analyzer::visit_stmt): New.
13350 (pointer_equiv_analyzer::visit_edge): New.
13351 (hybrid_folder::value_of_expr): Call PTA.
13352 (hybrid_folder::value_on_edge): Same.
13353 (hybrid_folder::pre_fold_bb): New.
13354 (hybrid_folder::post_fold_bb): New.
13355 (hybrid_folder::pre_fold_stmt): New.
13356 (rvrp_folder::pre_fold_bb): New.
13357 (rvrp_folder::post_fold_bb): New.
13358 (rvrp_folder::pre_fold_stmt): New.
13359 (rvrp_folder::value_of_expr): Call PTA.
13360 (rvrp_folder::value_on_edge): Same.
13362 2021-06-08 Jakub Jelinek <jakub@redhat.com>
13365 * tree-inline.c (copy_tree_body_r): For OMP_CLAUSE_DEPEND don't
13366 check TREE_CODE if OMP_CLAUSE_DECL is NULL.
13368 2021-06-08 Richard Biener <rguenther@suse.de>
13370 PR middle-end/100951
13371 * tree-vect-generic.c (expand_vector_piecewise): Build a
13372 VECTOR_CST if all elements are constant.
13373 (expand_vector_condition): Likewise.
13374 (lower_vec_perm): Likewise.
13375 (expand_vector_conversion): Likewise.
13377 2021-06-08 Martin Liska <mliska@suse.cz>
13379 * doc/invoke.texi: Document new param evrp-sparse-threshold.
13381 2021-06-08 Martin Liska <mliska@suse.cz>
13383 * genautomata.c (create_automata): Fix typo.
13385 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
13387 PR tree-optimization/100794
13388 * tree-predcom.c (tree_predictive_commoning_loop): Add parameter
13389 allow_unroll_p and only allow unrolling when it's true.
13390 (tree_predictive_commoning): Add parameter allow_unroll_p and
13392 (run_tree_predictive_commoning): Likewise.
13393 (pass_predcom::gate): Check flag_tree_loop_vectorize and
13394 global_options_set.x_flag_predictive_commoning.
13395 (pass_predcom::execute): Adjust for allow_unroll_p.
13397 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
13399 * tree-predcom.c (execute_pred_commoning): Remove update_ssa call.
13400 (tree_predictive_commoning_loop): Factor some cleanup stuffs into
13401 lambda function cleanup, remove scev_reset call, and adjust return
13403 (tree_predictive_commoning): Adjust for different changed values,
13404 only set flag TODO_update_ssa_only_virtuals if changed.
13405 (pass_data pass_data_predcom): Remove TODO_update_ssa_only_virtuals
13406 from todo_flags_finish.
13408 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
13410 * gimple-range-cache.cc (class sbr_sparse_bitmap): New.
13411 (sbr_sparse_bitmap::sbr_sparse_bitmap): New.
13412 (sbr_sparse_bitmap::bitmap_set_quad): New.
13413 (sbr_sparse_bitmap::bitmap_get_quad): New.
13414 (sbr_sparse_bitmap::set_bb_range): New.
13415 (sbr_sparse_bitmap::get_bb_range): New.
13416 (sbr_sparse_bitmap::bb_range_p): New.
13417 (block_range_cache::block_range_cache): initialize bitmap obstack.
13418 (block_range_cache::~block_range_cache): Destruct obstack.
13419 (block_range_cache::set_bb_range): Decide when to utilze the
13420 sparse on entry cache.
13421 * gimple-range-cache.h (block_range_cache): Add bitmap obstack.
13422 * params.opt (-param=evrp-sparse-threshold): New.
13424 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
13426 * bitmap.c (bitmap_set_aligned_chunk): New.
13427 (bitmap_get_aligned_chunk): New.
13428 (test_aligned_chunk): New.
13429 (bitmap_c_tests): Call test_aligned_chunk.
13430 * bitmap.h (bitmap_set_aligned_chunk, bitmap_get_aligned_chunk): New.
13432 2021-06-07 Uroš Bizjak <ubizjak@gmail.com>
13435 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
13437 (ix86_expand_vector_init_one_nonzero): Ditto.
13438 (ix86_expand_vector_init_one_var): Ditto.
13439 (ix86_expand_vector_init_general): Ditto.
13440 * config/i386/mmx.md (vec_initv4qiqi): New expander.
13442 2021-06-07 Jeff Law <jeffreyalaw@gmail.com>
13444 * config/h8300/movepush.md: Change most _clobber_flags
13445 patterns to instead use <cczn> subst.
13446 (movsi_cczn): New pattern with usable CC cases split out.
13447 (movsi_h8sx_cczn): Likewise.
13449 2021-06-07 Martin Liska <mliska@suse.cz>
13451 * common/common-target.def: Split long lines and replace them
13453 * target.def: Likewise.
13454 * doc/tm.texi: Re-generated.
13456 2021-06-07 Jakub Jelinek <jakub@redhat.com>
13459 * fold-const.c (fold_read_from_vector): Return NULL if trying to
13460 read from a CONSTRUCTOR with vector type elements.
13462 2021-06-07 Jakub Jelinek <jakub@redhat.com>
13464 PR middle-end/100898
13465 * tree-inline.c (copy_bb): Only use gimple_call_arg_ptr if memcpy
13466 should copy any arguments. Don't call gimple_call_num_args
13467 on id->call_stmt or call_stmt more than once.
13469 2021-06-07 liuhongt <hongtao.liu@intel.com>
13472 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): Refine
13474 (<insn>v4siv4di2): Delete constraints for define_expand.
13476 2021-06-07 liuhongt <hongtao.liu@intel.com>
13479 * config/i386/i386-expand.c (ix86_expand_builtin): Remove
13480 assignment of cfun->machine->has_explicit_vzeroupper.
13481 * config/i386/i386-features.c
13482 (ix86_add_reg_usage_to_vzerouppers): Delete.
13483 (ix86_add_reg_usage_to_vzeroupper): Ditto.
13484 (rest_of_handle_insert_vzeroupper): Remove
13485 ix86_add_reg_usage_to_vzerouppers, add df_analyze at the end
13487 (gate): Remove cfun->machine->has_explicit_vzeroupper.
13488 * config/i386/i386-protos.h (ix86_expand_avx_vzeroupper):
13490 * config/i386/i386.c (ix86_insn_callee_abi): New function.
13491 (ix86_initialize_callee_abi): Ditto.
13492 (ix86_expand_avx_vzeroupper): Ditto.
13493 (ix86_hard_regno_call_part_clobbered): Adjust for vzeroupper
13495 (TARGET_INSN_CALLEE_ABI): Define as ix86_insn_callee_abi.
13496 (ix86_emit_mode_set): Call ix86_expand_avx_vzeroupper
13498 * config/i386/i386.h (struct GTY(()) machine_function): Delete
13499 has_explicit_vzeroupper.
13500 * config/i386/i386.md (enum unspec): New member
13502 (ABI_DEFAULT,ABI_VZEROUPPER,ABI_UNKNOWN): New
13503 define_constants for insn callee abi index.
13504 * config/i386/predicates.md (vzeroupper_pattern): Adjust.
13505 * config/i386/sse.md (UNSPECV_VZEROUPPER): Deleted.
13506 (avx_vzeroupper): Call ix86_expand_avx_vzeroupper.
13507 (*avx_vzeroupper): Rename to ..
13508 (avx_vzeroupper_callee_abi): .. this, and adjust pattern as
13509 call_insn which has a special vzeroupper ABI.
13510 (*avx_vzeroupper_1): Deleted.
13512 2021-06-07 liuhongt <hongtao.liu@intel.com>
13515 * df-scan.c (df_get_call_refs): When call_insn is a fake call,
13516 it won't use stack pointer reg.
13517 * final.c (leaf_function_p): When call_insn is a fake call, it
13518 won't affect caller as a leaf function.
13519 * reg-stack.c (callee_clobbers_any_stack_reg): New.
13520 (subst_stack_regs): When call_insn doesn't clobber any stack
13521 reg, don't clear the arguments.
13522 * rtl.c (shallow_copy_rtx): Don't clear flag used when orig is
13524 * shrink-wrap.c (requires_stack_frame_p): No need for stack
13525 frame for a fake call.
13526 * rtl.h (FAKE_CALL_P): New macro.
13528 2021-06-06 Eric Botcazou <ebotcazou@adacore.com>
13530 * config/sparc/sparc-protos.h (order_regs_for_local_alloc): Rename
13532 (sparc_order_regs_for_local_alloc): ...this.
13533 (sparc_leaf_reg_remap): Declare.
13534 * config/sparc/sparc.h (ADJUST_REG_ALLOC_ORDER): Adjust.
13535 (LEAF_REG_REMAP): Reimplement as call to sparc_leaf_reg_remap.
13536 * config/sparc/sparc.c (leaf_reg_remap): Delete.
13537 (order_regs_for_local_alloc): Rename to...
13538 (sparc_order_regs_for_local_alloc): ...this.
13539 (sparc_leaf_reg_remap): New function.
13540 (sparc_conditional_register_usage): Do not modify leaf_reg_remap.
13542 2021-06-06 David Edelsohn <dje.gcc@gmail.com>
13544 * config/rs6000/rs6000.c (rs6000_xcoff_asm_output_aligned_decl_common):
13545 Use assemble_name to output BSS section name.
13547 2021-06-06 Uroš Bizjak <ubizjak@gmail.com>
13549 * config/i386/constraints.md (Bs):
13550 Remove boolean operators from match_test RTX.
13553 (M): Use "mode" variable instead of GET_MODE (op) in match_test RTX.
13556 2021-06-06 Martin Liska <mliska@suse.cz>
13558 * doc/extend.texi: Add missing @headitem.
13559 * doc/invoke.texi: Likewise.
13560 * doc/objc.texi: Likewise.
13562 2021-06-06 Martin Liska <mliska@suse.cz>
13564 * genhooks.c (emit_findices): Remove unused function.
13565 (emit_documentation): Do not call emit_findices
13566 and do not search for @Fcode directives.
13568 2021-06-06 Martin Liska <mliska@suse.cz>
13570 * doc/invoke.texi: Remove extra character.
13572 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
13574 * config/sh/sh.md (doloop_end_split): Fix empty split condition.
13576 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
13578 * config/sparc/sparc.md (*snedi<W:mode>_zero_vis3,
13579 *neg_snedi<W:mode>_zero_subxc, *plus_snedi<W:mode>_zero,
13580 *plus_plus_snedi<W:mode>_zero, *minus_snedi<W:mode>_zero,
13581 *minus_minus_snedi<W:mode>_zero): Fix empty split condition.
13583 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
13585 * config/or1k/or1k.md (*movdi): Fix empty split condition.
13587 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
13589 * config/mips/mips.md (<anonymous>, bswapsi2, bswapdi2): Fix empty
13592 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
13594 * config/m68k/m68k.md (*zero_extend_inc, *zero_extend_dec,
13595 *zero_extendsidi2): Fix empty split condition.
13597 2021-06-05 Jeff Law <jeffreyalaw@gmail.com>
13599 * config/h8300/addsub.md: Fix split condition in define_insn_and_split
13601 * config/h8300/bitfield.md: Likewise.
13602 * config/h8300/combiner.md: Likewise.
13603 * config/h8300/divmod.md: Likewise.
13604 * config/h8300/extensions.md: Likewise.
13605 * config/h8300/jumpcall.md: Likewise.
13606 * config/h8300/movepush.md: Likewise.
13607 * config/h8300/multiply.md: Likewise.
13608 * config/h8300/other.md: Likewise.
13609 * config/h8300/shiftrotate.md: Likewise.
13610 * config/h8300/logical.md: Likewise. Fix split pattern to use
13611 code iterator that somehow slipped through.
13613 2021-06-04 Tobias Burnus <tobias@codesourcery.com>
13615 PR middle-end/100905
13616 * tree-nested.c (convert_nonlocal_omp_clauses,
13617 convert_local_omp_clauses): Handle OMP_CLAUSE_BIND.
13619 2021-06-04 Martin Sebor <msebor@redhat.com>
13621 PR middle-end/100732
13622 * gimple-fold.c (gimple_fold_builtin_sprintf): Avoid folding calls
13623 with either source or destination argument of invalid type.
13624 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Avoid checking
13625 calls with arguments of invalid type.
13627 2021-06-04 Martin Sebor <msebor@redhat.com>
13629 * attribs.c (init_attr_rdwr_indices): Use VLA bounds in the expected
13631 (attr_access::vla_bounds): Also handle VLA bounds.
13633 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
13635 * config/i386/predicates.md (GOT_memory_operand):
13636 Implement using match_code RTXes.
13637 (GOT32_symbol_operand): Ditto.
13639 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
13642 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
13644 (ix86_expand_vector_init_general): Ditto.
13645 Use SImode instead of word_mode for logic operations
13646 when GET_MODE_SIZE (mode) < UNITS_PER_WORD.
13647 (expand_vec_perm_even_odd_1): Assert that V2HI mode should be
13648 implemented by expand_vec_perm_1.
13649 (expand_vec_perm_broadcast_1): Assert that V2HI and V4HI modes
13650 should be implemented using standard shuffle patterns.
13651 (ix86_vectorize_vec_perm_const): Handle V2HImode. Add V4HI and
13652 V2HI modes to modes, implementable with shuffle for one operand.
13653 * config/i386/mmx.md (*punpckwd): New insn_and_split pattern.
13654 (*pshufw_1): New insn pattern.
13655 (*vec_dupv2hi): Ditto.
13656 (vec_initv2hihi): New expander.
13658 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
13660 * config/arm/vfp.md (no_literal_pool_df_immediate,
13661 no_literal_pool_sf_immediate): Fix empty split condition.
13663 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
13665 * config/i386/i386.md (*load_tp_x32_zext, *add_tp_x32_zext,
13666 *tls_dynamic_gnu2_combine_32): Fix empty split condition.
13667 * config/i386/sse.md (*<sse2_avx2>_pmovmskb_lt,
13668 *<sse2_avx2>_pmovmskb_zext_lt, *sse2_pmovmskb_ext_lt,
13669 *<sse4_1_avx2>_pblendvb_lt): Likewise.
13671 2021-06-04 Jakub Jelinek <jakub@redhat.com>
13674 * config/i386/i386-expand.c (ix86_expand_vector_init): Handle
13675 concatenation from half-sized modes with TImode elements.
13677 2021-06-04 Claudiu Zissulescu <claziss@synopsys.com>
13679 * config/arc/arc.c (arc_override_options): Disable millicode
13680 thunks when RF16 is on.
13682 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
13684 * config/rs6000/rs6000.h (PROMOTE_MODE): Remove.
13686 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
13688 * config/rs6000/rs6000-call.c (rs6000_promote_function_mode):
13689 Replace PROMOTE_MODE marco with its content.
13691 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
13693 * config/cris/cris.md (*addi_reload): Fix empty split condition.
13695 2021-06-03 Jim Wilson <jimw@sifive.com>
13697 * config.gcc (riscv*-*-*): If --with-riscv-attribute not used,
13698 turn it on for all riscv targets.
13700 2021-06-03 Uroš Bizjak <ubizjak@gmail.com>
13703 * config/i386/i386-expand.c (ix86_expand_vector_set):
13704 Handle V2HI and V4QI modes.
13705 (ix86_expand_vector_extract): Ditto.
13706 * config/i386/mmx.md (*pinsrw): New insn pattern.
13709 (*pextrw_zext): Ditto.
13711 (*pextrb_zext): Ditto.
13712 (vec_setv2hi): New expander.
13713 (vec_extractv2hihi): Ditto.
13714 (vec_setv4qi): Ditto.
13715 (vec_extractv4qiqi): Ditto.
13716 (vec_setv8qi): Enable only for TARGET_SSE4_1.
13717 (vec_extractv8qiqi): Ditto.
13719 2021-06-03 Aaron Sawdey <acsawdey@linux.ibm.com>
13721 * config/rs6000/genfusion.pl (gen_logical_addsubf): Fix input
13722 order to subf instruction.
13723 * config/rs6000/fusion.md: Regenerate.
13725 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
13727 * calls.c (get_size_range): Use range_of_expr instead of
13728 determine_value_range.
13729 * tree-affine.c (expr_to_aff_combination): Same.
13730 * tree-data-ref.c (split_constant_offset): Same.
13731 * tree-vrp.c (determine_value_range_1): Remove.
13732 (determine_value_range): Remove.
13733 * tree-vrp.h (determine_value_range): Remove.
13735 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
13737 * function-tests.c (test_ranges): Call gimple_range_tests.
13738 * gimple-range-cache.cc (ranger_cache::range_of_expr): Pass stmt
13740 * gimple-range.cc (fur_source::get_operand): Do not call
13741 get_tree_range or gimple_range_global.
13743 (get_tree_range): Move to value-query.cc.
13744 Call get_arith_expr_range.
13745 (gimple_ranger::range_of_expr): Add argument to get_tree_range.
13746 Include gimple-range-tests.cc.
13747 * gimple-range.h (fold_range): Add argument.
13748 (get_tree_range): Remove.
13749 * selftest.h (gimple_range_tests): New.
13750 * value-query.cc (global_range_query::range_of_expr): Add
13752 (range_query::get_tree_range): Move from gimple-range.cc.
13753 * value-query.h (class range_query): Add get_tree_range and
13754 get_arith_expr_range. Make fur_source a friend.
13755 * vr-values.c (vr_values::range_of_expr): Pass stmt to
13757 * gimple-range-tests.cc: New file.
13759 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
13761 * gimple-range.cc (gimple_ranger::export_global_ranges): Call
13762 update_global_range.
13763 * value-query.cc (update_global_range): New.
13764 * value-query.h (update_global_range): New.
13766 2021-06-03 David Malcolm <dmalcolm@redhat.com>
13768 * diagnostic-show-locus.c (diagnostic_show_locus): Don't reject
13769 printing the same location twice if there are fix-it hints,
13770 multiple locations, or a label.
13772 2021-06-03 Andre Vieira <andre.simoesdiasvieira@arm.com>
13774 * tree-vect-loop.c (vect_transform_loop): Use main loop's various'
13775 thresholds to narrow the upper bound on epilogue iterations.
13777 2021-06-03 Christophe Lyon <christophe.lyon@linaro.org>
13779 * config/arm/mve.md (mve_vabsq_f<mode>): Use 'abs' instead of unspec.
13780 (mve_vabsq_s<mode>): Likewise.
13781 * config/arm/neon.md (abs<mode>2): Rename to neon_abs<mode>2.
13782 * config/arm/unspecs.md (VABSQ_F, VABSQ_S): Delete.
13783 * config/arm/vec-common.md (neg<mode>2): Rename to
13784 <absneg_str><mode>2.
13786 2021-06-03 Claudiu Zissulescu <claziss@synopsys.com>
13788 * common/config/arc/arc-common.c (arc_option_optimization_table):
13789 Remove malign-call.
13790 * config/arc/arc.c (arc_unalign_branch_p): Remove unused function.
13791 * config/arc/arc.h (TARGET_MIXED_CODE): Remove macro.
13792 (INDEX_REG_CLASS): Only refer to GENERAL_REGS.
13793 * config/arc/arc.md (abssi2_mixed): Remove pattern.
13794 * config/arc/arc.opt (munalign-prob-threshold): Mark it obsolete.
13795 (malign-call): Likewise.
13796 (mmixed-code): Likewise.
13797 * doc/invoke.texi (ARC): Update doc.
13799 2021-06-03 Martin Liska <mliska@suse.cz>
13801 * common.opt: Use proper Enum values.
13802 * opts.c (COVERAGE_SANITIZER_OPT): Remove.
13803 (parse_sanitizer_options): Handle only sanitizer_opts.
13804 (common_handle_option): Just assign value.
13806 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
13809 * tree-inline.c (inline_forbidden_p): Remove test on return type.
13811 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
13813 * dwarf2out.c (loc_list_from_tree_1) <FUNCTION_DECL>: Also generate
13814 DW_OP_GNU_variable_value referencing an existing DIE at file scope.
13815 (type_byte_size): Inline into...
13816 (add_byte_size_attribute): ...this and call add_scalar_info.
13818 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
13820 * dwarf2out.c (mem_loc_descriptor) <UDIV>: Fix typo.
13821 (typed_binop_from_tree): New function.
13822 (loc_list_from_tree_1) <EXACT_DIV_EXPR>: For an unsigned type,
13823 turn a divide by a power of 2 into a shift.
13824 <CEIL_DIV_EXPR>: For an unsigned type, use a signed divide if the
13825 size of the mode is lower than DWARF2_ADDR_SIZE; otherwise, do a
13826 typed divide by calling typed_binop_from_tree.
13828 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
13830 * dwarf2out.c (scompare_loc_descriptor): Fix head comment.
13831 (is_handled_procedure_type): Likewise.
13832 (struct loc_descr_context): Add strict_signedness field.
13833 (resolve_args_picking_1): Deal with DW_OP_[GNU_]deref_type,
13834 DW_OP_[GNU_]convert and DW_OP_[GNU_]reinterpret.
13835 (resolve_args_picking): Minor tweak.
13836 (function_to_dwarf_procedure): Initialize strict_signedness field.
13837 (type_byte_size): Likewise.
13838 (field_byte_offset): Likewise.
13839 (gen_descr_array_type_die): Likewise.
13840 (gen_variant_part): Likewise.
13841 (loc_list_from_tree_1) <CALL_EXPR>: Tidy up and set strict_signedness
13842 to true when a context is present before evaluating the arguments.
13843 <COND_EXPR>: Do not generate a useless comparison with zero.
13844 When dereferencing an address, if strict_signedness is true and the
13845 type is small and signed, use DW_OP_deref_type to do the dereference
13846 and then DW_OP_convert to convert back to the generic type.
13848 2021-06-03 Jakub Jelinek <jakub@redhat.com>
13851 * tree-inline.c (copy_tree_body_r): Handle iterators on
13852 OMP_CLAUSE_AFFINITY or OMP_CLAUSE_DEPEND.
13854 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
13856 * config/arc/arc.md (*bbit_di): Remove.
13858 2021-06-02 Christoph Muellner <cmuellner@gcc.gnu.org>
13860 PR rtl-optimization/100264
13861 * ree.c (get_sub_rtx): Ignore SET expressions without register
13862 destinations and remove assertion, as it is not valid anymore
13863 with this new behaviour.
13864 (merge_def_and_ext): Eliminate destination check for register
13865 as such SET expressions can't occur anymore.
13866 (combine_reaching_defs): Likewise.
13868 2021-06-02 Jakub Jelinek <jakub@redhat.com>
13871 * config/xtensa/xtensa.h (LEAF_REG_REMAP): Cast REGNO to int to avoid
13872 -Wtype-limits warnings.
13873 (DWARF_FRAME_REGISTER): Rewrite into ternary operator with addition
13874 in operands to avoid -Wsign-compare warnings.
13876 2021-06-02 Pat Haugen <pthaugen@linux.ibm.com>
13878 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): Use
13881 2021-06-02 Vineet Gupta <vgupta@synopsys.com>
13883 * config/arc/arc.h (TARGET_CPU_DEFAULT): Change to hs38_linux.
13885 2021-06-02 Ilya Leoshkevich <iii@linux.ibm.com>
13887 * config/s390/s390.md(*ashrdi3_31<setcc><cconly>): Use a single
13889 * config/s390/subst.md(cconly_subst): Use a single constraint
13890 in (match_scratch).
13892 2021-06-02 Martin Liska <mliska@suse.cz>
13894 * ipa-icf.h: Use auto_vec for memory_access_types.
13896 2021-06-02 Jeff Law <jeffreyalaw@gmail.com>
13898 * config/h8300/h8300-protos.h (compute_a_shift_length): Drop unused
13899 argument from prototype.
13900 (output_logical_op): Add rtx_code argument.
13901 (compute_logical_op_length): Likewise.
13902 * config/h8300/h8300.c (h8300_and_costs): Pass additional argument
13903 to compute_a_shift_length.
13904 (output_logical_op); New argument with the rtx code rather than
13905 extracting it from an operand. Handle QImode too.
13906 (compute_logical_op_length): Similary.
13907 (compute_a_shift_length): Drop unused argument.
13908 * config/h8300/h8300.md (logicals): New code iterator.
13909 * config/h8300/logical.md (<code><mode>3 expander): Combine
13910 the "and" expander with the "ior"/"xor" expander.
13911 (bclr<mode>msx): Combine the QI/HI mode patterns.
13912 (<logical><mode>3 insns): Use code iterator rather than match_operator.
13913 Handle QImode as well. Update call to output_logical_op and
13914 compute_logical_op_length to pass in rtx_code
13915 Fix split condition on all define_insn_and_split patterns.
13916 (one_cmpl<mode>2<cczn>): Use <cczn> to support both clobbering
13917 the flags and setting ZN via existing define_subst.
13918 * config/h8300/shiftrotate.md: Drop unused argument from
13919 calls to compute_a_shift_length.
13920 Signed-off-by: Jeff Law <jeffreyalaw@gmail.com>
13922 2021-06-01 Andrew Pinski <apinski@marvell.com>
13924 PR tree-optimization/25290
13925 * tree-ssa-phiopt.c (match_simplify_replacement):
13927 (tree_ssa_phiopt_worker): Use match_simplify_replacement.
13928 (two_value_replacement): Change the comment about
13929 conditional_replacement.
13930 (conditional_replacement): Delete.
13932 2021-06-01 Andrew Pinski <apinski@marvell.com>
13934 PR tree-optimization/95481
13935 * tree-tailcall.c (find_tail_calls): Handle empty typed
13938 2021-06-01 Andrew Pinski <apinski@marvell.com>
13940 * gimplify.c (zero_sized_field_decl): Delete
13941 (zero_sized_type): Delete
13942 (gimplify_init_ctor_eval): Use is_empty_type instead
13943 of zero_sized_field_decl.
13944 (gimplify_modify_expr): Use is_empty_type instead of
13947 2021-06-01 Jason Merrill <jason@redhat.com>
13950 * tree.h (CALL_FROM_NEW_OR_DELETE_P): Adjust comment.
13952 2021-06-01 Jason Merrill <jason@redhat.com>
13955 * diagnostic.h (warning_enabled_at): Declare.
13956 * diagnostic.c (diagnostic_enabled): Factor out from...
13957 (diagnostic_report_diagnostic): ...here.
13958 (warning_enabled_at): New.
13960 2021-06-01 Aldy Hernandez <aldyh@redhat.com>
13962 * gimple-ssa-evrp.c: Enable exporting of global ranges.
13964 2021-06-01 Martin Liska <mliska@suse.cz>
13967 * doc/invoke.texi: Mention that -fgcse-after-reload
13968 is enabled with -O3.
13970 2021-06-01 liuhongt <hongtao.liu@intel.com>
13972 PR tree-optimization/98365
13973 * tree-if-conv.c (strip_nop_cond_scalar_reduction): New function.
13974 (is_cond_scalar_reduction): Handle nop_expr in cond scalar reduction.
13975 (convert_scalar_cond_reduction): Ditto.
13976 (predicate_scalar_phi): Ditto.
13978 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
13980 PR tree-optimization/100781
13981 * gimple-range-cache.cc (ranger_cache::ranger_cache): Enable new
13982 value calculation by default.
13983 (ranger_cache::enable_new_values): New.
13984 (ranger_cache::disable_new_values): New.
13985 (ranger_cache::push_poor_value): Check if new values are allowed.
13986 * gimple-range-cache.h (class ranger_cache): New member/methods.
13987 * gimple-range.cc (gimple_ranger::range_of_expr): Check for debug
13988 statement, and disable/renable new value calculation.
13990 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
13992 * gimple-range-cache.cc (ranger_cache::ssa_range_in_bb): Delete.
13993 (ranger_cache::range_of_def): New.
13994 (ranger_cache::entry_range): New.
13995 (ranger_cache::exit_range): New.
13996 (ranger_cache::range_of_expr): Adjust.
13997 (ranger_cache::range_on_edge): Adjust.
13998 (ranger_cache::propagate_cache): Call exit_range directly.
13999 * gimple-range-cache.h (class ranger_cache): Adjust.
14001 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
14003 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust for
14004 gori_compute being a member rather than base class.
14005 dervied call to member call.
14006 (ranger_cache::dump): No longer dump gori_map.
14007 (ranger_cache::dump_bb): New.
14008 (ranger_cache::get_non_stale_global_range): Adjust for gori_compute
14009 being a member rather than base class.
14010 (ranger_cache::set_global_range): Ditto.
14011 (ranger_cache::ssa_range_in_bb): Ditto.
14012 (ranger_cache::range_of_expr): New.
14013 (ranger_cache::range_on_edge): New.
14014 (ranger_cache::block_range): Adjust for gori_computes. Debug changes.
14015 (ranger_cache::propagate_cache): Adjust debugging output.
14016 (ranger_cache::fill_block_cache): Adjust for gori_computes. Debug
14018 * gimple-range-cache.h (class ranger_cache): Make gori_compute a
14019 member, and inherit from range_query instead.
14020 (ranger_cache::dump_bb): New. split from dump.
14021 * gimple-range-gori.cc (gori_compute::ssa_range_in_bb): Delete.
14022 (gori_compute::expr_range_at_stmt): Delete.
14023 (gori_compute::compute_name_range_op): Delete.
14024 (gori_compute::compute_operand_range_switch): Add fur_source.
14025 (gori_compute::compute_operand_range): Add fur_source param, inline
14026 old compute_name_range_op and optimize_logical_operands.
14027 (struct tf_range): Delete.
14028 (gori_compute::logical_combine): Adjust
14029 (gori_compute::optimize_logical_operands): Delete.
14030 (gori_compute::compute_logical_operands_in_chain): Delete.
14031 (gori_compute::compute_logical_operands): Adjust.
14032 (gori_compute::compute_operand1_range): Adjust to fur_source.
14033 (gori_compute::compute_operand2_range): Ditto.
14034 (gori_compute::compute_operand1_and_operand2_range): Ditto.
14035 (gori_compute::outgoing_edge_range_p): Add range_query parameter,
14036 and adjust to fur_source.
14037 * gimple-range-gori.h (class gori_compute): Simplify and adjust to
14038 range_query and fur_source.
14039 * gimple-range.cc (gimple_ranger::range_on_edge): Query range_on_edge
14040 from the ranger_cache..
14041 (gimple_ranger::fold_range_internal): Adjust to base class change of
14043 (gimple_ranger::dump_bb): Adjust dump.
14044 * gimple-range.h (gimple_ranger):export gori computes object.
14046 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
14048 PR tree-optimization/100774
14049 * gimple-range-cache.cc (ranger_cache::get_non_stale_global_range):
14050 Constant values are also not stale.
14051 (ranger_cache::set_global_range): Range invariant values should also
14052 have the correct timestamp.
14054 2021-05-31 Martin Liska <mliska@suse.cz>
14056 * tree-streamer-in.c (unpack_ts_function_decl_value_fields):
14057 Unpack FUNCTION_DECL_DECL_TYPE.
14058 * tree-streamer-out.c (pack_ts_function_decl_value_fields):
14059 Stream FUNCTION_DECL_DECL_TYPE instead of
14060 DECL_IS_OPERATOR_NEW_P.
14061 * tree.h (set_function_decl_type): Use FUNCTION_DECL_DECL_TYPE
14063 (DECL_IS_OPERATOR_NEW_P): Likewise.
14064 (DECL_IS_OPERATOR_DELETE_P): Likewise.
14065 (DECL_LAMBDA_FUNCTION_P): Likewise.
14067 2021-05-31 Richard Biener <rguenther@suse.de>
14070 * internal-fn.c (expand_SHUFFLEVECTOR): Define.
14071 * internal-fn.def (SHUFFLEVECTOR): New.
14072 * internal-fn.h (expand_SHUFFLEVECTOR): Declare.
14073 * doc/extend.texi: Document __builtin_shufflevector.
14075 2021-05-31 Peter Bergner <bergner@linux.ibm.com>
14078 * config/rs6000/predicates.md(mma_assemble_input_operand): Allow
14079 indexed form addresses.
14081 2021-05-29 Jeff Law <jlaw@tachyum.com>
14083 * config/h8300/h8300.c (h8300_emit_stack_adjustment): Drop unused
14084 parameter. Call callers fixed.
14086 (output_plussi): Add FALLTHRU markers.
14087 (h8300_shift_needs_scratch_p): Add gcc_unreachable marker.
14089 2021-05-29 Jakub Jelinek <jakub@redhat.com>
14091 PR middle-end/99928
14092 * gimplify.c (gimplify_scan_omp_clauses): For taskloop simd
14093 combined with parallel, make sure to add shared clause to
14094 parallel for explicit linear clause.
14096 2021-05-29 Aldy Hernandez <aldyh@redhat.com>
14098 PR tree-optimization/100787
14099 * gimple-ssa-evrp.c: Disable exporting of global ranges.
14101 2021-05-28 Jason Merrill <jason@redhat.com>
14103 * tree-iterator.h (struct tree_stmt_iterator): Add operator++,
14104 operator--, operator*, operator==, and operator!=.
14105 (class tsi_range): New.
14107 2021-05-28 Richard Biener <rguenther@suse.de>
14109 PR tree-optimization/100778
14110 * tree-vect-slp.c (vect_build_slp_tree_1): Prevent possibly
14111 trapping ops in different BBs.
14113 2021-05-28 Richard Biener <rguenther@suse.de>
14116 * tree-inline.c (copy_bb): When processing __builtin_va_arg_pack
14117 copy fntype from original call.
14119 2021-05-28 Martin Liska <mliska@suse.cz>
14121 PR gcov-profile/100751
14122 * doc/gcov.texi: Revert partially a hunk that was wrong.
14124 2021-05-28 Cooper Qu <cooper.qu@linux.alibaba.com>
14126 * config/csky/csky-linux-elf.h (HAVE_sync_compare_and_swapqi):
14128 (HAVE_sync_compare_and_swaphi): Likewise.
14129 (HAVE_sync_compare_and_swapsi): Likewise.
14131 2021-05-28 Jakub Jelinek <jakub@redhat.com>
14133 PR middle-end/99928
14134 * tree.h (OMP_CLAUSE_MAP_IMPLICIT): Define.
14136 2021-05-28 Tobias Burnus <tobias@codesourcery.com>
14138 * gimplify.c (gimplify_omp_affinity): New.
14139 (gimplify_scan_omp_clauses): Call it; remove affinity clause afterwards.
14140 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_AFFINITY.
14141 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_AFFINITY.
14142 * tree.c (omp_clause_num_ops, omp_clause_code_name): Add clause.
14143 (walk_tree_1): Handle OMP_CLAUSE_AFFINITY.
14145 2021-05-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
14146 Richard Biener <rguenther@suse.de>
14148 * match.pd <popcount & / + pattern matching>:
14149 When generating popcount directly fails, try doing it in two halves.
14151 2021-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
14153 * Makefile.in (generated_files): Add gimple-match.c and
14156 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
14158 * gensupport.c (alter_predicate_for_insn): Handle MATCH_DUP.
14160 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
14162 * gensupport.c (alter_constraints): Add MATCH_SCRATCH case.
14164 2021-05-28 Kewen Lin <linkw@linux.ibm.com>
14166 PR tree-optimization/99398
14167 * tree-ssa-forwprop.c (simplify_permutation): Optimize some cases
14168 where the fed operands are CTOR/CST and propagated through
14169 VIEW_CONVERT_EXPR. Call vec_perm_indices::new_shrunk_vector.
14170 * vec-perm-indices.c (vec_perm_indices::new_shrunk_vector): New
14172 * vec-perm-indices.h (vec_perm_indices::new_shrunk_vector): New
14175 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
14177 * config/i386/mmx.md (addv2sf3): Do not call
14178 ix86_fixup_binary_operands_no_copy.
14181 (<smaxmin:code>v2sf3): Ditto.
14182 (<plusminus:insn><MMXMODEI:mode>3): Ditto.
14183 (<plusminus:insn><VI_32:mode>3): Remove expander.
14184 (<plusminus:insn><VI_32:mode>3): Rename from
14185 "*<plusminus:insn><VI_32:mode>3".
14186 (mulv4hi): Do not call ix86_fixup_binary_operands_no_copy.
14187 (mulv2hi3): Remove expander.
14188 (mulv2hi3): Rename from *mulv2hi3.
14189 (<s>mulv2hi3_highpart): Remove expander.
14190 (<s>mulv2hi3_highpart): Rename from *<s>mulv2hi3_highpart.
14191 (<smaxmin:code><MMXMODE14:mode>3): Rename from
14192 "*mmx_<smaxmin:code><MMXMODE14:mode>3".
14193 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Remove expander.
14194 (SMAXMIN_MMXMODEI): Remove mode iterator.
14195 (<smaxmin:code>v4hi3): New expander.
14196 (<smaxmin:code>v4qi3): Rename from *<smaxmin:code>v4qi3.
14197 (<smaxmin:code>v2hi3): Rename from *<smaxmin:code>v2hi3.
14198 (<smaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
14199 (SMAXMIN_VI_32): Remove mode iterator.
14200 (<umaxmin:code><MMXMODE24:mode>3): Rename from
14201 "*mmx_<umaxmin:code><MMXMODE24:mode>3".
14202 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Remove expander.
14203 (UMAXMIN_MMXMODEI): Remove mode iterator.
14204 (<umaxmin:code>v8qi3): New expander.
14205 (<umaxmin:code>v4qi3): Rename from *<umaxmin:code>v4qi3.
14206 (<umaxmin:code>v2hi3): Rename from *<umaxmin:code>v2hi3.
14207 (<umaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
14208 (UMAXMIN_VI_32): Remove mode iterator.
14209 (<any_shift:insn>v2hi3): Remove expander.
14210 (<any_shift:insn>v2hi3): Rename from *<any_shift:insn>v2hi3.
14211 (<any_logic:code><MMXMODEI:mode>3): Do not call
14212 ix86_fixup_binary_operands_no_copy.
14213 (<any_logic:code><VI_32:mode>3): Remove expander.
14214 (<any_logic:code><VI_32:mode>3): Rename from
14215 "*<any_logic:code><VI_32:mode>3".
14216 (uavg<mode>3_ceil): Do not call ix86_fixup_binary_operands_no_copy.
14217 * config/i386/sse.md (div<VF2:mode>3): Do not call
14218 ix86_fixup_binary_operands_no_copy.
14219 (div<VF1:mode>3): Ditto.
14220 (<maxmin:code><VI8_AVX2_AVX512F:mode>3): Ditto.
14221 (smulhrsv4hi3): Ditto.
14222 (smulhrsv2hi3): Ditto.
14224 2021-05-27 Martin Sebor <msebor@redhat.com>
14226 * ggc.h (gt_ggc_mx): Add overloads for all integers.
14228 * hash-map.h (class hash_map): Add pch_nx_helper overloads for all
14230 (hash_map::operator==): New function.
14232 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
14235 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
14236 For TARGET_XOP bypass SSE comparisons for all supported vector modes.
14237 * config/i386/mmx.md (*xop_maskcmp<MMXMODEI:mode>3): New insn pattern.
14238 (*xop_maskcmp<VI_32:mode>3): Ditto.
14239 (*xop_maskcmp_uns<MMXMODEI:mode>3): Ditto.
14240 (*xop_maskcmp_uns<VI_32:mode>3): Ditto.
14242 2021-05-27 Richard Earnshaw <rearnsha@arm.com>
14245 * config/arm/arm.c (arm_configure_build_target): Remove parameter
14246 opts_set, directly check opts parameters for being non-null.
14247 (arm_option_restore): Update call to arm_configure_build_target.
14248 (arm_option_override): Likewise.
14249 (arm_can_inline_p): Likewise.
14250 (arm_valid_target_attribute_tree): Likewise.
14251 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
14252 * config/arm/arm-protos.h (arm_configure_build_target): Adjust
14255 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
14257 * vr-values.c (simplify_conversion_using_ranges): Use
14258 get_range_query instead of get_global_range_query.
14260 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
14262 * gimple-range.cc (get_range_global): Move to value-query.cc.
14263 (gimple_range_global): Same.
14264 (get_global_range_query): Same.
14265 (global_range_query::range_of_expr): Same.
14266 * gimple-range.h (class global_range_query): Move to
14268 (gimple_range_global): Same.
14269 * tree-ssanames.c (get_range_info): Move to value-query.cc.
14270 (get_ptr_nonnull): Same.
14271 * tree-ssanames.h (get_range_info): Remove.
14272 (get_ptr_nonnull): Remove.
14273 * value-query.cc (get_ssa_name_range_info): Move from
14275 (get_ssa_name_ptr_info_nonnull): Same.
14276 (get_range_global): Move from gimple-range.cc.
14277 (gimple_range_global): Same.
14278 (get_global_range_query): Same.
14279 (global_range_query::range_of_expr): Same.
14280 * value-query.h (class global_range_query): Move from
14282 (gimple_range_global): Same.
14284 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
14287 * config/i386/mmx.md (uavgv4qi3_ceil): New insn pattern.
14288 (uavgv2hi3_ceil): Ditto.
14290 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
14293 * doc/extend.texi (scalar_storage_order): Rephrase slightly.
14295 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
14297 * tree-ssanames.c (get_range_info): Merge both copies of
14298 get_range_info into one that works with irange.
14299 * tree-ssanames.h (get_range_info): Remove version that works on
14302 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
14304 * builtins.c (check_nul_terminated_array): Convert to get_range_query.
14305 (expand_builtin_strnlen): Same.
14306 (determine_block_size): Same.
14307 * fold-const.c (expr_not_equal_to): Same.
14308 * gimple-fold.c (size_must_be_zero_p): Same.
14309 * gimple-match-head.c: Include gimple-range.h.
14310 * gimple-pretty-print.c (dump_ssaname_info): Convert to get_range_query.
14311 * gimple-ssa-warn-restrict.c
14312 (builtin_memref::extend_offset_range): Same.
14313 * graphite-sese-to-poly.c (add_param_constraints): Same.
14314 * internal-fn.c (get_min_precision): Same.
14315 * ipa-fnsummary.c (set_switch_stmt_execution_predicate): Same.
14316 * ipa-prop.c (ipa_compute_jump_functions_for_edge): Same.
14318 * tree-data-ref.c (split_constant_offset): Same.
14319 (dr_step_indicator): Same.
14320 * tree-dfa.c (get_ref_base_and_extent): Same.
14321 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
14322 * tree-ssa-loop-niter.c (refine_value_range_using_guard): Same.
14323 (determine_value_range): Same.
14324 (record_nonwrapping_iv): Same.
14325 (infer_loop_bounds_from_signedness): Same.
14326 (scev_var_range_cant_overflow): Same.
14327 * tree-ssa-phiopt.c (two_value_replacement): Same.
14328 * tree-ssa-pre.c (insert_into_preds_of_block): Same.
14329 * tree-ssa-reassoc.c (optimize_range_tests_to_bit_test): Same.
14330 * tree-ssa-strlen.c (handle_builtin_stxncpy_strncat): Same.
14332 (dump_strlen_info): Same.
14333 (set_strlen_range): Same.
14334 (maybe_diag_stxncpy_trunc): Same.
14335 (get_len_or_size): Same.
14336 (handle_integral_assign): Same.
14337 * tree-ssa-structalias.c (find_what_p_points_to): Same.
14338 * tree-ssa-uninit.c (find_var_cmp_const): Same.
14339 * tree-switch-conversion.c (bit_test_cluster::emit): Same.
14340 * tree-vect-patterns.c (vect_get_range_info): Same.
14341 (vect_recog_divmod_pattern): Same.
14342 * tree-vrp.c (intersect_range_with_nonzero_bits): Same.
14343 (register_edge_assert_for_2): Same.
14344 (determine_value_range_1): Same.
14345 * tree.c (get_range_pos_neg): Same.
14346 * vr-values.c (vr_values::get_lattice_entry): Same.
14347 (vr_values::update_value_range): Same.
14348 (simplify_conversion_using_ranges): Same.
14350 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
14352 * gimple-ssa-warn-alloca.c (alloca_call_type): Use
14353 get_range_query instead of query argument.
14354 (pass_walloca::execute): Enable and disable global ranger.
14356 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
14358 * gimple-ssa-evrp.c (rvrp_folder::rvrp_folder): Call
14360 (rvrp_folder::~rvrp_folder): Call disable_ranger.
14361 (hybrid_folder::hybrid_folder): Call enable_ranger.
14362 (hybrid_folder::~hybrid_folder): Call disable_ranger.
14364 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
14366 * function.c (allocate_struct_function): Set cfun->x_range_query.
14367 * function.h (struct function): Declare x_range_query.
14368 (get_range_query): New.
14369 (get_global_range_query): New.
14370 * gimple-range-cache.cc (ssa_global_cache::ssa_global_cache):
14371 Remove call to safe_grow_cleared.
14372 * gimple-range.cc (get_range_global): New.
14373 (gimple_range_global): Move from gimple-range.h.
14374 (get_global_range_query): New.
14375 (global_range_query::range_of_expr): New.
14376 (enable_ranger): New.
14377 (disable_ranger): New.
14378 * gimple-range.h (gimple_range_global): Move to gimple-range.cc.
14379 (class global_range_query): New.
14380 (enable_ranger): New.
14381 (disable_ranger): New.
14382 * gimple-ssa-evrp.c (evrp_folder::~evrp_folder): Rename
14383 dump_all_value_ranges to dump.
14384 * tree-vrp.c (vrp_prop::finalize): Same.
14385 * value-query.cc (range_query::dump): New.
14386 * value-query.h (range_query::dump): New.
14387 * vr-values.c (vr_values::dump_all_value_ranges): Rename to...
14388 (vr_values::dump): ...this.
14389 * vr-values.h (class vr_values): Rename dump_all_value_ranges to
14390 dump and make virtual.
14392 2021-05-26 Uroš Bizjak <ubizjak@gmail.com>
14394 * config/i386/i386.c (ix86_autovectorize_vector_modes):
14395 Add V4QImode and V16QImode for TARGET_SSE2.
14396 * doc/sourcebuild.texi (Vector-specific attributes):
14397 Add vect64 and vect32 description.
14399 2021-05-26 Bernd Edlinger <bernd.edlinger@hotmail.de>
14401 * gimple-range-gori.cc (range_def_chain::register_dependency):
14402 Resize m_def_chain when needed.
14404 2021-05-26 Christophe Lyon <christophe.lyon@linaro.org>
14406 * config/arm/mve.md (mve_vaddvq_<supf><mode>): Prefix with '@'.
14407 * config/arm/neon.md (reduc_plus_scal_<mode>): Move to ..
14408 * config/arm/vec-common.md: .. here. Add support for MVE.
14410 2021-05-26 Jakub Jelinek <jakub@redhat.com>
14412 * config/epiphany/epiphany.c (epiphany_print_operand_address): Remove
14414 * config/microblaze/microblaze.c (microblaze_legitimize_address,
14416 microblaze_option_override, print_operand): Likewise.
14417 * config/microblaze/microblaze.md (call_internal_plt,
14418 call_value_intern_plt, call_value_intern): Likewise.
14419 * config/arm/aout.h (ASM_OUTPUT_ALIGN): Likewise.
14420 * config/iq2000/iq2000.md (call_internal1, call_value_internal1,
14421 call_value_multiple_internal1): Likewise.
14422 * config/bfin/bfin.c (symbolic_reference_mentioned_p): Likewise.
14424 2021-05-26 Jan-Benedict Glaw <jbglaw@lug-owl.de>
14426 * config/arc/arc.c (arc_address_cost, arc_print_operand_address,
14427 arc_ccfsm_advance, symbolic_reference_mentioned_p,
14428 arc_raw_symbolic_reference_mentioned_p): Remove register
14431 2021-05-26 Jakub Jelinek <jakub@redhat.com>
14434 * omp-low.c: Include omp-offload.h.
14435 (create_omp_child_function): If current_function_decl has
14436 "omp declare target" attribute and is_gimple_omp_offloaded,
14437 remove that attribute from the copy of attribute list and
14438 add "omp target entrypoint" attribute instead.
14439 (lower_omp_target): Mark .omp_data_sizes.* and .omp_data_kinds.*
14440 variables for offloading if in omp_maybe_offloaded_ctx.
14441 * omp-offload.c (pass_omp_target_link::execute): Nullify second
14442 argument to GOMP_target_data_ext in offloaded code.
14444 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
14446 * config/csky/csky.c (csky_can_change_mode_class): Delete.
14447 For csky, HF/SF mode use the low bits of VREGS.
14449 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
14451 * gimplify.c (gimplify_decl_expr): Do not clear TREE_READONLY on a
14452 DECL which is a reference for OMP.
14454 2021-05-26 Martin Liska <mliska@suse.cz>
14456 PR gcov-profile/100751
14457 * doc/gcov.texi: Document that __gcov_dump can be called just
14458 once and that __gcov_reset resets run-time counters.
14460 2021-05-26 Martin Liska <mliska@suse.cz>
14462 * doc/install.texi: Port relevant part from install-old.texi
14463 and re-generate list of CPUs and systems.
14465 2021-05-26 Martin Liska <mliska@suse.cz>
14467 * Makefile.in: Remove it.
14468 * doc/include/fdl.texi: Update next/previous chapters.
14469 * doc/install.texi: Likewise.
14470 * doc/install-old.texi: Removed.
14472 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
14474 * config/csky/csky.c (ck810_legitimate_index_p): Support
14475 "base + index" with DF mode.
14476 * config/csky/constraints.md ("Y"): New constraint for memory operands
14477 without index register.
14478 * config/csky/csky_insn_fpuv2.md (fpuv3_movdf): Use "Y" instead of "m"
14479 when mov between memory and general registers, and lower their priority.
14480 * config/csky/csky_insn_fpuv3.md (fpuv2_movdf): Likewise.
14482 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
14484 * config/csky/csky.c (TARGET_PROMOTE_PROTOTYPES): Delete.
14486 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
14488 * config/csky/csky.md (untyped_call): Emit clobber for return
14489 registers to mark them used.
14491 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
14493 * config/csky/csky.md (cskyv2_sextend_ldbs): New.
14495 2021-05-26 Andrew Pinski <apinski@marvell.com>
14497 * match.pd (x < 0 ? ~y : y): New patterns.
14499 2021-05-26 Andrew Pinski <apinski@marvell.com>
14501 * match.pd (A?CST1:CST2): Add simplifcations for A?0:+-1, A?+-1:0,
14502 A?POW2:0 and A?0:POW2.
14504 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14506 * gimple-range-gori.cc (class logical_stmt_cache): Delete
14507 (logical_stmt_cache::logical_stmt_cache ): Delete.
14508 (logical_stmt_cache::~logical_stmt_cache): Delete.
14509 (logical_stmt_cache::cache_entry::dump): Delete.
14510 (logical_stmt_cache::get_range): Delete.
14511 (logical_stmt_cache::cached_name ): Delete.
14512 (logical_stmt_cache::same_cached_name): Delete.
14513 (logical_stmt_cache::cacheable_p): Delete.
14514 (logical_stmt_cache::slot_diagnostics ): Delete.
14515 (logical_stmt_cache::dump): Delete.
14516 (gori_compute_cache::gori_compute_cache): Delete.
14517 (gori_compute_cache::~gori_compute_cache): Delete.
14518 (gori_compute_cache::compute_operand_range): Delete.
14519 (gori_compute_cache::cache_stmt): Delete.
14520 * gimple-range-gori.h (gori_compute::compute_operand_range): Remove
14522 (class gori_compute_cache): Delete.
14524 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14526 * gimple-range.cc (fold_using_range::range_of_range_op): Use m_gori
14528 (fold_using_range::range_of_address): Adjust.
14529 (fold_using_range::range_of_phi): Adjust.
14530 * gimple-range.h (class fur_source): Adjust.
14531 (fur_source::fur_source): Adjust.
14533 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14535 * gimple-range-gori.cc (gori_compute::expr_range_at_stmt): Rename
14536 from expr_range_in_bb and adjust.
14537 (gori_compute::compute_name_range_op): Adjust.
14538 (gori_compute::optimize_logical_operands): Adjust.
14539 (gori_compute::compute_logical_operands_in_chain): Adjust.
14540 (gori_compute::compute_operand1_range): Adjust.
14541 (gori_compute::compute_operand2_range): Adjust.
14542 (ori_compute_cache::cache_stmt): Adjust.
14543 * gimple-range-gori.h (gori_compute): Rename prototype.
14545 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14547 * gimple-range.cc (gimple_ranger::range_of_expr): Non-null should be
14548 checked only after range_of_stmt, not range_on_entry.
14549 (gimple_ranger::range_on_entry): Check for non-null in any
14550 predecessor block, if it is not already non-null.
14551 (gimple_ranger::range_on_exit): DOnt check for non-null after
14552 range on entry call.
14553 (gimple_ranger::dump_bb): New. Split from dump.
14554 (gimple_ranger::dump): Adjust.
14555 * gimple-range.h (class gimple_ranger): Adjust.
14557 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14559 * gimple-range-cache.cc (struct range_timestamp): Delete.
14560 (class temporal_cache): Adjust.
14561 (temporal_cache::get_timestamp): Delete.
14562 (temporal_cache::set_dependency): Delete.
14563 (temporal_cache::temporal_value): Adjust.
14564 (temporal_cache::current_p): Take dependencies as params.
14565 (temporal_cache::set_timestamp): Adjust.
14566 (temporal_cache::set_always_current): Adjust.
14567 (ranger_cache::get_non_stale_global_range): Adjust.
14568 (ranger_cache::register_dependency): Delete.
14569 * gimple-range-cache.h (class range_cache): Adjust.
14571 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14573 * gimple-range-gori.cc (range_def_chain::range_def_chain): init
14575 (range_def_chain::~range_def_chain): Dispose of obstack rather than
14576 each individual bitmap.
14577 (range_def_chain::set_import): New.
14578 (range_def_chain::get_imports): New.
14579 (range_def_chain::chain_import_p): New.
14580 (range_def_chain::register_dependency): Rename from build_def_chain
14582 (range_def_chain::def_chain_in_bitmap_p): New.
14583 (range_def_chain::add_def_chain_to_bitmap): New.
14584 (range_def_chain::has_def_chain): Just check first depenedence.
14585 (range_def_chain::get_def_chain): Process imports, use generic
14586 register_dependency routine.
14587 (range_def_chain::dump): New.
14588 (gori_map::gori_map): Allocate import list.
14589 (gori_map::~gori_map): Release imports.
14590 (gori_map::exports): Check for past allocated block size.
14591 (gori_map::imports): New.
14592 (gori_map::def_chain_in_export_p): Delete.
14593 (gori_map::is_import_p): New.
14594 (gori_map::maybe_add_gori): Handle imports.
14595 (gori_map::dump): Adjust output, add imports.
14596 (gori_compute::has_edge_range_p): Remove def_chain_in_export call.
14597 (gori_export_iterator::gori_export_iterator): New.
14598 (gori_export_iterator::next): New.
14599 (gori_export_iterator::get_name): New.
14600 * gimple-range-gori.h (range_def_chain): Add imports and direct
14601 dependecies via struct rdc.
14602 (range_def_chain::depend1): New.
14603 (range_def_chain::depend2): New.
14604 (class gori_map): Adjust.
14605 (FOR_EACH_GORI_IMPORT_NAME): New.
14606 (FOR_EACH_GORI_EXPORT_NAME): New.
14607 (class gori_export_iterator): New.
14609 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14611 * gimple-range-cache.cc (ranger_cache::ranger_cache): Move initial
14612 export cache filling to here.
14613 * gimple-range-gori.cc (gori_compute::gori_compute) : From Here.
14615 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14617 * gimple-range-gori.cc (range_def_chain): Move to gimple-range-gori.h.
14618 (gori_map): Move to gimple-range-gori.h.
14619 (gori_compute::gori_compute): Adjust.
14620 (gori_compute::~gori_compute): Delete.
14621 (gori_compute::compute_operand_range_switch): Adjust.
14622 (gori_compute::compute_operand_range): Adjust.
14623 (gori_compute::compute_logical_operands): Adjust.
14624 (gori_compute::has_edge_range_p ): Adjust.
14625 (gori_compute::set_range_invariant): Delete.
14626 (gori_compute::dump): Adjust.
14627 (gori_compute::outgoing_edge_range_p): Adjust.
14628 * gimple-range-gori.h (class range_def_chain): Relocate here.
14629 (class gori_map): Relocate here.
14630 (class gori_compute): Inherit from gori_map, and adjust.
14632 2021-05-25 Aldy Hernandez <aldyh@redhat.com>
14634 * value-range.cc (range_tests_legacy): Use
14635 build_nonstandard_integer_type instead of int and short.
14637 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
14639 * gimplify.c (gimplify_decl_expr): Clear TREE_READONLY on the DECL
14640 when really creating an initialization statement for it.
14642 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
14644 * tree-inline.c (setup_one_parameter): Fix thinko in new condition.
14646 2021-05-25 Kito Cheng <kito.cheng@sifive.com>
14648 * config/riscv/riscv.h (ASM_SPEC): Pass -mno-relax.
14650 2021-05-25 Martin Liska <mliska@suse.cz>
14652 PR tree-optimization/92860
14654 * optc-save-gen.awk: Remove exceptions.
14656 2021-05-25 Martin Liska <mliska@suse.cz>
14658 * asan.h (sanitize_coverage_p): New function.
14659 * doc/extend.texi: Document it.
14660 * fold-const.c (fold_range_test): Use sanitize_flags_p
14661 instead of flag_sanitize_coverage.
14662 (fold_truth_andor): Likewise.
14663 * sancov.c: Likewise.
14664 * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
14665 * ipa-inline.c (sanitize_attrs_match_for_inline_p): Handle
14666 -fsanitize-coverage when inlining.
14668 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
14670 * config/csky/csky-modes.def : Fix copyright.
14672 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
14674 * config/csky/csky-modes.def : Amend copyright.
14675 * config/csky/csky_insn_fpuv2.md : Likewise.
14676 * config/csky/csky_insn_fpuv3.md : Likewise.
14678 2021-05-25 Richard Biener <rguenther@suse.de>
14680 PR middle-end/100727
14681 * calls.c (initialize_argument_information): Explicitely test
14682 for WITH_SIZE_EXPR.
14683 * gimple-expr.c (mark_addressable): Skip outer WITH_SIZE_EXPR.
14685 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
14687 * config/csky/csky.h (FRAME_POINTER_REGNUM): Use
14688 HARD_FRAME_POINTER_REGNUM and FRAME_POINTER_REGNUM instead of
14689 the signle definition. The signle definition may not work well
14690 at simplify_subreg_regno().
14691 (HARD_FRAME_POINTER_REGNUM): New.
14692 (ELIMINABLE_REGS): Add for HARD_FRAME_POINTER_REGNUM.
14693 * config/csky/csky.c (get_csky_live_regs, csky_can_eliminate,
14694 csky_initial_elimination_offset, csky_expand_prologue,
14695 csky_expand_epilogue): Add for HARD_FRAME_POINTER_REGNUM.
14697 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
14699 * config/csky/csky.c (csky_option_override):
14700 Init csky_arch_isa_features[] in advance, so TARGET_DSP
14701 and TARGET_DIV can be set well.
14703 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
14705 * config/csky/constraints.md ("l", "h"): Delete.
14706 * config/csky/csky.h (reg_class, REG_CLASS_NAMES,
14707 REG_CLASS_CONTENTS): Delete LO_REGS and HI_REGS.
14708 * config/csky/csky.c (regno_reg_classm,
14709 csky_secondary_reload, csky_register_move_cost):
14710 Use HILO_REGS instead of LO_REGS and HI_REGS.
14712 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
14714 * config/csky/constraints.md ("W"): New constriant for mem operand
14715 with base reg, index register.
14716 ("Q"): Renamed and modified "csky_valid_fpuv2_mem_operand" to
14717 "csky_valid_mem_constraint_operand" to deal with both "Q" and "W"
14719 ("Dv"): New constraint for const double value that can be used at
14721 * config/csky/csky-modes.def (HFmode): New mode.
14722 * config/csky/csky-protos.h (csky_valid_fpuv2_mem_operand): Rename
14723 to "csky_valid_mem_constraint_operand" and support new constraint
14725 (csky_get_movedouble_length): New.
14726 (fpuv3_output_move): New.
14727 (fpuv3_const_double): New.
14728 * config/csky/csky.c (csky_option_override): New arch CK860 with fpv3.
14729 (decompose_csky_address): Refine.
14730 (csky_print_operand): New "CONST_DOUBLE" operand.
14731 (csky_output_move): Support fpv3 instructions.
14732 (csky_get_movedouble_length): New.
14733 (fpuv3_output_move): New.
14734 (fpuv3_const_double): New.
14735 (csky_emit_compare): Cover float comparsion.
14736 (csky_emit_compare_float): Refine.
14737 (csky_vaild_fpuv2_mem_operand): Rename to
14738 "csky_valid_mem_constraint_operand" and support new constraint "W".
14739 (ck860_rtx_costs): New.
14740 (csky_rtx_costs): Add the cost calculation of CK860.
14741 (regno_reg_class): New vregs for fpuv3.
14742 (csky_dbx_regno): Likewise.
14743 (csky_cpu_cpp_builtins): New builtin macro for fpuv3.
14744 (csky_conditional_register_usage): Suporrot fpuv3.
14745 (csky_dwarf_register_span): Suporrot fpuv3.
14746 (csky_init_builtins, csky_mangle_type): Support "__fp16" type.
14747 (ck810_legitimate_index_p): Support fp16.
14748 * config/csky/csky.h (TARGET_TLS): ADD CK860.
14749 (CSKY_VREG_P, CSKY_VREG_LO_P, CSKY_VREG_HI_P): Support fpuv3.
14750 (TARGET_SINGLE_FPU): Support fpuv3.
14751 (TARGET_SUPPORT_FPV3): New.
14752 (FIRST_PSEUDO_REGISTER): Change to 202 to hold the new fpuv3 registers.
14753 (FIXED_REGISTERS, CALL_REALLY_USED_REGISTERS, REGISTER_NAMES,
14754 REG_CLASS_CONTENTS): Support fpuv3.
14755 * config/csky/csky.md (movsf): Move to cksy_insn_fpu.md and refine.
14756 (csky_movsf_fpv2): Likewise.
14757 (ck801_movsf): Likewise.
14758 (csky_movsf): Likewise.
14760 (csky_movdf_fpv2): Likewise.
14761 (ck801_movdf): Likewise.
14762 (csky_movdf): Likewise.
14763 (movsicc): Refine. Use "comparison_operatior" instead of
14764 "ordered_comparison_operatior".
14765 (addsicc): Likewise.
14766 (CSKY_FIRST_VFP3_REGNUM, CSKY_LAST_VFP3_REGNUM): New constant.
14767 (call_value_internal_vh): New.
14768 * config/csky/csky_cores.def (CK860): New arch and cpu.
14773 * config/csky/csky_insn_fpu.md: Refactor. Separate all float patterns
14774 into emit-patterns and match-patterns, remain the emit-patterns here,
14775 and move the match-patterns to csky_insn_fpuv2.md or
14776 csky_insn_fpuv3.md.
14777 * config/csky/csky_insn_fpuv2.md: New file for fpuv2 instructions.
14778 * config/csky/csky_insn_fpuv3.md: New file and new patterns for fpuv3
14780 * config/csky/csky_isa.def (fcr): New.
14785 (CK860): New definition for ck860.
14786 * config/csky/csky_tables.opt (ck860): New processors ck860,
14787 ck860f. And new arch ck860.
14792 * config/csky/predicates.md (csky_float_comparsion_operator): Delete
14793 "geu", "gtu", "leu", "ltu", which will never appear at float comparison.
14794 * config/csky/t-csky-elf: Support 860.
14795 * config/csky/t-csky-linux: Likewise.
14796 * doc/md.texi: Add "Q" and "W" constraints for C-SKY.
14798 2021-05-24 Aaron Sawdey <acsawdey@linux.ibm.com>
14800 * config/rs6000/genfusion.pl (gen_logical_addsubf): Refactor to
14801 add generation of logical-add and add-logical fusion pairs.
14802 * config/rs6000/rs6000-cpus.def: Add new fusion to ISA 3.1 mask
14804 * config/rs6000/rs6000.c (rs6000_option_override_internal): Turn on
14805 logical-add and add-logical fusion by default.
14806 * config/rs6000/rs6000.opt: Add -mpower10-fusion-logical-add and
14807 -mpower10-fusion-add-logical options.
14808 * config/rs6000/fusion.md: Regenerate file.
14810 2021-05-24 Aldy Hernandez <aldyh@redhat.com>
14812 * value-range.cc (irange::legacy_equal_p): Check type when
14813 comparing VR_VARYING types.
14814 (range_tests_legacy): Test comparing VARYING ranges of different
14817 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
14819 * config/aarch64/aarch64.c (neoversen1_tunings):
14820 Enable AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
14822 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
14824 * config/aarch64/aarch64.c (aarch64_classify_symbol): Use GOT for
14825 extern weak symbols. Limit symbol offsets for non-GOT symbols with
14828 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
14830 * config/arm/neon.md (vec_load_lanesxi<mode>)
14831 (vec_store_lanexoi<mode>): Move ...
14832 * config/arm/vec-common.md: here.
14834 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
14836 * config/arm/neon.md (vec_load_lanesoi<mode>)
14837 (vec_store_lanesoi<mode>): Move ...
14838 * config/arm/vec-common.md: here.
14840 2021-05-24 liuhongt <hongtao.liu@intel.com>
14843 * config/i386/i386.c (ix86_gimple_fold_builtin): Replacing
14844 stmt with GIMPLE_NOP when lhs doesn't exist.
14846 2021-05-23 Uroš Bizjak <ubizjak@gmail.com>
14849 * config/i386/mmx.md (*push<VI_32:mode>2_rex64):
14850 New instruction pattern.
14851 (*push<VI_32:mode>2): Ditto.
14852 (push splitter for SSE registers): New splitter.
14854 2021-05-23 Andrew Pinski <apinski@marvell.com>
14856 * match.pd ((A & C) != 0 ? D : 0): Limit to non pointer types.
14858 2021-05-22 Aaron Sawdey <acsawdey@linux.ibm.com>
14860 * config/rs6000/genfusion.pl (gen_addadd): Fix incorrect attr types.
14861 * config/rs6000/fusion.md: Regenerate file.
14863 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
14865 * config/rs6000/genfusion.pl (gen_addadd): New function.
14866 * config/rs6000/fusion.md: Regenerate file.
14867 * config/rs6000/rs6000-cpus.def: Add
14868 OPTION_MASK_P10_FUSION_2ADD to masks.
14869 * config/rs6000/rs6000.c (rs6000_option_override_internal):
14870 Handle default value of OPTION_MASK_P10_FUSION_2ADD.
14871 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2add.
14873 2021-05-21 Jakub Jelinek <jakub@redhat.com>
14875 PR middle-end/99928
14876 * tree.h (OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET): Define.
14877 * gimplify.c (enum gimplify_omp_var_data): Fix up
14878 GOVD_MAP_HAS_ATTACHMENTS value, add GOVD_FIRSTPRIVATE_IMPLICIT.
14879 (omp_lastprivate_for_combined_outer_constructs): If combined target
14880 has GOVD_FIRSTPRIVATE_IMPLICIT set for the decl, change it to
14881 GOVD_MAP | GOVD_SEEN.
14882 (gimplify_scan_omp_clauses): Set GOVD_FIRSTPRIVATE_IMPLICIT for
14883 firstprivate clauses with OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT.
14884 (gimplify_adjust_omp_clauses): For firstprivate clauses with
14885 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT either clear that bit and
14886 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET too, or remove it and
14887 let it be replaced by implicit map clause.
14889 2021-05-21 Jakub Jelinek <jakub@redhat.com>
14891 PR middle-end/99928
14892 * gimplify.c (omp_lastprivate_for_combined_outer_constructs): New
14894 (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LASTPRIVATE>: Use it.
14895 (gimplify_omp_for): Likewise.
14897 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14899 PR middle-end/90115
14900 * omp-low.c (oacc_privatization_candidate_p): Reject 'static',
14901 'external' in blocks.
14903 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14905 PR middle-end/90115
14906 * flag-types.h (enum openacc_privatization): New.
14907 * params.opt (-param=openacc-privatization): New.
14908 * doc/invoke.texi (openacc-privatization): Document it.
14909 * omp-general.h (get_openacc_privatization_dump_flags): New
14911 * omp-low.c (oacc_privatization_candidate_p): Add diagnostics.
14912 * omp-offload.c (execute_oacc_device_lower)
14913 <IFN_UNIQUE_OACC_PRIVATE>: Re-work diagnostics.
14914 * target.def (goacc.adjust_private_decl): Add 'location_t'
14916 * doc/tm.texi: Regenerate.
14917 * config/gcn/gcn-protos.h (gcn_goacc_adjust_private_decl): Adjust.
14918 * config/gcn/gcn-tree.c (gcn_goacc_adjust_private_decl): Likewise.
14919 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl):
14920 Likewise. Preserve it for...
14921 (nvptx_goacc_expand_var_decl): ... use here.
14923 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14925 * doc/sourcebuild.texi (Other attributes): Document '__OPTIMIZE__'
14928 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14930 PR middle-end/90115
14931 * omp-low.c (oacc_privatization_candidate_p): New function.
14932 (oacc_privatization_scan_clause_chain)
14933 (oacc_privatization_scan_decl_chain): Use it. Also
14934 'gcc_checking_assert' that we're not seeing duplicates.
14936 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14938 PR middle-end/90115
14939 * omp-offload.c (execute_oacc_device_lower): Skip processing if no
14942 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14944 PR middle-end/90115
14945 * omp-offload.c (execute_oacc_device_lower): Explain.
14947 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14949 PR middle-end/90115
14950 * omp-offload.c (execute_oacc_device_lower)
14951 <IFN_UNIQUE_OACC_PRIVATE>: Diagnose and handle for 'level == -1'
14953 * internal-fn.c (expand_UNIQUE): Don't expect
14954 'IFN_UNIQUE_OACC_PRIVATE'.
14956 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14958 PR middle-end/90115
14959 * omp-low.c (lower_omp_for): Don't evaluate OpenMP 'for' clauses.
14961 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14963 PR middle-end/90115
14964 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl)
14965 (nvptx_goacc_expand_var_decl): Tighten.
14967 2021-05-21 Julian Brown <julian@codesourcery.com>
14968 Chung-Lin Tang <cltang@codesourcery.com>
14969 Thomas Schwinge <thomas@codesourcery.com>
14971 PR middle-end/90115
14972 * doc/tm.texi.in (TARGET_GOACC_EXPAND_VAR_DECL)
14973 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Add documentation hooks.
14974 * doc/tm.texi: Regenerate.
14975 * expr.c (expand_expr_real_1): Expand decls using the
14976 expand_var_decl OpenACC hook if defined.
14977 * internal-fn.c (expand_UNIQUE): Handle IFN_UNIQUE_OACC_PRIVATE.
14978 * internal-fn.h (IFN_UNIQUE_CODES): Add OACC_PRIVATE.
14979 * omp-low.c (omp_context): Add oacc_privatization_candidates
14981 (lower_oacc_reductions): Add PRIVATE_MARKER parameter. Insert
14983 (lower_oacc_head_tail): Add PRIVATE_MARKER parameter. Modify
14984 private marker's gimple call arguments, and pass it to
14985 lower_oacc_reductions.
14986 (oacc_privatization_scan_clause_chain)
14987 (oacc_privatization_scan_decl_chain, lower_oacc_private_marker):
14989 (lower_omp_for, lower_omp_target, lower_omp_1): Use these.
14990 * omp-offload.c (convert.h): Include.
14991 (oacc_loop_xform_head_tail): Treat private-variable markers like
14992 fork/join when transforming head/tail sequences.
14993 (struct var_decl_rewrite_info): Add struct.
14994 (oacc_rewrite_var_decl, is_sync_builtin_call): New functions.
14995 (execute_oacc_device_lower): Support rewriting gang-private
14996 variables using target hook, and fix up addr_expr and var_decl
14998 * target.def (adjust_private_decl, expand_var_decl): New hooks.
14999 * config/gcn/gcn-protos.h (gcn_goacc_adjust_gangprivate_decl):
15001 (gcn_goacc_adjust_private_decl): ...this.
15002 * config/gcn/gcn-tree.c (gcn_goacc_adjust_gangprivate_decl):
15004 (gcn_goacc_adjust_private_decl): ...this. Add LEVEL parameter.
15005 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_GANGPRIVATE_DECL): Rename
15006 definition using gcn_goacc_adjust_gangprivate_decl...
15007 (TARGET_GOACC_ADJUST_PRIVATE_DECL): ...to this, using
15008 gcn_goacc_adjust_private_decl.
15009 * config/nvptx/nvptx.c (tree-pretty-print.h): Include.
15010 (gang_private_shared_size): New global variable.
15011 (gang_private_shared_align): Likewise.
15012 (gang_private_shared_sym): Likewise.
15013 (gang_private_shared_hmap): Likewise.
15014 (nvptx_option_override): Initialize these.
15015 (nvptx_file_end): Output gang_private_shared_sym.
15016 (nvptx_goacc_adjust_private_decl, nvptx_goacc_expand_var_decl):
15018 (nvptx_set_current_function): Clear gang_private_shared_hmap.
15019 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Define hook.
15020 (TARGET_GOACC_EXPAND_VAR_DECL): Likewise.
15022 2021-05-21 H.J. Lu <hjl.tools@gmail.com>
15024 * config/i386/i386-modes.def (MAX_BITSIZE_MODE_ANY_INT): Removed.
15026 2021-05-21 Richard Biener <rguenther@suse.de>
15027 H.J. Lu <hjl.tools@gmail.com>
15029 PR middle-end/90773
15030 * expr.c (expand_constructor): Elide expand_constructor if
15031 move by pieces is preferred.
15033 2021-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15035 * config/aarch64/aarch64-builtins.c (aarch64_call_properties):
15036 Take a flag and mode value as arguments.
15037 (aarch64_modifies_global_state_p): Likewise.
15038 (aarch64_reads_global_state_p): Likewise.
15039 (aarch64_could_trap_p): Likewise.
15040 (aarch64_get_attributes): Likewise.
15041 (aarch64_init_simd_builtins): Adjust callsite of above.
15042 (aarch64_init_fcmla_laneq_builtins): Use aarch64_get_attributes to get
15043 function attributes to apply to builtins.
15044 (aarch64_init_crc32_builtins): Likewise.
15045 (aarch64_init_builtin_rsqrt): Likewise.
15047 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
15049 * config/rs6000/rs6000.md (define_attr "type"): Add types for fusion.
15050 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10): Use new fusion types.
15051 (gen_2logical): Use new fusion types.
15052 * config/rs6000/fusion.md: Regenerate.
15054 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
15057 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
15058 Handle V4QI and V2HI modes.
15059 (ix86_expand_sse_movcc): Ditto.
15060 * config/i386/mmx.md (*<sat_plusminus:insn><VI_32:mode>3):
15061 New instruction pattern.
15062 (*eq<VI_32:mode>3): Ditto.
15063 (*gt<VI_32:mode>3): Ditto.
15064 (*xop_pcmov_<VI_32:mode>): Ditto.
15065 (mmx_pblendvb32): Ditto.
15066 (mmx_pblendvb64): Rename from mmx_pblendvb.
15067 (vec_cmp<VI_32:mode><VI_32:mode>): New expander.
15068 (vec_cmpu<VI_32:mode><VI_32:mode>): Ditto.
15069 (vcond<VI_32:mode><VI_32:mode>): Ditto.
15070 (vcondu<VI_32:mode><VI_32:mode>): Ditto.
15071 (vcond_mask_<VI_32:mode><VI_32:mode>): Ditto.
15073 2021-05-21 Jakub Jelinek <jakub@redhat.com>
15075 PR tree-optimization/94589
15076 * tree-ssa-phiopt.c (spaceship_replacement): For integral rhs1 and
15077 rhs2, treat x <= 4 equivalently to x < 5 etc. In cmp1 and cmp2 (if
15078 not the same as cmp3) treat <= the same as < and >= the same as >.
15079 Don't require that cond2_phi_edge is true edge, instead take
15080 false/true edges into account based on cmp1/cmp2 comparison kinds.
15082 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
15085 * config/i386/mmx.md (SMAXMIN_MMXMODEI): New mode iterator.
15086 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Macroize expander
15087 from <smaxmin:code>v4hi3> and <smaxmin:code><MMXMODE14:mode>3
15088 using SMAXMIN_MMXMODEI mode iterator.
15089 (*<smaxmin:code>v4qi3): New insn pattern.
15090 (*<smaxmin:code>v2hi3): Ditto.
15091 (SMAXMIN_VI_32): New mode iterator.
15092 (<smaxmin:code><SMAXMIN_VI_32>mode3): New expander.
15093 (UMAXMIN_MMXMODEI): New mode iterator.
15094 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Macroize expander
15095 from <umaxmin:code>v8qi3> and <umaxmin:code><MMXMODE24:mode>3
15096 using UMAXMIN_MMXMODEI mode iterator.
15097 (*<umaxmin:code>v4qi3): New insn pattern.
15098 (*<umaxmin:code>v2hi3): Ditto.
15099 (UMAXMIN_VI_32): New mode iterator.
15100 (<umaxmin:code><UMAXMIN_VI_32>mode3): New expander.
15101 (abs<VI_32:mode>2): New insn pattern.
15102 (ssse3_abs<MMXMODEI:mode>2, abs<MMXMODEI:mode>2): Move from ...
15103 * config/i386/sse.md: ... here.
15105 2021-05-20 Clement Chigot <clement.chigot@atos.net>
15106 David Edelsohn <dje.gcc@gmail.com>
15108 * collect2.c (scan_prog_file): Issue non-fatal warning for
15111 2021-05-20 Jonathan Wakely <jwakely@redhat.com>
15113 * doc/invoke.texi (-Wno-c++11-extensions)
15114 (-Wno-c++14-extensions, -Wno-c++17-extensions)
15115 (-Wno-c++20-extensions, -Wno-c++23-extensions): Document
15118 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
15120 * config/c6x/c6x.c (c6x_output_file_unwind): Use dwarf_debuginfo_p.
15121 * config/darwin.c (darwin_override_options): Likewise.
15122 * config/i386/cygming.h (DBX_REGISTER_NUMBER): Likewise.
15123 * config/i386/darwin.h (DBX_REGISTER_NUMBER): Likewise.
15124 (DWARF2_FRAME_REG_OUT): Likewise.
15125 * config/mips/mips.c (mips_output_filename): Likewise.
15126 * config/rs6000/rs6000.c (rs6000_xcoff_declare_function_name):
15128 (rs6000_dbx_register_number): Likewise.
15129 * dbxout.c: Include flags.h.
15130 * dwarf2cfi.c (cfi_label_required_p): Likewise.
15131 (dwarf2out_do_frame): Likewise.
15132 * except.c: Include flags.h.
15133 * final.c (dwarf2_debug_info_emitted_p): Likewise.
15134 (final_scan_insn_1): Likewise.
15135 * flags.h (dwarf_debuginfo_p): New function declaration.
15136 * opts.c (dwarf_debuginfo_p): New function definition.
15137 * targhooks.c (default_debug_unwind_info): Use dwarf_debuginfo_p.
15138 * toplev.c (process_options): Likewise.
15140 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
15142 * common.opt: Change type to support bitmasks.
15143 * flag-types.h (enum debug_info_type): Rename enumerator constants.
15144 (NO_DEBUG): New bitmask.
15145 (DBX_DEBUG): Likewise.
15146 (DWARF2_DEBUG): Likewise.
15147 (XCOFF_DEBUG): Likewise.
15148 (VMS_DEBUG): Likewise.
15149 (VMS_AND_DWARF2_DEBUG): Likewise.
15150 * flags.h (debug_set_to_format): New function declaration.
15151 (debug_set_count): Likewise.
15152 (debug_set_names): Likewise.
15153 * opts.c (debug_type_masks): Array of bitmasks for debug formats.
15154 (debug_set_to_format): New function definition.
15155 (debug_set_count): Likewise.
15156 (debug_set_names): Likewise.
15157 (set_debug_level): Update access to debug_type_names.
15158 * toplev.c: Likewise.
15160 2021-05-20 Martin Sebor <msebor@redhat.com>
15162 PR middle-end/100684
15163 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Handle C++ lambda.
15165 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
15168 * config/i386/i386.md (isa): Remove x64_bmi.
15169 (enabled): Remove x64_bmi.
15170 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
15171 Remove general register alternative.
15172 (*andnot<VI_32:mode>3): Ditto.
15173 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
15174 (*<any_logic:code><VI_32:mode>3): Ditto.
15176 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
15178 * config/arm/arm.c: Include head files tree-vectorizer.h and
15181 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
15184 * config/i386/mmx.md (Yv_Yw): Revert adding V4QI and V2HI modes.
15185 (*<plusminus:insn><VI_32:mode>3): Use Yw instad of <Yv_Yw> constrint.
15186 (<s>mulv4hi3_highpart): New expander.
15187 (*<s>mulv2hi3_highpart): New insn pattern.
15188 (<s>mulv2hi3_higpart): New expander.
15189 (*<any_shift:insn>v2hi3): New insn pattern.
15190 (<any_shift:insn>v2hi3): New expander.
15191 * config/i386/sse.md (smulhrsv2hi3): New expander.
15192 (*smulhrsv2hi3): New insn pattern.
15194 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
15196 * doc/invoke.texi (vect-inner-loop-cost-factor): Document new
15198 * params.opt (vect-inner-loop-cost-factor): New.
15199 * targhooks.c (default_add_stmt_cost): Replace hardcoded factor
15200 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR, include head file
15201 tree-vectorizer.h and its required ones.
15202 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Replace
15203 hardcoded factor 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR.
15204 * config/arm/arm.c (arm_add_stmt_cost): Likewise.
15205 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
15206 * config/rs6000/rs6000.c (rs6000_add_stmt_cost): Likewise.
15207 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
15209 (_loop_vec_info::_loop_vec_info): Init inner_loop_cost_factor.
15210 * tree-vectorizer.h (_loop_vec_info): Add inner_loop_cost_factor.
15211 (LOOP_VINFO_INNER_LOOP_COST_FACTOR): New macro.
15213 2021-05-20 Christophe Lyon <christophe.lyon@linaro.org>
15214 Torbjörn Svensson <torbjorn.svensson@st.com>
15217 * doc/cpp.texi (Common Predefined Macros): Document __FILE_NAME__.
15219 2021-05-20 Jakub Jelinek <jakub@redhat.com>
15221 PR middle-end/99928
15222 * gimplify.c (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LINEAR>: For
15223 explicit linear clause when combined with target, make it map(tofrom:)
15224 instead of no clause or firstprivate.
15226 2021-05-20 Jakub Jelinek <jakub@redhat.com>
15228 PR tree-optimization/94589
15229 * match.pd ((X & Y) == X -> (X & ~Y) == 0): Simplify even in presence
15230 of integral conversions.
15232 2021-05-19 Andrew MacLeod <amacleod@redhat.com>
15234 * gimple-range.cc (fur_source::get_operand): New.
15235 (gimple_range_fold): Delete.
15236 (fold_using_range::fold_stmt): Move from gimple_ranger::calc_stmt.
15237 (fold_using_range::range_of_range_op): Move from gimple_ranger.
15238 (fold_using_range::range_of_address): Ditto.
15239 (fold_using_range::range_of_phi): Ditto.
15240 (fold_using_range::range_of_call): Ditto.
15241 (fold_using_range::range_of_builtin_ubsan_call): Move from
15242 range_of_builtin_ubsan_call.
15243 (fold_using_range::range_of_builtin_call): Move from
15244 range_of_builtin_call.
15245 (gimple_ranger::range_of_builtin_call): Delete.
15246 (fold_using_range::range_of_cond_expr): Move from gimple_ranger.
15247 (gimple_ranger::fold_range_internal): New.
15248 (gimple_ranger::range_of_stmt): Use new fold_using_range API.
15249 (fold_using_range::range_of_ssa_name_with_loop_info): Move from
15250 gimple_ranger. Improve ranges of SSA_NAMES when possible.
15251 * gimple-range.h (gimple_ranger): Remove various range_of routines.
15252 (class fur_source): New.
15253 (class fold_using_range): New.
15254 (fur_source::fur_source): New.
15256 * vr-values.c (vr_values::extract_range_basic): Use fold_using_range
15257 instead of range_of_builtin_call.
15259 2021-05-19 Jonathan Wakely <jwakely@redhat.com>
15261 * doc/cpp.texi (Common Predefined Macros): Update documentation
15262 for the __GXX_EXPERIMENTAL_CXX0X__ macro.
15264 2021-05-19 Alex Coplan <alex.coplan@arm.com>
15267 * config/arm/arm.md (nonsecure_call_internal): Always ensure
15268 callee's address is in a register.
15270 2021-05-19 Geng Qi <gengqi@linux.alibaba.com>
15272 * common/config/riscv/riscv-common.c
15273 (riscv_subset_list::parsing_subset_version): Properly parse the letter
15275 (riscv_subset_list::parse_std_ext,
15276 riscv_subset_list::parse_multiletter_ext): To handle errors generated
15277 in riscv_subset_list::parsing_subset_version.
15279 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
15281 * config/aarch64/aarch64-simd.md: Use "neon_move_narrow_q"
15282 type attribute in patterns generating XTN(2).
15284 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
15286 * config/aarch64/aarch64-simd.md (aarch64_simd_vec_pack_trunc_<mode>):
15287 Remove as duplicate of...
15288 (aarch64_xtn<mode>): This.
15289 (aarch64_xtn2<mode>_le): Move position in file.
15290 (aarch64_xtn2<mode>_be): Move position in file.
15291 (aarch64_xtn2<mode>): Move position in file.
15292 (vec_pack_trunc_<mode>): Define as an expander.
15294 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
15296 * config/aarch64/aarch64-simd-builtins.def: Split builtin
15297 generation for aarch64_<sur>q<r>shr<u>n_n<mode> pattern into
15298 separate scalar and vector generators.
15299 * config/aarch64/aarch64-simd.md
15300 (aarch64_<sur>q<r>shr<u>n_n<mode>): Define as an expander and
15302 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_le): This and...
15303 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_be): This.
15304 * config/aarch64/iterators.md: Define SD_HSDI iterator.
15306 2021-05-19 Jonathn Wright <jonathan.wright@arm.com>
15308 * config/aarch64/aarch64-simd.md: Use UNSPEC_SQXTUN instead
15310 * config/aarch64/iterators.md: Remove UNSPEC_SQXTUN2.
15312 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
15314 * config/aarch64/aarch64-simd.md (aarch64_<sur>q<r>shr<u>n2_n<mode>):
15315 Implement as an expand emitting a big/little endian
15316 instruction pattern.
15317 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_le): Define.
15318 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_be): Define.
15320 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
15322 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn2<mode>):
15323 Implement as an expand emitting a big/little endian
15324 instruction pattern.
15325 (aarch64_<sur><addsub>hn2<mode>_insn_le): Define.
15326 (aarch64_<sur><addsub>hn2<mode>_insn_be): Define.
15327 * config/aarch64/iterators.md: Remove UNSPEC_[R]ADDHN2 and
15328 UNSPEC_[R]SUBHN2 unspecs and ADDSUBHN2 iterator.
15330 2021-05-19 Richard Biener <rguenther@suse.de>
15332 PR middle-end/100672
15333 * fold-const.c (fold_negate_expr_1): Use element_precision.
15334 (negate_expr_p): Likewise.
15336 2021-05-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
15338 * config/aarch64/iterators.md (SVE_PRED_LOAD): New iterator.
15339 (pred_load): New int attribute.
15340 * config/aarch64/aarch64-sve.md
15341 (aarch64_load_<ANY_EXTEND:optab><SVE_HSDI:mode><SVE_PARTIAL_I:mode>): Use
15342 SVE_PRED_LOAD enum iterator and corresponding pred_load attribute.
15343 * config/aarch64/aarch64-sve-builtins-base.cc (expand): Update call to
15344 code_for_aarch64_load.
15346 2021-05-19 Richard Biener <rguenther@suse.de>
15348 * cfgexpand.c (discover_nonconstant_array_refs_r): Make
15349 sure TARGET_MEM_REF bases are expanded as memory.
15350 * tree-ssa-operands.c (operands_scanner::get_tmr_operands):
15351 Do not mark TARGET_MEM_REF bases addressable.
15352 * tree-ssa.c (non_rewritable_mem_ref_base): Handle
15353 TARGET_MEM_REF bases as never rewritable.
15354 * gimple-walk.c (walk_stmt_load_store_addr_ops): Do not
15355 walk TARGET_MEM_REF bases as address-takens.
15356 * tree-ssa-dce.c (ref_may_be_aliased): Handle TARGET_MEM_REF.
15358 2021-05-19 Richard Biener <rguenther@suse.de>
15360 * builtins.c (get_object_alignment_1): Strip outer
15362 * tree-dfa.c (get_ref_base_and_extent): Handle outer
15363 WITH_SIZE_EXPR for size processing and process the
15365 * tree-ssa-alias.c (ao_ref_base_alias_set): Strip
15366 outer WITH_SIZE_EXPR.
15367 (ao_ref_base_alias_ptr_type): Likewise.
15368 (refs_may_alias_p_2): Allow WITH_SIZE_EXPR in ref->ref
15369 and handle that accordingly, stripping it for the
15370 core alias workers.
15371 * tree.c (get_base_address): Handle WITH_SIZE_EXPR by
15372 looking through it instead of returning NULL.
15374 2021-05-19 Jakub Jelinek <jakub@redhat.com>
15376 PR middle-end/100576
15377 * builtins.c (check_read_access): Convert bound to size_type_node if
15380 2021-05-19 Richard Biener <rguenther@suse.de>
15382 * tree-cfg.c (verify_types_in_gimple_min_lval): Inline...
15383 (verify_types_in_gimple_reference): ... here. Sanitize.
15384 (verify_gimple_call): Verify references in LHS and arguments.
15385 (verify_gimple_assign_single): Reject WITH_SIZE_EXPR.
15387 2021-05-19 Uroš Bizjak <ubizjak@gmail.com>
15389 * config/i386/i386.h (VALID_INT_MODE_P):
15390 Add V8QI, V4HI and V2SI modes for TARGET_64BIT.
15391 * config/i386/i386.md (isa): Add x64_bmi.
15392 (enabled): Handle x64_bmi.
15393 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
15394 Add alternative using 64bit general registers.
15395 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
15397 2021-05-19 Jakub Jelinek <jakub@redhat.com>
15399 PR middle-end/99928
15400 * tree.h (OMP_MASTER_COMBINED): Define.
15401 * gimplify.c (gimplify_scan_omp_clauses): Rewrite lastprivate
15402 handling for outer combined/composite constructs to a loop.
15403 Handle lastprivate on combined target.
15404 (gimplify_expr): Formatting fix.
15406 2021-05-19 Xionghu Luo <luoxhu@linux.ibm.com>
15408 * passes.def: Add sink_code pass before store_merging.
15409 * tree-ssa-sink.c (pass_sink_code:clone): New.
15411 2021-05-18 Bill Schmidt <wschmidt@linux.ibm.com>
15413 * config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Remove call to
15414 rs6000_special_adjust_field_align_p.
15415 * config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
15416 * config/rs6000/rs6000-call.c (rs6000_function_arg_boundary):
15417 Remove ABI warning.
15418 (rs6000_function_arg): Likewise.
15419 * config/rs6000/rs6000-protos.h
15420 (rs6000_special_adjust_field_align_p): Remove prototype.
15421 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
15423 * config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Remove call to
15424 rs6000_special_adjust_field_align_p.
15426 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
15429 * config/i386/i386.h (VALID_SSE2_REG_MODE):
15430 Add V4QI and V2HI modes.
15431 (VALID_INT_MODE_P): Ditto.
15432 * config/i386/mmx.md (VI_32): New mode iterator.
15433 (mmxvecsize): Handle V4QI and V2HI.
15435 (mov<VI_32:mode>): New expander.
15436 (*mov<mode>_internal): New insn pattern.
15437 (movmisalign<VI_32:mode>): New expander.
15438 (neg<VI_32:mode>): New expander.
15439 (<plusminus:insn><VI_32:mode>3): New expander.
15440 (*<plusminus:insn><VI_32:mode>3): New insn pattern.
15441 (mulv2hi3): New expander.
15442 (*mulv2hi3): New insn pattern.
15443 (one_cmpl<VI_32:mode>2): New expander.
15444 (*andnot<VI_32:mode>3): New insn pattern.
15445 (<any_logic:code><VI_32:mode>3): New expander.
15446 (*<any_logic:code><VI_32:mode>3): New insn pattern.
15448 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
15450 * config/i386/sse.md (<any_extend:insn>v4qiv4di2):
15451 Fix a mode mismatch with operand 1.
15453 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
15456 * config/i386/i386-expand.c (split_double_mode): Return
15457 temporary register when simplify_gen_subreg fails with
15458 the high half od the paradoxical subreg.
15460 2021-05-18 Richard Biener <rguenther@suse.de>
15462 * cfgexpand.c (expand_one_var): Pass in forced_stack_var
15463 and honor it when expanding.
15464 (expand_used_vars_for_block): Pass through forced_stack_var.
15465 (expand_used_vars): Likewise.
15466 (discover_nonconstant_array_refs_r): Set bits in
15467 forced_stack_vars instead of marking vars TREE_ADDRESSABLE.
15468 (avoid_type_punning_on_regs): Likewise.
15469 (discover_nonconstant_array_refs): Likewise.
15470 (pass_expand::execute): Create and pass down forced_stack_var
15471 bitmap. For parameters and returns temporarily set
15472 TREE_ADDRESSABLE when expand_function_start.
15474 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
15476 * doc/sourcebuild.texi: Document 'dg-note'.
15478 2021-05-18 Tobias Burnus <tobias@codesourcery.com>
15481 * configure: Regenerate.
15482 * configure.ac (BUILD_CFLAG, BUILD_CXXFLAGS): Add $(CFLAGS-$@).
15484 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
15486 * gimple.h (is_gimple_omp_oacc): Tighten.
15487 * omp-low.c (check_omp_nesting_restrictions): Adjust.
15489 2021-05-18 Richard Biener <rguenther@suse.de>
15491 * tree-ssa-operands.c (mark_address_taken): Simplify.
15493 2021-05-18 Martin Liska <mliska@suse.cz>
15495 * config/gcn/mkoffload.c (STR): Redefine.
15496 * config/i386/intelmic-mkoffload.c (STR): Likewise.
15497 * config/nvptx/mkoffload.c (STR): Likewise.
15499 2021-05-18 Martin Liska <mliska@suse.cz>
15501 * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
15502 Use startswith function instead of strncmp.
15503 * common/config/bfin/bfin-common.c (bfin_handle_option): Likewise.
15504 * common/config/riscv/riscv-common.c (riscv_subset_list::parse): Likewise.
15505 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_type): Likewise.
15506 * config/aarch64/aarch64.c (aarch64_process_one_target_attr): Likewise.
15507 * config/alpha/alpha.c (alpha_elf_section_type_flags): Likewise.
15508 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
15509 * config/arm/arm.c (arm_file_start): Likewise.
15510 (arm_valid_target_attribute_rec): Likewise.
15511 (thumb1_md_asm_adjust): Likewise.
15512 * config/arm/driver-arm.c (host_detect_local_cpu): Likewise.
15513 * config/avr/avr.c (STR_PREFIX_P): Likewise.
15514 (avr_set_current_function): Likewise.
15515 (avr_handle_addr_attribute): Likewise.
15516 (avr_asm_output_aligned_decl_common): Likewise.
15517 (avr_asm_named_section): Likewise.
15518 (avr_section_type_flags): Likewise.
15519 (avr_asm_select_section): Likewise.
15520 * config/c6x/c6x.c (c6x_in_small_data_p): Likewise.
15521 (c6x_section_type_flags): Likewise.
15522 * config/darwin-c.c (darwin_cfstring_ref_p): Likewise.
15523 (darwin_objc_declare_unresolved_class_reference): Likewise.
15524 (darwin_objc_declare_class_definition): Likewise.
15525 * config/darwin.c (indirect_data): Likewise.
15526 (darwin_encode_section_info): Likewise.
15527 (darwin_objc2_section): Likewise.
15528 (darwin_objc1_section): Likewise.
15529 (machopic_select_section): Likewise.
15530 (darwin_globalize_label): Likewise.
15531 (darwin_label_is_anonymous_local_objc_name): Likewise.
15532 (darwin_asm_named_section): Likewise.
15533 (darwin_asm_output_dwarf_offset): Likewise.
15534 * config/frv/frv.c (frv_string_begins_with): Likewise.
15535 (frv_in_small_data_p): Likewise.
15536 * config/gcn/mkoffload.c (STR): Likewise.
15538 * config/i386/i386-builtins.c (get_builtin_code_for_version): Likewise.
15539 * config/i386/i386-options.c (ix86_option_override_internal): Likewise.
15540 * config/i386/i386.c (x86_64_elf_section_type_flags): Likewise.
15541 (ix86_md_asm_adjust): Likewise.
15542 * config/i386/intelmic-mkoffload.c (STR): Likewise.
15543 * config/i386/winnt.c (i386_pe_asm_named_section): Likewise.
15544 (i386_pe_file_end): Likewise.
15545 * config/ia64/ia64.c (ia64_in_small_data_p): Likewise.
15546 (ia64_section_type_flags): Likewise.
15547 * config/mips/driver-native.c (host_detect_local_cpu): Likewise.
15548 * config/mips/mips.c (mips_handle_interrupt_attr): Likewise.
15549 (mips16_stub_function_p): Likewise.
15550 (mips_function_rodata_section): Likewise.
15551 * config/msp430/msp430.c (msp430_mcu_name): Likewise.
15552 (msp430_function_section): Likewise.
15553 (msp430_section_type_flags): Likewise.
15554 (msp430_expand_helper): Likewise.
15555 * config/nios2/nios2.c (nios2_small_section_name_p): Likewise.
15556 (nios2_valid_target_attribute_rec): Likewise.
15557 * config/nvptx/mkoffload.c (process): Likewise.
15559 * config/pa/som.h: Likewise.
15560 * config/pdp11/pdp11.c (pdp11_output_ident): Likewise.
15561 * config/riscv/riscv.c (riscv_elf_select_rtx_section): Likewise.
15562 * config/rs6000/rs6000.c (VTABLE_NAME_P): Likewise.
15563 (rs6000_inner_target_options): Likewise.
15564 * config/s390/driver-native.c (s390_host_detect_local_cpu): Likewise.
15565 * config/sparc/driver-sparc.c (host_detect_local_cpu): Likewise.
15566 * config/vax/vax.c (vax_output_int_move): Likewise.
15567 * config/vms/vms-ld.c (startswith): Likewise.
15568 (process_args): Likewise.
15570 * config/vms/vms.c: Likewise.
15572 2021-05-18 Jakub Jelinek <jakub@redhat.com>
15574 PR rtl-optimization/100590
15575 * regcprop.c (copyprop_hardreg_forward_1): Only DCE dead sets if
15576 they are NONJUMP_INSN_P.
15578 2021-05-18 Jakub Jelinek <jakub@redhat.com>
15581 * function.c (push_dummy_function): Set DECL_ARTIFICIAL and
15582 DECL_ASSEMBLER_NAME on the fn_decl.
15584 2021-05-18 Jakub Jelinek <jakub@redhat.com>
15586 PR tree-optimization/94589
15587 * tree-ssa-phiopt.c (spaceship_replacement): Pattern match
15588 phi result used in (res & ~1) == 0 comparison as res >= 0 as
15589 res == 2 would be UB with -ffinite-math-only.
15591 2021-05-18 Martin Liska <mliska@suse.cz>
15593 * Makefile.in: genversion.o should depend on DATESTAMP.
15595 2021-05-18 Claudiu Zissulescu <claziss@synopsys.com>
15597 * config/arc/simdext.md (negv2si2): Remove round bracket.
15599 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
15601 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Define
15602 _Bool as macro expanding to _Bool.
15604 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
15607 * tree.c (build_reference_type_for_mode)
15608 (build_pointer_type_for_mode): Pick pointer mode if MODE argument
15610 (build_reference_type, build_pointer_type): Invoke
15611 build_*_type_for_mode with VOIDmode.
15613 2021-05-17 Andrew MacLeod <amacleod@redhat.com>
15615 PR tree-optimization/100512
15616 * gimple-range-cache.cc (ranger_cache::set_global_range): Mark const
15617 and non-zero pointer ranges as invariant.
15618 * gimple-range.cc (gimple_ranger::range_of_stmt): Remove pointer
15619 processing from here.
15621 2021-05-17 Tom de Vries <tdevries@suse.de>
15624 * config/nvptx/nvptx-protos.h (nvptx_output_atomic_insn): Declare
15625 * config/nvptx/nvptx.c (nvptx_output_barrier)
15626 (nvptx_output_atomic_insn): New function.
15627 (nvptx_print_operand): Add support for 'B'.
15628 * config/nvptx/nvptx.md: Use nvptx_output_atomic_insn for atomic
15631 2021-05-17 Aldy Hernandez <aldyh@redhat.com>
15633 PR tree-optimization/100349
15634 * vr-values.c (bounds_of_var_in_loop): Bail if scev returns
15637 2021-05-17 Tamar Christina <tamar.christina@arm.com>
15639 * config/aarch64/driver-aarch64.c (DEFAULT_ARCH): New.
15640 (host_detect_local_cpu): Use it.
15642 2021-05-17 Martin Liska <mliska@suse.cz>
15644 * doc/invoke.texi: Add 2 missing dots.
15646 2021-05-17 Marius Hillenbrand <mhillen@linux.ibm.com>
15648 PR bootstrap/100552
15649 * configure.ac: Replace pattern substitution with call to sed.
15650 * configure: Regenerate.
15652 2021-05-17 Richard Biener <rguenther@suse.de>
15654 PR middle-end/100582
15655 * tree.c (array_at_struct_end_p): Get to the base of the
15656 reference before looking for the underlying decl.
15658 2021-05-17 Joern Rennecke <joern.rennecke@embecosm.com>
15660 * genoutput.c (validate_insn_alternatives) Make "wrong number of
15661 alternatives" message more specific, and remove assumption on where
15664 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
15666 * config/arm/iterators.md (V16): New iterator.
15667 (VH_cvtto): New iterator.
15668 (v_cmp_result): Added V4HF and V8HF support.
15669 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result>): Use VDQWH.
15670 (vcond<mode><mode>): Likewise.
15671 (vcond_mask_<mode><v_cmp_result>): Likewise.
15672 (vcond<VH_cvtto><mode>): New expander.
15674 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
15676 * config/arm/arm-protos.h (arm_expand_vector_compare): Update
15678 * config/arm/arm.c (arm_expand_vector_compare): Add support for
15680 (arm_expand_vcond): Likewise.
15681 * config/arm/iterators.md (supf): Remove VCMPNEQ_S, VCMPEQQ_S,
15682 VCMPEQQ_N_S, VCMPNEQ_N_S.
15683 (VCMPNEQ, VCMPEQQ, VCMPEQQ_N, VCMPNEQ_N): Remove.
15684 * config/arm/mve.md (@mve_vcmp<mve_cmp_op>q_<mode>): Add '@' prefix.
15685 (@mve_vcmp<mve_cmp_op>q_f<mode>): Likewise.
15686 (@mve_vcmp<mve_cmp_op>q_n_f<mode>): Likewise.
15687 (@mve_vpselq_<supf><mode>): Likewise.
15688 (@mve_vpselq_f<mode>"): Likewise.
15689 * config/arm/neon.md (vec_cmp<mode><v_cmp_result): Enable for MVE
15690 and move to vec-common.md.
15691 (vec_cmpu<mode><mode>): Likewise.
15692 (vcond<mode><mode>): Likewise.
15693 (vcond<V_cvtto><mode>): Likewise.
15694 (vcondu<mode><v_cmp_result>): Likewise.
15695 (vcond_mask_<mode><v_cmp_result>): Likewise.
15696 * config/arm/unspecs.md (VCMPNEQ_U, VCMPNEQ_S, VCMPEQQ_S)
15697 (VCMPEQQ_N_S, VCMPNEQ_N_S, VCMPEQQ_U, CMPEQQ_N_U, VCMPNEQ_N_U)
15698 (VCMPGEQ_N_S, VCMPGEQ_S, VCMPGTQ_N_S, VCMPGTQ_S, VCMPLEQ_N_S)
15699 (VCMPLEQ_S, VCMPLTQ_N_S, VCMPLTQ_S, VCMPCSQ_N_U, VCMPCSQ_U)
15700 (VCMPHIQ_N_U, VCMPHIQ_U): Remove.
15701 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result): Moved
15703 (vec_cmpu<mode><mode>): Likewise.
15704 (vcond<mode><mode>): Likewise.
15705 (vcond<V_cvtto><mode>): Likewise.
15706 (vcondu<mode><v_cmp_result>): Likewise.
15707 (vcond_mask_<mode><v_cmp_result>): Likewise. Added unsafe math
15710 2021-05-17 liuhongt <hongtao.liu@intel.com>
15713 * config/i386/i386.c (ix86_gimple_fold_builtin): Use
15714 gsi_insert_seq_before instead.
15716 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
15718 * doc/sourcebuild.texi (arm_qbit_ok): Rename into...
15719 (arm_sat_ok): ...this.
15721 2021-05-17 Martin Liska <mliska@suse.cz>
15723 * lto-wrapper.c (merge_flto_options): Factor out a new function.
15724 (merge_and_complain): Use it.
15725 (run_gcc): Merge also linker command line -flto=foo argument
15728 2021-05-16 Christophe Lyon <christophe.lyon@linaro.org>
15730 * config/arm/arm.h (CPP_SPEC): Remove error message about
15731 -mlittle-endian/-mbig-endian conflict.
15733 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
15735 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
15736 __ROP_PROTECT__ if -mrop-protect is selected.
15738 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
15740 * config/rs6000/rs6000-internal.h (rs6000_stack): Add
15741 rop_hash_save_offset and rop_hash_size.
15742 * config/rs6000/rs6000-logue.c (rs6000_stack_info): Compute
15743 rop_hash_size and rop_hash_save_offset.
15744 (debug_stack_info): Dump rop_hash_save_offset and rop_hash_size.
15745 (rs6000_emit_prologue): Emit hashst[p] in prologue.
15746 (rs6000_emit_epilogue): Emit hashchk[p] in epilogue.
15747 * config/rs6000/rs6000.md (unspec): Add UNSPEC_HASHST and
15749 (hashst): New define_insn.
15750 (hashchk): Likewise.
15752 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
15754 * config/rs6000/rs6000.c (rs6000_option_override_internal):
15755 Disable shrink wrap when inserting ROP-protect instructions.
15756 * config/rs6000/rs6000.opt (mrop-protect): New option.
15757 (mprivileged): Likewise.
15758 * doc/invoke.texi: Document mrop-protect and mprivileged.
15760 2021-05-15 Hans-Peter Nilsson <hp@axis.com>
15762 * reorg.c (fill_slots_from_thread): Reinstate code typoed out in
15765 2021-05-15 Martin Jambor <mjambor@suse.cz>
15768 2021-05-13 Martin Jambor <mjambor@suse.cz>
15770 PR tree-optimization/100453
15771 * tree-sra.c (sra_modify_assign): All const base accesses do not
15772 need refreshing, not just those from decl_pool.
15773 (sra_modify_assign): Do not refresh into a const base decl.
15775 2021-05-15 Jakub Jelinek <jakub@redhat.com>
15777 PR rtl-optimization/100342
15778 * regcprop.c (copy_value): When copying a source reg in a wider
15779 mode than it has recorded for the value, adjust recorded destination
15780 mode too or punt if !REG_CAN_CHANGE_MODE_P.
15782 2021-05-14 Jason Merrill <jason@redhat.com>
15784 * intl.h: Add comments.
15786 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15788 * config/aarch64/aarch64-simd.md
15789 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Split into...
15790 (aarch64_sqdmlsl2_lane<mode>_internal): ... This...
15791 (aarch64_sqdmlal2_lane<mode>_internal): ... And this.
15792 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>_internal): Split into ...
15793 (aarch64_sqdmlsl2_laneq<mode>_internal): ... This...
15794 (aarch64_sqdmlal2_laneq<mode>_internal): ... And this.
15795 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>_internal): Split into...
15796 (aarch64_sqdmlsl2_n<mode>_internal): ... This...
15797 (aarch64_sqdmlal2_n<mode>_internal): ... And this.
15799 2021-05-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15802 * config/arm/arm_neon.h (vtst_s8): Replace call to vtst builtin with it's
15803 boolean logic equivalent.
15804 (vtst_s16): Likewise.
15805 (vtst_s32): Likewise.
15806 (vtst_u8): Likewise.
15807 (vtst_u16): Likewise.
15808 (vtst_u32): Likewise.
15809 (vtst_p8): Likewise.
15810 (vtst_p16): Likewise.
15811 (vtstq_s8): Likewise.
15812 (vtstq_s16): Likewise.
15813 (vtstq_s32): Likewise.
15814 (vtstq_u8): Likewise.
15815 (vtstq_u16): Likewise.
15816 (vtstq_u32): Likewise.
15817 (vtstq_p8): Likewise.
15818 (vtstq_p16): Likewise.
15819 * config/arm/arm_neon_builtins.def: Remove entry for vtst.
15820 * config/arm/neon.md (neon_vtst<mode>): Remove pattern.
15822 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15824 * config/aarch64/aarch64-simd.md (aarch64_sqdmlal2<mode>): Merge into...
15825 (aarch64_sqdml<SBINQOPS:as>l2<mode>): ... This.
15826 (aarch64_sqdmlsl2<mode>): Delete.
15827 (aarch64_sqdmlal2_lane<mode>): Merge this...
15828 (aarch64_sqdmlsl2_lane<mode>): ... And this...
15829 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>): ... Into this.
15830 (aarch64_sqdmlal2_laneq<mode>): Merge this...
15831 (aarch64_sqdmlsl2_laneq<mode>): ... And this...
15832 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>): ... Into this.
15833 (aarch64_sqdmlal2_n<mode>): Merge this...
15834 (aarch64_sqdmlsl2_n<mode>): ... And this...
15835 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>): ... Into this.
15837 2021-05-13 Martin Sebor <msebor@redhat.com>
15839 PR middle-end/100574
15840 * builtins.c (access_ref::get_ref): Improve detection of PHIs with
15841 all null arguments.
15843 2021-05-13 Martin Sebor <msebor@redhat.com>
15845 PR tree-optimization/93100
15846 PR middle-end/98583
15847 * tree-ssa-uninit.c (check_defs): Exclude intrinsic functions that
15848 don't modify referenced objects.
15850 2021-05-13 Martin Jambor <mjambor@suse.cz>
15852 PR tree-optimization/100453
15853 * tree-sra.c (sra_modify_assign): All const base accesses do not
15854 need refreshing, not just those from decl_pool.
15855 (sra_modify_assign): Do not refresh into a const base decl.
15857 2021-05-13 Martin Liska <mliska@suse.cz>
15859 * tree-ssa-dom.c: Remove m_simplifier.
15861 2021-05-13 Richard Earnshaw <rearnsha@arm.com>
15864 * config/arm/arm.c (arm_canonicalize_comparison): Correctly
15865 canonicalize DImode inequality comparisons against the
15866 maximum integral value.
15868 2021-05-13 Jakub Jelinek <jakub@redhat.com>
15870 PR tree-optimization/98856
15871 * config/i386/i386.c (ix86_shift_rotate_cost): Add CODE argument.
15872 Expect V2DI and V4DI arithmetic right shifts to be emulated.
15873 (ix86_rtx_costs, ix86_add_stmt_cost): Adjust ix86_shift_rotate_cost
15875 * config/i386/i386-expand.c (expand_vec_perm_2perm_interleave,
15876 expand_vec_perm_2perm_pblendv): New functions.
15877 (ix86_expand_vec_perm_const_1): Use them.
15878 * config/i386/sse.md (ashr<mode>3<mask_name>): Rename to ...
15879 (<mask_codefor>ashr<mode>3<mask_name>): ... this.
15880 (ashr<mode>3): New define_expand with VI248_AVX512BW iterator.
15881 (ashrv4di3): New define_expand.
15882 (ashrv2di3): Change condition to TARGET_SSE2, handle !TARGET_XOP
15883 and !TARGET_AVX512VL expansion.
15885 2021-05-13 Uroš Bizjak <ubizjak@gmail.com>
15888 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Force mode
15889 sizes < 16 to a register when constructing vpcmov pattern.
15890 * config/i386/mmx.md (*xop_pcmov_<mode>): Use MMXMODE124 mode.
15892 2021-05-13 Martin Liska <mliska@suse.cz>
15894 * gcov-io.c (gcov_write_block): Remove.
15895 (gcov_write_words): Likewise.
15896 (gcov_read_words): Re-implement using gcov_read_bytes.
15897 (gcov_allocate): Remove.
15898 (GCOV_BLOCK_SIZE): Likewise.
15899 (struct gcov_var): Remove most of the fields.
15900 (gcov_position): Implement with ftell.
15901 (gcov_rewrite): Remove setting of start and offset fields.
15902 (from_file): Re-format.
15903 (gcov_open): Remove setbuf call. It should not be needed.
15904 (gcov_close): Remove internal buffer handling.
15905 (gcov_magic): Use __builtin_bswap32.
15906 (gcov_write_counter): Use directly gcov_write_unsigned.
15907 (gcov_write_string): Use direct fwrite and do not round
15909 (gcov_seek): Use directly fseek.
15910 (gcov_write_tag): Use gcov_write_unsigned directly.
15911 (gcov_write_length): Likewise.
15912 (gcov_write_tag_length): Likewise.
15913 (gcov_read_bytes): Use directly fread.
15914 (gcov_read_unsigned): Use gcov_read_words.
15915 (gcov_read_counter): Likewise.
15916 (gcov_read_string): Use gcov_read_bytes.
15917 * gcov-io.h (GCOV_WORD_SIZE): Adjust to reflect
15918 that size is not in bytes, but words (4B).
15919 (GCOV_TAG_FUNCTION_LENGTH): Likewise.
15920 (GCOV_TAG_ARCS_LENGTH): Likewise.
15921 (GCOV_TAG_ARCS_NUM): Likewise.
15922 (GCOV_TAG_COUNTER_LENGTH): Likewise.
15923 (GCOV_TAG_COUNTER_NUM): Likewise.
15924 (GCOV_TAG_SUMMARY_LENGTH): Likewise.
15926 2021-05-13 liuhongt <hongtao.liu@intel.com>
15929 * config/i386/sse.md (ssedoublevecmode): Add attribute for
15930 V64QI/V32HI/V16SI/V4DI.
15931 (ssehalfvecmode): Add attribute for V2DI/V2DF.
15932 (*vec_concatv4si_0): Extend to VI124_128.
15933 (*vec_concat<mode>_0): New pre-reload splitter.
15934 * config/i386/predicates.md (movq_parallel): New predicate.
15936 2021-05-13 Alexandre Oliva <oliva@adacore.com>
15938 * targhooks.c (default_zero_call_used_regs): Retry using
15939 successfully-zeroed registers as sources.
15941 2021-05-12 Tobias Burnus <tobias@codesourcery.com>
15943 * omp-low.c (finish_taskreg_scan): Use the proper detach decl.
15945 2021-05-12 Aldy Hernandez <aldyh@redhat.com>
15948 * gimple-range.cc (range_of_builtin_call): Skip out on
15949 processing __builtin_clz when varying.
15951 2021-05-12 Tom de Vries <tdevries@suse.de>
15954 * config/nvptx/nvptx-opts.h (enum ptx_version): New enum.
15955 * config/nvptx/nvptx.c (nvptx_file_start): Print .version according
15956 to ptx_version_option.
15957 * config/nvptx/nvptx.h (TARGET_PTX_6_3): Define.
15958 * config/nvptx/nvptx.md (define_insn "nvptx_shuffle<mode>")
15959 (define_insn "nvptx_vote_ballot"): Use sync variant for
15961 * config/nvptx/nvptx.opt (ptx_version): Add enum.
15962 (mptx): Add option.
15963 * doc/invoke.texi (Nvidia PTX Options): Add mptx item.
15965 2021-05-12 Richard Biener <rguenther@suse.de>
15967 PR tree-optimization/100566
15968 * tree-ssa-sccvn.c (dominated_by_p_w_unex): Properly handle
15969 allow_back for all edge queries.
15971 2021-05-12 liuhongt <hongtao.liu@intel.com>
15974 * config/i386/sse.md (<sse4_1_avx2>_pblendvb): Add
15975 splitters for pblendvb of NOT mask register.
15977 2021-05-12 Richard Biener <rguenther@suse.de>
15979 PR tree-optimization/100519
15980 * tree-ssa-reassoc.c (can_associate_p): Split into...
15981 (can_associate_op_p): ... this
15982 (can_associate_type_p): ... and this.
15983 (is_reassociable_op): Call can_associate_op_p.
15984 (break_up_subtract_bb): Call the appropriate predicates.
15985 (reassociate_bb): Likewise.
15987 2021-05-12 Martin Liska <mliska@suse.cz>
15989 * lto-wrapper.c (merge_and_complain): Merge -flto=arg options.
15990 (run_gcc): Use -flto argument detection for merged
15993 2021-05-12 Martin Liska <mliska@suse.cz>
15995 * lto-wrapper.c (print_lto_docs_link): New function.
15996 (run_gcc): Print warning about missing job server detection
15997 after we know NR of partitions. Do the same for -flto{,=1}.
15998 * opts.c (get_option_html_page): Support -flto option.
16000 2021-05-12 Martin Liska <mliska@suse.cz>
16002 * lto-wrapper.c (get_options_from_collect_gcc_options): Change
16004 (append_option): Remove.
16005 (find_option): Rework to use the vector type.
16006 (remove_option): Remove.
16007 (merge_and_complain): Use vectors for cl_decoded_option data
16009 (append_compiler_options): Likewise.
16010 (append_diag_options): Likewise.
16011 (append_linker_options): Likewise.
16012 (append_offload_options): Likewise.
16013 (compile_offload_image): Likewise.
16014 (compile_images_for_offload_targets): Likewise.
16015 (find_and_merge_options): Likewise.
16016 (run_gcc): Likewise.
16018 2021-05-12 Bernd Edlinger <bernd.edlinger@hotmail.de>
16021 * dwarf2out.c (dwarf2out_finish): Set
16022 have_multiple_function_sections with multi-range text_section.
16024 2021-05-12 Martin Liska <mliska@suse.cz>
16026 PR bootstrap/100560
16027 * Makefile.in: Remove version.h from linker command line.
16029 2021-05-12 Richard Biener <rguenther@suse.de>
16031 PR middle-end/100547
16032 * rtl.h (rtvec_alloc): Make argument size_t.
16033 * rtl.c (rtvec_alloc): Verify the count is less than INT_MAX.
16035 2021-05-12 Jakub Jelinek <jakub@redhat.com>
16037 PR middle-end/100508
16038 * cfgexpand.c (expand_debug_expr): For DEBUG_EXPR_DECL with vector
16039 type, don't reuse DECL_RTL if it has different mode, instead force
16040 creation of a new DEBUG_EXPR.
16042 2021-05-12 Jakub Jelinek <jakub@redhat.com>
16043 Marc Glisse <marc.glisse@inria.fr>
16045 PR tree-optimization/94589
16046 * match.pd ((X & Y) == X -> (X & ~Y) == 0,
16047 (X | Y) == Y -> (X & ~Y) == 0): New GIMPLE simplifications.
16049 2021-05-12 Uroš Bizjak <ubizjak@gmail.com>
16052 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Handle V2SF mode.
16053 * config/i386/mmx.md (MMXMODE124): New mode iterator.
16055 (mmxintvecmode): New mode attribute.
16056 (mmxintvecmodelower): Ditto.
16057 (*mmx_maskcmpv2sf3_comm): New insn pattern.
16058 (*mmx_maskcmpv2sf3): Ditto.
16059 (vec_cmpv2sfv2si): New expander.
16060 (vcond<V2FI:mode>v2si): Ditto.
16061 (mmx_vlendvps): New insn pattern.
16062 (vcond<MMXMODE124:mode><MMXMODEI:mode>): Also handle V2SFmode.
16063 (vcondu<MMXMODE124:mode><MMXMODEI:mode>): Ditto.
16064 (vcond_mask_<mode><mmxintvecmodelower>): Ditto.
16066 2021-05-11 Martin Sebor <msebor@redhat.com>
16068 PR middle-end/21433
16069 * expr.c (expand_expr_real_1): Replace unreachable code with an assert.
16071 2021-05-11 Richard Biener <rguenther@suse.de>
16073 * gimple-fold.c (gimple_fold_call): Do not call
16074 maybe_fold_reference on call arguments or the static chain.
16075 (fold_stmt_1): Do not call maybe_fold_reference on GIMPLE_ASM
16078 2021-05-11 Martin Liska <mliska@suse.cz>
16080 * builtins.def (DEF_HSAIL_BUILTIN): Remove.
16081 (DEF_HSAIL_ATOMIC_BUILTIN): Likewise.
16082 (DEF_HSAIL_SAT_BUILTIN): Likewise.
16083 (DEF_HSAIL_INTR_BUILTIN): Likewise.
16084 (DEF_HSAIL_CVT_ZEROI_SAT_BUILTIN): Likewise.
16085 * doc/frontends.texi: Remove BRIG.
16086 * doc/install.texi: Likewise.
16087 * doc/invoke.texi: Likewise.
16088 * doc/standards.texi: Likewise.
16089 * brig-builtins.def: Removed.
16090 * brig/ChangeLog: Removed.
16091 * brig/Make-lang.in: Removed.
16092 * brig/brig-builtins.h: Removed.
16093 * brig/brig-c.h: Removed.
16094 * brig/brig-lang.c: Removed.
16095 * brig/brigfrontend/brig-arg-block-handler.cc: Removed.
16096 * brig/brigfrontend/brig-atomic-inst-handler.cc: Removed.
16097 * brig/brigfrontend/brig-basic-inst-handler.cc: Removed.
16098 * brig/brigfrontend/brig-branch-inst-handler.cc: Removed.
16099 * brig/brigfrontend/brig-cmp-inst-handler.cc: Removed.
16100 * brig/brigfrontend/brig-code-entry-handler.cc: Removed.
16101 * brig/brigfrontend/brig-code-entry-handler.h: Removed.
16102 * brig/brigfrontend/brig-comment-handler.cc: Removed.
16103 * brig/brigfrontend/brig-control-handler.cc: Removed.
16104 * brig/brigfrontend/brig-copy-move-inst-handler.cc: Removed.
16105 * brig/brigfrontend/brig-cvt-inst-handler.cc: Removed.
16106 * brig/brigfrontend/brig-fbarrier-handler.cc: Removed.
16107 * brig/brigfrontend/brig-function-handler.cc: Removed.
16108 * brig/brigfrontend/brig-function.cc: Removed.
16109 * brig/brigfrontend/brig-function.h: Removed.
16110 * brig/brigfrontend/brig-inst-mod-handler.cc: Removed.
16111 * brig/brigfrontend/brig-label-handler.cc: Removed.
16112 * brig/brigfrontend/brig-lane-inst-handler.cc: Removed.
16113 * brig/brigfrontend/brig-machine.c: Removed.
16114 * brig/brigfrontend/brig-machine.h: Removed.
16115 * brig/brigfrontend/brig-mem-inst-handler.cc: Removed.
16116 * brig/brigfrontend/brig-module-handler.cc: Removed.
16117 * brig/brigfrontend/brig-queue-inst-handler.cc: Removed.
16118 * brig/brigfrontend/brig-seg-inst-handler.cc: Removed.
16119 * brig/brigfrontend/brig-signal-inst-handler.cc: Removed.
16120 * brig/brigfrontend/brig-to-generic.cc: Removed.
16121 * brig/brigfrontend/brig-to-generic.h: Removed.
16122 * brig/brigfrontend/brig-util.cc: Removed.
16123 * brig/brigfrontend/brig-util.h: Removed.
16124 * brig/brigfrontend/brig-variable-handler.cc: Removed.
16125 * brig/brigfrontend/hsa-brig-format.h: Removed.
16126 * brig/brigfrontend/phsa.h: Removed.
16127 * brig/brigspec.c: Removed.
16128 * brig/config-lang.in: Removed.
16129 * brig/gccbrig.texi: Removed.
16130 * brig/lang-specs.h: Removed.
16131 * brig/lang.opt: Removed.
16133 2021-05-11 Richard Biener <rguenther@suse.de>
16136 * ipa-param-manipulation.c
16137 (ipa_param_body_adjustments::modify_call_stmt): Avoid
16138 altering SSA_NAME_DEF_STMT by adjusting the calls LHS
16139 via gimple_call_lhs_ptr.
16141 2021-05-11 Alex Coplan <alex.coplan@arm.com>
16144 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear):
16145 Avoid emitting CFA adjusts on the sp if we have the fp.
16147 2021-05-11 Richard Sandiford <richard.sandiford@arm.com>
16149 * config/aarch64/iterators.md (VMUL_CHANGE_NLANES): Delete.
16150 (VMULD): New iterator.
16151 (VCOND): Handle V4HF and V8HF.
16152 (VCONQ): Fix entry for V2SF.
16153 * config/aarch64/aarch64-simd.md (mul_lane<mode>3): Use VMULD
16154 instead of VMUL. Use a 64-bit vector mode for the indexed operand.
16155 (*aarch64_mul3_elt_<vswap_width_name><mode>): Merge with...
16156 (mul_laneq<mode>3): ...this define_insn. Use VMUL instead of VDQSF.
16157 Use a 128-bit vector mode for the indexed operand. Use stype for
16158 the scheduling type.
16160 2021-05-11 Richard Biener <rguenther@suse.de>
16162 * gimple-fold.c (maybe_fold_reference): Only return
16163 is_gimple_min_invariant values.
16165 2021-05-11 Richard Biener <rguenther@suse.de>
16167 PR middle-end/100509
16168 * gimple-fold.c (fold_gimple_assign): Only call
16169 get_symbol_constant_value on register type symbols.
16171 2021-05-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
16172 Joe Ramsay <joe.ramsay@arm.com>
16175 * config/arm/arm_mve.h (__arm_vstrwq_scatter_offset): Fix wrong arguments.
16176 (__arm_vcmpneq): Remove duplicate definition.
16177 (__arm_vstrwq_scatter_offset_p): Likewise.
16178 (__arm_vmaxq_x): Likewise.
16179 (__arm_vmlsdavaq): Likewise.
16180 (__arm_vmlsdavaxq): Likewise.
16181 (__arm_vmlsdavq_p): Likewise.
16182 (__arm_vmlsdavxq_p): Likewise.
16183 (__arm_vrmlaldavhaq): Likewise.
16184 (__arm_vstrbq_p): Likewise.
16185 (__arm_vstrbq_scatter_offset): Likewise.
16186 (__arm_vstrbq_scatter_offset_p): Likewise.
16187 (__arm_vstrdq_scatter_offset): Likewise.
16188 (__arm_vstrdq_scatter_offset_p): Likewise.
16189 (__arm_vstrdq_scatter_shifted_offset): Likewise.
16190 (__arm_vstrdq_scatter_shifted_offset_p): Likewise.
16192 2021-05-11 Jakub Jelinek <jakub@redhat.com>
16194 PR middle-end/100471
16195 * omp-low.c (lower_omp_task_reductions): For OMP_TASKLOOP, if data
16196 is 0, bypass the reduction loop including
16197 GOMP_taskgroup_reduction_unregister call.
16199 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
16201 * config/rs6000/rs6000.c (struct rs6000_cost_data): New member
16202 costing_for_scalar.
16203 (rs6000_density_test): Early return if costing_for_scalar is true.
16204 (rs6000_init_cost): Init costing_for_scalar of rs6000_cost_data.
16206 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
16208 * doc/tm.texi: Regenerated.
16209 * target.def (init_cost): Add new parameter costing_for_scalar.
16210 * targhooks.c (default_init_cost): Adjust for new parameter.
16211 * targhooks.h (default_init_cost): Likewise.
16212 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
16213 (vect_compute_single_scalar_iteration_cost): Likewise.
16214 (vect_analyze_loop_2): Likewise.
16215 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
16216 (vect_bb_vectorization_profitable_p): Likewise.
16217 * tree-vectorizer.h (init_cost): Likewise.
16218 * config/aarch64/aarch64.c (aarch64_init_cost): Likewise.
16219 * config/i386/i386.c (ix86_init_cost): Likewise.
16220 * config/rs6000/rs6000.c (rs6000_init_cost): Likewise.
16222 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
16224 * config/rs6000/rs6000.c (rs6000_vect_nonmem): Renamed to
16225 vect_nonmem and moved into...
16226 (struct rs6000_cost_data): ...here.
16227 (rs6000_init_cost): Use vect_nonmem of cost_data instead.
16228 (rs6000_add_stmt_cost): Likewise.
16229 (rs6000_finish_cost): Likewise.
16231 2021-05-10 Eric Botcazou <ebotcazou@adacore.com>
16233 * range-op.cc (get_bool_state): Adjust head comment.
16234 (operator_not_equal::op1_range): Fix comment.
16235 (operator_bitwise_xor::op1_range): Remove call to gcc_unreachable.
16237 2021-05-10 Martin Sebor <msebor@redhat.com>
16239 PR middle-end/100425
16240 PR middle-end/100510
16241 * gimple-ssa-warn-alloca.c (pass_walloca::firast_time_p): Rename...
16242 (pass_walloca::xlimit_certain_p): ...to this.
16243 (pass_walloca::gate): Execute for any kind of handled warning.
16244 (pass_walloca::execute): Avoid issuing "maybe" and "unbounded"
16245 warnings when xlimit_certain_p is set.
16247 2021-05-10 Pat Haugen <pthaugen@linux.ibm.com>
16249 * config/rs6000/rs6000.c (rs6000_ira_change_pseudo_allocno_class):
16250 Return ALTIVEC_REGS if that is best_class.
16251 (rs6000_compute_pressure_classes): Add ALTIVEC_REGS.
16253 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16255 * config/arm/arm.h (CPP_SPEC): Remove error message about
16258 2021-05-10 Martin Jambor <mjambor@suse.cz>
16260 * ipa-prop.h (IPA_NODE_REF): Removed.
16261 (IPA_NODE_REF_GET_CREATE): Likewise.
16262 (IPA_EDGE_REF): Likewise.
16263 (IPA_EDGE_REF_GET_CREATE): Likewise.
16264 (IS_VALID_JUMP_FUNC_INDEX): Likewise.
16265 * ipa-cp.c (print_all_lattices): Replaced IPA_NODE_REF with a direct
16266 use of ipa_node_params_sum.
16267 (ipcp_versionable_function_p): Likewise.
16268 (push_node_to_stack): Likewise.
16269 (pop_node_from_stack): Likewise.
16270 (set_single_call_flag): Replaced two IPA_NODE_REF with one single
16271 direct use of ipa_node_params_sum.
16272 (initialize_node_lattices): Replaced IPA_NODE_REF with a direct use of
16273 ipa_node_params_sum.
16274 (ipa_context_from_jfunc): Replaced IPA_EDGE_REF with a direct use of
16276 (ipcp_verify_propagated_values): Replaced IPA_NODE_REF with a direct
16277 use of ipa_node_params_sum.
16278 (self_recursively_generated_p): Likewise.
16279 (propagate_scalar_across_jump_function): Likewise.
16280 (propagate_context_across_jump_function): Replaced IPA_EDGE_REF with a
16281 direct use of ipa_edge_args_sum, moved the lookup after the early
16282 exit. Replaced IPA_NODE_REF with a direct use of ipa_node_params_sum.
16283 (propagate_bits_across_jump_function): Replaced IPA_NODE_REF with
16284 direct uses of ipa_node_params_sum.
16285 (propagate_vr_across_jump_function): Likewise.
16286 (propagate_aggregate_lattice): Likewise.
16287 (propagate_aggs_across_jump_function): Likewise.
16288 (propagate_constants_across_call): Likewise, also replaced
16289 IPA_EDGE_REF with a direct use of ipa_edge_args_sum.
16290 (good_cloning_opportunity_p): Replaced IPA_NODE_REF with a direct use
16291 of ipa_node_params_sum.
16292 (estimate_local_effects): Likewise.
16293 (add_all_node_vals_to_toposort): Likewise.
16294 (propagate_constants_topo): Likewise.
16295 (ipcp_propagate_stage): Likewise.
16296 (ipcp_discover_new_direct_edges): Likewise.
16297 (calls_same_node_or_its_all_contexts_clone_p): Likewise.
16298 (cgraph_edge_brings_value_p): Likewise (in both overloaded functions).
16299 (get_info_about_necessary_edges): Likewise.
16300 (want_remove_some_param_p): Likewise.
16301 (create_specialized_node): Likewise.
16302 (self_recursive_pass_through_p): Likewise.
16303 (self_recursive_agg_pass_through_p): Likewise.
16304 (find_more_scalar_values_for_callers_subset): Likewise and also
16305 replaced IPA_EDGE_REF with direct uses of ipa_edge_args_sum, in one
16306 case replacing two of those with a single query.
16307 (find_more_contexts_for_caller_subset): Likewise for the
16308 ipa_polymorphic_call_context overload.
16309 (intersect_aggregates_with_edge): Replaced IPA_EDGE_REF with a direct
16310 use of ipa_edge_args_sum. Replaced IPA_NODE_REF with direct uses of
16311 ipa_node_params_sum.
16312 (find_aggregate_values_for_callers_subset): Likewise, also reusing
16313 results of ipa_edge_args_sum->get.
16314 (cgraph_edge_brings_all_scalars_for_node): Replaced IPA_NODE_REF with
16315 direct uses of ipa_node_params_sum, replaced IPA_EDGE_REF with a
16316 direct use of ipa_edge_args_sum.
16317 (cgraph_edge_brings_all_agg_vals_for_node): Likewise, moved node
16318 summary query after the early exit and reused the result later.
16319 (decide_about_value): Replaced IPA_NODE_REF with a direct use of
16320 ipa_node_params_sum.
16321 (decide_whether_version_node): Likewise. Removed re-querying for
16322 summaries after cloning.
16323 (spread_undeadness): Replaced IPA_NODE_REF with a direct use of
16324 ipa_node_params_sum.
16325 (has_undead_caller_from_outside_scc_p): Likewise, reusing results of
16327 (identify_dead_nodes): Likewise.
16328 (ipcp_store_bits_results): Replaced IPA_NODE_REF with direct uses of
16329 ipa_node_params_sum.
16330 (ipcp_store_vr_results): Likewise.
16331 * ipa-fnsummary.c (evaluate_properties_for_edge): Likewise.
16332 (ipa_fn_summary_t::duplicate): Likewise.
16333 (analyze_function_body): Likewise.
16334 (estimate_calls_size_and_time): Likewise.
16335 (ipa_cached_call_context::duplicate_from): Likewise.
16336 (ipa_call_context::equal_to): Likewise.
16337 (remap_edge_params): Likewise.
16338 (ipa_merge_fn_summary_after_inlining): Likewise.
16339 (inline_read_section): Likewise.
16340 * ipa-icf.c (sem_function::param_used_p): Likewise.
16341 * ipa-modref.c (compute_parm_map): Likewise.
16342 (compute_parm_map): Replaced IPA_EDGE_REF with a direct use of
16344 (get_access_for_fnspec): Replaced IPA_NODE_REF with a direct use of
16345 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
16347 * ipa-profile.c (check_argument_count): Likewise.
16348 * ipa-prop.c (ipa_alloc_node_params): Replaced IPA_NODE_REF_GET_CREATE
16349 with a direct use of ipa_node_params_sum.
16350 (ipa_initialize_node_params): Likewise.
16351 (ipa_print_node_jump_functions_for_edge): Replaced IPA_EDGE_REF with a
16352 direct use of ipa_edge_args_sum and reused the query result.
16353 (ipa_compute_jump_functions_for_edge): Replaced IPA_NODE_REF with a
16354 direct use of ipa_node_params_sum and replaced IPA_EDGE_REF with a
16355 direct use of ipa_edge_args_sum.
16356 (ipa_note_param_call): Replaced IPA_NODE_REF with a direct use of
16357 ipa_node_params_sum and reused the result of the query.
16358 (ipa_analyze_node): Likewise.
16359 (ipa_analyze_controlled_uses): Replaced IPA_NODE_REF with a direct use
16360 of ipa_node_params_sum.
16361 (update_jump_functions_after_inlining): Replaced IPA_EDGE_REF with
16362 direct uses of ipa_edge_args_sum.
16363 (update_indirect_edges_after_inlining): Replaced IPA_NODE_REF with
16364 direct uses of ipa_node_params_sum and replaced IPA_EDGE_REF with a
16365 direct use of ipa_edge_args_sum. Removed superficial re-querying the
16367 (propagate_controlled_uses): Replaced IPA_NODE_REF with direct uses of
16368 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
16370 (ipa_propagate_indirect_call_infos): Replaced IPA_EDGE_REF with a
16371 direct use of ipa_edge_args_sum.
16372 (ipa_edge_args_sum_t::duplicate): Replaced IPA_NODE_REF with a direct
16373 use of ipa_node_params_sum.
16374 (ipa_print_node_params): Likewise.
16375 (ipa_write_node_info): Likewise and also replaced IPA_EDGE_REF with
16376 direct uses of ipa_edge_args_sum.
16377 (ipa_read_edge_info): Replaced IPA_EDGE_REF with a direct use of
16379 (ipa_read_node_info): Replaced IPA_NODE_REF with a direct use of
16380 ipa_node_params_sum.
16381 (ipa_prop_write_jump_functions): Likewise. Move variable node to the
16382 scopes where it is used.
16384 2021-05-10 Uroš Bizjak <ubizjak@gmail.com>
16386 * config/i386/i386-expand.c (ix86_expand_sse_movcc)
16387 <case E_V2SImode>: Force op_true to register.
16389 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16391 * config/arm/iterators.md (MVE_FP_COMPARISONS): New.
16392 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_f<mode>)
16393 (mve_vcmp<mve_cmp_op>q_n_f<mode>): New, merge all vcmp_*f*
16395 (mve_vcmpeqq_f<mode>, mve_vcmpeqq_n_f<mode>, mve_vcmpgeq_f<mode>)
16396 (mve_vcmpgeq_n_f<mode>, mve_vcmpgtq_f<mode>)
16397 (mve_vcmpgtq_n_f<mode>, mve_vcmpleq_f<mode>)
16398 (mve_vcmpleq_n_f<mode>, mve_vcmpltq_f<mode>)
16399 (mve_vcmpltq_n_f<mode>, mve_vcmpneq_f<mode>)
16400 (mve_vcmpneq_n_f<mode>): Remove.
16401 * config/arm/unspecs.md (VCMPEQQ_F, VCMPEQQ_N_F, VCMPGEQ_F)
16402 (VCMPGEQ_N_F, VCMPGTQ_F, VCMPGTQ_N_F, VCMPLEQ_F, VCMPLEQ_N_F)
16403 (VCMPLTQ_F, VCMPLTQ_N_F, VCMPNEQ_F, VCMPNEQ_N_F): Remove.
16405 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16407 * config/arm/iterators.md (MVE_COMPARISONS): New.
16409 (mve_cmp_type): New.
16410 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_<mode>): New, merge all
16412 (mve_vcmpneq_<mode>, mve_vcmpcsq_n_<mode>, mve_vcmpcsq_<mode>)
16413 (mve_vcmpeqq_n_<mode>, mve_vcmpeqq_<mode>, mve_vcmpgeq_n_<mode>)
16414 (mve_vcmpgeq_<mode>, mve_vcmpgtq_n_<mode>, mve_vcmpgtq_<mode>)
16415 (mve_vcmphiq_n_<mode>, mve_vcmphiq_<mode>, mve_vcmpleq_n_<mode>)
16416 (mve_vcmpleq_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
16417 (mve_vcmpneq_n_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
16418 (mve_vcmpneq_n_<mode>): Remove.
16420 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16422 * config/arm/arm_mve.h (__arm_vcmp*): Remove 's' suffix.
16423 * config/arm/arm_mve_builtins.def (vcmp*): Remove 's' suffix.
16424 * config/arm/mve.md (mve_vcmp*): Remove 's' suffix in pattern
16427 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16429 * config/arm/arm_mve_builtins.def (vcmpneq_u): Remove.
16430 (vcmpneq_n_u): Likewise.
16431 (vcmpeqq_u,): Likewise.
16432 (vcmpeqq_n_u): Likewise.
16433 * config/arm/iterators.md (supf): Remove VCMPNEQ_U, VCMPEQQ_U,
16434 VCMPEQQ_N_U and VCMPNEQ_N_U.
16435 * config/arm/mve.md (mve_vcmpneq): Remove <supf> iteration.
16436 (mve_vcmpeqq_n): Likewise.
16437 (mve_vcmpeqq): Likewise.
16438 (mve_vcmpneq_n): Likewise.
16440 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16442 * config/arm/arm_mve.h (__arm_vcmpeq*u*, __arm_vcmpne*u*): Call
16443 the 's' version of the builtin.
16445 2021-05-10 Richard Biener <rguenther@suse.de>
16447 PR tree-optimization/100492
16448 * tree-loop-distribution.c (find_seed_stmts_for_distribution):
16449 Find nothing when the loop contains an irreducible region.
16451 2021-05-10 Richard Biener <rguenther@suse.de>
16453 PR middle-end/100464
16455 * gimple-fold.c (canonicalize_constructor_val): Do not set
16458 2021-05-10 Richard Biener <rguenther@suse.de>
16460 PR tree-optimization/100434
16461 * tree-ssa-dse.c (initialize_ao_ref_for_dse): Handle
16463 (dse_optimize_stmt): Handle call LHS by dropping the
16464 LHS or the whole call if it doesn't have other
16466 (pass_dse::execute): Adjust.
16468 2021-05-10 Martin Liska <mliska@suse.cz>
16470 * Makefile.in: Add missing genversion rule.
16472 2021-05-10 Alex Coplan <alex.coplan@arm.com>
16475 * config/arm/mve.md (*mve_mov<mode>): Simplify output code. Use
16476 vldrw.u32 and vstrw.32 for V2D[IF]mode loads and stores.
16478 2021-05-10 Martin Liska <mliska@suse.cz>
16480 * builtins.c (is_builtin_name): Use startswith
16481 function instead of strncmp.
16482 * collect2.c (main): Likewise.
16483 (has_lto_section): Likewise.
16484 (scan_libraries): Likewise.
16485 * coverage.c (coverage_checksum_string): Likewise.
16486 (coverage_init): Likewise.
16487 * dwarf2out.c (is_cxx): Likewise.
16488 (gen_compile_unit_die): Likewise.
16489 * gcc-ar.c (main): Likewise.
16490 * gcc.c (init_spec): Likewise.
16491 (read_specs): Likewise.
16492 (execute): Likewise.
16493 (check_live_switch): Likewise.
16494 * genattrtab.c (write_attr_case): Likewise.
16495 (IS_ATTR_GROUP): Likewise.
16496 * gencfn-macros.c (main): Likewise.
16497 * gengtype.c (type_for_name): Likewise.
16498 (gen_rtx_next): Likewise.
16499 (get_file_langdir): Likewise.
16500 (write_local): Likewise.
16501 * genmatch.c (get_operator): Likewise.
16502 (get_operand_type): Likewise.
16503 (expr::gen_transform): Likewise.
16504 * genoutput.c (validate_optab_operands): Likewise.
16505 * incpath.c (add_sysroot_to_chain): Likewise.
16506 * langhooks.c (lang_GNU_C): Likewise.
16507 (lang_GNU_CXX): Likewise.
16508 (lang_GNU_Fortran): Likewise.
16509 (lang_GNU_OBJC): Likewise.
16510 * lto-wrapper.c (run_gcc): Likewise.
16511 * omp-general.c (omp_max_simt_vf): Likewise.
16512 * omp-low.c (omp_runtime_api_call): Likewise.
16513 * opts-common.c (parse_options_from_collect_gcc_options): Likewise.
16514 * read-rtl-function.c (function_reader::read_rtx_operand_r): Likewise.
16515 * real.c (real_from_string): Likewise.
16516 * selftest.c (assert_str_startswith): Likewise.
16517 * timevar.c (timer::validate_phases): Likewise.
16518 * tree.c (get_file_function_name): Likewise.
16519 * ubsan.c (ubsan_use_new_style_p): Likewise.
16520 * varasm.c (default_function_rodata_section): Likewise.
16521 (incorporeal_function_p): Likewise.
16522 (default_section_type_flags): Likewise.
16523 * system.h (startswith): Define startswith.
16525 2021-05-10 Martin Liska <mliska@suse.cz>
16527 * bitmap.h (class auto_bitmap): Remove
16528 __cplusplus >= 201103.
16529 * config/aarch64/aarch64.c: Likewise.
16530 * gimple-ssa-store-merging.c (store_immediate_info::store_immediate_info):
16532 * sbitmap.h: Likewise.
16534 2021-05-10 Martin Liska <mliska@suse.cz>
16536 * Makefile.in: Rename gcov-iov to genversion and depend
16537 on version.h (instead of gcov-iov.h).
16538 * gcov-io.h: Include version.h instread of gcov-iov.h.
16539 * gengtype-state.c (read_state_version): Likewise.
16540 * gcov-iov.c: Moved to...
16541 * genversion.c: ...here.
16542 * lto-streamer.h (LTO_major_version): Define it with
16544 * version.c: Removed.
16545 * version.h: Removed.
16547 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16549 * config/arc/arc.md (UNSPEC_ARC_DMPYWH): Define.
16550 * config/arc/simdext.md (VCT): Add predicates for iterator
16553 (voptab): Likewise.
16554 (vec_widen_<V_US>mult_hi_v4hi): Change pattern predicate.
16555 (<voptab>v2si3): New patterns.
16557 (reduc_plus_scal_v4hi): Likewise.
16558 (reduc_plus_scal_v2si): Likewise.
16559 (vec_duplicatev2si): Likewise.
16560 (vec_duplicatev4hi): Likewise.
16562 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16564 * config/arc/simdext.md: Format and cleanup file.
16566 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16568 * config/arc/simdext.md (movmisalignv2hi): Allow misaligned access
16569 only when munaligned-access option is on.
16570 (movmisalign<mode>): Likewise.
16572 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16574 * common/config/arc/arc-common.c (arc_handle_option): Remove dot
16576 * config/arc/arc.c (arc_reorg): Remove underscore from string.
16578 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16580 * config/arc/arc.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
16581 (CTZ_DEFINED_VALUE_AT_ZERO): Likewise.
16582 * config/arc/arc.md (clrsbsi2): Cleanup pattern.
16583 (norm_f): Likewise.
16586 (clzsi2): Use fls instruction when available.
16587 (arc_clzsi2): Likewise.
16589 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16591 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add r26 and r27.
16593 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16595 * doc/extend.texi (__builtin_arc_sr): Swap arguments.
16597 2021-05-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
16599 PR middle-end/100467
16600 * toplev.c (compile_file): Call insn_locations_init before
16601 targetm.asm_out.code_end.
16603 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
16606 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
16608 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
16610 2021-05-07 Jakub Jelinek <jakub@redhat.com>
16611 Andrew Stubbs <amd@codesourcery.com>
16614 * builtins.c (try_store_by_multiple_pieces): Use force_operand for
16615 emit_move_insn operands.
16617 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
16619 * cfgexpand.c (expand_gimple_basic_block): Do not inherit a current
16620 location for the outgoing edges of an empty block.
16621 * dwarf2out.c (add_subscript_info): Retrieve the bounds and index
16622 type by means of the get_array_descr_info langhook, if it is set and
16623 returns true. Remove obsolete code dealing with unnamed subtypes.
16625 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16627 * gimple-range-cache.cc (ssa_block_ranges): Virtualize.
16628 (sbr_vector): Renamed from ssa_block_cache.
16629 (sbr_vector::sbr_vector): Allocate from obstack abd initialize.
16630 (ssa_block_ranges::~ssa_block_ranges): Remove.
16631 (sbr_vector::set_bb_range): Use varying and undefined cached values.
16632 (ssa_block_ranges::set_bb_varying): Remove.
16633 (sbr_vector::get_bb_range): Adjust assert.
16634 (sbr_vector::bb_range_p): Adjust assert.
16635 (~block_range_cache): No freeing loop required.
16636 (block_range_cache::get_block_ranges): Remove.
16637 (block_range_cache::set_bb_range): Inline get_block_ranges.
16638 (block_range_cache::set_bb_varying): Remove.
16639 * gimple-range-cache.h (set_bb_varying): Remove prototype.
16640 * value-range.h (irange_allocator::get_memory): New.
16642 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16644 * gimple-range-cache.cc (non_null_ref::non_null_deref_p): Search
16645 dominator tree is available and requested.
16646 (ranger_cache::ssa_range_in_bb): Don't search dom tree here.
16647 (ranger_cache::fill_block_cache): Don't search dom tree here either.
16648 * gimple-range-cache.h (non_null_deref_p): Add dom_search param.
16650 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16652 * gimple-range.cc (gimple_ranger::range_on_exit): Handle block with
16653 only PHI nodes better.
16655 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16657 * gimple-range-edge.h (gimple_outgoing_range): Rename from
16659 (gcond_edge_range): Export prototype.
16660 * gimple-range-edge.cc (gcond_edge_range): New.
16661 (gimple_outgoing_range::edge_range_p): Use gcond_edge_range.
16662 * gimple-range-gori.h (gori_compute): Use gimple_outgoing_range.
16664 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16666 * gimple-range-edge.cc (outgoing_range::calc_switch_ranges): Compute
16667 default range into a temp and allocate only what is needed.
16669 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16671 * range-op.cc (operator_trunc_mod::wi_fold): x % 0 is UNDEFINED.
16673 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16675 * gimple-range.h (gimple_range_global): Pick up parameter initial
16676 values, and use-before defined locals are UNDEFINED.
16678 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
16680 * doc/extend.texi (scalar_storage_order): Mention effect on pointer
16682 * tree.h (reverse_storage_order_for_component_p): Return false if
16683 the type is a pointer.
16685 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
16687 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
16689 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
16692 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
16693 Handle V8QI, V4HI and V2SI modes.
16694 * config/i386/mmx.md (mmx_pblendvb): New insn pattern.
16695 * config/i386/sse.md (unspec): Move UNSPEC_BLENDV ...
16696 * config/i386/i386.md (unspec): ... here.
16698 2021-05-07 Tobias Burnus <tobias@codesourcery.com>
16699 Tom de Vries <tdevries@suse.de>
16701 * omp-low.c (lower_rec_simd_input_clauses): Set max_vf = 1 if
16702 a truth_value_p reduction variable is nonintegral.
16704 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
16707 * config/i386/i386-expand.c (ix86_use_mask_cmp_p):
16708 Return false for mode sizes < 16.
16710 2021-05-07 Jakub Jelinek <jakub@redhat.com>
16713 * config/i386/mmx.md (*xop_pcmov_<mode>): New define_insn.
16715 2021-05-06 Martin Jambor <mjambor@suse.cz>
16717 * ipa-sra.c (ipa_sra_dump_all_summaries): Dump edge summaries even
16718 when there is no function summary.
16719 (ipa_sra_summarize_function): produce edge summaries even when
16722 2021-05-06 Tom Tromey <tom@tromey.com>
16724 * godump.c (string_hash_eq): Remove.
16725 (go_finish): Use htab_eq_string.
16727 2021-05-06 Tom Tromey <tom@tromey.com>
16729 * gengtype-state.c (read_state): Use htab_eq_string.
16730 (string_eq): Remove.
16732 2021-05-06 Tom Tromey <tom@tromey.com>
16734 * gensupport.c (htab_eq_string): Remove.
16736 2021-05-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
16739 * debug.h (gcc_debug_hooks): Add set_ignored_loc function pointer.
16740 * dwarf2out.h (dw_fde_node::ignored_debug): New data item.
16741 * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Add dummy
16742 set_ignored_loc callbacks.
16743 * debug.c (do_nothing_debug_hooks): Likewise.
16744 * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
16745 * dwarf2out.c (text_section_used, cold_text_section_used): Remove.
16746 (in_text_section_p, last_text_label, last_cold_label,
16747 switch_text_ranges, switch_cold_ranges): New data items.
16748 (dwarf2out_note_section_used): Remove.
16749 (dwarf2out_begin_prologue): Set fde->ignored_debug and
16751 (mark_ignored_debug_section): New helper function.
16752 (dwarf2out_end_epilogue, dwarf2out_switch_text_section): Call
16753 mark_ignored_debug_section.
16754 (dwarf2_debug_hooks): Use dwarf2out_set_ignored_loc.
16755 (dwarf2_lineno_debug_hooks): Use dummy for set_ignored_loc.
16756 (size_of_aranges): Adjust formula for multi-part text ranges size.
16757 (output_aranges): Output multi-part text ranges.
16758 (dwarf2out_set_ignored_loc): New callback function.
16759 (dwarf2out_finish): Output multi-part text ranges.
16760 (dwarf2out_c_finalize): Clear new data items.
16761 * final.c (final_start_function_1): Call set_ignored_loc callback.
16762 (final_scan_insn_1): Likewise.
16763 * ggc-page.c (gt_ggc_mx): New helper function.
16764 * stringpool.c (gt_pch_nx): Likewise.
16766 2021-05-06 Richard Biener <rguenther@suse.de>
16768 * timevar.def (TV_TREE_INSERT_PHI_NODES): Remove.
16769 (TV_TREE_SSA_REWRITE_BLOCKS): Likewise.
16770 (TV_TREE_INTO_SSA): New.
16771 * tree-into-ssa.c (insert_phi_nodes): Do not account separately.
16772 (rewrite_blocks): Likewise.
16773 (pass_data_build_ssa): Account to TV_TREE_INTO_SSA.
16775 2021-05-06 Jakub Jelinek <jakub@redhat.com>
16777 * tree-ssa-phiopt.c (value_replacement, minmax_replacement,
16778 abs_replacement, xor_replacement,
16779 cond_removal_in_popcount_clz_ctz_pattern,
16780 replace_phi_edge_with_variable): Change type of phi argument from
16781 gimple * to gphi *.
16783 2021-05-06 Richard Biener <rguenther@suse.de>
16785 * tree-ssa-loop-split.c (split_loop): Delay updating SSA form.
16786 Output an opt-info message.
16787 (do_split_loop_on_cond): Likewise.
16788 (tree_ssa_split_loops): Update SSA form here.
16790 2021-05-06 Richard Biener <rguenther@suse.de>
16792 * tree-inline.c (tree_function_versioning): Fix DECL_BY_REFERENCE
16793 return variable removal.
16795 2021-05-06 Marius Hillenbrand <mhillen@linux.ibm.com>
16797 * config/s390/s390-builtins.def (O_M5, O1_M5, ...): Remove unused macros.
16798 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
16799 (s390_vec_permi_dbl, s390_vpdi): Use the O3_U2 type for the immediate
16801 * config/s390/s390.c (s390_const_operand_ok): Remove unused
16804 2021-05-06 Jakub Jelinek <jakub@redhat.com>
16806 PR tree-optimization/94589
16807 * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Call
16808 spaceship_replacement.
16809 (cond_only_block_p, spaceship_replacement): New functions.
16811 2021-05-06 Richard Biener <rguenther@suse.de>
16814 * tree-emutls.c (gen_emutls_addr): Pass in whether we're
16815 dealing with a debug use and only query existing addresses
16817 (lower_emutls_1): Avoid splitting out addresses for debug
16818 stmts, reset the debug stmt when we fail to find existing
16820 (lower_emutls_phi_arg): Set wi.stmt.
16822 2021-05-06 Christoph Muellner <cmuellner@gcc.gnu.org>
16825 * config/riscv/riscv.c (riscv_block_move_loop): Use cbranch helper.
16826 * config/riscv/riscv.md (cbranch<mode>4): Generate helpers.
16827 (stack_protect_test): Use cbranch helper.
16829 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
16832 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
16833 always return the establisher frame for __builtin_frame_address (0).
16835 2021-05-05 Ivan Sorokin <vanyacpp@gmail.com>
16838 * config/i386/i386-builtins.c (ix86_cpu_model_type_node): New.
16839 (ix86_cpu_model_var): Likewise.
16840 (ix86_cpu_features2_type_node): Likewise.
16841 (ix86_cpu_features2_var): Likewise.
16842 (fold_builtin_cpu): Cache __cpu_model and __cpu_features2 with
16845 2021-05-05 Martin Sebor <msebor@redhat.com>
16847 * passes.def (pass_warn_printf): Run after SSA.
16849 2021-05-05 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16851 * config/arm/neon.md (neon_vtst_combine<mode>): New pattern.
16852 * config/arm/predicates.md (minus_one_operand): New predicate.
16854 2021-05-05 Jeff Law <jlaw@tachyum.com>
16856 * config/avr/avr.md: Remove references to CC_STATUS_INIT.
16858 2021-05-05 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
16860 PR rtl-optimization/100263
16861 * postreload.c (move2add_valid_value_p): Ensure register can
16864 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
16866 PR rtl-optimization/100411
16867 * cfgcleanup.c (try_crossjump_to_edge): Also skip end of prologue
16868 and beginning of function markers.
16870 2021-05-05 Jeff Law <jlaw@tachyum.com>
16872 * config/cr16/cr16.h (NOTICE_UPDATE_CC): Remove.
16873 * config/cr16/cr16.c (notice_update_cc): Remove.
16874 * config/cr16/cr16-protos.h (notice_update_cc): Remove.
16876 2021-05-05 Uroš Bizjak <ubizjak@gmail.com>
16879 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
16880 Handle V8QI, V4HI and V2SI modes.
16881 * config/i386/i386.c (ix86_build_const_vector): Handle V2SImode.
16882 (ix86_build_signbit_mask): Ditto.
16883 * config/i386/mmx.md (MMXMODE14): New mode iterator.
16884 (<smaxmin:code><MMXMODE14:mode>3): New expander.
16885 (*mmx_<smaxmin:code><MMXMODE14:mode>3): New insn pattern.
16886 (<umaxmin:code><MMXMODE24:mode>3): New expander.
16887 (*mmx_<umaxmin:code><MMXMODE24:mode>3): New insn pattern.
16888 (vec_cmp<MMXMODEI:mode><MMXMODEI:mode>): New expander.
16889 (vec_cmpu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
16890 (vcond<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
16891 (vcondu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
16892 (vcond_mask_<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
16894 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
16896 * dwarf2out.c (loc_list_from_tree_1) <DECL>: During early DWARF, do
16897 not expand the VALUE_EXPR of variables put in the non-local frame.
16898 * gimplify.c (gimplify_type_sizes) <RECORD_TYPE>: If the type is not
16899 to be ignored for debug info, ensure its variable offsets are not.
16901 2021-05-05 Richard Biener <rguenther@suse.de>
16903 PR tree-optimization/79333
16904 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt):
16905 Fold stmt following SSA edges.
16907 2021-05-05 Richard Biener <rguenther@suse.de>
16909 PR middle-end/100394
16910 * calls.c (expand_call): Preserve possibly throwing calls.
16911 * cfgexpand.c (expand_call_stmt): When a call can throw signal
16912 RTL expansion there are side-effects.
16913 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Simplify,
16914 mark all possibly throwing stmts necessary unless we can elide
16916 * tree-ssa-dse.c (pass_dse::execute): Preserve exceptions unless
16917 -fdelete-dead-exceptions.
16918 * tree.h (DECL_PURE_P): Add note about exceptions.
16920 2021-05-05 Alexandre Oliva <oliva@adacore.com>
16922 * config/i386/vxworks.h (DBX_REGISTER_NUMBER): Make it
16925 2021-05-04 David Edelsohn <dje.gcc@gmail.com>
16927 * config/rs6000/rs6000-call.c (rs6000_output_mi_thunk): Use
16928 get_fnname_from_decl for name of thunk.
16929 * config/rs6000/rs6000.c (rs6000_declare_alias): Use assemble_name
16930 and ASM_OUTPUT_LABEL.
16931 (rs6000_xcoff_declare_function_name): Use assemble_name and
16933 (rs6000_xcoff_declare_object_name): Use ASM_OUTPUT_LABEL.
16934 (rs6000_xcoff_encode_section_info): Don't add mapping class
16935 for aliases. Always add [DS] mapping class to primary
16937 (rs6000_asm_weaken_decl): Don't explicitly add [DS].
16939 2021-05-04 Martin Sebor <msebor@redhat.com>
16941 PR middle-end/100307
16942 * builtins.c (compute_objsize_r): Clear base0 for pointers.
16944 2021-05-04 Jeff Law <jlaw@tachyum.com>
16946 * config/bfin/bfin.h (NOTICE_UPDATE_CC): Remove.
16948 2021-05-04 Segher Boessenkool <segher@kernel.crashing.org>
16950 * caller-save.c: Remove CC0.
16951 * cfgcleanup.c: Remove CC0.
16952 * cfgrtl.c: Remove CC0.
16953 * combine.c: Remove CC0.
16954 * compare-elim.c: Remove CC0.
16955 * conditions.h: Remove CC0.
16956 * config/h8300/h8300.h: Remove CC0.
16957 * config/h8300/h8300-protos.h: Remove CC0.
16958 * config/h8300/peepholes.md: Remove CC0.
16959 * config/i386/x86-tune-sched.c: Remove CC0.
16960 * config/m68k/m68k.c: Remove CC0.
16961 * config/rl78/rl78.c: Remove CC0.
16962 * config/sparc/sparc.c: Remove CC0.
16963 * config/xtensa/xtensa.c: Remove CC0.
16964 (gen_conditional_move): Use pc_rtx instead of cc0_rtx in a piece of
16965 RTL where that is used as a placeholder only.
16966 * cprop.c: Remove CC0.
16967 * cse.c: Remove CC0.
16968 * cselib.c: Remove CC0.
16969 * df-problems.c: Remove CC0.
16970 * df-scan.c: Remove CC0.
16971 * doc/md.texi: Remove CC0. Adjust an example.
16972 * doc/rtl.texi: Remove CC0. Adjust an example.
16973 * doc/tm.texi: Regenerate.
16974 * doc/tm.texi.in: Remove CC0.
16975 * emit-rtl.c: Remove CC0.
16976 * final.c: Remove CC0.
16977 * fwprop.c: Remove CC0.
16978 * gcse-common.c: Remove CC0.
16979 * gcse.c: Remove CC0.
16980 * genattrtab.c: Remove CC0.
16981 * genconfig.c: Remove CC0.
16982 * genemit.c: Remove CC0.
16983 * genextract.c: Remove CC0.
16984 * gengenrtl.c: Remove CC0.
16985 * genrecog.c: Remove CC0.
16986 * haifa-sched.c: Remove CC0.
16987 * ifcvt.c: Remove CC0.
16988 * ira-costs.c: Remove CC0.
16989 * ira.c: Remove CC0.
16990 * jump.c: Remove CC0.
16991 * loop-invariant.c: Remove CC0.
16992 * lra-constraints.c: Remove CC0.
16993 * lra-eliminations.c: Remove CC0.
16994 * optabs.c: Remove CC0.
16995 * postreload-gcse.c: Remove CC0.
16996 * postreload.c: Remove CC0.
16997 * print-rtl.c: Remove CC0.
16998 * read-rtl-function.c: Remove CC0.
16999 * reg-notes.def: Remove CC0.
17000 * reg-stack.c: Remove CC0.
17001 * reginfo.c: Remove CC0.
17002 * regrename.c: Remove CC0.
17003 * reload.c: Remove CC0.
17004 * reload1.c: Remove CC0.
17005 * reorg.c: Remove CC0.
17006 * resource.c: Remove CC0.
17007 * rtl.c: Remove CC0.
17008 * rtl.def: Remove CC0.
17009 * rtl.h: Remove CC0.
17010 * rtlanal.c: Remove CC0.
17011 * sched-deps.c: Remove CC0.
17012 * sched-rgn.c: Remove CC0.
17013 * shrink-wrap.c: Remove CC0.
17014 * simplify-rtx.c: Remove CC0.
17015 * system.h: Remove CC0. Poison NOTICE_UPDATE_CC, CC_STATUS_MDEP_INIT,
17016 CC_STATUS_MDEP, and CC_STATUS.
17017 * target.def: Remove CC0.
17018 * valtrack.c: Remove CC0.
17019 * var-tracking.c: Remove CC0.
17021 2021-05-04 Richard Biener <rguenther@suse.de>
17023 PR tree-optimization/100414
17024 * tree-ssa-phiopt.c (get_non_trapping): Do not compute dominance
17026 (tree_ssa_phiopt_worker): But unconditionally here.
17028 2021-05-04 Tobias Burnus <tobias@codesourcery.com>
17030 * omp-low.c (lower_rec_input_clauses, lower_reduction_clauses): Handle
17031 && and || with floating-point and complex arguments.
17033 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
17035 * tree-inline.c (insert_debug_decl_map): Delete.
17036 (copy_debug_stmt): Minor tweak.
17037 (setup_one_parameter): Do not use a variable if the value is either
17038 a read-only DECL or a non-addressable local variable in the caller.
17039 In this case, insert the debug-only variable in the map manually.
17040 (expand_call_inline): Do not generate a CLOBBER for these values.
17041 * tree-inline.h (debug_map): Minor tweak.
17043 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
17045 * builtins.c (builtin_with_linkage_p): Return true for stp[n]cpy.
17046 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Tidy up.
17048 2021-05-04 Richard Biener <rguenther@suse.de>
17050 PR tree-optimization/100329
17051 * tree-ssa-reassoc.c (can_reassociate_p): Do not reassociate
17053 (insert_stmt_after): Assert we're not running into asm goto.
17055 2021-05-04 Richard Biener <rguenther@suse.de>
17057 PR tree-optimization/100398
17058 * tree-ssa-dse.c (pass_dse::execute): Preserve control
17061 2021-05-04 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17063 * builtins.c (try_store_by_multiple_pieces): Fix constfun's prototype.
17065 2021-05-04 Alexandre Oliva <oliva@adacore.com>
17067 * builtins.c (try_store_by_multiple_pieces): New.
17068 (expand_builtin_memset_args): Use it. If target_char_cast
17069 fails, proceed as for non-constant val. Pass len's ctz to...
17070 * expr.c (clear_storage_hints): ... this. Try store by
17071 multiple pieces after setmem.
17072 (clear_storage): Adjust.
17073 * expr.h (clear_storage_hints): Likewise.
17074 (try_store_by_multiple_pieces): Declare.
17075 * passes.def: Replace the last copy_prop with ccp.
17077 2021-05-03 Tom de Vries <tdevries@suse.de>
17080 * omp-low.c (lower_rec_input_clauses): Disable SIMT for user-defined
17083 2021-05-03 Richard Biener <rguenther@suse.de>
17085 * tree-ssa-dse.c (dse_classify_store): Track two PHI defs.
17087 2021-05-03 Richard Biener <rguenther@suse.de>
17089 * tree-ssa-dse.c: Do not include domwalk.h but cfganal.h.
17090 (dse_dom_walker): Remove.
17091 (dse_dom_walker::dse_optimize_stmt): Rename...
17092 (dse_optimize_stmt): ... to this, pass in live_bytes sbitmap.
17093 (dse_dom_walker::before_dom_children): Inline ...
17094 (pass_dse::execute): ... here. Perform a reverse program
17097 2021-05-03 H.J. Lu <hjl.tools@gmail.com>
17100 * configure: Regenerated.
17102 2021-05-03 Ilya Leoshkevich <iii@linux.ibm.com>
17105 * config/s390/s390.c (s390_hard_fp_reg_p): New function.
17106 (s390_md_asm_adjust): Handle hard registers.
17108 2021-05-03 Jakub Jelinek <jakub@redhat.com>
17110 PR tree-optimization/100382
17111 * tree-ssa-dse.c: Include tree-eh.h.
17112 (dse_dom_walker::before_dom_children): Don't remove stmts if
17113 stmt_unremovable_because_of_non_call_eh_p is true.
17115 2021-05-02 David Edelsohn <dje.gcc@gmail.com>
17117 * varasm.c (compute_reloc_for_var): Split out from...
17118 (get_variable_section): Use it.
17119 * output.h (compute_reloc_for_var): Declare.
17120 * config/rs6000/rs6000-protos.h
17121 (rs6000_xcoff_asm_output_aligned_decl_common): Change alignment to
17123 * config/rs6000/rs6000.c (rs6000_legitimize_tls_address_aix):
17124 Don't append storage mapping class to symbol.
17125 (rs6000_xcoff_asm_named_section): Add BS and UL mapping classes.
17126 Don't convert TLS BSS to common.
17127 (rs6000_xcoff_unique_section): Don't fall back to select_secton.
17128 (rs6000_xcoff_section_type_flags): Add SECTION_BSS if DECL is
17130 (rs6000_xcoff_asm_globalize_decl_name): Don't strip storage
17132 (rs6000_xcoff_asm_output_aligned_decl_common): Align is unsigned int.
17133 If align is 0 from TLS class, use the same rules as varasm.c
17134 If not common, switch to BSS section manually.
17135 If common, emit appropriate comm or lcomm directive.
17136 (rs6000_xcoff_encode_section_info): Add logic to append all
17137 storage mapping classes.
17138 (rs6000_asm_weaken_decl): Adjust for qualname symbols.
17139 * config/rs6000/xcoff.h (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
17140 rs6000_xcoff_asm_output_aligned_decl_common.
17141 (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
17142 rs6000_xcoff_asm_output_aligned_decl_common.
17143 (ASM_OUTPUT_TLS_COMMON): Use
17144 rs6000_xcoff_asm_output_aligned_decl_common.
17146 2021-05-02 Jakub Jelinek <jakub@redhat.com>
17149 * config/nvptx/nvptx.c (nvptx_sese_pseudo): Use nullptr instead of 0
17150 as first argument of pseudo_node_t constructors.
17152 2021-05-02 Jakub Jelinek <jakub@redhat.com>
17155 * config/i386/t-i386 (TM_H): Add $(srcdir)/config/i386/i386-isa.def.
17157 2021-05-01 Aldy Hernandez <aldyh@redhat.com>
17159 * value-range.cc (DEFINE_INT_RANGE_GC_STUBS): Remove.
17160 (gt_pch_nx (int_range<1> *&)): New.
17161 (gt_ggc_mx (int_range<1> *&)): New.
17162 * value-range.h (class irange): Add GTY support for
17165 2021-05-01 Geng Qi <gengqi@linux.alibaba.com>
17167 * doc/options.texi (Negative): Change either or to both and.
17169 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
17171 * config/aarch64/aarch64-simd-builtins.def: Add
17172 float_ml[as][q]_laneq builtin generator macros.
17173 * config/aarch64/aarch64-simd.md (mul_laneq<mode>3): Define.
17174 (aarch64_float_mla_laneq<mode>): Define.
17175 (aarch64_float_mls_laneq<mode>): Define.
17176 * config/aarch64/arm_neon.h (vmla_laneq_f32): Use RTL builtin
17177 instead of GCC vector extensions.
17178 (vmlaq_laneq_f32): Likewise.
17179 (vmls_laneq_f32): Likewise.
17180 (vmlsq_laneq_f32): Likewise.
17182 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
17184 * config/aarch64/aarch64-simd-builtins.def: Add
17185 float_ml[as]_lane builtin generator macros.
17186 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt<mode>):
17188 (mul_lane<mode>3): This, and re-order arguments.
17189 (aarch64_float_mla_lane<mode>): Define.
17190 (aarch64_float_mls_lane<mode>): Define.
17191 * config/aarch64/arm_neon.h (vmla_lane_f32): Use RTL builtin
17192 instead of GCC vector extensions.
17193 (vmlaq_lane_f32): Likewise.
17194 (vmls_lane_f32): Likewise.
17195 (vmlsq_lane_f32): Likewise.
17197 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
17199 * config/aarch64/aarch64-simd-builtins.def: Add float_ml[as]
17200 builtin generator macros.
17201 * config/aarch64/aarch64-simd.md (aarch64_float_mla<mode>):
17203 (aarch64_float_mls<mode>): Define.
17204 * config/aarch64/arm_neon.h (vmla_f32): Use RTL builtin
17205 instead of relying on GCC vector extensions.
17206 (vmla_f64): Likewise.
17207 (vmlaq_f32): Likewise.
17208 (vmlaq_f64): Likewise.
17209 (vmls_f32): Likewise.
17210 (vmls_f64): Likewise.
17211 (vmlsq_f32): Likewise.
17212 (vmlsq_f64): Likewise.
17213 * config/aarch64/iterators.md: Define VDQF_DF mode iterator.
17215 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
17217 * config/aarch64/aarch64-simd-builtins.def: Add
17218 float_ml[as]_n_builtin generator macros.
17219 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt_from_dup<mode>):
17221 (mul_n<mode>3): This, and re-order arguments.
17222 (aarch64_float_mla_n<mode>): Define.
17223 (aarch64_float_mls_n<mode>): Define.
17224 * config/aarch64/arm_neon.h (vmla_n_f32): Use RTL builtin
17225 instead of inline asm.
17226 (vmlaq_n_f32): Likewise.
17227 (vmls_n_f32): Likewise.
17228 (vmlsq_n_f32): Likewise.
17230 2021-04-30 Jonathan Wright <joanthan.wright@arm.com>
17232 * config/aarch64/aarch64-simd-builtins.def: Add pmull[2]
17233 builtin generator macros.
17234 * config/aarch64/aarch64-simd.md (aarch64_pmullv8qi): Define.
17235 (aarch64_pmull_hiv16qi_insn): Define.
17236 (aarch64_pmull_hiv16qi): Define.
17237 * config/aarch64/arm_neon.h (vmull_high_p8): Use RTL builtin
17238 instead of inline asm.
17239 (vmull_p8): Likewise.
17241 2021-04-30 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
17243 * config/avr/avr.md: Adjust peepholes to match and
17244 generate parallels with clobber of REG_CC.
17245 (mov<mode>_insn): Rename to mov<mode>_insn_split.
17246 (*mov<mode>_insn): Rename to mov<mode>_insn.
17248 2021-04-30 David Edelsohn <dje.gcc@gmail.com>
17250 * varasm.c (use_blocks_for_decl_p): Don't use section anchors
17251 for VAR_DECLs if -fdata-sections enabled.
17253 2021-04-30 Michael Meissner <meissner@linux.ibm.com>
17255 PR bootstrap/100327
17256 * config/rs6000/rs6000.c
17257 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Define.
17258 (rs6000_libgcc_floating_mode_supported_p): New target hook.
17260 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
17262 * tree-ssa-threadbackward.c (class thread_jumps): Split out code
17264 (class back_threader_registry): ...to here...
17265 (class back_threader_profitability): ...and here...
17266 (thread_jumps::thread_through_all_blocks): Remove argument.
17267 (back_threader_registry::back_threader_registry): New.
17268 (back_threader_registry::~back_threader_registry): New.
17269 (back_threader_registry::thread_through_all_blocks): New.
17270 (thread_jumps::profitable_jump_thread_path): Move from here...
17271 (back_threader_profitability::profitable_path_p): ...to here.
17272 (thread_jumps::find_taken_edge): New.
17273 (thread_jumps::convert_and_register_current_path): Move...
17274 (back_threader_registry::register_path): ...to here.
17275 (thread_jumps::register_jump_thread_path_if_profitable): Move...
17276 (thread_jumps::maybe_register_path): ...to here.
17277 (thread_jumps::handle_phi): Call find_taken_edge and
17278 maybe_register_path.
17279 (thread_jumps::handle_assignment): Same.
17280 (thread_jumps::fsm_find_control_statement_thread_paths): Remove
17281 tree argument to handle_phi and handle_assignment.
17282 (thread_jumps::find_jump_threads_backwards): Set m_name. Remove
17283 set of m_speed_p and m_max_threaded_paths.
17284 (pass_thread_jumps::execute): Remove second argument from
17285 find_jump_threads_backwards.
17286 (pass_early_thread_jumps::execute): Same.
17288 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
17290 * tree-ssa-dom.c (class dom_jump_threader_simplifier): New.
17291 (class dom_opt_dom_walker): Initialize some class variables.
17292 (pass_dominator::execute): Pass evrp_range_analyzer and
17293 dom_jump_threader_simplifier to dom_opt_dom_walker.
17294 Adjust for some functions moving into classes.
17295 (simplify_stmt_for_jump_threading): Adjust and move to...
17296 (jump_threader_simplifier::simplify): ...here.
17297 (dom_opt_dom_walker::before_dom_children): Adjust for
17298 m_evrp_range_analyzer.
17299 (dom_opt_dom_walker::after_dom_children): Remove x_vr_values hack.
17300 (test_for_singularity): Place in dom_opt_dom_walker class.
17301 (dom_opt_dom_walker::optimize_stmt): The argument
17302 evrp_range_analyzer is now a class field.
17303 * tree-ssa-threadbackward.c (class thread_jumps): Add m_registry.
17304 (thread_jumps::thread_through_all_blocks): New.
17305 (thread_jumps::convert_and_register_current_path): Use m_registry.
17306 (pass_thread_jumps::execute): Adjust for thread_through_all_blocks
17307 being in the threader class.
17308 (pass_early_thread_jumps::execute): Same.
17309 * tree-ssa-threadedge.c (threadedge_initialize_values): Move...
17310 (jump_threader::jump_threader): ...here.
17311 (threadedge_finalize_values): Move...
17312 (jump_threader::~jump_threader): ...here.
17313 (jump_threader::remove_jump_threads_including): New.
17314 (jump_threader::thread_through_all_blocks): New.
17315 (record_temporary_equivalences_from_phis): Move...
17316 (jump_threader::record_temporary_equivalences_from_phis): ...here.
17317 (record_temporary_equivalences_from_stmts_at_dest): Move...
17318 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
17320 (simplify_control_stmt_condition_1): Move to jump_threader class.
17321 (simplify_control_stmt_condition): Move...
17322 (jump_threader::simplify_control_stmt_condition): ...here.
17323 (thread_around_empty_blocks): Move...
17324 (jump_threader::thread_around_empty_blocks): ...here.
17325 (thread_through_normal_block): Move...
17326 (jump_threader::thread_through_normal_block): ...here.
17327 (thread_across_edge): Move...
17328 (jump_threader::thread_across_edge): ...here.
17329 (thread_outgoing_edges): Move...
17330 (jump_threader::thread_outgoing_edges): ...here.
17331 * tree-ssa-threadedge.h: Move externally facing functings...
17332 (class jump_threader): ...here...
17333 (class jump_threader_simplifier): ...and here.
17334 * tree-ssa-threadupdate.c (struct redirection_data): Remove comment.
17335 (jump_thread_path_allocator::jump_thread_path_allocator): New.
17336 (jump_thread_path_allocator::~jump_thread_path_allocator): New.
17337 (jump_thread_path_allocator::allocate_thread_edge): New.
17338 (jump_thread_path_allocator::allocate_thread_path): New.
17339 (jump_thread_path_registry::jump_thread_path_registry): New.
17340 (jump_thread_path_registry::~jump_thread_path_registry): New.
17341 (jump_thread_path_registry::allocate_thread_edge): New.
17342 (jump_thread_path_registry::allocate_thread_path): New.
17343 (dump_jump_thread_path): Make extern.
17344 (debug (const vec<jump_thread_edge *> &path)): New.
17345 (struct removed_edges): Move to tree-ssa-threadupdate.h.
17346 (struct thread_stats_d): Remove.
17347 (remove_ctrl_stmt_and_useless_edges): Make static.
17348 (lookup_redirection_data): Move...
17349 (jump_thread_path_registry::lookup_redirection_data): ...here.
17350 (ssa_redirect_edges): Make static.
17351 (thread_block_1): Move...
17352 (jump_thread_path_registry::thread_block_1): ...here.
17353 (thread_block): Move...
17354 (jump_thread_path_registry::thread_block): ...here.
17355 (thread_through_loop_header): Move...
17356 (jump_thread_path_registry::thread_through_loop_header): ...here.
17357 (mark_threaded_blocks): Move...
17358 (jump_thread_path_registry::mark_threaded_blocks): ...here.
17359 (debug_path): Move...
17360 (jump_thread_path_registry::debug_path): ...here.
17361 (debug_all_paths): Move...
17362 (jump_thread_path_registry::dump): ..here.
17363 (rewire_first_differing_edge): Move...
17364 (jump_thread_path_registry::rewire_first_differing_edge): ...here.
17365 (adjust_paths_after_duplication): Move...
17366 (jump_thread_path_registry::adjust_paths_after_duplication): ...here.
17367 (duplicate_thread_path): Move...
17368 (jump_thread_path_registry::duplicate_thread_path): ..here.
17369 (remove_jump_threads_including): Move...
17370 (jump_thread_path_registry::remove_jump_threads_including): ...here.
17371 (thread_through_all_blocks): Move to...
17372 (jump_thread_path_registry::thread_through_all_blocks): ...here.
17373 (delete_jump_thread_path): Remove.
17374 (register_jump_thread): Move...
17375 (jump_thread_path_registry::register_jump_thread): ...here.
17376 * tree-ssa-threadupdate.h: Move externally facing functions...
17377 (class jump_thread_path_allocator): ...here...
17378 (class jump_thread_path_registry): ...and here.
17379 (thread_through_all_blocks): Remove.
17380 (struct removed_edges): New.
17381 (register_jump_thread): Remove.
17382 (remove_jump_threads_including): Remove.
17383 (delete_jump_thread_path): Remove.
17384 (remove_ctrl_stmt_and_useless_edges): Remove.
17385 (free_dom_edge_info): New prototype.
17386 * tree-vrp.c: Remove x_vr_values hack.
17387 (class vrp_jump_threader_simplifier): New.
17388 (vrp_jump_threader_simplifier::simplify): New.
17389 (vrp_jump_threader::vrp_jump_threader): Adjust method signature.
17390 Remove m_dummy_cond.
17391 Instantiate m_simplifier and m_threader.
17392 (vrp_jump_threader::thread_through_all_blocks): New.
17393 (vrp_jump_threader::simplify_stmt): Remove.
17394 (vrp_jump_threader::after_dom_children): Do not set m_dummy_cond.
17395 Remove x_vr_values hack.
17396 (execute_vrp): Adjust for thread_through_all_blocks being in a
17399 2021-04-30 Christophe Lyon <christophe.lyon@linaro.org>
17401 * genflags.c (gen_insn): Print failed expansion string.
17403 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
17405 * expr.c (alignment_for_piecewise_move): Call mode_for_size
17406 without limit to MAX_FIXED_MODE_SIZE.
17408 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
17410 PR middle-end/90773
17411 * builtins.c (builtin_memset_gen_str): Don't use return from
17412 simplify_gen_subreg.
17414 2021-04-30 Uroš Bizjak <ubizjak@gmail.com>
17417 * config/i386/i386.md (*add<mode>3_carry_0r): New insn pattern.
17418 (*addsi3_carry_zext_0r): Ditto.
17419 (*sub<mode>3_carry_0): Ditto.
17420 (*subsi3_carry_zext_0r): Ditto.
17421 * config/i386/predicates.md (ix86_carry_flag_unset_operator):
17423 * config/i386/i386.c (ix86_rtx_costs) <case PLUS, case MINUS>:
17424 Also consider ix86_carry_flag_unset_operator to calculate
17425 the cost of adc/sbb insn.
17427 2021-04-30 Roman Zhuykov <zhroma@ispras.ru>
17429 PR rtl-optimization/100225
17430 PR rtl-optimization/84878
17431 * modulo-sched.c (sms_schedule): Use note_stores to skip loops
17432 where we have an instruction which touches (writes) any hard
17433 register from df->regular_block_artificial_uses set.
17434 Allow not-single-set instruction only right before basic block
17437 2021-04-30 Geng Qi <gengqi@linux.alibaba.com>
17439 * config/riscv/riscv.opt (march=,mabi=): Negative itself.
17441 2021-04-30 LevyHsu <admin@levyhsu.com>
17443 * config/riscv/riscv.c (riscv_min_arithmetic_precision): New.
17444 * config/riscv/riscv.h (TARGET_MIN_ARITHMETIC_PRECISION): New.
17445 * config/riscv/riscv.md (addv<mode>4, uaddv<mode>4): New.
17446 (subv<mode>4, usubv<mode>4, mulv<mode>4, umulv<mode>4): New.
17448 2021-04-29 Alexandre Oliva <oliva@adacore.com>
17450 * config.gcc: Merged x86 and x86_64 cpu_type-setting cases.
17452 2021-04-29 Alexandre Oliva <oliva@adacore.com>
17454 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Rename to...
17455 (ASM_OUTPUT_MAX_SKIP_ALIGN): ... this. Enclose in do/while(0).
17456 * config/i386/i386.c: Adjust.
17457 * config/i386/i386.md: Adjust.
17458 * config/i386/darwin.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Drop.
17459 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17460 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17461 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17462 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17463 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17464 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17465 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17466 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17467 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17468 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
17470 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
17472 * config/i386/i386-expand.c (ix86_expand_int_compare):
17473 Swap operands of GTU and LEU comparison to emit carry flag comparison.
17474 * config/i386/i386.md (*add<mode>3_carry_0): Change insn
17475 predicate to allow more combine opportunities with memory operands.
17476 (*sub<mode>3_carry_0): Ditto.
17478 2021-04-29 Richard Sandiford <richard.sandiford@arm.com>
17480 PR rtl-optimization/100303
17481 * rtl-ssa/accesses.cc (function_info::make_use_available): Take a
17482 boolean that indicates whether the use will only be used in
17483 debug instructions. Treat it in the same way that existing
17484 cross-EBB debug references would be handled if so.
17485 (function_info::make_uses_available): Likewise.
17486 * rtl-ssa/functions.h (function_info::make_uses_available): Update
17487 prototype accordingly.
17488 (function_info::make_uses_available): Likewise.
17489 * fwprop.c (try_fwprop_subst): Update call accordingly.
17491 2021-04-29 Jeff Law <jlaw@tachyum.com>
17493 * config/nios2/nios2-protos.h (nios2_fpu_insn_enabled): Move outside
17496 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
17497 Richard Biener <rguenther@suse.de>
17500 * config/i386/i386-builtin.def (IX86_BUILTIN_MASKLOADPD)
17501 (IX86_BUILTIN_MASKLOADPS, IX86_BUILTIN_MASKLOADPD256)
17502 (IX86_BUILTIN_MASKLOADPS256, IX86_BUILTIN_MASKLOADD)
17503 (IX86_BUILTIN_MASKLOADQ, IX86_BUILTIN_MASKLOADD256)
17504 (IX86_BUILTIN_MASKLOADQ256): Move from SPECIAL_ARGS
17505 to PURE_ARGS category.
17506 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
17507 Handle PURE_ARGS category.
17508 * config/i386/i386-expand.c (ix86_expand_builtin): Ditto.
17510 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
17512 * configure.ac: Check for the presence of sys/locking.h header and
17513 for whether _LK_LOCK is supported by _locking.
17514 * configure: Regenerate.
17515 * config.in: Likewise.
17516 * gcov-io.h: Define GCOV_LOCKED_WITH_LOCKING if HOST_HAS_LK_LOCK.
17517 * gcov-io.c (gcov_open): Add support for GCOV_LOCKED_WITH_LOCKING.
17518 * system.h: Include <sys/locking.h> if HAVE_SYS_LOCKING_H.
17520 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
17522 * config/i386/predicates.md (fcmov_comparison_operator):
17523 Do not check for trivial FP comparison operator.
17524 <case GEU, case LTU>: Allow CCGZmode.
17525 <case GTU, case LEU>: Do not allow CCCmode.
17526 (ix86_comparison_operator) <case GTU, case LEU>: Allow only CCmode.
17527 (ix86_carry_flag_operator): Match only LTU and UNLT code.
17528 Do not check for trivial FP comparison operator. Allow CCGZmode.
17530 2021-04-29 Tom de Vries <tdevries@suse.de>
17532 * omp-expand.c (expand_omp_simd): Add step_orig, and replace uses of
17533 fd->loop.step by either step or orig_step.
17535 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
17537 * config/sparc/sparc.c (gen_load_pcrel_sym): Delete.
17538 (load_got_register): Do the PIC dance here.
17539 (sparc_legitimize_tls_address): Simplify.
17540 (sparc_emit_probe_stack_range): Likewise.
17541 (sparc32_initialize_trampoline): Likewise.
17542 (sparc64_initialize_trampoline): Likewise.
17543 * config/sparc/sparc.md (load_pcrel_sym<P:mode>): Add @ marker.
17544 (probe_stack_range<P:mode>): Likewise.
17545 (flush<P:mode>): Likewise.
17546 (tgd_hi22<P:mode>): Likewise.
17547 (tgd_lo10<P:mode>): Likewise.
17548 (tgd_add<P:mode>): Likewise.
17549 (tgd_call<P:mode>): Likewise.
17550 (tldm_hi22<P:mode>): Likewise.
17551 (tldm_lo10<P:mode>): Likewise.
17552 (tldm_add<P:mode>): Likewise.
17553 (tldm_call<P:mode>): Likewise.
17554 (tldo_hix22<P:mode>): Likewise.
17555 (tldo_lox10<P:mode>): Likewise.
17556 (tldo_add<P:mode>): Likewise.
17557 (tie_hi22<P:mode>): Likewise.
17558 (tie_lo10<P:mode>): Likewise.
17559 (tie_add<P:mode>): Likewise.
17560 (tle_hix22<P:mode>): Likewise.
17561 (tle_lox10<P:mode>): Likewise.
17562 (stack_protect_setsi): Rename to...
17563 (stack_protect_set32): ...this.
17564 (stack_protect_setdi): Rename to...
17565 (stack_protect_set64): ...this.
17566 (stack_protect_set): Adjust calls to above.
17567 (stack_protect_testsi): Rename to...
17568 (stack_protect_test32): ...this.
17569 (stack_protect_testdi): Rename to...
17570 (stack_protect_test64): ...this.
17571 (stack_protect_test): Adjust calls to above.
17573 2021-04-29 H.J. Lu <hjl.tools@gmail.com>
17575 PR middle-end/90773
17576 * builtins.c (builtin_memcpy_read_str): Add a dummy argument.
17577 (builtin_strncpy_read_str): Likewise.
17578 (builtin_memset_read_str): Add an argument for the previous RTL
17579 information and generate the new RTL from the previous RTL info.
17580 (builtin_memset_gen_str): Likewise.
17581 * builtins.h (builtin_strncpy_read_str): Update the prototype.
17582 (builtin_memset_read_str): Likewise.
17583 * expr.c (by_pieces_ninsns): If targetm.overlap_op_by_pieces_p()
17584 returns true, round up size and alignment to the widest integer
17585 mode for maximum size.
17586 (pieces_addr::adjust): Add a pointer to by_pieces_prev argument
17587 and pass it to m_constfn.
17588 (op_by_pieces_d): Add m_push and m_overlap_op_by_pieces.
17589 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
17590 initialize m_push. Initialize m_overlap_op_by_pieces with
17591 targetm.overlap_op_by_pieces_p ().
17592 (op_by_pieces_d::run): Pass the previous RTL information to
17593 pieces_addr::adjust and generate overlapping operations if
17594 m_overlap_op_by_pieces is true.
17596 (move_by_pieces_d::move_by_pieces_d): Updated for op_by_pieces_d
17598 (store_by_pieces_d::store_by_pieces_d): Updated for op_by_pieces_d
17600 (can_store_by_pieces): Use by_pieces_constfn on constfun.
17601 (store_by_pieces): Use by_pieces_constfn on constfun. Updated
17602 for op_by_pieces_d change.
17603 (clear_by_pieces_1): Add a dummy argument.
17604 (clear_by_pieces): Updated for op_by_pieces_d change.
17605 (compare_by_pieces_d::compare_by_pieces_d): Likewise.
17606 (string_cst_read_str): Add a dummy argument.
17607 * expr.h (by_pieces_constfn): Add a dummy argument.
17608 (by_pieces_prev): New.
17609 * target.def (overlap_op_by_pieces_p): New target hook.
17610 * config/i386/i386.c (TARGET_OVERLAP_OP_BY_PIECES_P): New.
17611 * doc/tm.texi.in: Add TARGET_OVERLAP_OP_BY_PIECES_P.
17612 * doc/tm.texi: Regenerated.
17614 2021-04-29 Richard Biener <rguenther@suse.de>
17616 PR tree-optimization/100253
17617 * tree-vect-stmts.c (vectorizable_load): Do not assume
17618 element alignment when DR_MISALIGNMENT is -1.
17619 (vectorizable_store): Likewise.
17621 2021-04-29 Jakub Jelinek <jakub@redhat.com>
17624 * config/aarch64/aarch64.c (aarch64_add_offset_1_temporaries): Use
17625 absu_hwi instead of abs_hwi.
17627 2021-04-29 Richard Biener <rguenther@suse.de>
17629 PR middle-end/38474
17630 * tree-ssa-structalias.c (add_graph_edge): Avoid direct
17631 forwarding when indirect forwarding through ESCAPED
17634 2021-04-29 Tom de Vries <tdevries@suse.de>
17637 * internal-fn.c (expand_GOMP_SIMT_ENTER_ALLOC)
17638 (expand_GOMP_SIMT_LAST_LANE, expand_GOMP_SIMT_ORDERED_PRED)
17639 (expand_GOMP_SIMT_VOTE_ANY, expand_GOMP_SIMT_XCHG_BFLY)
17640 (expand_GOMP_SIMT_XCHG_IDX): Ensure target is assigned to.
17642 2021-04-29 Richard Biener <rguenther@suse.de>
17644 PR tree-optimization/99912
17645 * tree-ssa-dse.c (dse_dom_walker::m_need_cfg_cleanup): New.
17646 (dse_dom_walker::todo): Likewise.
17647 (dse_dom_walker::dse_optimize_stmt): Move VDEF check to the
17649 (dse_dom_walker::before_dom_children): Remove trivially
17650 dead SSA defs and schedule CFG cleanup if we removed all
17652 (pass_dse::execute): Get TODO as computed by the DOM walker
17653 and return it. Wipe dominator info earlier.
17655 2021-04-29 Richard Biener <rguenther@suse.de>
17658 * ipa-prop.c (ipcp_modif_dom_walker::before_dom_children):
17659 Track blocks to cleanup EH in new m_need_eh_cleanup.
17660 (ipcp_modif_dom_walker::cleanup_eh): New.
17661 (ipcp_transform_function): Release dominator info before
17664 2021-04-29 Martin Sebor <msebor@redhat.com>
17666 PR middle-end/100250
17667 * attribs.c (attr_access::array_as_string): Avoid dereferencing
17668 a pointer when it's null.
17670 2021-04-29 Martin Sebor <msebor@redhat.com>
17672 * Makefile.in (OBJS): Add ipa-free-lang-data.o.
17673 * ipa-free-lang-data.cc: New file.
17674 * tree.c: Move pass free_lang_data to file above.
17675 (build_array_type_1): Declare extern.
17676 * tree.h (build_array_type_1): Declare.
17678 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17680 * config/aarch64/aarch64-simd-builtins.def: Modify comment to
17681 make consistent with updated RTL pattern.
17682 * config/aarch64/aarch64-simd.md (aarch64_<sur>qmovn<mode>):
17683 Implement using ss_truncate and us_truncate rather than
17685 * config/aarch64/iterators.md: Remove redundant unspecs and
17686 iterator: UNSPEC_[SU]QXTN and SUQMOVN respectively.
17688 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17690 * config/aarch64/arm_acle.h (__attribute__): Make intrinsic
17691 attributes consistent with those defined in arm_neon.h.
17693 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17695 * config/aarch64/arm_fp16.h (__attribute__): Make intrinsic
17696 attributes consistent with those defined in arm_neon.h.
17698 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17700 * config/aarch64/aarch64-simd-builtins.def: Add
17701 float_trunc_rodd builtin generator macros.
17702 * config/aarch64/aarch64-simd.md (aarch64_float_trunc_rodd_df):
17704 (aarch64_float_trunc_rodd_lo_v2sf): Define.
17705 (aarch64_float_trunc_rodd_hi_v4sf_le): Define.
17706 (aarch64_float_trunc_rodd_hi_v4sf_be): Define.
17707 (aarch64_float_trunc_rodd_hi_v4sf): Define.
17708 * config/aarch64/arm_neon.h (vcvtx_f32_f64): Use RTL builtin
17709 instead of inline asm.
17710 (vcvtx_high_f32_f64): Likewise.
17711 (vcvtxd_f32_f64): Likewise.
17712 * config/aarch64/iterators.md: Add FCVTXN unspec.
17714 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17716 * config/aarch64/aarch64-simd-builtins.def: Add tbx1 builtin
17718 * config/aarch64/aarch64-simd.md (aarch64_tbx1<mode>):
17720 * config/aarch64/arm_neon.h (vqtbx1_s8): USE RTL builtin
17721 instead of inline asm.
17722 (vqtbx1_u8): Likewise.
17723 (vqtbx1_p8): Likewise.
17724 (vqtbx1q_s8): Likewise.
17725 (vqtbx1q_u8): Likewise.
17726 (vqtbx1q_p8): Likewise.
17727 (vtbx2_s8): Likewise.
17728 (vtbx2_u8): Likewise.
17729 (vtbx2_p8): Likewise.
17731 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17733 * config/aarch64/aarch64-simd-builtins.def: Add tbl1 builtin
17735 * config/aarch64/arm_neon.h (vqtbl1_p8): Use RTL builtin
17736 instead of inline asm.
17737 (vqtbl1_s8): Likewise.
17738 (vqtbl1_u8): Likewise.
17739 (vqtbl1q_p8): Likewise.
17740 (vqtbl1q_s8): Likewise.
17741 (vqtbl1q_u8): Likewise.
17742 (vtbl1_s8): Likewise.
17743 (vtbl1_u8): Likewise.
17744 (vtbl1_p8): Likewise.
17745 (vtbl2_s8): Likewise.
17746 (vtbl2_u8): Likewise.
17747 (vtbl2_p8): Likewise.
17749 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17751 * config/aarch64/aarch64-simd-builtins.def: Add polynomial
17752 ssri_n buitin generator macro.
17753 * config/aarch64/arm_neon.h (vsri_n_p8): Use RTL builtin
17754 instead of inline asm.
17755 (vsri_n_p16): Likewise.
17756 (vsri_n_p64): Likewise.
17757 (vsriq_n_p8): Likewise.
17758 (vsriq_n_p16): Likewise.
17759 (vsriq_n_p64): Likewise.
17761 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17763 * config/aarch64/aarch64-simd-builtins.def: Use VALLP mode
17764 iterator for polynomial ssli_n builtin generator macro.
17765 * config/aarch64/arm_neon.h (vsli_n_p8): Use RTL builtin
17766 instead of inline asm.
17767 (vsli_n_p16): Likewise.
17768 (vsliq_n_p8): Likewise.
17769 (vsliq_n_p16): Likewise.
17770 * config/aarch64/iterators.md: Define VALLP mode iterator.
17772 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17774 * config/aarch64/aarch64-simd-builtins.def: Use VDQV_L
17775 iterator to generate [su]adalp RTL builtins.
17776 * config/aarch64/aarch64-simd.md: Use VDQV_L iterator in
17777 [su]adalp RTL pattern.
17778 * config/aarch64/arm_neon.h (vpadal_s32): Use RTL builtin
17779 instead of inline asm.
17780 (vpadal_u32): Likewise.
17782 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17784 * config/aarch64/aarch64-simd-builtins.def: Add [su]addlp
17785 builtin generator macros.
17786 * config/aarch64/aarch64-simd.md (aarch64_<su>addlp<mode>):
17788 * config/aarch64/arm_neon.h (vpaddl_s8): Use RTL builtin
17789 instead of inline asm.
17790 (vpaddl_s16): Likewise.
17791 (vpaddl_s32): Likewise.
17792 (vpaddl_u8): Likewise.
17793 (vpaddl_u16): Likewise.
17794 (vpaddl_u32): Likewise.
17795 (vpaddlq_s8): Likewise.
17796 (vpaddlq_s16): Likewise.
17797 (vpaddlq_s32): Likewise.
17798 (vpaddlq_u8): Likewise.
17799 (vpaddlq_u16): Likewise.
17800 (vpaddlq_u32): Liwewise.
17801 * config/aarch64/iterators.md: Define [SU]ADDLP unspecs with
17802 appropriate attributes.
17804 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17806 * config/aarch64/aarch64-simd-builtins.def: Use VDQ_I iterator
17807 for aarch64_addp<mode> builtin macro generator.
17808 * config/aarch64/aarch64-simd.md: Use VDQ_I iterator in
17809 aarch64_addp<mode> RTL pattern.
17810 * config/aarch64/arm_neon.h (vpaddq_s8): Use RTL builtin
17811 instead of inline asm.
17812 (vpaddq_s16): Likewise.
17813 (vpaddq_s32): Likewise.
17814 (vpaddq_s64): Likewise.
17815 (vpaddq_u8): Likewise.
17816 (vpaddq_u16): Likewise.
17817 (vpaddq_u32): Likewise.
17818 (vpaddq_u64): Likewise.
17820 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17822 * config/aarch64/aarch64-simd-builtins.def: Add sq[r]dmulh_n
17823 builtin generator macros.
17824 * config/aarch64/aarch64-simd.md (aarch64_sq<r>dmulh_n<mode>):
17826 * config/aarch64/arm_neon.h (vqdmulh_n_s16): Use RTL builtin
17827 instead of inline asm.
17828 (vqdmulh_n_s32): Likewise.
17829 (vqdmulhq_n_s16): Likewise.
17830 (vqdmulhq_n_s32): Likewise.
17831 (vqrdmulh_n_s16): Likewise.
17832 (vqrdmulh_n_s32): Likewise.
17833 (vqrdmulhq_n_s16): Likewise.
17834 (vqrdmulhq_n_s32): Likewise.
17836 2021-04-28 Tobias Burnus <tobias@codesourcery.com>
17838 * doc/install.texi (--enable-offload-defaulted): Document.
17840 2021-04-28 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
17842 * config/avr/avr-dimode.md: Turn existing patterns into
17843 define_insn_and_split style patterns where the splitter
17844 adds a clobber of the condition code register. Drop "cc"
17845 attribute. Add new patterns to match output of
17847 * config/avr/avr-fixed.md: Likewise.
17848 * config/avr/avr.c (cc_reg_rtx): New.
17849 (avr_parallel_insn_from_insns): Adjust insn count
17850 for removal of set of cc0.
17851 (avr_is_casesi_sequence): Likewise.
17852 (avr_casei_sequence_check_operands): Likewise.
17853 (avr_optimize_casesi): Likewise. Also insert
17854 new insns after jump_insn.
17855 (avr_pass_casesi::avr_rest_of_handle_casesi): Adjust
17856 for removal of set of cc0.
17857 (avr_init_expanders): Initialize cc_reg_rtx.
17858 (avr_regno_reg_class): Handle REG_CC.
17859 (cond_string): Remove usage of CC_OVERFLOW_UNUSABLE.
17860 (avr_notice_update_cc): Remove function.
17861 (ret_cond_branch): Remove usage of CC_OVERFLOW_UNUSABLE.
17862 (compare_condition): Adjust for PARALLEL with
17864 (out_shift_with_cnt): Likewise.
17865 (ashlhi3_out): Likewise.
17866 (ashrhi3_out): Likewise.
17867 (lshrhi3_out): Likewise.
17868 (avr_class_max_nregs): Return single reg for REG_CC.
17869 (avr_compare_pattern): Check for REG_CC instead
17871 (avr_reorg_remove_redundant_compare): Likewise.
17872 (avr_reorg):Adjust for PARALLEL with REG_CC clobber.
17873 (avr_hard_regno_nregs): Return single reg for REG_CC.
17874 (avr_hard_regno_mode_ok): Allow only CCmode for REG_CC.
17875 (avr_md_asm_adjust): Clobber REG_CC.
17876 (TARGET_HARD_REGNO_NREGS): Define.
17877 (TARGET_CLASS_MAX_NREGS): Define.
17878 (TARGET_MD_ASM_ADJUST): Define.
17879 * config/avr/avr.h (FIRST_PSEUDO_REGISTER): Adjust
17881 (enum reg_class): Add CC_REG class.
17882 (NOTICE_UPDATE_CC): Remove.
17883 (CC_OVERFLOW_UNUSABLE): Remove.
17884 (CC_NO_CARRY): Remove.
17885 * config/avr/avr.md: Turn existing patterns into
17886 define_insn_and_split style patterns where the splitter
17887 adds a clobber of the condition code register. Drop "cc"
17888 attribute. Add new patterns to match output of
17890 (sez): Remove unused pattern.
17892 2021-04-28 Richard Earnshaw <rearnsha@arm.com>
17895 * config/arm/arm.c (arm_hard_regno_mode_ok): Only allow VPR to be
17898 2021-04-28 Richard Sandiford <richard.sandiford@arm.com>
17901 * config/aarch64/constraints.md (Utq): Require the address to
17902 be valid for both the element mode and for V2DImode.
17904 2021-04-28 Jakub Jelinek <jakub@redhat.com>
17905 Tobias Burnus <tobias@codesourcery.com>
17907 * configure.ac (OFFLOAD_DEFAULTED): AC_DEFINE if offload-defaulted.
17908 * gcc.c (process_command): New variable.
17909 (driver::maybe_putenv_OFFLOAD_TARGETS): If OFFLOAD_DEFAULTED,
17910 set it if -foffload is defaulted.
17911 * lto-wrapper.c (OFFLOAD_TARGET_DEFAULT_ENV): Define.
17912 (compile_offload_image): If OFFLOAD_DEFAULTED and
17913 OFFLOAD_TARGET_DEFAULT is in the environment, don't fail
17914 if corresponding mkoffload can't be found.
17915 (compile_images_for_offload_targets): Likewise. Free and clear
17916 offload_names if no valid offload is found.
17917 * config.in: Regenerate.
17918 * configure: Regenerate.
17920 2021-04-28 Richard Biener <rguenther@suse.de>
17922 PR tree-optimization/100292
17923 * tree-vect-generic.c (expand_vector_condition): Do not fold
17926 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
17928 * config/rs6000/aix.h (SUBTARGET_DRIVER_SELF_SPECS): New.
17929 * config/rs6000/aix64.opt (m64): New.
17932 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
17934 * config/vax/vax.c (print_operand_address, vax_address_cost_1)
17935 (index_term_p): Handle ASHIFT too.
17937 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
17939 * config/vax/builtins.md (jbb<ccss>i<mode>): Remove operand #3.
17940 (sync_lock_test_and_set<mode>): Adjust accordingly.
17941 (sync_lock_release<mode>): Likewise.
17943 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
17945 * config/vax/vax-protos.h (adjacent_operands_p): Remove
17947 * config/vax/vax.c (adjacent_operands_p): Remove.
17949 2021-04-27 Maciej W. Rozycki <macro@linux-mips.org>
17951 * ifcvt.c (dead_or_predicable) [!IFCVT_MODIFY_TESTS]: Fall
17952 through to the non-conditional execution case if getting the
17953 condition for conditional execution has failed.
17955 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
17957 PR middle-end/100284
17958 * gimple.c (gimple_could_trap_p_1): Remove VEC_COND_EXPR test.
17959 * tree-eh.c (operation_could_trap_p): Handle VEC_COND_EXPR rather
17960 than asserting on it.
17962 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
17964 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): Protect
17965 with TARGET_AIX_OS.
17967 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
17970 * calls.c (precompute_register_parameters): Additionally test
17971 targetm.precompute_tls_p to pre-compute argument.
17972 * config/rs6000/aix.h (TARGET_PRECOMPUTE_TLS_P): Define.
17973 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): New.
17974 * target.def (precompute_tls_p): New.
17975 * doc/tm.texi.in (TARGET_PRECOMPUTE_TLS_P): Add hook documentation.
17976 * doc/tm.texi: Regenerated.
17978 2021-04-27 Jakub Jelinek <jakub@redhat.com>
17981 * config/aarch64/aarch64.c (aarch64_print_operand): Cast -UINTVAL
17982 back to HOST_WIDE_INT.
17984 2021-04-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
17987 * simplify-rtx.c (simplify_context::simplify_subreg): Check the
17988 memory alignment for the outer mode.
17990 2021-04-27 H.J. Lu <hjl.tools@gmail.com>
17992 PR middle-end/90773
17993 * expr.c (op_by_pieces_d::get_usable_mode): New member function.
17994 (op_by_pieces_d::run): Cange a while loop to a do-while loop.
17996 2021-04-27 Alex Coplan <alex.coplan@arm.com>
17999 * config/arm/arm.c (arm_split_compare_and_swap): Fix up codegen
18000 with negative immediates: ensure we expand cbranchsi4_scratch
18001 correctly and ensure we satisfy its constraints.
18002 * config/arm/sync.md
18003 (@atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1): Don't
18004 attempt to tie two output operands together with constraints;
18005 collapse two alternatives.
18006 (@atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): Likewise.
18007 * config/arm/thumb1.md (cbranchsi4_neg_late): New.
18009 2021-04-27 Jakub Jelinek <jakub@redhat.com>
18012 * config/aarch64/predicates.md (aarch64_sub_immediate,
18013 aarch64_plus_immediate): Use -UINTVAL instead of -INTVAL.
18014 * config/aarch64/aarch64.md (casesi, rotl<mode>3): Likewise.
18015 * config/aarch64/aarch64.c (aarch64_print_operand,
18016 aarch64_split_atomic_op, aarch64_expand_subvti): Likewise.
18018 2021-04-27 Jakub Jelinek <jakub@redhat.com>
18020 PR tree-optimization/100239
18021 * tree-vect-generic.c (lower_vec_perm): Don't accept constant
18022 permutations with all indices from the first zero element as vec_shl.
18024 2021-04-27 Jakub Jelinek <jakub@redhat.com>
18026 PR rtl-optimization/100254
18027 * cfgcleanup.c (outgoing_edges_match): Check REG_EH_REGION on
18028 last1 and last2 insns rather than BB_END (bb1) and BB_END (bb2) insns.
18030 2021-04-27 Richard Biener <rguenther@suse.de>
18032 PR tree-optimization/99912
18033 * passes.def: Add comment about new TODO_remove_unused_locals.
18034 * tree-stdarg.c (pass_data_stdarg): Run TODO_remove_unused_locals
18037 2021-04-27 Richard Biener <rguenther@suse.de>
18039 PR tree-optimization/99912
18040 * passes.def (pass_all_optimizations): Add pass_dse before
18041 the first pass_dce, move the first pass_dse before the
18042 pass_dce following pass_pre.
18044 2021-04-27 Jakub Jelinek <jakub@redhat.com>
18046 PR tree-optimization/95527
18047 * generic-match-head.c: Include tm.h.
18048 * gimple-match-head.c: Include tm.h.
18049 * match.pd (CLZ == INTEGER_CST): Don't use
18050 #ifdef CLZ_DEFINED_VALUE_AT_ZERO, only test CLZ_DEFINED_VALUE_AT_ZERO
18051 if clz == CFN_CLZ. Add missing val declaration.
18052 (CTZ cmp CST): New simplifications.
18054 2021-04-27 Jakub Jelinek <jakub@redhat.com>
18056 PR tree-optimization/96696
18057 * expr.c (expand_expr_divmod): New function.
18058 (expand_expr_real_2) <case TRUNC_DIV_EXPR>: Use it for truncations and
18059 divisions. Formatting fixes.
18060 <case MULT_EXPR>: Optimize x / y * y as x - x % y if the latter is
18063 2021-04-27 Martin Jambor <mjambor@suse.cz>
18066 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call):
18067 If removing a call statement LHS SSA name, release it.
18069 2021-04-27 Richard Earnshaw <rearnsha@arm.com>
18072 * config/arm/arm.c (THUMB2_WORK_REGS): Check PIC_OFFSET_TABLE_REGNUM
18073 is valid before including it in the mask.
18075 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
18078 * config/aarch64/aarch64.c (aarch64_comp_type_attributes): Handle
18081 2021-04-27 Richard Biener <rguenther@suse.de>
18083 PR tree-optimization/100051
18084 * tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Add
18085 disambiguator based on access size vs. decl size.
18087 2021-04-27 Richard Biener <rguenther@suse.de>
18089 PR tree-optimization/100278
18090 * tree-ssa-pre.c (compute_avail): Give up when we cannot
18091 adjust TBAA beacuse of mismatching bases.
18093 2021-04-27 Jakub Jelinek <jakub@redhat.com>
18096 * config/i386/i386.md (*<insn><mode>3_mask, *<insn><mode>3_mask_1):
18097 For any_rotate define_insn_split and following splitters, use
18098 SWI iterator instead of SWI48.
18100 2021-04-27 Richard Biener <rguenther@suse.de>
18102 PR tree-optimization/99776
18103 * match.pd (bit_field_ref (ctor)): Relax element extract
18104 type compatibility checks.
18106 2021-04-27 Cui,Lili <lili.cui@intel.com>
18108 * common/config/i386/i386-common.c (processor_names):
18109 Sync processor_names with processor_type.
18110 * config/i386/i386-options.c (processor_cost_table):
18111 Sync processor_cost_table with processor_type.
18113 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
18115 * value-range.cc (irange::irange_set_1bit_anti_range): Add assert.
18116 (irange::set): Call irange_set_1bit_anti_range for handling all
18117 1-bit ranges. Fall through on ~[MIN,MAX].
18119 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
18121 * value-range.cc (irange::legacy_num_pairs): Remove.
18122 (irange::invert): Change gcc_assert to gcc_checking_assert.
18123 * value-range.h (irange::num_pairs): Adjust for a cached
18124 num_pairs(). Also, rename all gcc_assert's to
18125 gcc_checking_assert's.
18127 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
18129 * value-range.cc (irange::operator=): Set m_kind.
18130 (irange::copy_to_legacy): Handle varying and undefined sources
18131 as a legacy copy since they can be easily copied.
18132 (irange::irange_set): Set m_kind.
18133 (irange::irange_set_anti_range): Same.
18134 (irange::set): Rename normalize_min_max to normalize_kind.
18135 (irange::verify_range): Adjust for multi-ranges having the
18137 (irange::irange_union): Set m_kind.
18138 (irange::irange_intersect): Same.
18139 (irange::invert): Same.
18140 * value-range.h (irange::kind): Always return m_kind.
18141 (irange::varying_p): Rename to...
18142 (irange::varying_comptaible_p): ...this.
18143 (irange::undefined_p): Only look at m_kind.
18144 (irange::irange): Always set VR_UNDEFINED if applicable.
18145 (irange::set_undefined): Always set VR_UNDEFINED.
18146 (irange::set_varying): Always set m_kind to VR_VARYING.
18147 (irange::normalize_min_max): Rename to...
18148 (irange::normalize_kind): ...this.
18150 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
18152 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::set_ssa_range_info):
18153 Adjust for constant_p including varying_p.
18154 * tree-vrp.c (vrp_prop::finalize): Same.
18155 (determine_value_range): Same.
18156 * vr-values.c (vr_values::range_of_expr): Same.
18157 * value-range.cc (irange::symbolic_p): Do not check varying_p.
18158 (irange::constant_p): Same.
18160 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
18162 * value-range.cc (irange::legacy_lower_bound): Replace
18163 !undefined_p check with num_ranges > 0.
18164 (irange::legacy_upper_bound): Same.
18165 * value-range.h (irange::type): Same.
18166 (irange::lower_bound): Same.
18167 (irange::upper_bound): Same.
18169 2021-04-26 Richard Biener <rguenther@suse.de>
18171 PR tree-optimization/99956
18172 * gimple-loop-interchange.cc (compute_access_stride):
18173 Try instantiating the access in a shallower loop nest
18174 if instantiating failed.
18175 (compute_access_strides): Pass adjustable loop_nest
18176 to compute_access_stride.
18178 2021-04-26 Christophe Lyon <christophe.lyon@linaro.org>
18180 * doc/sourcebuild.texi (arm_cmse_hw): Document.
18182 2021-04-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18184 * config/aarch64/iterators.md (vwcore): Handle V4BF, V8BF.
18186 2021-04-26 Thomas Schwinge <thomas@codesourcery.com>
18187 Nathan Sidwell <nathan@codesourcery.com>
18188 Tom de Vries <vries@codesourcery.com>
18189 Julian Brown <julian@codesourcery.com>
18190 Kwok Cheung Yeung <kcy@codesourcery.com>
18192 * omp-offload.c (oacc_validate_dims): Implement
18193 '-Wopenacc-parallelism'.
18194 * doc/invoke.texi (-Wopenacc-parallelism): Document.
18196 2021-04-26 Richard Biener <rguenther@suse.de>
18198 * tree-cfg.h (gimplify_build1): Remove.
18199 (gimplify_build2): Likewise.
18200 (gimplify_build3): Likewise.
18201 * tree-cfg.c (gimplify_build1): Move to tree-vect-generic.c.
18202 (gimplify_build2): Likewise.
18203 (gimplify_build3): Likewise.
18204 * tree-vect-generic.c (gimplify_build1): Move from tree-cfg.c.
18206 (gimplify_build2): Likewise.
18207 (gimplify_build3): Likewise.
18208 (tree_vec_extract): Use resimplify with following SSA edges.
18209 (expand_vector_parallel): Avoid passing NULL size/bitpos
18210 to tree_vec_extract.
18211 * expr.c (store_constructor): Deal with zero-element CTORs.
18212 * match.pd (bit_field_ref <vector CTOR>): Make sure to
18213 produce vector constants when possible.
18215 2021-04-26 Richard Biener <rguenther@suse.de>
18217 * tree-complex.c: Include gimple-fold.h.
18218 (expand_complex_addition): Use gimple_build.
18219 (expand_complex_multiplication_components): Likewise.
18220 (expand_complex_multiplication): Likewise.
18221 (expand_complex_div_straight): Likewise.
18222 (expand_complex_div_wide): Likewise.
18223 (expand_complex_division): Likewise.
18224 (expand_complex_conjugate): Likewise.
18225 (expand_complex_comparison): Likewise.
18227 2021-04-26 Richard Biener <rguenther@suse.de>
18229 * tree-ssa-phiopt.c (two_value_replacement): Remove use
18230 of legacy gimplify_buildN API.
18232 2021-04-26 Richard Biener <rguenther@suse.de>
18234 PR tree-optimization/99473
18235 * tree-ssa-phiopt.c (cond_store_replacement): Handle all
18238 2021-04-26 Richard Biener <rguenther@suse.de>
18240 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_builtin):
18241 Use replace_call_with_value.
18243 2021-04-26 Richard Biener <rguenther@suse.de>
18245 * tree-ssa-propagate.h (valid_gimple_rhs_p): Remove.
18246 (update_gimple_call): Likewise.
18247 (update_call_from_tree): Likewise.
18248 * tree-ssa-propagate.c (valid_gimple_rhs_p): Remove.
18249 (valid_gimple_call_p): Likewise.
18250 (move_ssa_defining_stmt_for_defs): Likewise.
18251 (finish_update_gimple_call): Likewise.
18252 (update_gimple_call): Likewise.
18253 (update_call_from_tree): Likewise.
18254 (propagate_tree_value_into_stmt): Use replace_call_with_value.
18255 * gimple-fold.h (update_gimple_call): Declare.
18256 * gimple-fold.c (valid_gimple_rhs_p): Move here from
18257 tree-ssa-propagate.c.
18258 (update_gimple_call): Likewise.
18259 (valid_gimple_call_p): Likewise.
18260 (finish_update_gimple_call): Likewise, and simplify.
18261 (gimplify_and_update_call_from_tree): Implement
18262 update_call_from_tree functionality, avoid excessive
18263 push/pop_gimplify_context.
18264 (gimple_fold_builtin): Use only gimplify_and_update_call_from_tree.
18265 (gimple_fold_call): Likewise.
18266 * gimple-ssa-sprintf.c (try_substitute_return_value): Likewise.
18267 * tree-ssa-ccp.c (ccp_folder::fold_stmt): Likewise.
18268 (pass_fold_builtins::execute): Likewise.
18269 (optimize_stack_restore): Use replace_call_with_value.
18270 * tree-cfg.c (fold_loop_internal_call): Likewise.
18271 * tree-ssa-dce.c (maybe_optimize_arith_overflow): Use
18272 only gimplify_and_update_call_from_tree.
18273 * tree-ssa-strlen.c (handle_builtin_strlen): Likewise.
18274 (handle_builtin_strchr): Likewise.
18275 * tsan.c: Include gimple-fold.h instead of tree-ssa-propagate.h.
18277 2021-04-26 Jakub Jelinek <jakub@redhat.com>
18280 * vmsdbgout.c (ASM_OUTPUT_DEBUG_STRING, vmsdbgout_begin_block,
18281 vmsdbgout_end_block, lookup_filename, vmsdbgout_source_line): Remove
18284 2021-04-25 liuhongt <hongtao.liu@intel.com>
18287 * config/i386/i386-builtin.def (BDESC): Change the icode of
18288 the following builtins to CODE_FOR_nothing.
18289 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
18290 IX86_BUILTIN_PCMPEQB128, IX86_BUILTIN_PCMPEQW128,
18291 IX86_BUILTIN_PCMPEQD128, IX86_BUILTIN_PCMPEQQ,
18292 IX86_BUILTIN_PCMPEQB256, IX86_BUILTIN_PCMPEQW256,
18293 IX86_BUILTIN_PCMPEQD256, IX86_BUILTIN_PCMPEQQ256,
18294 IX86_BUILTIN_PCMPGTB128, IX86_BUILTIN_PCMPGTW128,
18295 IX86_BUILTIN_PCMPGTD128, IX86_BUILTIN_PCMPGTQ,
18296 IX86_BUILTIN_PCMPGTB256, IX86_BUILTIN_PCMPGTW256,
18297 IX86_BUILTIN_PCMPGTD256, IX86_BUILTIN_PCMPGTQ256.
18298 * config/i386/sse.md (avx2_eq<mode>3): Deleted.
18299 (sse2_eq<mode>3): Ditto.
18300 (sse4_1_eqv2di3): Ditto.
18301 (sse2_gt<mode>3): Rename to ..
18302 (*sse2_gt<mode>3): .. this.
18304 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
18307 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
18310 * config/darwin.c (darwin_binds_local_p): Assume that any
18311 public symbol might be interposed for PIC code. Update function
18312 header comment to reflect current Darwin capability.
18314 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
18317 * config/darwin.c (darwin_binds_local_p): Assume that any
18318 public symbol might be interposed for PIC code. Update function
18319 header comment to reflect current Darwin capability.
18321 2021-04-24 Richard Sandiford <richard.sandiford@arm.com>
18323 * doc/sourcebuild.texi: Document no-opts and any-opts target
18326 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
18328 * config/bpf/bpf.h (ASM_OUTPUT_ALIGNED_BSS): Use .type and .lcomm.
18330 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
18332 * config/bpf/bpf.h (FUNCTION_BOUNDARY): Set to 64.
18334 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
18337 * config/i386/i386-options.c (ix86_option_override_internal):
18338 Error out when -m96bit-long-double is used with 64bit targets.
18339 * config/i386/i386.md (*pushxf_rounded): Remove pattern.
18341 2021-04-23 Martin Liska <mliska@suse.cz>
18343 * lto-wrapper.c: Remove FIXME about usage of
18344 hardware_concurrency. The function is not on par with
18347 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
18350 * config/i386/sync.md (FILD_ATOMIC/FIST_ATOMIC FP load peephole2):
18351 Copy operand 3 to operand 4. Use sse_reg_operand
18352 as operand 3 predicate.
18353 (FILD_ATOMIC/FIST_ATOMIC FP load peephole2 with mem blockage): Ditto.
18354 (LDX_ATOMIC/STX_ATOMIC FP load peephole2): Ditto.
18355 (LDX_ATOMIC/LDX_ATOMIC FP load peephole2 with mem blockage): Ditto.
18356 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2):
18357 Copy operand 1 to operand 0.
18358 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2 with mem blockage): Ditto.
18359 (LDX_ATOMIC/STX_ATOMIC FP store peephole2): Ditto.
18360 (LDX_ATOMIC/LDX_ATOMIC FP store peephole2 with mem blockage): Ditto.
18362 2021-04-23 Alex Coplan <alex.coplan@arm.com>
18364 PR rtl-optimization/100230
18365 * early-remat.c (early_remat::sort_candidates): Use delete[]
18366 instead of delete for array allocated with new[].
18368 2021-04-23 Richard Biener <rguenther@suse.de>
18370 * genmatch.c (lower_cond): Remove VEC_COND_EXPR special-casing.
18371 (capture_info::capture_info): Likewise.
18372 (capture_info::walk_match): Likewise.
18373 (expr::gen_transform): Likewise.
18374 (dt_simplify::gen_1): Likewise.
18375 * gimple-match-head.c (maybe_resimplify_conditional_op):
18376 Remove VEC_COND_EXPR special-casing.
18377 (gimple_simplify): Likewise.
18378 * gimple.c (gimple_could_trap_p_1): Adjust.
18379 * tree-ssa-pre.c (compute_avail): Allow VEC_COND_EXPR
18380 to participate in PRE.
18382 2021-04-23 Richard Biener <rguenther@suse.de>
18384 * cfganal.c (connect_infinite_loops_to_exit): First call
18385 add_noreturn_fake_exit_edges.
18386 * ipa-sra.c (process_scan_results): Do not call the now redundant
18387 add_noreturn_fake_exit_edges.
18388 * predict.c (tree_estimate_probability): Likewise.
18389 (rebuild_frequencies): Likewise.
18390 * store-motion.c (one_store_motion_pass): Likewise.
18392 2021-04-23 Richard Biener <rguenther@suse.de>
18394 PR tree-optimization/100222
18395 * predict.c (pass_profile::execute): Remove redundant call to
18396 mark_irreducible_loops.
18397 (report_predictor_hitrates): Likewise.
18399 2021-04-23 Richard Biener <rguenther@suse.de>
18401 * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Avoid
18402 valid_gimple_rhs_p by instead gimplifying to one.
18404 2021-04-23 Richard Biener <rguenther@suse.de>
18406 PR tree-optimization/99971
18407 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
18408 Always use TBAA for loads.
18410 2021-04-23 liuhongt <hongtao.liu@intel.com>
18413 * config/i386/i386-options.c (ix86_option_override_internal):
18414 Clear MASK_AVX256_SPLIT_UNALIGNED_LOAD/STORE in x_target_flags
18415 when X86_TUNE_AVX256_UNALIGNED_LOAD/STORE_OPTIMAL is enabled
18416 by target attribute.
18418 2021-04-23 David Edelsohn <dje.gcc@gmail.com>
18420 * config/rs6000/aix71.h (PREFERRED_DEBUGGING_TYPE): Change to
18422 * config/rs6000/aix72.h (PREFERRED_DEBUGGING_TYPE): Same.
18424 2021-04-22 David Edelsohn <dje.gcc@gmail.com>
18426 * config.gcc (powerpc-ibm-aix6.*): Remove.
18427 * config/rs6000/aix61.h: Delete.
18429 2021-04-22 Martin Liska <mliska@suse.cz>
18431 PR testsuite/100159
18432 PR testsuite/100192
18433 * builtins.c (expand_builtin): Fix typos and missing comments.
18434 * dwarf2out.c (gen_subprogram_die): Likewise.
18435 (gen_struct_or_union_type_die): Likewise.
18437 2021-04-22 Uroš Bizjak <ubizjak@gmail.com>
18440 * config/i386/i386-expand.c (ix86_expand_convert_uns_sidf_sse):
18441 Remove the sign with FE_DOWNWARD, where x - x = -0.0.
18443 2021-04-21 Iain Sandoe <iain@sandoe.co.uk>
18445 * config/i386/darwin.h (TARGET_64BIT): Remove definition
18446 based on TARGET_ISA_64BIT.
18447 (TARGET_64BIT_P): Remove definition based on
18448 TARGET_ISA_64BIT_P().
18450 2021-04-21 Martin Liska <mliska@suse.cz>
18453 2021-04-21 Martin Liska <mliska@suse.cz>
18455 * lto-wrapper.c (cpuset_popcount): Remove.
18456 (init_num_threads): Remove and use hardware_concurrency.
18458 2021-04-21 Martin Liska <mliska@suse.cz>
18461 * main.c (main): Call toplev::finalize in CHECKING_P mode.
18462 * ipa-modref.c (ipa_modref_c_finalize): summaries are NULL
18463 when incremental LTO linking happens.
18465 2021-04-21 Martin Liska <mliska@suse.cz>
18467 * lto-wrapper.c (run_gcc): When -flto=jobserver is used, but the
18468 makeserver cannot be detected, then use -flto=N fallback.
18470 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
18472 * acinclude.m4 (gcc_AC_INITFINI_ARRAY): When cross-compiling,
18473 default to yes for aarch64-linux-gnu.
18474 * configure: Regenerate.
18476 2021-04-21 Martin Liska <mliska@suse.cz>
18478 * lto-wrapper.c (cpuset_popcount): Remove.
18479 (init_num_threads): Remove and use hardware_concurrency.
18481 2021-04-21 Martin Liska <mliska@suse.cz>
18483 * config/i386/i386.c: Remove superfluous || TARGET_MACHO
18484 which remains to be '(... || 0)' and clang complains about it.
18485 * dwarf2out.c (AT_vms_delta): Declare conditionally.
18486 (add_AT_vms_delta): Likewise.
18487 * tree.c (fld_simplified_type): Use rather more common pattern
18488 for disabling of something (#if 0).
18489 (get_tree_code_name): Likewise.
18490 (verify_type_variant): Likewise.
18492 2021-04-21 Martin Liska <mliska@suse.cz>
18494 * config/i386/i386-expand.c (decide_alignment): Use newly named
18495 macro TARGET_CPU_P.
18496 * config/i386/i386.c (ix86_decompose_address): Likewise.
18497 (ix86_address_cost): Likewise.
18498 (ix86_lea_outperforms): Likewise.
18499 (ix86_avoid_lea_for_addr): Likewise.
18500 (ix86_add_stmt_cost): Likewise.
18501 * config/i386/i386.h (TARGET_*): Remove.
18502 (TARGET_CPU_P): New macro.
18503 * config/i386/i386.md: Use newly named macro TARGET_CPU_P.
18504 * config/i386/x86-tune-sched-atom.c (do_reorder_for_imul): Likewise.
18505 (swap_top_of_ready_list): Likewise.
18506 (ix86_atom_sched_reorder): Likewise.
18507 * config/i386/x86-tune-sched-bd.c (ix86_bd_has_dispatch): Likewise.
18508 * config/i386/x86-tune-sched.c (ix86_adjust_cost): Likewise.
18510 2021-04-21 Martin Liska <mliska@suse.cz>
18512 * config/i386/i386-options.c (TARGET_EXPLICIT_NO_SAHF_P):
18514 (SET_TARGET_NO_SAHF): Likewise.
18515 (TARGET_EXPLICIT_PREFETCH_SSE_P): Likewise.
18516 (SET_TARGET_PREFETCH_SSE): Likewise.
18517 (TARGET_EXPLICIT_NO_TUNE_P): Likewise.
18518 (SET_TARGET_NO_TUNE): Likewise.
18519 (TARGET_EXPLICIT_NO_80387_P): Likewise.
18520 (SET_TARGET_NO_80387): Likewise.
18522 * config/i386/i386.h (TARGET_*): Remove.
18523 * opth-gen.awk: Generate new used macros.
18525 2021-04-21 Martin Liska <mliska@suse.cz>
18527 * config/i386/i386.h (PTA_*): Remove.
18528 (enum pta_flag): New.
18529 (DEF_PTA): Generate PTA_* values from i386-isa.def.
18530 * config/i386/i386-isa.def: New file.
18532 2021-04-21 Alex Coplan <alex.coplan@arm.com>
18535 * config/aarch64/aarch64-bti-insert.c (aarch64_bti_j_insn_p): New.
18536 (rest_of_insert_bti): Avoid inserting duplicate bti j insns for
18537 jump table targets.
18539 2021-04-21 H.J. Lu <hjl.tools@gmail.com>
18541 * config.gcc: Install mwaitintrin.h for i[34567]86-*-* and
18542 x86_64-*-* targets.
18543 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_MWAIT_SET):
18545 (OPTION_MASK_ISA2_MWAIT_UNSET): Likewise.
18546 (ix86_handle_option): Handle -mmwait.
18547 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
18548 Replace OPTION_MASK_ISA_SSE3 with OPTION_MASK_ISA2_MWAIT on
18549 __builtin_ia32_monitor and __builtin_ia32_mwait.
18550 * config/i386/i386-options.c (isa2_opts): Add -mmwait.
18551 (ix86_valid_target_attribute_inner_p): Likewise.
18552 (ix86_option_override_internal): Enable mwait/monitor
18553 instructions for -msse3.
18554 * config/i386/i386.h (TARGET_MWAIT): New.
18555 (TARGET_MWAIT_P): Likewise.
18556 * config/i386/i386.opt: Add -mmwait.
18557 * config/i386/mwaitintrin.h: New file.
18558 * config/i386/pmmintrin.h: Include <mwaitintrin.h>.
18559 * config/i386/sse.md (sse3_mwait): Replace TARGET_SSE3 with
18561 (@sse3_monitor_<mode>): Likewise.
18562 * config/i386/x86gprintrin.h: Include <mwaitintrin.h>.
18563 * doc/extend.texi: Document mwait target attribute.
18564 * doc/invoke.texi: Document -mmwait.
18566 2021-04-21 Martin Liska <mliska@suse.cz>
18568 * config/i386/i386-options.c (DEF_ENUM): Remove it.
18569 * config/i386/i386-opts.h (DEF_ENUM): Likewise.
18570 * config/i386/stringop.def (DEF_ENUM): Likewise.
18572 2021-04-21 Martin Liska <mliska@suse.cz>
18574 * tree-cfg.c (gimple_verify_flow_info): Use qD instead
18575 of print_generic_expr.
18577 2021-04-21 Jakub Jelinek <jakub@redhat.com>
18579 PR rtl-optimization/100148
18580 * cprop.c (constprop_register): Use next_nondebug_insn instead of
18583 2021-04-21 Martin Liska <mliska@suse.cz>
18586 * cgraphunit.c (cgraph_node::analyze): Remove duplicate
18587 free_dominance_info calls.
18589 2021-04-21 Richard Biener <rguenther@suse.de>
18591 * gimple-fold.c (maybe_fold_reference): Remove is_lhs
18592 parameter (and assume it to be false).
18593 (fold_gimple_assign): Adjust, remove all callers of
18594 maybe_fold_reference calling it with is_lhs true.
18595 (gimple_fold_call): Likewise.
18596 (fold_stmt_1): Likewise.
18598 2021-04-21 Richard Biener <rguenther@suse.de>
18600 * fold-const.c (pedantic_non_lvalue_loc): Remove.
18601 (fold_binary_loc): Adjust.
18602 (fold_ternary_loc): Likewise.
18604 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
18606 PR middle-end/100130
18607 * varasm.c (get_block_for_decl): Make sure that any use of the
18608 retain attribute matches the section's retain flag.
18609 (switch_to_section): Check for retain mismatches even when
18610 changing sections, but do not warn if the given decl is the
18611 section's named.decl.
18612 (output_object_block): Pass the first decl in the block (if any)
18613 to switch_to_section.
18615 2021-04-20 H.J. Lu <hjl.tools@gmail.com>
18617 * config/i386/i386-c.c (ix86_target_macros_internal): Define
18618 __CRC32__ for -mcrc32.
18619 * config/i386/i386-options.c (ix86_option_override_internal):
18620 Enable crc32 instruction for -msse4.2.
18621 * config/i386/i386.md (sse4_2_crc32<mode>): Remove TARGET_SSE4_2
18623 (sse4_2_crc32di): Likewise.
18624 * config/i386/ia32intrin.h: Use crc32 target option for CRC32
18627 2021-04-20 Segher Boessenkool <segher@kernel.crashing.org>
18630 * config/rs6000/rs6000.c (rs6000_machine_from_flags): Do not consider
18633 2021-04-20 Martin Liska <mliska@suse.cz>
18635 * doc/invoke.texi: Fix typo.
18636 * params.opt: Likewise.
18638 2021-04-20 Martin Liska <mliska@suse.cz>
18640 * doc/invoke.texi: Document new param.
18642 2021-04-19 Andrew MacLeod <amacleod@redhat.com>
18644 PR tree-optimization/100081
18645 * gimple-range-cache.h (ranger_cache): Inherit from gori_compute
18646 rather than gori_compute_cache.
18647 * gimple-range-gori.cc (is_gimple_logical_p): Move to top of file.
18648 (range_def_chain::m_logical_depth): New member.
18649 (range_def_chain::range_def_chain): Initialize m_logical_depth.
18650 (range_def_chain::get_def_chain): Don't build defchains through more
18651 than LOGICAL_LIMIT logical expressions.
18652 * params.opt (param_ranger_logical_depth): New.
18654 2021-04-19 Richard Earnshaw <rearnsha@arm.com>
18657 * config/arm/arm.c (arm_configure_build_target): Do not strip
18658 extended FPU/SIMD feature bits from the target ISA when -mfpu
18659 is specified (partial revert of r11-8168).
18661 2021-04-19 Thomas Schwinge <thomas@codesourcery.com>
18663 * params.opt (-param=openacc-kernels=): Add.
18664 * omp-oacc-kernels-decompose.cc
18665 (pass_omp_oacc_kernels_decompose::gate): Use it.
18666 * doc/invoke.texi (-fopenacc-kernels=@var{mode}): Move...
18667 (--param): ... here, 'openacc-kernels'.
18669 2021-04-19 Martin Liska <mliska@suse.cz>
18672 * gengtype.c (finish_root_table): Align function arguments
18673 in between declaration and definition.
18675 2021-04-19 Eric Botcazou <ebotcazou@adacore.com>
18677 * config/i386/winnt.c (i386_pe_seh_cold_init): Properly deal with
18678 frames larger than the SEH maximum frame size.
18680 2021-04-18 Segher Boessenkool <segher@kernel.crashing.org>
18682 PR rtl-optimization/99927
18683 * combine.c (distribute_notes) [REG_UNUSED]: If the register already
18684 is dead, just drop it.
18686 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
18689 * config/i386/winnt-d.c (TARGET_D_TEMPLATES_ALWAYS_COMDAT): Define.
18690 * doc/tm.texi: Regenerate.
18691 * doc/tm.texi.in (D language and ABI): Add @hook for
18692 TARGET_D_TEMPLATES_ALWAYS_COMDAT.
18694 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
18696 * config/darwin-d.c (darwin_d_handle_target_object_format): New
18698 (darwin_d_register_target_info): New function.
18699 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18700 * config/dragonfly-d.c (dragonfly_d_handle_target_object_format): New
18702 (dragonfly_d_register_target_info): New function.
18703 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18704 * config/freebsd-d.c (freebsd_d_handle_target_object_format): New
18706 (freebsd_d_register_target_info): New function.
18707 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18708 * config/glibc-d.c (glibc_d_handle_target_object_format): New
18710 (glibc_d_register_target_info): New function.
18711 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18712 * config/i386/i386-d.c (ix86_d_handle_target_object_format): New
18714 (ix86_d_register_target_info): Add ix86_d_handle_target_object_format
18715 as handler for objectFormat key.
18716 * config/i386/winnt-d.c (winnt_d_handle_target_object_format): New
18718 (winnt_d_register_target_info): New function.
18719 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18720 * config/netbsd-d.c (netbsd_d_handle_target_object_format): New
18722 (netbsd_d_register_target_info): New function.
18723 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18724 * config/openbsd-d.c (openbsd_d_handle_target_object_format): New
18726 (openbsd_d_register_target_info): New function.
18727 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18728 * config/pa/pa-d.c (pa_d_handle_target_object_format): New function.
18729 (pa_d_register_target_info): Add pa_d_handle_target_object_format as
18730 handler for objectFormat key.
18731 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_object_format): New
18733 (rs6000_d_register_target_info): Add
18734 rs6000_d_handle_target_object_format as handler for objectFormat key.
18735 * config/sol2-d.c (solaris_d_handle_target_object_format): New
18737 (solaris_d_register_target_info): New function.
18738 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18740 2021-04-16 Jakub Jelinek <jakub@redhat.com>
18743 * config/aarch64/aarch64.c (aarch64_function_arg_alignment): Change
18744 abi_break argument from bool * to unsigned *, store there the pre-GCC 9
18746 (aarch64_layout_arg, aarch64_gimplify_va_arg_expr): Adjust callers.
18747 (aarch64_function_arg_regno_p): Likewise. Only emit -Wpsabi note if
18748 the old and new alignment after applying MIN/MAX to it is different.
18750 2021-04-16 Tamar Christina <tamar.christina@arm.com>
18753 * config/aarch64/aarch64-sve.md (@aarch64_sve_trn1_conv<mode>): New.
18754 * config/aarch64/aarch64.c (aarch64_expand_sve_const_pred_trn): Use new
18756 * config/aarch64/iterators.md (UNSPEC_TRN1_CONV): New.
18758 2021-04-16 Bill Schmidt <wschmidt@linux.ibm.com>
18760 * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions): Revise
18761 this section and its subsections.
18763 2021-04-16 Jakub Jelinek <jakub@redhat.com>
18766 * config/aarch64/aarch64.md (*neg_asr_si2_extr, *extrsi5_insn_di): New
18767 define_insn patterns.
18769 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
18771 PR rtl-optimization/98689
18772 * reg-notes.def (UNTYPED_CALL): New note.
18773 * combine.c (distribute_notes): Handle it.
18774 * emit-rtl.c (try_split): Likewise.
18775 * rtlanal.c (rtx_properties::try_to_add_insn): Likewise. Assume
18776 that calls with the note implicitly set all return value registers.
18777 * builtins.c (expand_builtin_apply): Add a REG_UNTYPED_CALL
18780 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
18782 PR rtl-optimization/99596
18783 * rtlanal.c (rtx_properties::try_to_add_insn): Don't add global
18784 register accesses for const calls. Assume that pure functions
18785 can only read from global registers. Ignore cases in which
18786 the stack pointer has been marked global.
18788 2021-04-16 Jakub Jelinek <jakub@redhat.com>
18791 * tree-vect-loop.c (vect_transform_loop): Don't remove just
18792 dead scalar .MASK_LOAD calls, but also dead .COND_* calls - replace
18793 them by their last argument.
18795 2021-04-15 Martin Liska <mliska@suse.cz>
18797 * doc/invoke.texi: Other params don't use it, remove it.
18799 2021-04-15 Richard Biener <rguenther@suse.de>
18801 * gimple-builder.h: Add deprecation note.
18803 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
18806 * attribs.h (restrict_type_identity_attributes_to): Declare.
18807 * attribs.c (restrict_type_identity_attributes_to): New function.
18809 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
18812 * attribs.h (affects_type_identity_attributes): Declare.
18813 * attribs.c (remove_attributes_matching): New function.
18814 (affects_type_identity_attributes): Likewise.
18816 2021-04-15 Jakub Jelinek <jakub@redhat.com>
18819 * config/aarch64/aarch64.md (*<LOGICAL:optab>_<SHIFT:optab><mode>3):
18820 Add combine splitters for *<LOGICAL:optab>_ashl<mode>3 with
18821 ZERO_EXTEND, SIGN_EXTEND or AND.
18823 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
18825 PR rtl-optimization/99929
18826 * rtl.h (same_vector_encodings_p): New function.
18827 * cse.c (exp_equiv_p): Check that CONST_VECTORs have the same encoding.
18828 * cselib.c (rtx_equal_for_cselib_1): Likewise.
18829 * jump.c (rtx_renumbered_equal_p): Likewise.
18830 * lra-constraints.c (operands_match_p): Likewise.
18831 * reload.c (operands_match_p): Likewise.
18832 * rtl.c (rtx_equal_p_cb, rtx_equal_p): Likewise.
18834 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
18836 * print-rtl.c (rtx_writer::print_rtx_operand_codes_E_and_V): Print
18837 more information about variable-length CONST_VECTORs.
18839 2021-04-14 Vladimir N. Makarov <vmakarov@redhat.com>
18841 PR rtl-optimization/100066
18842 * lra-constraints.c (split_reg): Check paradoxical_subreg_p for
18843 ordered modes when choosing splitting mode for hard reg.
18845 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
18848 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector_sel):
18850 (aarch64_expand_sve_const_vector): Use it for nelts_per_pattern==2.
18852 2021-04-14 Andreas Krebbel <krebbel@linux.ibm.com>
18854 * config/s390/s390-builtins.def (O_M5, O_M12, ...): Add new macros
18855 for mask operand types.
18856 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
18857 (s390_vec_permi_dbl, s390_vpdi): Use the M5 type for the immediate
18859 (s390_vec_msum_u128, s390_vmslg): Use the M12 type for the
18861 * config/s390/s390.c (s390_const_operand_ok): Check the new
18862 operand types and generate a list of valid values.
18864 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
18866 * doc/tm.texi: Regenerate.
18867 * doc/tm.texi.in (D language and ABI): Add @hook for
18868 TARGET_D_REGISTER_OS_TARGET_INFO.
18870 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
18872 * config/aarch64/aarch64-d.c (aarch64_d_handle_target_float_abi): New
18874 (aarch64_d_register_target_info): New function.
18875 * config/aarch64/aarch64-protos.h (aarch64_d_register_target_info):
18877 * config/aarch64/aarch64.h (TARGET_D_REGISTER_CPU_TARGET_INFO):
18879 * config/arm/arm-d.c (arm_d_handle_target_float_abi): New function.
18880 (arm_d_register_target_info): New function.
18881 * config/arm/arm-protos.h (arm_d_register_target_info): Declare.
18882 * config/arm/arm.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18883 * config/i386/i386-d.c (ix86_d_handle_target_float_abi): New function.
18884 (ix86_d_register_target_info): New function.
18885 * config/i386/i386-protos.h (ix86_d_register_target_info): Declare.
18886 * config/i386/i386.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18887 * config/mips/mips-d.c (mips_d_handle_target_float_abi): New function.
18888 (mips_d_register_target_info): New function.
18889 * config/mips/mips-protos.h (mips_d_register_target_info): Declare.
18890 * config/mips/mips.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18891 * config/pa/pa-d.c (pa_d_handle_target_float_abi): New function.
18892 (pa_d_register_target_info): New function.
18893 * config/pa/pa-protos.h (pa_d_register_target_info): Declare.
18894 * config/pa/pa.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18895 * config/riscv/riscv-d.c (riscv_d_handle_target_float_abi): New
18897 (riscv_d_register_target_info): New function.
18898 * config/riscv/riscv-protos.h (riscv_d_register_target_info): Declare.
18899 * config/riscv/riscv.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18900 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_float_abi): New
18902 (rs6000_d_register_target_info): New function.
18903 * config/rs6000/rs6000-protos.h (rs6000_d_register_target_info):
18905 * config/rs6000/rs6000.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18906 * config/s390/s390-d.c (s390_d_handle_target_float_abi): New function.
18907 (s390_d_register_target_info): New function.
18908 * config/s390/s390-protos.h (s390_d_register_target_info): Declare.
18909 * config/s390/s390.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18910 * config/sparc/sparc-d.c (sparc_d_handle_target_float_abi): New
18912 (sparc_d_register_target_info): New function.
18913 * config/sparc/sparc-protos.h (sparc_d_register_target_info): Declare.
18914 * config/sparc/sparc.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18915 * doc/tm.texi: Regenerate.
18916 * doc/tm.texi.in (D language and ABI): Add @hook for
18917 TARGET_D_REGISTER_CPU_TARGET_INFO.
18919 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
18921 * config/i386/i386-d.c (ix86_d_has_stdcall_convention): New function.
18922 * config/i386/i386-protos.h (ix86_d_has_stdcall_convention): Declare.
18923 * config/i386/i386.h (TARGET_D_HAS_STDCALL_CONVENTION): Define.
18924 * doc/tm.texi: Regenerate.
18925 * doc/tm.texi.in (D language and ABI): Add @hook for
18926 TARGET_D_HAS_STDCALL_CONVENTION.
18928 2021-04-14 Richard Biener <rguenther@suse.de>
18930 * tree-cfg.c (verify_gimple_assign_ternary): Verify that
18931 VEC_COND_EXPRs have a gimple_val condition.
18932 * tree-ssa-propagate.c (valid_gimple_rhs_p): VEC_COND_EXPR
18933 can no longer have a GENERIC condition.
18935 2021-04-14 Richard Earnshaw <rearnsha@arm.com>
18938 * config/arm/arm.c (arm_configure_build_target): Strip isa_all_fpbits
18939 from the isa_delta when -mfpu has been used.
18940 (arm_options_perform_arch_sanity_checks): It's the architecture that
18941 lacks an FPU not the processor.
18943 2021-04-13 Richard Biener <rguenther@suse.de>
18945 PR tree-optimization/100053
18946 * tree-ssa-sccvn.c (vn_nary_op_get_predicated_value): Do
18947 not use optimistic dominance queries for backedges to validate
18949 (dominated_by_p_w_unex): Add parameter to ignore executable
18950 state on backedges.
18951 (rpo_elim::eliminate_avail): Adjust.
18953 2021-04-13 Jakub Jelinek <jakub@redhat.com>
18956 * config/aarch64/aarch64.md (*aarch64_bfxil<mode>_extr,
18957 *aarch64_bfxilsi_extrdi): New define_insn patterns.
18959 2021-04-13 Jakub Jelinek <jakub@redhat.com>
18962 * simplify-rtx.c (simplify_immed_subreg): For MODE_COMPOSITE_P
18963 outermode, return NULL if the result doesn't encode back to the
18964 original byte sequence.
18965 (simplify_gen_subreg): Don't create SUBREGs from constants to
18966 MODE_COMPOSITE_P outermode.
18968 2021-04-12 Jakub Jelinek <jakub@redhat.com>
18970 PR rtl-optimization/99905
18971 * combine.c (expand_compound_operation): If pos + len > modewidth,
18972 perform the right shift by pos in inner_mode and then convert to mode,
18973 instead of trying to simplify a shift of rtx with inner_mode by pos
18974 as if it was a shift in mode.
18976 2021-04-12 Jakub Jelinek <jakub@redhat.com>
18979 * combine.c (simplify_and_const_int_1): Don't optimize varop
18980 away if it has side-effects.
18982 2021-04-12 Martin Liska <mliska@suse.cz>
18984 * doc/extend.texi: Escape @smallexample content.
18986 2021-04-12 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
18988 * config/s390/s390.md ("*movdi_31", "*movdi_64"): Add
18989 alternative in order to load a DFP zero.
18991 2021-04-12 Martin Liska <mliska@suse.cz>
18993 * doc/extend.texi: Be more precise in documentation
18994 of symver attribute.
18996 2021-04-12 Martin Liska <mliska@suse.cz>
18999 * gimplify.c (gimplify_expr): Right now, we unpoison all
19000 variables before a goto <dest>. We should not do it if we are
19003 2021-04-12 Cui,Lili <lili.cui@intel.com>
19005 * common/config/i386/cpuinfo.h (get_intel_cpu): Handle
19007 * common/config/i386/i386-common.c (processor_names): Add
19009 (processor_alias_table): Add rocketlake.
19010 * common/config/i386/i386-cpuinfo.h (processor_subtypes): Add
19011 INTEL_COREI7_ROCKETLAKE.
19012 * config.gcc: Add -march=rocketlake.
19013 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19015 * config/i386/i386-options.c (m_ROCKETLAKE) : Define.
19016 (processor_cost_table): Add rocketlake cost.
19017 * config/i386/i386.h (ix86_size_cost) : Define
19019 (processor_type) : Add PROCESSOR_ROCKETLAKE.
19020 (PTA_ROCKETLAKE): Ditto.
19021 * doc/extend.texi: Add rocketlake.
19022 * doc/invoke.texi: Add rocketlake.
19024 2021-04-12 Cui,Lili <lili.cui@intel.com>
19026 * config/i386/i386.h (PTA_ALDERLAKE): Change alderlake ISA list.
19027 * config/i386/i386-options.c (m_CORE_AVX2): Add m_ALDERLAKE.
19028 * common/config/i386/cpuinfo.h (get_intel_cpu): Add AlderLake model.
19029 * doc/invoke.texi: Change alderlake ISA list.
19031 2021-04-11 Hafiz Abid Qadeer <abidh@codesourcery.com>
19033 PR middle-end/98088
19034 * omp-expand.c (expand_oacc_collapse_init): Update condition in
19037 2021-04-10 H.J. Lu <hjl.tools@gmail.com>
19040 * config/i386/serializeintrin.h (_serialize): Defined as macro.
19042 2021-04-10 Jakub Jelinek <jakub@redhat.com>
19045 * expr.c (expand_expr_addr_expr_1): Test is_global_var rather than
19046 just TREE_STATIC on COMPOUND_LITERAL_EXPR_DECLs.
19048 2021-04-10 Jakub Jelinek <jakub@redhat.com>
19050 PR middle-end/99989
19051 * gimple-ssa-warn-alloca.c
19052 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit to
19053 0 with integer precision unconditionally.
19055 2021-04-10 Jakub Jelinek <jakub@redhat.com>
19057 PR rtl-optimization/98601
19058 * rtlanal.c (rtx_addr_can_trap_p_1): Allow in assert unknown size
19059 not just for BLKmode, but also for VOIDmode. For STRICT_ALIGNMENT
19060 unaligned_mems handle VOIDmode like BLKmode.
19062 2021-04-10 Jan Hubicka <hubicka@ucw.cz>
19065 * tree.c (free_lang_data_in_decl): Do not release body of
19066 declare_variant_alt.
19068 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
19070 * config/aarch64/aarch64.c (aarch64_option_restore): If the
19071 architecture was specified explicitly and the tuning wasn't,
19072 tune for the architecture rather than the configured default CPU.
19074 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
19076 * config/aarch64/aarch64.md (tlsdesc_small_sve_<mode>): Use X30
19077 as the temporary register.
19079 2021-04-09 Martin Liska <mliska@suse.cz>
19081 * doc/extend.texi: Move non-target attributes on the top level.
19083 2021-04-09 Martin Liska <mliska@suse.cz>
19085 * doc/invoke.texi: Document minimum and maximum value of the
19086 argument for both supported compression algorithms.
19088 2021-04-08 David Edelsohn <dje.gcc@gmail.com>
19090 * config/rs6000/rs6000.c (rs6000_xcoff_select_section): Select
19091 TLS BSS before TLS data.
19092 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .comm.
19094 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
19096 * doc/sourcebuild.texi (stdint_types_mbig_endian): Document.
19098 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
19100 * match.pd: Extend vec_cond folds to handle shifts.
19102 2021-04-08 Maciej W. Rozycki <macro@orcam.me.uk>
19104 * config/vax/vax.md: Fix comment for `*bit<mode>' pattern's
19107 2021-04-08 Alex Coplan <alex.coplan@arm.com>
19110 * config/arm/iterators.md (MVE_vecs): New.
19111 (V_elem): Also handle V2DF.
19112 * config/arm/mve.md (*mve_mov<mode>): Rename to ...
19113 (*mve_vdup<mode>): ... this. Remove second alternative since
19114 vec_duplicate of const_int is not canonical RTL, and we don't
19115 want to match symbol_refs.
19116 (*mve_vec_duplicate<mode>): Delete (pattern is redundant).
19118 2021-04-08 Xionghu Luo <luoxhu@linux.ibm.com>
19120 * fold-const.c (fold_single_bit_test): Fix typo.
19121 * print-rtl.c (print_rtx_insn_vec): Call print_rtl_single
19124 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
19126 PR tree-optimization/97513
19127 * tree-vect-slp.c (vect_add_slp_permutation): New function,
19129 (vectorizable_slp_permutation): ...here. Detect cases in which
19130 all VEC_PERM_EXPRs are guaranteed to have the same stepped
19131 permute vector and only generate one permute vector for that case.
19132 Extend that case to handle variable-length vectors.
19134 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
19136 PR tree-optimization/99873
19137 * tree-vect-slp.c (vect_slp_prefer_store_lanes_p): New function.
19138 (vect_build_slp_instance): Don't split store groups that could
19139 use IFN_STORE_LANES.
19141 2021-04-07 Jakub Jelinek <jakub@redhat.com>
19144 * varasm.c (output_constant_pool_contents): Don't strip name encoding
19145 from XSTR (desc->sym, 0) or from label before passing those to
19148 2021-04-07 Richard Biener <rguenther@suse.de>
19150 PR tree-optimization/99954
19151 * tree-loop-distribution.c: Include tree-affine.h.
19152 (generate_memcpy_builtin): Try using tree-affine to prove
19154 (loop_distribution::classify_builtin_ldst): Always classify
19157 2021-04-07 Richard Biener <rguenther@suse.de>
19159 PR tree-optimization/99947
19160 * tree-vect-loop.c (vectorizable_induction): Pre-allocate
19161 steps vector to avoid pushing elements from the reallocated
19164 2021-04-07 Richard Biener <rguenther@suse.de>
19166 * tree-ssa-sccvn.h (print_vn_reference_ops): Declare.
19167 * tree-ssa-pre.c (print_pre_expr): Factor out VN reference operand
19169 * tree-ssa-sccvn.c (print_vn_reference_ops): ... into this new
19171 (debug_vn_reference_ops): New.
19173 2021-04-07 Bin Cheng <bin.cheng@linux.alibaba.com>
19175 PR tree-optimization/98736
19176 * tree-loop-distribution.c
19177 * (loop_distribution::bb_top_order_init):
19178 Compute RPO with programing order preserved by calling function
19179 rev_post_order_and_mark_dfs_back_seme.
19181 2021-04-06 Vladimir N. Makarov <vmakarov@redhat.com>
19184 * lra-constraints.c (split_reg): Don't check paradoxical_subreg_p.
19185 * lra-lives.c (clear_sparseset_regnos, regnos_in_sparseset_p): New
19187 (process_bb_lives): Don't update biggest mode of hard reg for
19188 implicit in multi-register group. Use the new functions for
19189 updating dead_set and unused_set by register notes.
19191 2021-04-06 Xianmiao Qu <xianmiao_qu@c-sky.com>
19193 * config/csky/csky_pipeline_ck802.md : Use insn reservation name
19196 2021-04-06 H.J. Lu <hjl.tools@gmail.com>
19198 * config/i386/x86-tune-costs.h (skylake_memcpy): Updated.
19199 (skylake_memset): Likewise.
19200 (skylake_cost): Change CLEAR_RATIO to 17.
19201 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
19202 Replace m_CANNONLAKE, m_ICELAKE_CLIENT, m_ICELAKE_SERVER,
19203 m_TIGERLAKE and m_SAPPHIRERAPIDS with m_SKYLAKE and m_CORE_AVX512.
19205 2021-04-06 Richard Biener <rguenther@suse.de>
19207 PR tree-optimization/99880
19208 * tree-vect-loop.c (maybe_set_vectorized_backedge_value): Only
19209 set vectorized defs of relevant PHIs.
19211 2021-04-06 Richard Biener <rguenther@suse.de>
19213 PR tree-optimization/99924
19214 * tree-vect-slp.c (vect_bb_partition_graph_r): Do not mark
19215 nodes w/o scalar stmts as visited.
19217 2021-04-06 Alex Coplan <alex.coplan@arm.com>
19220 * config/arm/arm.c (arm_libcall_uses_aapcs_base): Also use base
19221 PCS for [su]fix_optab.
19223 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
19225 * config/darwin.c (machopic_legitimize_pic_address): Check
19226 that the current pic register is one of the hard reg set
19227 before setting liveness.
19229 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
19231 * config/darwin.c (machopic_legitimize_pic_address): Fix
19232 whitespace, remove unused code.
19234 2021-04-03 Jakub Jelinek <jakub@redhat.com>
19236 PR tree-optimization/99882
19237 * gimple-ssa-store-merging.c (bswap_view_convert): Handle val with
19240 2021-04-03 Jakub Jelinek <jakub@redhat.com>
19242 PR rtl-optimization/99863
19243 * dse.c (replace_read): Drop regs_live argument. Instead of
19244 regs_live, use store_insn->fixed_regs_live if non-NULL,
19245 otherwise punt if insns sequence clobbers or sets any hard
19248 2021-04-03 Jakub Jelinek <jakub@redhat.com>
19251 * targhooks.h (default_print_patchable_function_entry_1): Declare.
19252 * targhooks.c (default_print_patchable_function_entry_1): New function,
19253 copied from default_print_patchable_function_entry with an added flags
19255 (default_print_patchable_function_entry): Rewritten into a small
19256 wrapper around default_print_patchable_function_entry_1.
19257 * config/rs6000/rs6000.c (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY):
19259 (rs6000_print_patchable_function_entry): New function.
19261 2021-04-02 Eric Botcazou <ebotcazou@adacore.com>
19263 * doc/invoke.texi (fdelete-dead-exceptions): Minor tweak.
19265 2021-04-01 Jason Merrill <jason@redhat.com>
19268 * common.opt: Document v15 and v16.
19270 2021-04-01 Richard Biener <rguenther@suse.de>
19272 PR tree-optimization/99863
19273 * gimplify.c (gimplify_init_constructor): Recompute vector
19276 2021-04-01 Jakub Jelinek <jakub@redhat.com>
19278 * doc/extend.texi (symver attribute): Fix up syntax errors
19281 2021-04-01 Jakub Jelinek <jakub@redhat.com>
19283 PR tree-optimization/96573
19284 * gimple-ssa-store-merging.c (init_symbolic_number): Handle
19285 also pointer types.
19287 2021-04-01 Richard Biener <rguenther@suse.de>
19289 PR tree-optimization/99856
19290 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Promote
19291 precision to vector element precision.
19293 2021-04-01 Martin Jambor <mjambor@suse.cz>
19295 PR tree-optimization/97009
19296 * tree-sra.c (access_or_its_child_written): New function.
19297 (propagate_subaccesses_from_rhs): Use it instead of a simple grp_write
19300 2021-03-31 Jan Hubicka <hubicka@ucw.cz>
19303 * cif-code.def (USES_COMDAT_LOCAL): Make CIF_FINAL_NORMAL.
19305 2021-03-31 Pat Haugen <pthaugen@linux.ibm.com>
19308 * config/rs6000/altivec.md (xxspltiw_v4si, xxspltiw_v4sf_inst,
19309 xxspltidp_v2df_inst, xxsplti32dx_v4si_inst, xxsplti32dx_v4sf_inst,
19310 xxblend_<mode>, xxpermx_inst, xxeval): Mark prefixed.
19311 * config/rs6000/mma.md (mma_<vvi4i4i8>, mma_<avvi4i4i8>,
19312 mma_<vvi4i4i2>, mma_<avvi4i4i2>, mma_<vvi4i4>, mma_<avvi4i4>,
19313 mma_<pvi4i2>, mma_<apvi4i2>, mma_<vvi4i4i4>, mma_<avvi4i4i4>):
19315 * config/rs6000/rs6000.c (rs6000_final_prescan_insn): Adjust test.
19316 * config/rs6000/rs6000.md (define_attr "maybe_prefixed"): New.
19317 (define_attr "prefixed"): Update initializer.
19319 2021-03-31 Jakub Jelinek <jakub@redhat.com>
19322 * dwarf2out.c (debug_ranges_dwo_section): New variable.
19323 (DW_RANGES_IDX_SKELETON): Define.
19324 (struct dw_ranges): Add begin_entry and end_entry members.
19325 (DEBUG_DWO_RNGLISTS_SECTION): Define.
19326 (add_ranges_num): Adjust r initializer for addition of *_entry
19328 (add_ranges_by_labels): For -gsplit-dwarf and force_direct,
19329 set idx to DW_RANGES_IDX_SKELETON.
19330 (use_distinct_base_address_for_range): New function.
19331 (index_rnglists): Don't set r->idx if it is equal to
19332 DW_RANGES_IDX_SKELETON. Initialize r->begin_entry and
19333 r->end_entry for -gsplit-dwarf if those will be needed by
19335 (output_rnglists): Add DWO argument. If true, switch to
19336 debug_ranges_dwo_section rather than debug_ranges_section.
19337 Adjust l1/l2 label indexes. Only output the offset table when
19338 dwo is true and don't include in there the skeleton range
19339 entry if present. For -gsplit-dwarf, skip ranges that belong
19340 to the other rnglists section. Change return type from void
19341 to bool and return true if there are any range entries for
19342 the other section. For dwarf_split_debug_info use
19343 DW_RLE_startx_endx, DW_RLE_startx_length and DW_RLE_base_addressx
19344 entries instead of DW_RLE_start_end, DW_RLE_start_length and
19345 DW_RLE_base_address. Use use_distinct_base_address_for_range.
19346 (init_sections_and_labels): Initialize debug_ranges_dwo_section
19347 if -gsplit-dwarf and DWARF >= 5. Adjust ranges_section_label
19348 and range_base_label indexes.
19349 (dwarf2out_finish): Call index_rnglists earlier before finalizing
19350 .debug_addr. Never emit DW_AT_rnglists_base attribute. For
19351 -gsplit-dwarf and DWARF >= 5 call output_rnglists up to twice
19352 with different dwo arguments.
19353 (dwarf2out_c_finalize): Clear debug_ranges_dwo_section.
19355 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
19357 PR tree-optimization/98268
19358 * gimple-fold.c (maybe_canonicalize_mem_ref_addr): Call
19359 recompute_tree_invariant_for_addr_expr after successfully
19360 folding a TARGET_MEM_REF that occurs inside an ADDR_EXPR.
19362 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
19364 PR tree-optimization/99726
19365 * tree-data-ref.c (create_intersect_range_checks_index): Bail
19366 out if there is more than one access function SCEV for the loop
19369 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
19371 PR rtl-optimization/97141
19372 PR rtl-optimization/98726
19373 * emit-rtl.c (valid_for_const_vector_p): Return true for
19375 * rtx-vector-builder.h (rtx_vector_builder::step): Return a
19376 poly_wide_int instead of a wide_int.
19377 (rtx_vector_builder::apply_set): Take a poly_wide_int instead
19379 * rtx-vector-builder.c (rtx_vector_builder::apply_set): Likewise.
19380 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Return
19381 false for CONST_VECTORs that cannot be forced to memory.
19382 * config/aarch64/aarch64-simd.md (mov<mode>): If a CONST_VECTOR
19383 is too complex to force to memory, build it up from individual
19386 2021-03-31 Jan Hubicka <jh@suse.cz>
19389 * cgraph.c (cgraph_node::release_body): Fix overactive check.
19391 2021-03-31 Christophe Lyon <christophe.lyon@linaro.org>
19394 * config/arm/vec-common.md (mul<mode>3): Disable on iwMMXT, expect
19397 2021-03-31 H.J. Lu <hjl.tools@gmail.com>
19399 * config/i386/i386-expand.c (expand_set_or_cpymem_via_rep):
19400 For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, don't convert QImode
19402 (decide_alg): For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, use
19403 "rep movsb/stosb" only for known sizes.
19404 * config/i386/i386-options.c (processor_cost_table): Use Ice
19405 Lake cost for Cannon Lake, Ice Lake, Tiger Lake, Sapphire
19406 Rapids and Alder Lake.
19407 * config/i386/i386.h (TARGET_PREFER_KNOWN_REP_MOVSB_STOSB): New.
19408 * config/i386/x86-tune-costs.h (icelake_memcpy): New.
19409 (icelake_memset): Likewise.
19410 (icelake_cost): Likewise.
19411 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
19414 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
19417 * config/aarch64/aarch64.c
19418 (aarch64_vectorize_preferred_vector_alignment): Query the size
19419 of the provided SVE vector; do not assume that all SVE vectors
19420 have the same size.
19422 2021-03-31 Jan Hubicka <jh@suse.cz>
19425 * cgraph.c (cgraph_node::release_body): Remove all callers and
19427 * cgraphclones.c (cgraph_node::materialize_clone): Do not do it here.
19428 * cgraphunit.c (cgraph_node::expand): And here.
19430 2021-03-31 Martin Liska <mliska@suse.cz>
19432 * ipa-modref.c (analyze_ssa_name_flags): Fix coding style
19433 and one negated condition.
19435 2021-03-31 Jakub Jelinek <jakub@redhat.com>
19436 Richard Sandiford <richard.sandiford@arm.com>
19439 * config/aarch64/aarch64.md (*add<mode>3_poly_1): Swap Uai and Uav
19440 constraints on operands[2] and similarly 0 and rk constraints
19441 on operands[1] corresponding to that.
19443 2021-03-31 Jakub Jelinek <jakub@redhat.com>
19446 * configure.ac (HAVE_LD_BROKEN_PE_DWARF5): New AC_DEFINE if PECOFF
19447 linker doesn't support DWARF sections new in DWARF5.
19448 * config/i386/i386-options.c (ix86_option_override_internal): Default
19449 to dwarf_version 4 if HAVE_LD_BROKEN_PE_DWARF5 for TARGET_PECOFF
19451 * config.in: Regenerated.
19452 * configure: Regenerated.
19454 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19457 * config/aarch64/aarch64.c (aarch64_analyze_loop_vinfo): Check for
19458 available issue_info before using it.
19460 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19463 * config/aarch64/aarch64.md (sub<mode>3_compare1_imm): Do not allow zero
19466 2021-03-30 Xionghu Luo <luoxhu@linux.ibm.com>
19469 * config/rs6000/altivec.md (altivec_lvsl_reg): Change to ...
19470 (altivec_lvsl_reg_<mode>): ... this.
19471 (altivec_lvsr_reg): Change to ...
19472 (altivec_lvsr_reg_<mode>): ... this.
19473 * config/rs6000/predicates.md (vec_set_index_operand): New.
19474 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
19475 Enable 32bit variable vec_insert for all TARGET_VSX.
19476 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
19477 Enable 32bit variable vec_insert for p9 and above.
19478 (rs6000_expand_vector_set_var_p8): Rename to ...
19479 (rs6000_expand_vector_set_var_p7): ... this.
19480 (rs6000_expand_vector_set): Use TARGET_VSX and adjust assert
19482 * config/rs6000/vector.md (vec_set<mode>): Use vec_set_index_operand.
19483 * config/rs6000/vsx.md (xl_len_r): Use gen_altivec_lvsl_reg_di and
19484 gen_altivec_lvsr_reg_di.
19486 2021-03-30 H.J. Lu <hjl.tools@gmail.com>
19489 * config/i386/ia32intrin.h (__rdtsc): Defined as macro.
19490 (__rdtscp): Likewise.
19492 2021-03-30 Tamar Christina <tamar.christina@arm.com>
19494 PR tree-optimization/99825
19495 * tree-vect-slp-patterns.c (vect_check_evenodd_blend):
19496 Reject non-mult 2 lanes.
19498 2021-03-30 Richard Earnshaw <rearnsha@arm.com>
19501 * config/arm/arm.c (arm_file_start): Fix emission of
19502 Tag_ABI_VFP_args attribute.
19504 2021-03-30 Richard Biener <rguenther@suse.de>
19506 PR tree-optimization/99824
19507 * stor-layout.c (set_min_and_max_values_for_integral_type):
19508 Assert the precision is within the bounds of
19509 WIDE_INT_MAX_PRECISION.
19510 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
19511 the outermost component ref only to lower the access size
19512 and initialize that from the access type.
19514 2021-03-30 Richard Sandiford <richard.sandiford@arm.com>
19517 * config/aarch64/aarch64.md (mov<mode>): Pass multi-instruction
19518 CONST_INTs to aarch64_expand_mov_immediate when called after RA.
19520 2021-03-30 Mihailo Stojanovic <mihailo.stojanovic@typhoon-hil.com>
19522 * config/aarch64/aarch64.md
19523 (<optab>_trunc<fcvt_target><GPI:mode>2): Set the "arch"
19524 attribute to disambiguate between SIMD and FP variants of the
19527 2021-03-29 Jan Hubicka <hubicka@ucw.cz>
19529 * ipa-modref.c (merge_call_lhs_flags): Correct handling of deref.
19530 (analyze_ssa_name_flags): Fix typo in comment.
19532 2021-03-29 Alex Coplan <alex.coplan@arm.com>
19535 * config/aarch64/aarch64-sve-builtins.cc
19536 (function_builder::add_function): Add placeholder_p argument, use
19537 placeholder decls if this is set.
19538 (function_builder::add_unique_function): Instead of conditionally adding
19539 direct overloads, unconditionally add either a direct overload or a
19541 (function_builder::add_overloaded_function): Set placeholder_p if we're
19542 using C++ overloads. Use the obstack for string storage instead
19543 of relying on the tree nodes.
19544 (function_builder::add_overloaded_functions): Don't return early for
19545 m_direct_overloads: we need to add placeholders.
19546 * config/aarch64/aarch64-sve-builtins.h
19547 (function_builder::add_function): Add placeholder_p argument.
19549 2021-03-29 Richard Biener <rguenther@suse.de>
19551 PR tree-optimization/99807
19552 * tree-vect-slp.c (vect_slp_analyze_node_operations_1): Move
19553 assert below VEC_PERM handling.
19555 2021-03-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19558 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>): Use
19559 aarch64_simd_or_scalar_imm_zero to match zeroes. Remove pattern
19560 matching const_int 0.
19561 (move_lo_quad_internal_be_<mode>): Likewise.
19562 (move_lo_quad_<mode>): Update for the above.
19563 * config/aarch64/iterators.md (VQ_2E): Delete.
19565 2021-03-29 Jakub Jelinek <jakub@redhat.com>
19567 PR tree-optimization/99777
19568 * fold-const.c (extract_muldiv_1): For conversions, punt on casts from
19569 types other than scalar integral types.
19571 2021-03-28 David Edelsohn <dje.gcc@gmail.com>
19573 * config/rs6000/rs6000.c (rs6000_output_dwarf_dtprel): Do not add
19574 XCOFF TLS reloc decorations.
19576 2021-03-28 Gerald Pfeifer <gerald@pfeifer.com>
19578 * doc/analyzer.texi (Analyzer Internals): Update link to
19579 "A Memory Model for Static Analysis of C Programs".
19581 2021-03-26 David Edelsohn <dje.gcc@gmail.com>
19583 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Call function.
19584 * config/rs6000/rs6000-protos.h (rs6000_special_adjust_field_align):
19586 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align): New.
19587 (rs6000_special_round_type_align): Recursively check innermost first
19590 2021-03-26 Jakub Jelinek <jakub@redhat.com>
19593 * dwarf2out.h (struct dw_fde_node): Add rule18 member.
19594 * dwarf2cfi.c (dwarf2out_frame_debug_expr): When handling (set hfp sp)
19595 assignment with drap_reg active, queue reg save for hfp with offset 0
19596 and flush queued reg saves. When handling a push with rule18,
19597 defer queueing reg save for hfp and just assert the offset is 0.
19598 (scan_trace): Assert that fde->rule18 is false.
19600 2021-03-26 Vladimir Makarov <vmakarov@redhat.com>
19603 * ira-costs.c (record_reg_classes): Put case with
19604 CT_RELAXED_MEMORY adjacent to one with CT_MEMORY.
19605 * ira.c (ira_setup_alts): Ditto.
19606 * lra-constraints.c (process_alt_operands): Ditto.
19607 * recog.c (asm_operand_ok): Ditto.
19608 * reload.c (find_reloads): Ditto.
19610 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19612 * config/aarch64/aarch64-protos.h
19613 (cpu_addrcost_table::post_modify_ld3_st3): New member variable.
19614 (cpu_addrcost_table::post_modify_ld4_st4): Likewise.
19615 * config/aarch64/aarch64.c (generic_addrcost_table): Update
19616 accordingly, using the same costs as for post_modify.
19617 (exynosm1_addrcost_table, xgene1_addrcost_table): Likewise.
19618 (thunderx2t99_addrcost_table, thunderx3t110_addrcost_table):
19619 (tsv110_addrcost_table, qdf24xx_addrcost_table): Likewise.
19620 (a64fx_addrcost_table): Likewise.
19621 (neoversev1_addrcost_table): New.
19622 (neoversev1_tunings): Use neoversev1_addrcost_table.
19623 (aarch64_address_cost): Use the new post_modify costs for CImode
19626 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19628 * config/aarch64/aarch64.opt
19629 (-param=aarch64-loop-vect-issue-rate-niters=): New parameter.
19630 * doc/invoke.texi: Document it.
19631 * config/aarch64/aarch64-protos.h (aarch64_base_vec_issue_info)
19632 (aarch64_scalar_vec_issue_info, aarch64_simd_vec_issue_info)
19633 (aarch64_advsimd_vec_issue_info, aarch64_sve_vec_issue_info)
19634 (aarch64_vec_issue_info): New structures.
19635 (cpu_vector_cost): Write comments above the variables rather
19637 (cpu_vector_cost::issue_info): New member variable.
19638 * config/aarch64/aarch64.c: Include gimple-pretty-print.h
19639 and tree-ssa-loop-niter.h.
19640 (generic_vector_cost, a64fx_vector_cost, qdf24xx_vector_cost)
19641 (thunderx_vector_cost, tsv110_vector_cost, cortexa57_vector_cost)
19642 (exynosm1_vector_cost, xgene1_vector_cost, thunderx2t99_vector_cost)
19643 (thunderx3t110_vector_cost): Initialize issue_info to null.
19644 (neoversev1_scalar_issue_info, neoversev1_advsimd_issue_info)
19645 (neoversev1_sve_issue_info, neoversev1_vec_issue_info): New structures.
19646 (neoversev1_vector_cost): Use them.
19647 (aarch64_vec_op_count, aarch64_sve_op_count): New structures.
19648 (aarch64_vector_costs::saw_sve_only_op): New member variable.
19649 (aarch64_vector_costs::num_vector_iterations): Likewise.
19650 (aarch64_vector_costs::scalar_ops): Likewise.
19651 (aarch64_vector_costs::advsimd_ops): Likewise.
19652 (aarch64_vector_costs::sve_ops): Likewise.
19653 (aarch64_vector_costs::seen_loads): Likewise.
19654 (aarch64_simd_vec_costs_for_flags): New function.
19655 (aarch64_analyze_loop_vinfo): Initialize num_vector_iterations.
19656 Count the number of predicate operations required by SVE WHILE
19658 (aarch64_comparison_type, aarch64_multiply_add_p): New functions.
19659 (aarch64_sve_only_stmt_p, aarch64_in_loop_reduction_latency): Likewise.
19660 (aarch64_count_ops): Likewise.
19661 (aarch64_add_stmt_cost): Record whether see an SVE operation
19662 that cannot currently be implementing using Advanced SIMD.
19663 Record issue information about the scalar, Advanced SIMD
19664 and (where relevant) SVE versions of a loop.
19665 (aarch64_vec_op_count::dump): New function.
19666 (aarch64_sve_op_count::dump): Likewise.
19667 (aarch64_estimate_min_cycles_per_iter): Likewise.
19668 (aarch64_adjust_body_cost): If issue information is available,
19669 try to compare the issue rates of the various loop implementations
19670 and increase or decrease the vector body cost accordingly.
19672 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19674 * config/aarch64/aarch64.c (aarch64_detect_vector_stmt_subtype):
19675 Assume a zero cost for induction phis.
19677 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19679 * config/aarch64/aarch64.c (aarch64_embedded_comparison_type): New
19681 (aarch64_adjust_stmt_cost): Add the costs of embedded scalar and
19682 vector comparisons.
19684 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19686 * config/aarch64/aarch64.c (aarch64_detect_scalar_stmt_subtype):
19688 (aarch64_add_stmt_cost): Call it.
19690 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19692 * config/aarch64/aarch64-tuning-flags.def (matched_vector_throughput):
19693 New tuning parameter.
19694 * config/aarch64/aarch64.c (neoversev1_tunings): Use it.
19695 (aarch64_estimated_sve_vq): New function.
19696 (aarch64_vector_costs::analyzed_vinfo): New member variable.
19697 (aarch64_vector_costs::is_loop): Likewise.
19698 (aarch64_vector_costs::unrolled_advsimd_niters): Likewise.
19699 (aarch64_vector_costs::unrolled_advsimd_stmts): Likewise.
19700 (aarch64_record_potential_advsimd_unrolling): New function.
19701 (aarch64_analyze_loop_vinfo, aarch64_analyze_bb_vinfo): Likewise.
19702 (aarch64_add_stmt_cost): Call aarch64_analyze_loop_vinfo or
19703 aarch64_analyze_bb_vinfo on the first use of a costs structure.
19704 Detect whether we're vectorizing a loop for SVE that might be
19705 completely unrolled if it used Advanced SIMD instead.
19706 (aarch64_adjust_body_cost_for_latency): New function.
19707 (aarch64_finish_cost): Call it.
19709 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19711 * config/aarch64/aarch64.c (aarch64_vector_costs): New structure.
19712 (aarch64_init_cost): New function.
19713 (aarch64_add_stmt_cost): Use aarch64_vector_costs instead of
19714 the default unsigned[3].
19715 (aarch64_finish_cost, aarch64_destroy_cost_data): New functions.
19716 (TARGET_VECTORIZE_INIT_COST): Override.
19717 (TARGET_VECTORIZE_FINISH_COST): Likewise.
19718 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
19720 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19722 * config/aarch64/aarch64.c (neoversev1_advsimd_vector_cost)
19723 (neoversev1_sve_vector_cost): New cost structures.
19724 (neoversev1_vector_cost): Likewise.
19725 (neoversev1_tunings): Use them. Enable use_new_vector_costs.
19727 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19729 * config/aarch64/aarch64-protos.h
19730 (sve_vec_cost::scatter_store_elt_cost): New member variable.
19731 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
19732 accordingly, taking the cost from the cost of a scalar_store.
19733 (a64fx_sve_vector_cost): Likewise.
19734 (aarch64_detect_vector_stmt_subtype): Detect scatter stores.
19736 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19738 * config/aarch64/aarch64-protos.h
19739 (simd_vec_cost::store_elt_extra_cost): New member variable.
19740 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
19741 accordingly, using the vec_to_scalar cost for the new field.
19742 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
19743 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
19744 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
19745 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
19746 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
19747 (thunderx3t110_advsimd_vector_cost): Likewise.
19748 (aarch64_detect_vector_stmt_subtype): Detect single-element stores.
19750 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19752 * config/aarch64/aarch64-protos.h (simd_vec_cost::ld2_st2_permute_cost)
19753 (simd_vec_cost::ld3_st3_permute_cost): New member variables.
19754 (simd_vec_cost::ld4_st4_permute_cost): Likewise.
19755 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
19756 accordingly, using zero for the new costs.
19757 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
19758 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
19759 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
19760 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
19761 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
19762 (thunderx3t110_advsimd_vector_cost): Likewise.
19763 (aarch64_ld234_st234_vectors): New function.
19764 (aarch64_adjust_stmt_cost): Likewise.
19765 (aarch64_add_stmt_cost): Call aarch64_adjust_stmt_cost if using
19766 the new vector costs.
19768 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19770 * config/aarch64/aarch64-protos.h (sve_vec_cost): Turn into a
19771 derived class of simd_vec_cost. Add information about CLAST[AB]
19772 and FADDA instructions.
19773 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
19774 accordingly, using the vec_to_scalar costs for the new fields.
19775 (a64fx_sve_vector_cost): Likewise.
19776 (aarch64_reduc_type): New function.
19777 (aarch64_sve_in_loop_reduction_latency): Likewise.
19778 (aarch64_detect_vector_stmt_subtype): Take a vinfo parameter.
19779 Use aarch64_sve_in_loop_reduction_latency to handle SVE reductions
19780 that occur in the loop body.
19781 (aarch64_add_stmt_cost): Update call accordingly.
19783 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19785 * config/aarch64/aarch64-tuning-flags.def (use_new_vector_costs):
19787 * config/aarch64/aarch64-protos.h (simd_vec_cost): Put comments
19788 above the fields rather than to the right.
19789 (simd_vec_cost::reduc_i8_cost): New member variable.
19790 (simd_vec_cost::reduc_i16_cost): Likewise.
19791 (simd_vec_cost::reduc_i32_cost): Likewise.
19792 (simd_vec_cost::reduc_i64_cost): Likewise.
19793 (simd_vec_cost::reduc_f16_cost): Likewise.
19794 (simd_vec_cost::reduc_f32_cost): Likewise.
19795 (simd_vec_cost::reduc_f64_cost): Likewise.
19796 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
19797 accordingly, using the vec_to_scalar_cost for the new fields.
19798 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
19799 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
19800 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
19801 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
19802 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
19803 (thunderx3t110_advsimd_vector_cost): Likewise.
19804 (aarch64_use_new_vector_costs_p): New function.
19805 (aarch64_simd_vec_costs): New function, split out from...
19806 (aarch64_builtin_vectorization_cost): ...here.
19807 (aarch64_is_reduction): New function.
19808 (aarch64_detect_vector_stmt_subtype): Likewise.
19809 (aarch64_add_stmt_cost): Call aarch64_detect_vector_stmt_subtype if
19810 using the new vector costs.
19812 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
19815 * tree-emutls.c (get_emutls_init_templ_addr): Mark initializer of weak
19816 TLS declarations as public.
19818 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
19820 * config/aarch64/aarch64-d.c (IN_TARGET_CODE): Define.
19821 * config/arm/arm-d.c (IN_TARGET_CODE): Likewise.
19822 * config/i386/i386-d.c (IN_TARGET_CODE): Likewise.
19823 * config/mips/mips-d.c (IN_TARGET_CODE): Likewise.
19824 * config/pa/pa-d.c (IN_TARGET_CODE): Likewise.
19825 * config/riscv/riscv-d.c (IN_TARGET_CODE): Likewise.
19826 * config/rs6000/rs6000-d.c (IN_TARGET_CODE): Likewise.
19827 * config/s390/s390-d.c (IN_TARGET_CODE): Likewise.
19828 * config/sparc/sparc-d.c (IN_TARGET_CODE): Likewise.
19830 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
19833 * config.gcc (*-*-cygwin*): Add winnt-d.o
19834 (*-*-mingw*): Likewise.
19835 * config/i386/cygwin.h (EXTRA_TARGET_D_OS_VERSIONS): New macro.
19836 * config/i386/mingw32.h (EXTRA_TARGET_D_OS_VERSIONS): Likewise.
19837 * config/i386/t-cygming: Add winnt-d.o.
19838 * config/i386/winnt-d.c: New file.
19840 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
19842 * config/freebsd-d.c: Include memmodel.h.
19844 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
19847 * config.gcc (*-*-openbsd*): Add openbsd-d.o.
19848 * config/t-openbsd: Add openbsd-d.o.
19849 * config/openbsd-d.c: New file.
19851 2021-03-25 Stam Markianos-Wright <stam.markianos-wright@arm.com>
19853 PR tree-optimization/96974
19854 * tree-vect-stmts.c (vect_get_vector_types_for_stmt): Replace assert
19855 with graceful exit.
19857 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
19860 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
19864 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
19865 always_inline in system headers.
19867 2021-03-25 Kewen Lin <linkw@linux.ibm.com>
19869 * tree-vect-loop.c (vect_model_reduction_cost): Init inside_cost.
19871 2021-03-25 Jakub Jelinek <jakub@redhat.com>
19874 * tree-core.h (enum operand_equal_flag): Add OEP_ADDRESS_OF_SAME_FIELD.
19875 * fold-const.c (operand_compare::operand_equal_p): Don't compare
19876 field offsets if OEP_ADDRESS_OF_SAME_FIELD.
19878 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
19882 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
19883 always_inline in system headers.
19885 2021-03-25 Richard Biener <rguenther@suse.de>
19887 PR tree-optimization/99746
19888 * tree-vect-slp-patterns.c (complex_pattern::build): Do not mark
19889 the scalar stmt as patterned. Instead set up required things
19892 2021-03-25 Xionghu Luo <luoxhu@linux.ibm.com>
19894 * config/rs6000/rs6000.c (power8_costs): Change l2 cache
19897 2021-03-24 Martin Liska <mliska@suse.cz>
19900 * common/config/i386/i386-common.c (ARRAY_SIZE): Fix off-by-one
19902 * config/i386/i386-options.c (ix86_option_override_internal):
19903 Add run-time assert.
19905 2021-03-24 Martin Jambor <mjambor@suse.cz>
19908 * ipa-cp.c (initialize_node_lattices): Mark as bottom all
19909 parameters with unknown type.
19910 (ipacp_value_safe_for_type): New function.
19911 (propagate_vals_across_arith_jfunc): Verify that the constant type
19912 can be used for a type of the formal parameter.
19913 (propagate_vals_across_ancestor): Likewise.
19914 (propagate_scalar_across_jump_function): Likewise. Pass the type
19915 also to propagate_vals_across_ancestor.
19917 2021-03-24 Christophe Lyon <christophe.lyon@linaro.org>
19920 * config/arm/mve.md (movmisalign<mode>_mve_store): Use Ux
19922 (movmisalign<mode>_mve_load): Likewise.
19924 2021-03-24 Jakub Jelinek <jakub@redhat.com>
19927 * config/arm/vec-common.md (one_cmpl<mode>2, neg<mode>2,
19928 movmisalign<mode>): Disable expanders for TARGET_REALLY_IWMMXT.
19930 2021-03-24 Alexandre Oliva <oliva@adacore.com>
19932 * doc/sourcebuild.texi (sysconf): New effective target.
19934 2021-03-24 Alexandre Oliva <oliva@adacore.com>
19936 * config/i386/predicates.md (reg_or_const_vec_operand): New.
19937 * config/i386/sse.md (ssse3_pshufbv8qi3): Add an expander for
19938 the now *-prefixed insn_and_split, turn the splitter const vec
19939 into an input for the insn, making it an ignored immediate for
19940 non-split cases, and loaded into the scratch register
19943 2021-03-23 Vladimir N. Makarov <vmakarov@redhat.com>
19946 * config/aarch64/constraints.md (Utq, UOb, UOh, UOw, UOd, UOty):
19947 Use define_relaxed_memory_constraint for them.
19949 2021-03-23 Iain Sandoe <iain@sandoe.co.uk>
19952 * config/host-darwin.c (darwin_gt_pch_use_address): Add a
19953 colon to the diagnostic message.
19955 2021-03-23 Ilya Leoshkevich <iii@linux.ibm.com>
19957 * fwprop.c (fwprop_propagation::fwprop_propagation): Look at
19959 (try_fwprop_subst_note): Use set_info instead of insn_info.
19960 (try_fwprop_subst_pattern): Likewise.
19961 (try_fwprop_subst_notes): Likewise.
19962 (try_fwprop_subst): Likewise.
19963 (forward_propagate_subreg): Likewise.
19964 (forward_propagate_and_simplify): Likewise.
19965 (forward_propagate_into): Likewise.
19966 * rtl-ssa/accesses.h (set_info::single_nondebug_use) New
19968 (set_info::single_nondebug_insn_use): Likewise.
19969 (set_info::single_phi_use): Likewise.
19970 * rtl-ssa/member-fns.inl (set_info::single_nondebug_use) New
19972 (set_info::single_nondebug_insn_use): Likewise.
19973 (set_info::single_phi_use): Likewise.
19975 2021-03-23 Christophe Lyon <christophe.lyon@linaro.org>
19977 * doc/sourcebuild.texi (arm_dsp_ok, arm_dsp): Document.
19979 2021-03-23 Jakub Jelinek <jakub@redhat.com>
19982 * config/aarch64/aarch64.c (aarch64_add_offset): Tell
19983 expand_mult to perform an unsigned rather than a signed
19986 2021-03-23 H.J. Lu <hjl.tools@gmail.com>
19989 * config/i386/cpuid.h (__cpuid): Add __volatile__.
19990 (__cpuid_count): Likewise.
19992 2021-03-23 Richard Biener <rguenther@suse.de>
19994 PR tree-optimization/99721
19995 * tree-vect-slp.c (vect_slp_analyze_node_operations):
19996 Make sure we can schedule the node.
19998 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
20000 * config/riscv/riscv.c (riscv_subword): Take endianness into
20001 account when calculating the byte offset.
20003 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
20005 * config/riscv/predicates.md (subreg_lowpart_operator): New predicate
20006 * config/riscv/riscv.md (*addsi3_extended2, *subsi3_extended2)
20007 (*negsi2_extended2, *mulsi3_extended2, *<optab>si3_mask)
20008 (*<optab>si3_mask_1, *<optab>di3_mask, *<optab>di3_mask_1)
20009 (*<optab>si3_extend_mask, *<optab>si3_extend_mask_1): Use
20010 new predicate "subreg_lowpart_operator"
20012 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
20014 * config/riscv/riscv.c (riscv_swap_instruction): New function
20015 to byteswap an SImode rtx containing an instruction.
20016 (riscv_trampoline_init): Byteswap the generated instructions
20019 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
20021 * common/config/riscv/riscv-common.c
20022 (TARGET_DEFAULT_TARGET_FLAGS): Set default endianness.
20023 * config.gcc (riscv32be-*, riscv64be-*): Set
20024 TARGET_BIG_ENDIAN_DEFAULT to 1.
20025 * config/riscv/elf.h (LINK_SPEC): Change -melf* value
20026 depending on default endianness.
20027 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
20028 * config/riscv/linux.h (LINK_SPEC): Likewise.
20029 * config/riscv/riscv.c (TARGET_DEFAULT_TARGET_FLAGS): Set
20030 default endianness.
20031 * config/riscv/riscv.h (DEFAULT_ENDIAN_SPEC): New macro.
20033 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
20035 * config/riscv/elf.h (LINK_SPEC): Pass linker endianness flag.
20036 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
20037 * config/riscv/linux.h (LINK_SPEC): Likewise.
20038 * config/riscv/riscv.h (ASM_SPEC): Pass -mbig-endian and
20040 (BYTES_BIG_ENDIAN): Handle big endian.
20041 (WORDS_BIG_ENDIAN): Define to BYTES_BIG_ENDIAN.
20042 * config/riscv/riscv.opt (-mbig-endian, -mlittle-endian): New
20044 * doc/invoke.texi (-mbig-endian, -mlittle-endian): Document.
20046 2021-03-23 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
20048 * regcprop.c (find_oldest_value_reg): Ask target whether
20049 different mode is fine for replacement register.
20051 2021-03-23 Aldy Hernandez <aldyh@redhat.com>
20053 PR tree-optimization/99296
20054 * value-range.cc (irange::irange_set_1bit_anti_range): New.
20055 (irange::irange_set_anti_range): Call irange_set_1bit_anti_range
20056 * value-range.h (irange::irange_set_1bit_anti_range): New.
20058 2021-03-22 Vladimir N. Makarov <vmakarov@redhat.com>
20061 * config/aarch64/constraints.md (UtQ): Use
20062 define_relaxed_memory_constraint for it.
20063 * doc/md.texi (define_relaxed_memory_constraint): Describe it.
20064 * genoutput.c (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
20065 * genpreds.c (constraint_data): Add bitfield is_relaxed_memory.
20066 (have_relaxed_memory_constraints): New static var.
20067 (relaxed_memory_start, relaxed_memory_end): Ditto.
20068 (add_constraint): Add arg is_relaxed_memory. Check name for
20069 relaxed memory. Set up is_relaxed_memory in constraint_data and
20070 have_relaxed_memory_constraints. Adjust calls.
20071 (choose_enum_order): Process relaxed memory.
20072 (write_tm_preds_h): Ditto.
20073 (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
20074 * gensupport.c (process_rtx): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
20075 * ira-costs.c (record_reg_classes): Process CT_RELAXED_MEMORY.
20076 * ira-lives.c (single_reg_class): Use
20077 insn_extra_relaxed_memory_constraint.
20078 * ira.c (ira_setup_alts): CT_RELAXED_MEMORY.
20079 * lra-constraints.c (valid_address_p): Use
20080 insn_extra_relaxed_memory_constraint instead of other memory
20082 (process_alt_operands): Process CT_RELAXED_MEMORY.
20083 (curr_insn_transform): Use insn_extra_relaxed_memory_constraint.
20084 * recog.c (asm_operand_ok, preprocess_constraints): Process
20086 * reload.c (find_reloads): Ditto.
20087 * rtl.def (DEFINE_RELAXED_MEMORY_CONSTRAINT): New.
20088 * stmt.c (parse_input_constraint): Use
20089 insn_extra_relaxed_memory_constraint.
20091 2021-03-22 Segher Boessenkool <segher@kernel.crashing.org>
20094 * ubsan.c (ubsan_instrument_float_cast): Don't test for unordered if
20097 2021-03-22 Alex Coplan <alex.coplan@arm.com>
20100 * config/arm/arm-protos.h (neon_make_constant): Add generate
20101 argument to guard emitting insns, default to true.
20102 * config/arm/arm.c (arm_legitimate_constant_p_1): Reject
20103 CONST_VECTORs which neon_make_constant can't handle.
20104 (neon_vdup_constant): Add generate argument, avoid emitting
20105 insns if it's not set.
20106 (neon_make_constant): Plumb new generate argument through.
20107 * config/arm/constraints.md (Ui): New. Use it...
20108 * config/arm/mve.md (*mve_mov<mode>): ... here.
20109 * config/arm/vec-common.md (movv8hf): Use neon_make_constant to
20110 synthesize constants.
20112 2021-03-22 Richard Biener <rguenther@suse.de>
20114 * debug.h: Add deprecation warning.
20116 2021-03-22 Richard Biener <rguenther@suse.de>
20118 PR tree-optimization/99694
20119 * tree-ssa-sccvn.c (visit_phi): Ignore edges with the
20122 2021-03-22 Kito Cheng <kito.cheng@sifive.com>
20125 * config/riscv/riscv.c (riscv_expand_block_move): Get RTL value
20126 after type checking.
20128 2021-03-22 Jakub Jelinek <jakub@redhat.com>
20132 * dwarf2out.c (get_full_len): Use get_precision rather than
20134 (add_const_value_attribute): Make sure add_AT_wide argument has
20135 precision prec rather than some very wide one.
20137 2021-03-22 Kewen Lin <linkw@linux.ibm.com>
20139 * config/rs6000/rs6000.md (*rotldi3_insert_sf,
20140 *mov<SFDF:mode><SFDF2:mode>cc_p9, floatsi<mode>2_lfiwax,
20141 floatsi<mode>2_lfiwax_mem, floatunssi<mode>2_lfiwzx,
20142 floatunssi<mode>2_lfiwzx_mem, *floatsidf2_internal,
20143 *floatunssidf2_internal, fix_trunc<mode>si2_stfiwx,
20144 fix_trunc<mode>si2_internal, fixuns_trunc<mode>si2_stfiwx,
20145 *round32<mode>2_fprs, *roundu32<mode>2_fprs,
20146 *fix_trunc<mode>si2_internal): Fix empty split condition.
20147 * config/rs6000/vsx.md (*vsx_le_undo_permute_<mode>,
20148 vsx_reduc_<VEC_reduc_name>_v2df, vsx_reduc_<VEC_reduc_name>_v4sf,
20149 *vsx_reduc_<VEC_reduc_name>_v2df_scalar,
20150 *vsx_reduc_<VEC_reduc_name>_v4sf_scalar): Likewise.
20152 2021-03-22 Xionghu Luo <luoxhu@linux.ibm.com>
20155 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
20156 Convert idx to DImode.
20157 (rs6000_expand_vector_set_var_p8): Likewise.
20159 2021-03-21 Jakub Jelinek <jakub@redhat.com>
20162 * dwarf2out.c (insert_float): Change return type from void to
20163 unsigned, handle GET_MODE_SIZE (mode) == 2 and return element size.
20164 (mem_loc_descriptor, loc_descriptor, add_const_value_attribute):
20167 2021-03-20 H.J. Lu <hjl.tools@gmail.com>
20170 * config/i386/i386.c (construct_container): Check cfun != NULL
20171 before accessing silent_p.
20173 2021-03-20 Ahamed Husni <ahamedhusni73@gmail.com>
20175 * asan.c: Fix typos in comments.
20177 2021-03-20 Vladimir N. Makarov <vmakarov@redhat.com>
20179 PR rtl-optimization/99680
20180 * lra-constraints.c (skip_contraint_modifiers): Rename to skip_constraint_modifiers.
20181 (process_address_1): Check empty constraint before using
20184 2021-03-19 Pat Haugen <pthaugen@linux.ibm.com>
20186 * config/rs6000/rs6000.c (power10_cost): New.
20187 (rs6000_option_override_internal): Set Power10 costs.
20188 (rs6000_issue_rate): Set Power10 issue rate.
20189 * config/rs6000/power10.md: Rewrite for Power10.
20191 2021-03-19 Vladimir N. Makarov <vmakarov@redhat.com>
20194 * lra-constraints.c (process_address_1): Don't use unknown
20195 constraint for address constraint.
20197 2021-03-19 Iain Sandoe <iain@sandoe.co.uk>
20200 * config.gcc (powerpc-*-darwin8): Delete the reference to
20201 the now removed darwin8.h.
20203 2021-03-19 Olivier Hainque <hainque@adacore.com>
20206 * config/vxworksae.h (VX_CPU_PREFIX): Define.
20208 2021-03-19 John David Anglin <danglin@gcc.gnu.org>
20210 * config/pa/pa.c (import_milli): Use memcpy instead of strncpy.
20212 2021-03-19 Tamar Christina <tamar.christina@arm.com>
20214 PR tree-optimization/99656
20215 * tree-vect-slp-patterns.c (linear_loads_p,
20216 complex_add_pattern::matches, is_eq_or_top,
20217 vect_validate_multiplication, complex_mul_pattern::matches,
20218 complex_fms_pattern::matches): Remove complex_perm_kinds_t.
20219 * tree-vectorizer.h: (complex_load_perm_t): Removed.
20220 (slp_tree_to_load_perm_map_t): Use complex_perm_kinds_t instead of
20221 complex_load_perm_t.
20223 2021-03-19 H.J. Lu <hjl.tools@gmail.com>
20226 * config/i386/i386-options.c (ix86_init_machine_status): Set
20228 * config/i386/i386.c (init_cumulative_args): Set silent_p to
20230 (construct_container): Return early for return and argument
20231 errors if silent_p is true.
20232 * config/i386/i386.h (machine_function): Add silent_p.
20234 2021-03-19 Jakub Jelinek <jakub@redhat.com>
20237 * config/arm/constraints.md (Ds): New constraint.
20238 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): Use w,Ds
20239 constraint instead of w,Dm.
20241 2021-03-19 Andrew Stubbs <ams@codesourcery.com>
20243 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Fix quotes
20246 2021-03-19 Eric Botcazou <ebotcazou@adacore.com>
20248 PR middle-end/99641
20249 * fold-const.c (native_encode_initializer) <CONSTRUCTOR>: For an
20250 array type, do the computation of the current position in sizetype.
20252 2021-03-18 Vladimir N. Makarov <vmakarov@redhat.com>
20255 * lra-constraints.c (process_address_1): Use lookup_constraint
20256 only for a single constraint.
20258 2021-03-18 Martin Sebor <msebor@redhat.com>
20260 PR middle-end/99502
20261 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): Rename...
20262 (inbounds_memaccess_p): ...to this. Check the ending offset of
20263 the accessed member.
20265 2021-03-18 Andrew Stubbs <ams@codesourcery.com>
20267 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Add %< and
20268 %> quote markers to error messages.
20269 (gcn_goacc_validate_dims): Likewise.
20270 (gcn_conditional_register_usage): Remove exclaimation mark from error
20272 (gcn_vectorize_vec_perm_const): Ensure perm is fully uninitialized.
20274 2021-03-18 Jan Hubicka <hubicka@ucw.cz>
20276 * config/i386/x86-tune-costs.h (struct processor_costs): Fix costs of
20279 2021-03-18 Sinan Lin <sinan@isrc.iscas.ac.cn>
20280 Kito Cheng <kito.cheng@sifive.com>
20282 * config/riscv/riscv.c (riscv_block_move_straight): Change type
20283 to unsigned HOST_WIDE_INT for parameter and local variable with
20284 HOST_WIDE_INT type.
20285 (riscv_adjust_block_mem): Ditto.
20286 (riscv_block_move_loop): Ditto.
20287 (riscv_expand_block_move): Ditto.
20289 2021-03-18 Nick Clifton <nickc@redhat.com>
20291 * config/v850/v850.c (construct_restore_jr): Increase static
20293 (construct_save_jarl): Likewise.
20294 * config/v850/v850.h (DWARF2_DEBUGGING_INFO): Define.
20296 2021-03-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20298 * config/aarch64/aarch64.c (aarch64_adjust_generic_arch_tuning): Define.
20299 (aarch64_override_options_internal): Use it.
20300 (generic_tunings): Add AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS to
20303 2021-03-17 Sandra Loosemore <sandra@codesourcery.com>
20305 * config/nios2/nios2.c (nios2_custom_check_insns): Clean up
20306 error message format issues.
20307 (nios2_option_override): Likewise.
20308 (nios2_expand_fpu_builtin): Likewise.
20309 (nios2_init_custom_builtins): Adjust to avoid bogus strncpy
20310 truncation warning.
20311 (nios2_expand_custom_builtin): More error message format fixes.
20312 (nios2_expand_rdwrctl_builtin): Likewise.
20313 (nios2_expand_rdprs_builtin): Likewise.
20314 (nios2_expand_eni_builtin): Likewise.
20315 (nios2_expand_builtin): Likewise.
20316 (nios2_register_custom_code): Likewise.
20317 (nios2_valid_target_attribute_rec): Likewise.
20318 (nios2_add_insn_asm): Fix uninitialized variable warning.
20320 2021-03-17 Jan Hubicka <jh@suse.cz>
20322 * config/i386/x86-tune-costs.h (struct processor_costs): Update costs
20323 of gather to match reality.
20324 * config/i386/x86-tune.def (X86_TUNE_USE_GATHER): Enable for znver3.
20326 2021-03-17 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20328 * config/aarch64/aarch64-builtins.c (aarch64_expand_rng_builtin): Use EQ
20329 to compare against CC_REG rather than NE.
20331 2021-03-17 H.J. Lu <hjl.tools@gmail.com>
20334 * config/i386/i386.c (ix86_force_load_from_GOT_p): Support
20335 inline assembly statements.
20336 (ix86_print_operand): Update 'P' handling for -fno-plt.
20338 2021-03-17 Tamar Christina <tamar.christina@arm.com>
20341 * config/aarch64/aarch64.c
20342 (aarch64_simd_clone_compute_vecsize_and_simdlen): Remove unused var.
20344 2021-03-16 Segher Boessenkool <segher@kernel.crashing.org>
20347 * config/rs6000/predicates.md (branch_comparison_operator): Allow
20348 ordered and unordered for CCFPmode, if flag_finite_math_only.
20350 2021-03-16 Jakub Jelinek <jakub@redhat.com>
20353 * config/i386/i386-expand.c (ix86_split_lea_for_addr): Emit a MULT
20354 rather than ASHIFT.
20355 * config/i386/i386.md (mult by 1248 into ashift): New splitter.
20357 2021-03-16 Martin Liska <mliska@suse.cz>
20360 * optc-save-gen.awk: Add flag_ipa_ra to exceptions for
20361 cl_optimization_compare function.
20363 2021-03-16 Ilya Leoshkevich <iii@linux.ibm.com>
20365 * config/s390/s390.c (f_constraint_p): Treat "fv" constraints
20368 2021-03-16 Jakub Jelinek <jakub@redhat.com>
20371 * config/i386/i386.h (struct machine_function): Add
20372 has_explicit_vzeroupper bitfield.
20373 * config/i386/i386-expand.c (ix86_expand_builtin): Set
20374 cfun->machine->has_explicit_vzeroupper when expanding
20375 IX86_BUILTIN_VZEROUPPER.
20376 * config/i386/i386-features.c (rest_of_handle_insert_vzeroupper):
20377 Do the mode switching only when TARGET_VZEROUPPER, expensive
20378 optimizations turned on and not optimizing for size.
20379 (pass_insert_vzeroupper::gate): Enable even when
20380 cfun->machine->has_explicit_vzeroupper is set.
20382 2021-03-16 Jakub Jelinek <jakub@redhat.com>
20385 * config/aarch64/aarch64.c
20386 (aarch64_simd_clone_compute_vecsize_and_simdlen): If not a function
20387 definition, walk TYPE_ARG_TYPES list if non-NULL for argument types
20388 instead of DECL_ARGUMENTS. Ignore types for uniform arguments.
20390 2021-03-15 Richard Biener <rguenther@suse.de>
20392 PR tree-optimization/98834
20393 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle missing
20394 subsetting by truncating the access size.
20396 2021-03-15 Jan Hubicka <hubicka@ucw.cz>
20398 * config/i386/i386-options.c (processor_cost_table): Add znver3_cost.
20399 * config/i386/x86-tune-costs.h (znver3_cost): New gobal variable; copy
20402 2021-03-15 Martin Liska <mliska@suse.cz>
20404 * spellcheck.c: Add missing comma in initialization.
20406 2021-03-14 Uroš Bizjak <ubizjak@gmail.com>
20408 * config/i386/sse.md (*vec_extract<mode>): Merge alternative 0 with
20409 alternative 2 and alternative 1 with alternative 3 using
20410 YW register constraint.
20411 (*vec_extract<PEXTR_MODE12:mode>_zext): Merge alternatives
20412 using YW register constraint.
20413 (*vec_extractv16qi_zext): Ditto.
20414 (*vec_extractv4si): Merge alternatives 4 and 5
20415 using Yw register constraint.
20416 (*ssse3_palignr<mode>_perm): Use Yw instead of v for alternative 3.
20418 2021-03-13 Martin Sebor <msebor@redhat.com>
20420 PR tree-optimization/99489
20421 * builtins.c (gimple_call_alloc_size): Fail gracefully when argument
20422 is not a call statement.
20424 2021-03-13 Jakub Jelinek <jakub@redhat.com>
20426 PR tree-optimization/99544
20427 * match.pd (X + (X << C) -> X * (1 + (1 << C))): Don't simplify
20428 if for vector types multiplication can't be done in type's mode.
20430 2021-03-12 Eric Botcazou <ebotcazou@adacore.com>
20433 * config/sparc/constraints.md (w): Rename to...
20434 (W): ... this and ditch previous implementation.
20435 * config/sparc/sparc.md (*movdi_insn_sp64): Replace W with m.
20436 (*movdf_insn_sp64): Likewise.
20437 (*mov<VM64:mode>_insn_sp64): Likewise.
20438 * config/sparc/sync.md (*atomic_compare_and_swap<mode>_1): Replace
20440 (atomic_compare_and_swap_leon3_1): Likewise.
20441 (*atomic_compare_and_swapdi_v8plus): Likewise.
20442 * config/sparc/sparc.c (memory_ok_for_ldd): Remove useless test on
20443 architecture and add missing address validity check during LRA.
20445 2021-03-12 Tobias Burnus <tobias@codesourcery.com>
20448 * gimplify.c (omp_add_variable): Handle NULL_TREE as size
20449 occuring for assumed-size arrays in use_device_{ptr,addr}.
20451 2021-03-12 Jakub Jelinek <jakub@redhat.com>
20454 * config/i386/constraints.md (YW): New internal constraint.
20455 * config/i386/sse.md (v_Yw): Add V4TI, V2TI, V1TI and TI cases.
20456 (*<sse2_avx2>_<insn><mode>3<mask_name>,
20457 *<sse2_avx2>_uavg<mode>3<mask_name>, *abs<mode>2,
20458 *<s>mul<mode>3_highpart<mask_name>): Use <v_Yw> instead of v in
20460 (<sse2_avx2>_psadbw): Use YW instead of v in constraints.
20461 (*avx2_pmaddwd, *sse2_pmaddwd, *<code>v8hi3, *<code>v16qi3,
20462 avx2_pmaddubsw256, ssse3_pmaddubsw128): Merge last two alternatives
20463 into one, use Yw instead of former x,v.
20464 (ashr<mode>3, <insn><mode>3): Use <v_Yw> instead of x in constraints of
20465 the last alternative.
20466 (<sse2_avx2>_packsswb<mask_name>, <sse2_avx2>_packssdw<mask_name>,
20467 <sse2_avx2>_packuswb<mask_name>, <sse4_1_avx2>_packusdw<mask_name>,
20468 *<ssse3_avx2>_pmulhrsw<mode>3<mask_name>, <ssse3_avx2>_palignr<mode>,
20469 <ssse3_avx2>_pshufb<mode>3<mask_name>): Merge last two alternatives
20470 into one, use <v_Yw> instead of former x,v.
20471 (avx2_interleave_highv32qi<mask_name>,
20472 vec_interleave_highv16qi<mask_name>): Use Yw instead of v in
20473 constraints. Add && <mask_avx512bw_condition> to condition.
20474 (avx2_interleave_lowv32qi<mask_name>,
20475 vec_interleave_lowv16qi<mask_name>,
20476 avx2_interleave_highv16hi<mask_name>,
20477 vec_interleave_highv8hi<mask_name>,
20478 avx2_interleave_lowv16hi<mask_name>, vec_interleave_lowv8hi<mask_name>,
20479 avx2_pshuflw_1<mask_name>, sse2_pshuflw_1<mask_name>,
20480 avx2_pshufhw_1<mask_name>, sse2_pshufhw_1<mask_name>,
20481 avx2_<code>v16qiv16hi2<mask_name>, sse4_1_<code>v8qiv8hi2<mask_name>,
20482 *sse4_1_<code>v8qiv8hi2<mask_name>_1, <sse2_avx2>_<insn><mode>3): Use
20483 Yw instead of v in constraints.
20484 * config/i386/mmx.md (Yv_Yw): New define_mode_attr.
20485 (*mmx_<insn><mode>3, mmx_ashr<mode>3, mmx_<insn><mode>3): Use <Yv_Yw>
20486 instead of Yv in constraints.
20487 (*mmx_<insn><mode>3, *mmx_mulv4hi3, *mmx_smulv4hi3_highpart,
20488 *mmx_umulv4hi3_highpart, *mmx_pmaddwd, *mmx_<code>v4hi3,
20489 *mmx_<code>v8qi3, mmx_pack<s_trunsuffix>swb, mmx_packssdw,
20490 mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
20491 *mmx_uavgv8qi3, *mmx_uavgv4hi3, mmx_psadbw): Use Yw instead of Yv in
20493 (*mmx_pinsrw, *mmx_pinsrb, *mmx_pextrw, *mmx_pextrw_zext, *mmx_pextrb,
20494 *mmx_pextrb_zext): Use YW instead of Yv in constraints.
20495 (*mmx_eq<mode>3, mmx_gt<mode>3): Use x instead of Yv in constraints.
20496 (mmx_andnot<mode>3, *mmx_<code><mode>3): Split last alternative into
20497 two, one with just x, another isa avx512vl with v.
20499 2021-03-12 Martin Liska <mliska@suse.cz>
20501 * doc/invoke.texi: Add missing param documentation.
20503 2021-03-11 David Malcolm <dmalcolm@redhat.com>
20506 * Makefile.in (ANALYZER_OBJS): Add analyzer/feasible-graph.o and
20507 analyzer/trimmed-graph.o.
20508 * doc/analyzer.texi (Analyzer Paths): Rewrite description of
20509 feasibility checking to reflect new implementation.
20510 * doc/invoke.texi (-fdump-analyzer-feasibility): Document new
20512 * shortest-paths.h (shortest_paths::get_shortest_distance): New.
20514 2021-03-11 David Malcolm <dmalcolm@redhat.com>
20516 * digraph.cc (selftest::test_shortest_paths): Update
20517 shortest_paths init for new param. Add test of
20518 SPS_TO_GIVEN_TARGET.
20519 * shortest-paths.h (enum shortest_path_sense): New.
20520 (shortest_paths::shortest_paths): Add "sense" param.
20521 Update for renamings. Generalize to use "sense" param.
20522 (shortest_paths::get_shortest_path): Rename param.
20523 (shortest_paths::m_sense): New field.
20524 (shortest_paths::m_prev): Rename...
20525 (shortest_paths::m_best_edge): ...to this.
20526 (shortest_paths::get_shortest_path): Update for renamings.
20527 Conditionalize flipping of path on sense of traversal.
20529 2021-03-11 David Malcolm <dmalcolm@redhat.com>
20531 * digraph.cc (selftest::test_shortest_paths): Add test coverage
20532 for paths from B and C.
20533 * shortest-paths.h (shortest_paths::shortest_paths): Handle
20534 unreachable nodes, rather than asserting.
20536 2021-03-11 David Edelsohn <dje.gcc@gmail.com>
20539 * config/rs6000/rs6000.c (rs6000_xcoff_file_start): Don't create
20540 xcoff_tbss_section_name.
20541 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .lcomm.
20542 * xcoffout.c (xcoff_tbss_section_name): Delete.
20543 * xcoffout.h (xcoff_tbss_section_name): Delete.
20545 2021-03-11 Richard Biener <rguenther@suse.de>
20547 PR tree-optimization/99523
20548 * tree-cfg.c (dump_function_to_file): Dump SSA names
20549 w/o identifier to the decls section as well, not only those
20550 without a VAR_DECL.
20552 2021-03-11 Jakub Jelinek <jakub@redhat.com>
20555 * ipa-icf-gimple.c (func_checker::compare_gimple_call): For internal
20556 function calls with lhs fail if the lhs don't have compatible types.
20558 2021-03-11 Hans-Peter Nilsson <hp@axis.com>
20560 * config/cris/cris.h (HARD_FRAME_POINTER_REGNUM): Define.
20561 Change FRAME_POINTER_REGNUM to correspond to a new faked
20562 register faked_fp, part of GENNONACR_REGS like faked_ap.
20563 (CRIS_FAKED_REGS_CONTENTS): New helper macro.
20564 (FIRST_PSEUDO_REGISTER, FIXED_REGISTERS, CALL_USED_REGISTERS):
20565 (REG_ALLOC_ORDER, REG_CLASS_CONTENTS, REGNO_OK_FOR_BASE_P)
20566 (ELIMINABLE_REGS, REGISTER_NAMES): Adjust accordingly.
20567 * config/cris/cris.md (CRIS_FP_REGNUM): Renumber to new faked
20569 (CRIS_REAL_FP_REGNUM): New constant.
20570 * config/cris/cris.c (cris_reg_saved_in_regsave_area): Check
20571 for HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM.
20572 (cris_initial_elimination_offset): Handle elimination changes
20573 to HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM
20574 and add one from FRAME_POINTER_REGNUM to
20575 HARD_FRAME_POINTER_REGNUM.
20576 (cris_expand_prologue, cris_expand_epilogue): Emit code for
20577 hard_frame_pointer_rtx instead of frame_pointer_rtx.
20579 2021-03-10 David Edelsohn <dje.gcc@gmail.com>
20582 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Add check for DCmode.
20583 * config/rs6000/rs6000.c (rs6000_special_round_type_align): Same.
20585 2021-03-10 Vladimir N. Makarov <vmakarov@redhat.com>
20588 * lra-constraints.c (process_address_1): Don't check unknown
20589 constraint, use X for empty constraint.
20591 2021-03-10 Alex Coplan <alex.coplan@arm.com>
20593 * config/aarch64/aarch64.c (aarch64_vfp_is_call_or_return_candidate):
20594 Fix typo in comment describing "is_ha" argument.
20596 2021-03-10 John David Anglin <danglin@gcc.gnu.org>
20598 * doc/sourcebuild.texi: Document LRA target selector.
20600 2021-03-10 David Malcolm <dmalcolm@redhat.com>
20602 * doc/ux.texi: Add subsection contrasting interactive versus
20603 batch usage of GCC.
20605 2021-03-10 Joel Hutton <joel.hutton@arm.com>
20608 * tree-vect-stmts.c (vectorizable_store): Fix scatter store mask
20610 (vectorizable_load): Fix gather load mask check condition.
20612 2021-03-10 Richard Biener <rguenther@suse.de>
20614 PR tree-optimization/99510
20615 * tree.c (check_aligned_type): Check that the candidate
20616 has TYPE_USER_ALIGN set instead of matching with the
20619 2021-03-10 Eric Botcazou <ebotcazou@adacore.com>
20621 * config/sparc/sparc.c (sparc_regmode_natural_size): Return 4 for
20622 float and vector integer modes only if the mode is not larger.
20624 2021-03-10 Hans-Peter Nilsson <hp@axis.com>
20626 * config/cris/cris.h (DWARF_FRAME_REGISTERS): Define.
20628 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
20630 * ira.c (ira_setup_alts, ira_get_dup_out_num): Process digital
20632 * ira-lives.c (single_reg_class): Ditto.
20634 2021-03-09 Sebastian Huber <sebastian.huber@embedded-brains.de>
20636 * config.gcc (aarch64-*-rtems*): Include general rtems.h after
20637 the architecture-specific rtems.h.
20638 (aarch64-*-rtems*): Likewise.
20639 (arm*-*-rtems*): Likewise.
20640 (epiphany-*-rtems*): Likewise.
20641 (riscv*-*-rtems*): Likewise.
20643 2021-03-09 Jakub Jelinek <jakub@redhat.com>
20645 PR tree-optimization/99305
20646 * tree-ssa-phiopt.c (conditional_replacement): Test integer_pow2p
20647 before integer_all_onesp instead of vice versa.
20649 2021-03-09 Richard Earnshaw <rearnsha@arm.com>
20651 * common/config/arm/arm-common.c (arm_config_default): Change type
20652 of 'i' to unsigned.
20654 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
20657 * lra-constraints.c (process_address_1): Process constraint 'g'
20658 separately and digital constraints containing more one digit.
20660 2021-03-09 Nick Clifton <nickc@redhat.com>
20662 * config/rx/rx.h (DBX_DEBUGGING_INFO): Define.
20663 (DWARF"_DEBUGGING_INFO): Define.
20665 2021-03-09 Eric Botcazou <ebotcazou@adacore.com>
20668 * calls.c (initialize_argument_information): When the argument
20669 is passed by reference, do not make a copy in a thunk only if
20670 the argument is already in memory. Remove redundant test for
20671 the case of callee copy.
20673 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
20676 * lra-constraints.c (process_address_1): Process 0..9 constraints
20677 in process_address_1.
20679 2021-03-09 Andreas Krebbel <krebbel@linux.ibm.com>
20681 * config/s390/s390.c (struct s390_processor processor_table):
20682 Binutils name string must not be empty.
20684 2021-03-09 Claudiu Zissulescu <claziss@synopsys.com>
20686 * config/arc/arc.c (arc_attr_type): Remove function.
20688 2021-03-09 Martin Liska <mliska@suse.cz>
20691 * config/i386/i386-options.c (ix86_option_override_internal):
20692 Set isa_flags for OPTS argument and not for the global
20695 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
20697 * config/rs6000/predicates.md (ds_form_mem_operand): Check
20700 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
20703 * config/rs6000/predicates.md (ds_form_mem_operand) New
20705 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10) Use
20706 ds_form_mem_operand in ld/lwa patterns.
20707 * config/rs6000/fusion.md: Regenerate file.
20709 2021-03-08 Martin Sebor <msebor@redhat.com>
20711 PR middle-end/98266
20712 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): New function.
20713 (array_bounds_checker::check_array_bounds): Call it.
20715 2021-03-08 Martin Sebor <msebor@redhat.com>
20717 PR middle-end/97631
20718 * tree-ssa-strlen.c (maybe_warn_overflow): Test rawmem.
20719 (handle_builtin_stxncpy_strncat): Rename locals. Determine
20720 destination size from allocation calls. Issue a more appropriate
20722 (handle_builtin_memcpy): Pass true as rawmem to maybe_warn_overflow.
20723 (handle_builtin_memset): Same.
20725 2021-03-08 Peter Bergner <bergner@linux.ibm.com>
20728 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_permute): Add an assert
20729 to ensure we do not have an Altivec style address.
20730 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): Disable if passed
20731 an Altivec style address.
20732 (*vsx_le_perm_store_<mode>): Likewise.
20733 (splitters after *vsx_le_perm_store_<mode>): Likewise.
20734 (vsx_load_<mode>): Disable special expander if passed an Altivec
20736 (vsx_store_<mode>): Likewise.
20738 2021-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20741 * config/aarch64/predicates.md (aarch64_simd_shift_imm_vec_qi): Define.
20742 (aarch64_simd_shift_imm_vec_hi): Likewise.
20743 (aarch64_simd_shift_imm_vec_si): Likewise.
20744 (aarch64_simd_shift_imm_vec_di): Likewise.
20745 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le): Use
20746 predicate from above.
20747 (aarch64_shrn<mode>_insn_be): Likewise.
20748 (aarch64_rshrn<mode>_insn_le): Likewise.
20749 (aarch64_rshrn<mode>_insn_be): Likewise.
20750 (aarch64_shrn2<mode>_insn_le): Likewise.
20751 (aarch64_shrn2<mode>_insn_be): Likewise.
20752 (aarch64_rshrn2<mode>_insn_le): Likewise.
20753 (aarch64_rshrn2<mode>_insn_be): Likewise.
20755 2021-03-08 Vladimir N. Makarov <vmakarov@redhat.com>
20758 * lra-constraints.c (skip_contraint_modifiers): New function.
20759 (process_address_1): Use it before lookup_constraint call.
20761 2021-03-08 Martin Liska <mliska@suse.cz>
20764 * config/i386/i386-options.c (ix86_option_override_internal):
20765 Enable UINTR and HRESET for -march that supports it.
20767 2021-03-08 Ilya Leoshkevich <iii@linux.ibm.com>
20769 * config/s390/s390.c (f_constraint_p): New function.
20770 (s390_md_asm_adjust): Implement TARGET_MD_ASM_ADJUST.
20771 (TARGET_MD_ASM_ADJUST): Likewise.
20773 2021-03-08 Tobias Burnus <tobias@codesourcery.com>
20776 * tree-nested.c (convert_local_reference_stmt): Avoid calling
20777 lookup_field_for_decl for Fortran module (= namespace context).
20779 2021-03-08 Andreas Krebbel <krebbel@linux.ibm.com>
20781 * config/s390/s390.c (s390_expand_vec_compare): Implement <0
20782 comparison with arithmetic right shift.
20783 (s390_expand_vcond): No need for a force_reg anymore.
20784 s390_vec_compare will do it.
20785 * config/s390/vector.md ("vec_cmp<mode><tointvec>"): Accept also
20786 immediate operands.
20788 2021-03-07 Jakub Jelinek <jakub@redhat.com>
20791 * config/i386/constraints.md (Yw): Use SSE_REGS if TARGET_SSE
20792 but TARGET_AVX512BW or TARGET_AVX512VL is not set. Adjust description
20794 * config/i386/sse.md (v_Yw): New define_mode_attr.
20795 (*<insn><mode>3, *mul<mode>3<mask_name>, *avx2_<code><mode>3,
20796 *sse4_1_<code><mode>3<mask_name>): Use <v_Yw> instead of v
20798 * config/i386/mmx.md (mmx_pshufw_1, *vec_dupv4hi): Use Yw instead of
20799 xYw in constraints.
20801 2021-03-06 Julian Brown <julian@codesourcery.com>
20803 * tree-pretty-print.c (dump_generic_node): Emit non-generic
20804 address space info for aggregates.
20806 2021-03-06 Hans-Peter Nilsson <hp@axis.com>
20808 * config/cris/cris.h (MAX_FIXED_MODE_SIZE): Don't define.
20810 2021-03-05 Jakub Jelinek <jakub@redhat.com>
20812 PR middle-end/99322
20813 * tree-cfg.c (bb_to_omp_idx): New variable.
20814 (execute_build_cfg): Release the bb_to_omp_idx vector after
20815 cleanup_tree_cfg returns.
20816 (handle_abnormal_edges): Remove bb_to_omp_idx argument, adjust
20817 for bb_to_omp_idx being a vec<int> instead of pointer to array
20819 (make_edges): Remove bb_to_omp_idx local variable, don't pass
20820 it to handle_abnormal_edges, adjust for bb_to_omp_idx being a
20821 vec<int> instead of pointer to array of ints and don't free/release
20823 (remove_bb): When removing a bb and placing forced label somewhere
20824 else, ensure it is put into the same OpenMP region during cfg
20825 pass if possible or to entry successor as fallback. Unregister
20826 bb from bb_to_omp_idx.
20828 2021-03-05 Vladimir N. Makarov <vmakarov@redhat.com>
20831 * lra-constraints.c (process_address_1): Skip decomposing address
20832 for asm insn operand with unknown constraint.
20834 2021-03-05 Martin Jambor <mjambor@suse.cz>
20837 * cgraph.c (cgraph_edge::set_call_stmt): Do not update all
20838 corresponding speculative edges if we are about to resolve
20839 sepculation. Make edge direct (and so resolve speculations) before
20840 removing it from call_site_hash.
20841 (cgraph_edge::make_direct): Relax the initial assert to allow calling
20842 the function on speculative direct edges.
20844 2021-03-05 Eric Botcazou <ebotcazou@adacore.com>
20846 PR rtl-optimization/99376
20847 * rtlanal.c (nonzero_bits1) <arithmetic operators>: If the number
20848 of low-order zero bits is too large, set the result to 0 directly.
20850 2021-03-04 Jakub Jelinek <jakub@redhat.com>
20852 PR middle-end/93235
20853 * expmed.c (store_bit_field_using_insv): Return false of xop0 is a
20854 SUBREG and a SUBREG to op_mode can't be created.
20856 2021-03-04 Alex Coplan <alex.coplan@arm.com>
20859 * config/aarch64/aarch64-sve-builtins.cc
20860 (function_resolver::require_vector_type): Handle error_mark_node.
20862 2021-03-04 Ilya Leoshkevich <iii@linux.ibm.com>
20864 * cfgexpand.c (expand_asm_loc): Pass new parameter.
20865 (expand_asm_stmt): Likewise.
20866 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Add new
20868 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
20869 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
20870 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
20871 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
20872 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
20873 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
20874 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
20875 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
20876 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
20877 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
20878 * doc/tm.texi (md_asm_adjust): Likewise.
20879 * target.def (md_asm_adjust): Likewise.
20881 2021-03-04 Richard Biener <rguenther@suse.de>
20883 PR middle-end/97855
20884 * tree-pretty-print.c: Poison pp_printf.
20885 (dump_decl_name): Avoid use of pp_printf.
20886 (dump_block_node): Likewise.
20887 (dump_generic_node): Likewise.
20889 2021-03-04 Martin Sebor <msebor@redhat.com>
20891 PR middle-end/96963
20892 PR middle-end/94655
20893 * builtins.c (handle_array_ref): New helper.
20894 (handle_mem_ref): New helper.
20895 (compute_objsize_r): Factor out ARRAY_REF and MEM_REF handling
20896 into new helper functions. Correct a workaround for vectorized
20899 2021-03-03 Pat Haugen <pthaugen@linux.ibm.com>
20901 * config/rs6000/dfp.md (extendddtd2, trunctddd2, *cmp<mode>_internal1,
20902 floatditd2, ftrunc<mode>2, fix<mode>di2, dfp_ddedpd_<mode>,
20903 dfp_denbcd_<mode>, dfp_dxex_<mode>, dfp_diex_<mode>,
20904 *dfp_sgnfcnc_<mode>, dfp_dscli_<mode>, dfp_dscri_<mode>): Update size
20905 attribute for Power10.
20906 * config/rs6000/mma.md (*movoo): Likewise.
20907 * config/rs6000/rs6000.md (define_attr "size"): Add 256.
20908 (define_mode_attr bits): Add DD/TD modes.
20909 * config/rs6000/sync.md (load_quadpti, store_quadpti, load_lockedpti,
20910 store_conditionalpti): Update size attribute for Power10.
20912 2021-03-03 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20915 * config/sparc/t-sparc (tree-ssanames.o-warn): Don't error for
20916 -Wuninitialized, -Wmaybe-uninitialized.
20917 (wide-int.o-warn): Likewise.
20919 2021-03-03 Richard Earnshaw <rearnsha@arm.com>
20921 * common/config/arm/arm-common.c: Include configargs.h.
20922 (arm_config_default): New function.
20923 (arm_target_mode): Renamed from arm_target_thumb_only. Handle
20924 processors that do not support Thumb. Take into account the
20925 --with-mode configuration setting for selecting the default.
20926 * config/arm/arm.h (OPTION_DEFAULT_SPECS): Remove entry for 'mode'.
20927 (TARGET_MODE_SPEC_FUNCTIONS): Update for function name change.
20929 2021-03-03 Martin Liska <mliska@suse.cz>
20931 PR gcov-profile/97461
20932 * gcov-io.h (GCOV_PREALLOCATED_KVP): Remove.
20934 2021-03-03 Eric Botcazou <ebotcazou@adacore.com>
20937 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
20938 point back the hard frame pointer to its default location when the
20939 frame is larger than SEH_MAX_FRAME_SIZE.
20941 2021-03-03 Jakub Jelinek <jakub@redhat.com>
20944 * config/i386/predicates.md (logic_operator): New define_predicate.
20945 * config/i386/i386.md (mov + mem using comm arith peephole2):
20946 Punt if operands[1] is EXT_REX_SSE_REGNO_P, AVX512BW is not enabled
20947 and the inner mode is [QH]Imode.
20949 2021-03-03 Jakub Jelinek <jakub@redhat.com>
20952 * dwarf2out.c (dw_loc_list_struct): Add end_entry member.
20953 (new_loc_list): Clear end_entry.
20954 (output_loc_list): Only use DW_LLE_startx_length for -gsplit-dwarf
20955 if HAVE_AS_LEB128, otherwise use DW_LLE_startx_endx. Fix comment
20957 (index_location_lists): For dwarf_version >= 5 without HAVE_AS_LEB128,
20958 initialize also end_entry.
20960 2021-03-03 Jakub Jelinek <jakub@redhat.com>
20963 * cfgrtl.c (fixup_partitions): When changing some bbs from hot to cold
20964 partitions, if in non-layout mode after reorder_blocks also move
20965 affected blocks to ensure a single partition transition.
20967 2021-03-03 Jason Merrill <jason@redhat.com>
20970 * cgraphunit.c (process_function_and_variable_attributes): Don't
20971 warn about flatten on an alias if the target also has it.
20972 * cgraph.h (symtab_node::get_alias_target_tree): New.
20974 2021-03-02 David Edelsohn <dje.gcc@gmail.com>
20976 * config/rs6000/rs6000.md (tls_get_tpointer_internal): Prepend
20977 period to symbol name.
20978 (tls_get_addr_internal<mode>): Same.
20980 2021-03-02 David Malcolm <dmalcolm@redhat.com>
20983 * diagnostic-show-locus.c
20984 (selftest::test_one_liner_many_fixits_2): Fix accidental usage of
20987 2021-03-02 Martin Sebor <msebor@redhat.com>
20989 PR middle-end/99276
20990 * builtins.c (warn_for_access): Remove stray warning text.
20992 2021-03-02 Martin Sebor <msebor@redhat.com>
20994 PR middle-end/99295
20995 * doc/extend.texi (attribute malloc): Reword and clarify nonaliasing
20998 2021-03-02 Jakub Jelinek <jakub@redhat.com>
21001 * dwarf2out.c (output_macinfo_op): Use DW_MACRO_*_str* even with
21002 -gdwarf-5 -gstrict-dwarf. For -gsplit-dwarf -gdwarf-5 use
21003 DW_MACRO_*_strx instead of DW_MACRO_*_strp. Handle
21004 DW_MACRO_define_strx and DW_MACRO_undef_strx.
21005 (save_macinfo_strings): Use DW_MACRO_*_str* even with
21006 -gdwarf-5 -gstrict-dwarf. Handle DW_MACRO_define_strx and
21007 DW_MACRO_undef_strx.
21009 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
21011 * config/s390/s390-builtin-types.def (BT_FN_V4SF_V8HI_UINT): New
21013 (BT_FN_V8HI_V8HI_UINT): Likewise.
21014 (BT_FN_V8HI_V4SF_V4SF_UINT): Likewise.
21015 * config/s390/s390-builtins.def (B_NNPA): New macro definition.
21016 (s390_vclfnhs, s390_vclfnls, s390_vcrnfs, s390_vcfn, s390_vcnf):
21017 New builtin definitions.
21018 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Bump
21019 vector extension version.
21020 * config/s390/s390.c (s390_expand_builtin): Check if builtins are
21021 available with current -march level.
21022 * config/s390/s390.md (UNSPEC_NNPA_VCLFNHS_V8HI)
21023 (UNSPEC_NNPA_VCLFNLS_V8HI, UNSPEC_NNPA_VCRNFS_V8HI)
21024 (UNSPEC_NNPA_VCFN_V8HI, UNSPEC_NNPA_VCNF_V8HI): New constants.
21025 * config/s390/vecintrin.h (vec_extend_to_fp32_hi): New macro.
21026 (vec_extend_to_fp32_lo): Likewise.
21027 (vec_round_from_fp32): Likewise.
21028 (vec_convert_to_fp16): Likewise.
21029 (vec_convert_from_fp16): Likewise.
21030 * config/s390/vx-builtins.md (vclfnhs_v8hi): New insn pattern.
21031 (vclfnls_v8hi): Likewise.
21032 (vcrnfs_v8hi): Likewise.
21033 (vcfn_v8hi): Likewise.
21034 (vcnf_v8hi): Likewise.
21036 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
21038 * common/config/s390/s390-common.c (processor_flags_table): New entry.
21039 * config.gcc: Enable arch14 for --with-arch and --with-tune.
21040 * config/s390/driver-native.c (s390_host_detect_local_cpu): Pick
21041 arch14 for unknown CPU models.
21042 * config/s390/s390-opts.h (enum processor_type): Add PROCESSOR_ARCH14.
21043 * config/s390/s390.c (s390_issue_rate): Add case for PROCESSOR_ARCH14.
21044 (s390_get_sched_attrmask): Likewise.
21045 (s390_get_unit_mask): Likewise.
21046 * config/s390/s390.h (enum processor_flags): Add PF_NNPA and PF_ARCH14.
21047 (TARGET_CPU_ARCH14, TARGET_CPU_ARCH14_P, TARGET_CPU_NNPA)
21048 (TARGET_CPU_NNPA_P, TARGET_ARCH14, TARGET_ARCH14_P, TARGET_NNPA)
21049 (TARGET_NNPA_P): New macro definitions.
21050 * config/s390/s390.md ("cpu_facility", "enabled"): Add arch14 and nnpa.
21051 * config/s390/s390.opt: Add PROCESSOR_ARCH14.
21053 2021-03-02 Jakub Jelinek <jakub@redhat.com>
21055 PR middle-end/95757
21056 * tree-vrp.c (register_edge_assert_for): Remove superfluous ()s around
21057 condition. Call register_edge_assert_for_1 for == 0, != 0, == 1 and
21058 != 1 comparisons if name is lhs of a comparison.
21060 2021-03-01 Iain Sandoe <iain@sandoe.co.uk>
21064 * config/darwin-protos.h (darwin_should_restore_cfa_state): New.
21065 * config/darwin.c (darwin_should_restore_cfa_state): New.
21066 * config/darwin.h (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New.
21067 * doc/tm.texi: Regenerated.
21068 * doc/tm.texi.in: Document TARGET_ASM_SHOULD_RESTORE_CFA_STATE.
21069 * dwarf2cfi.c (connect_traces): If the target requests, restore
21070 the CFA expression after a DW_CFA_restore.
21071 * target.def (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New hook.
21073 2021-03-01 Martin Liska <mliska@suse.cz>
21076 * optc-save-gen.awk: Add 4 more exceptions.
21078 2021-03-01 Nathan Sidwell <nathan@acm.org>
21081 * tree.h (TYPE_ALIGN_RAW): New accessor.
21082 (TYPE_ALIGN): Use it.
21084 2021-03-01 Jan Hubicka <jh@suse.cz>
21087 * ipa-fnsummary.c (compute_fn_summary): Fix sanity check.
21089 2021-03-01 Eric Botcazou <ebotcazou@adacore.com>
21092 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
21093 point the hard frame pointer to the SSE register save area instead
21094 of the general register save area. Perform only minimal adjustment
21095 for small frames if it is initially not correctly aligned.
21096 (ix86_expand_prologue): Remove early saves for a SEH target.
21097 * config/i386/winnt.c (struct seh_frame_state): Document constraint.
21099 2021-02-28 Jakub Jelinek <jakub@redhat.com>
21102 * ipa.c (symbol_table::remove_unreachable_nodes): Fix a comment
21103 typo - referneced -> referenced.
21104 * tree.c (component_ref_size): Fix comment typo -
21105 refernce -> reference.
21106 * tree-ssa-alias.c (access_path_may_continue_p): Fix comment typo -
21107 traling -> trailing.
21108 (aliasing_component_refs_p): Fix comment typos -
21109 refernce -> reference and refernece -> reference and
21110 traling -> trailing.
21111 (nonoverlapping_refs_since_match_p): Fix comment typo -
21112 referneces -> references.
21113 * doc/invoke.texi (--param modref-max-bases): Fix a typo -
21114 referneces -> references.
21116 2021-02-27 Iain Sandoe <iain@sandoe.co.uk>
21118 * config/host-darwin.c (darwin_gt_pch_use_address): Modify
21119 diagnostic message to avoid use of a contraction and format
21122 2021-02-27 Jakub Jelinek <jakub@redhat.com>
21125 * gcse.c (gcse_or_cprop_is_too_expensive): Use %wu instead of
21126 HOST_WIDE_INT_PRINT_UNSIGNED in warning format string.
21127 * ipa-devirt.c (ipa_odr_read_section): Use %wd instead of
21128 HOST_WIDE_INT_PRINT_DEC in inform format string. Fix comment
21131 2021-02-26 Richard Biener <rguenther@suse.de>
21133 PR middle-end/99281
21134 * expr.c (store_field): For calls with return-slot optimization
21135 and addressable return type expand the store directly.
21137 2021-02-26 Richard Biener <rguenther@suse.de>
21140 * builtins.c (warn_string_no_nul): Fix diagnostic formatting.
21142 2021-02-26 Peter Bergner <bergner@linux.ibm.com>
21145 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Replace assert
21148 2021-02-26 Aaron Sawdey <acsawdey@linux.ibm.com>
21150 * config.gcc: Add rs6000-pcrel-opt.o.
21151 * config/rs6000/rs6000-pcrel-opt.c: New file.
21152 * config/rs6000/pcrel-opt.md: New file.
21153 * config/rs6000/predicates.md: Add d_form_memory predicate.
21154 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_PCREL_OPT.
21155 * config/rs6000/rs6000-passes.def: Add pass_pcrel_opt.
21156 * config/rs6000/rs6000-protos.h: Add reg_to_non_prefixed(),
21157 pcrel_opt_valid_mem_p(), output_pcrel_opt_reloc(),
21158 and make_pass_pcrel_opt().
21159 * config/rs6000/rs6000.c (reg_to_non_prefixed): Make global.
21160 (rs6000_option_override_internal): Add pcrel-opt.
21161 (rs6000_delegitimize_address): Support pcrel-opt.
21162 (rs6000_opt_masks): Add pcrel-opt.
21163 (pcrel_opt_valid_mem_p): New function.
21164 (reg_to_non_prefixed): Make global.
21165 (rs6000_asm_output_opcode): Reset prepend_p_to_next_insn.
21166 (output_pcrel_opt_reloc): New function.
21167 * config/rs6000/rs6000.md (loads_extern_addr): New attr.
21168 (pcrel_extern_addr): Set loads_extern_addr.
21169 Add include for pcrel-opt.md.
21170 * config/rs6000/rs6000.opt: Add -mpcrel-opt.
21171 * config/rs6000/t-rs6000: Add rules for pcrel-opt.c and
21174 2021-02-26 YunQiang Su <yunqiang.su@cipunited.com>
21177 * config/mips/mips.c (mips_expand_ext_as_unaligned_load):
21178 If TARGET_64BIT and dest is SUBREG, we check the width, if it
21179 equal to SImode, we use SImode operation, just like what we are
21182 2021-02-26 Marek Polacek <polacek@redhat.com>
21184 * builtins.c (warn_for_access): Fix typos.
21186 2021-02-25 Iain Sandoe <iain@sandoe.co.uk>
21188 * config/aarch64/aarch64.md (<optab>_rol<mode>3): Add a '#'
21189 mark in front of the immediate quantity.
21190 (<optab>_rolsi3_uxtw): Likewise.
21192 2021-02-25 Richard Earnshaw <rearnsha@arm.com>
21195 * config/arm/thumb2.md (nonsecure_call_reg_thumb2_fpcxt): New pattern.
21196 (nonsecure_call_value_reg_thumb2_fpcxt): Likewise.
21197 (nonsecure_call_reg_thumb2): Restrict to using r4 for the callee
21198 address and disable when the FPCXT is not available.
21199 (nonsecure_call_value_reg_thumb2): Likewise.
21201 2021-02-25 Nathan Sidwell <nathan@acm.org>
21204 * doc/invoke.texi (flang-info-module-cmi): Renamed option.
21206 2021-02-25 Tamar Christina <tamar.christina@arm.com>
21208 * tree-vect-slp.c (optimize_load_redistribution_1): Abort on NULL nodes.
21210 2021-02-25 Richard Biener <rguenther@suse.de>
21212 PR tree-optimization/99253
21213 * tree-vect-loop.c (check_reduction_path): First compute
21214 code, then verify out-of-loop uses.
21216 2021-02-25 Jakub Jelinek <jakub@redhat.com>
21219 * match.pd ((T)(A) + CST -> (T)(A + CST)): Add :s to convert.
21221 2021-02-25 Jakub Jelinek <jakub@redhat.com>
21223 PR tree-optimization/80635
21224 * tree-vrp.c (vrp_simplify_cond_using_ranges): Also handle
21225 VIEW_CONVERT_EXPR if modes are the same, innerop is integral and
21226 has mode precision.
21228 2021-02-25 Richard Biener <rguenther@suse.de>
21230 * tree-vect-slp.c (optimize_load_redistribution_1): Delay
21231 load_map population.
21232 (vect_match_slp_patterns_2): Revert part of last change.
21233 (vect_analyze_slp): Do not interleave optimize_load_redistribution
21234 with pattern detection but do it afterwards. Dump the
21235 whole SLP graph after pattern recognition and load
21236 redistribution optimization finished.
21238 2021-02-24 Jakub Jelinek <jakub@redhat.com>
21241 * omp-low.c (struct omp_context): Add teams_nested_p and
21242 nonteams_nested_p members.
21243 (scan_omp_target): Diagnose teams nested inside of target with other
21244 directives strictly nested inside of the same target.
21245 (check_omp_nesting_restrictions): Set ctx->teams_nested_p or
21246 ctx->nonteams_nested_p as needed.
21248 2021-02-24 Vladimir N. Makarov <vmakarov@redhat.com>
21250 PR inline-asm/99123
21251 * lra-constraints.c (uses_hard_regs_p): Don't use decompose_mem_address.
21253 2021-02-24 Hans-Peter Nilsson <hp@axis.com>
21255 * config/cris/cris.c (cris_expand_prologue): Set
21256 current_function_static_stack_size, if flag_stack_usage_info.
21258 2021-02-24 Pat Haugen <pthaugen@linux.ibm.com>
21260 * config/rs6000/rs6000.c (next_insn_prefixed_p): Rename.
21261 (rs6000_final_prescan_insn): Adjust.
21262 (rs6000_asm_output_opcode): Likewise.
21264 2021-02-24 Martin Sebor <msebor@redhat.com>
21266 PR middle-end/97172
21267 * attribs.c (attr_access::free_lang_data): Clear attribute arg spec
21268 from function arguments.
21270 2021-02-24 Tamar Christina <tamar.christina@arm.com>
21272 PR tree-optimization/99220
21273 * tree-vect-slp.c (optimize_load_redistribution_1): Remove
21274 node from cache when it's about to be deleted.
21276 2021-02-24 Jakub Jelinek <jakub@redhat.com>
21278 PR tree-optimization/99225
21279 * fold-const.c (fold_binary_loc) <case NE_EXPR>: In (x & (1 << y)) != 0
21280 to ((x >> y) & 1) != 0 simplifications use build_one_cst instead of
21281 build_int_cst (..., 1). Formatting fixes.
21283 2021-02-24 Tamar Christina <tamar.christina@arm.com>
21285 PR tree-optimization/99149
21286 * tree-vect-slp-patterns.c (vect_detect_pair_op): Don't recreate the
21288 (vect_slp_reset_pattern): Remove.
21289 (complex_fma_pattern::matches): Remove call to vect_slp_reset_pattern.
21290 (complex_mul_pattern::build, complex_fma_pattern::build,
21291 complex_fms_pattern::build): Fix ref counts.
21292 * tree-vect-slp.c (vect_free_slp_tree): Undo SLP only pattern relevancy
21293 when node is being deleted.
21294 (vect_match_slp_patterns_2): Correct result of cache hit on patterns.
21295 (vect_schedule_slp): Invalidate SLP_TREE_REPRESENTATIVE of removed
21297 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize value.
21299 2021-02-24 Matthias Klose <doko@ubuntu.com>
21302 2020-12-07 Matthias Klose <doko@ubuntu.com>
21304 * genextract.c (print_header): Undefine ENABLE_RTL_CHECKING
21305 and ENABLE_RTL_FLAG_CHECKING.
21307 2021-02-24 Richard Biener <rguenther@suse.de>
21310 * builtins.c (fold_builtin_next_arg): Avoid NULL arg.
21312 2021-02-23 Peter Bergner <bergner@linux.ibm.com>
21314 * config/rs6000/mma.md (mma_assemble_pair): Rename from this...
21315 (vsx_assemble_pair): ...to this.
21316 (*mma_assemble_pair): Rename from this...
21317 (*vsx_assemble_pair): ...to this.
21318 (mma_disassemble_pair): Rename from this...
21319 (vsx_disassemble_pair): ...to this.
21320 (*mma_disassemble_pair): Rename from this...
21321 (*vsx_disassemble_pair): ...to this.
21322 * config/rs6000/rs6000-builtin.def (BU_MMA_V2, BU_MMA_V3,
21323 BU_COMPAT): New macros.
21324 (mma_assemble_pair): Rename from this...
21325 (vsx_assemble_pair): ...to this.
21326 (mma_disassemble_pair): Rename from this...
21327 (vsx_disassemble_pair): ...to this.
21328 (mma_assemble_pair): New compatibility built-in.
21329 (mma_disassemble_pair): Likewise.
21330 * config/rs6000/rs6000-call.c (struct builtin_compatibility): New.
21331 (RS6000_BUILTIN_COMPAT): Define.
21332 (bdesc_compat): New.
21333 (mma_expand_builtin): Use VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
21334 (rs6000_gimple_fold_mma_builtin): Use MMA_BUILTIN_DISASSEMBLE_PAIR
21335 and VSX_BUILTIN_ASSEMBLE_PAIR.
21336 (rs6000_init_builtins): Register compatibility built-ins.
21337 (mma_init_builtins): Use VSX_BUILTIN_ASSEMBLE_PAIR,
21338 VSX_BUILTIN_ASSEMBLE_PAIR_INTERNAL, VSX_BUILTIN_DISASSEMBLE_PAIR and
21339 VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
21340 * doc/extend.texi (__builtin_mma_assemble_pair): Rename from this...
21341 (__builtin_vsx_assemble_pair): ...to this.
21342 (__builtin_mma_disassemble_pair): Rename from this...
21343 (__builtin_vsx_disassemble_pair): ...to this.
21345 2021-02-23 Martin Liska <mliska@suse.cz>
21348 * ipa-icf.c (sem_variable::merge): Do not merge 2 variables
21349 with different alignment. That leads to an invalid red zone
21350 size allocated in runtime.
21352 2021-02-23 Jakub Jelinek <jakub@redhat.com>
21354 PR tree-optimization/99204
21355 * fold-const.c (fold_read_from_constant_string): Check that
21356 tree_fits_uhwi_p (index) rather than just that index is INTEGER_CST.
21358 2021-02-23 Segher Boessenkool <segher@kernel.crashing.org>
21359 Kewen Lin <linkw@gcc.gnu.org>
21361 * config/rs6000/rs6000.md (*rotl<mode>3_insert_3): Renamed to...
21362 (rotl<mode>3_insert_3): ...this.
21363 (plus_ior_xor): New code_iterator.
21364 (define_split for GPR rl*imi): New splitter.
21365 * config/rs6000/vsx.md (vsx_init_v4si): Use gen_rotldi3_insert_3
21366 for integer merging.
21368 2021-02-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21370 * config/aarch64/aarch64-tuning-flags.def (cse_sve_vl_constants):
21372 * config/aarch64/aarch64.md (add<mode>3): Force CONST_POLY_INT immediates
21373 into a register when the above is enabled.
21374 * config/aarch64/aarch64.c (neoversev1_tunings):
21375 AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
21376 (aarch64_rtx_costs): Use AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
21378 2021-02-22 Hans-Peter Nilsson <hp@axis.com>
21380 * config/cris/cris.c (cris_print_operand) <'T'>: Change
21381 valid operand from is now an addi mult-value to shift-value.
21382 * config/cris/cris.md (*addi): Change expression of scaled
21383 operand from mult to ashift.
21384 * config/cris/cris.md (*addi_reload): New insn_and_split.
21386 2021-02-22 John David Anglin <danglin@gcc.gnu.org>
21389 * config/pa/pa.c (TARGET_ASM_CAN_OUTPUT_MI_THUNK): Define as
21390 hook_bool_const_tree_hwi_hwi_const_tree_true.
21391 (pa_asm_output_mi_thunk): Add support for nonzero vcall_offset.
21393 2021-02-22 Andre Vieira <andre.simoesdiasvieira@arm.com>
21395 PR rtl-optimization/98791
21396 * ira-conflicts.c (process_regs_for_copy): Don't create allocno copies
21397 for unordered modes.
21399 2021-02-22 Martin Liska <mliska@suse.cz>
21401 * tree-inline.c (inline_forbidden_p): Set
21402 inline_forbidden_reason.
21404 2021-02-22 Richard Biener <rguenther@suse.de>
21406 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Dump
21409 2021-02-22 Richard Biener <rguenther@suse.de>
21411 PR tree-optimization/99165
21412 * gimple-ssa-store-merging.c (pass_store_merging::process_store):
21413 Accumulate changed to ret.
21415 2021-02-21 Uros Bizjak <ubizjak@gmail.com>
21418 2020-12-09 Uroš Bizjak <ubizjak@gmail.com>
21420 * config/i386/i386.h (REG_ALLOC_ORDER): Remove
21422 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
21425 * config/s390/vector.md (trunctf<DFP_ALL:mode>2_vr): New
21427 (trunctf<DFP_ALL:mode>2): Likewise.
21428 (trunctdtf2_vr): Likewise.
21429 (trunctdtf2): Likewise.
21430 (extend<DFP_ALL:mode>tf2_vr): Likewise.
21431 (extend<DFP_ALL:mode>tf2): Likewise.
21432 (extendtftd2_vr): Likewise.
21433 (extendtftd2): Likewise.
21435 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
21437 * config/s390/vector.md (*fprx2_to_tf): Rename to fprx2_to_tf,
21438 add memory alternative.
21439 (tf_to_fprx2): New pattern.
21441 2021-02-19 Martin Sebor <msebor@redhat.com>
21444 * attribs.c (init_attr_rdwr_indices): Guard vblist use.
21445 (attr_access::free_lang_data): Remove a spurious test.
21447 2021-02-19 Nathan Sidwell <nathan@acm.org>
21449 * doc/invoke.texi (flang-info-module-read): Document.
21451 2021-02-19 Martin Liska <mliska@suse.cz>
21453 PR translation/99167
21454 * params.opt: Fix typo.
21456 2021-02-19 Richard Biener <rguenther@suse.de>
21458 PR middle-end/99122
21459 * tree-inline.c (inline_forbidden_p): Do not inline functions
21460 with VLA arguments or return value.
21462 2021-02-19 Jakub Jelinek <jakub@redhat.com>
21465 * config/arm/arm.md (*stack_protect_combined_set_insn,
21466 *stack_protect_combined_test_insn): If force_const_mem result
21467 is not valid general operand, force its address into the destination
21470 2021-02-19 Jakub Jelinek <jakub@redhat.com>
21473 * tree-cfg.c (gimple_merge_blocks): If bb a starts with eh landing
21474 pad or non-local label, put FORCED_LABELs from bb b after that label
21475 rather than before it.
21477 2021-02-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
21480 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3): Use
21481 expand_vector_broadcast' to emit the vec_duplicate operand.
21483 2021-02-18 Vladimir N. Makarov <vmakarov@redhat.com>
21485 PR rtl-optimization/96264
21486 * lra-remat.c (reg_overlap_for_remat_p): Check also output insn
21489 2021-02-18 H.J. Lu <hjl.tools@gmail.com>
21492 * varasm.c (get_section): Replace SUPPORTS_SHF_GNU_RETAIN with
21493 looking up the retain attribute.
21494 (resolve_unique_section): Likewise.
21495 (get_variable_section): Likewise.
21496 (switch_to_section): Likewise. Warn when a symbol without the
21497 retain attribute and a symbol with the retain attribute are
21498 placed in the section with the same name, instead of the used
21500 * doc/extend.texi: Document the "retain" attribute.
21502 2021-02-18 Nathan Sidwell <nathan@acm.org>
21505 * doc/invoke.texi (flang-info-include-translate): Document header
21508 2021-02-18 Richard Biener <rguenther@suse.de>
21510 PR middle-end/99122
21511 * ipa-fnsummary.c (analyze_function_body): Set
21512 CIF_FUNCTION_NOT_INLINABLE for VLA parameter calls.
21513 * tree-inline.c (insert_init_debug_bind): Pass NULL for
21514 error_mark_node values.
21515 (force_value_to_type): Do not build V_C_Es for WITH_SIZE_EXPR
21517 (setup_one_parameter): Delay force_value_to_type until when
21520 2021-02-18 Hans-Peter Nilsson <hp@axis.com>
21522 PR tree-optimization/99142
21523 * match.pd (clz cmp 0): Gate replacement on single_use of clz result.
21525 2021-02-18 Jakub Jelinek <jakub@redhat.com>
21527 * wide-int-bitmask.h (wide_int_bitmask::wide_int_bitmask (),
21528 wide_int_bitmask::wide_int_bitmask (uint64_t),
21529 wide_int_bitmask::wide_int_bitmask (uint64_t, uint64_t),
21530 wide_int_bitmask::operator ~ () const,
21531 wide_int_bitmask::operator | (wide_int_bitmask) const,
21532 wide_int_bitmask::operator & (wide_int_bitmask) const): Use constexpr
21534 * config/i386/i386.h (PTA_3DNOW, PTA_3DNOW_A, PTA_64BIT, PTA_ABM,
21535 PTA_AES, PTA_AVX, PTA_BMI, PTA_CX16, PTA_F16C, PTA_FMA, PTA_FMA4,
21536 PTA_FSGSBASE, PTA_LWP, PTA_LZCNT, PTA_MMX, PTA_MOVBE, PTA_NO_SAHF,
21537 PTA_PCLMUL, PTA_POPCNT, PTA_PREFETCH_SSE, PTA_RDRND, PTA_SSE, PTA_SSE2,
21538 PTA_SSE3, PTA_SSE4_1, PTA_SSE4_2, PTA_SSE4A, PTA_SSSE3, PTA_TBM,
21539 PTA_XOP, PTA_AVX2, PTA_BMI2, PTA_RTM, PTA_HLE, PTA_PRFCHW, PTA_RDSEED,
21540 PTA_ADX, PTA_FXSR, PTA_XSAVE, PTA_XSAVEOPT, PTA_AVX512F, PTA_AVX512ER,
21541 PTA_AVX512PF, PTA_AVX512CD, PTA_NO_TUNE, PTA_SHA, PTA_PREFETCHWT1,
21542 PTA_CLFLUSHOPT, PTA_XSAVEC, PTA_XSAVES, PTA_AVX512DQ, PTA_AVX512BW,
21543 PTA_AVX512VL, PTA_AVX512IFMA, PTA_AVX512VBMI, PTA_CLWB, PTA_MWAITX,
21544 PTA_CLZERO, PTA_NO_80387, PTA_PKU, PTA_AVX5124VNNIW, PTA_AVX5124FMAPS,
21545 PTA_AVX512VPOPCNTDQ, PTA_SGX, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES,
21546 PTA_AVX512VBMI2, PTA_VPCLMULQDQ, PTA_AVX512BITALG, PTA_RDPID,
21547 PTA_PCONFIG, PTA_WBNOINVD, PTA_AVX512VP2INTERSECT, PTA_PTWRITE,
21548 PTA_AVX512BF16, PTA_WAITPKG, PTA_MOVDIRI, PTA_MOVDIR64B, PTA_ENQCMD,
21549 PTA_CLDEMOTE, PTA_SERIALIZE, PTA_TSXLDTRK, PTA_AMX_TILE, PTA_AMX_INT8,
21550 PTA_AMX_BF16, PTA_UINTR, PTA_HRESET, PTA_KL, PTA_WIDEKL, PTA_AVXVNNI,
21551 PTA_X86_64_BASELINE, PTA_X86_64_V2, PTA_X86_64_V3, PTA_X86_64_V4,
21552 PTA_CORE2, PTA_NEHALEM, PTA_WESTMERE, PTA_SANDYBRIDGE, PTA_IVYBRIDGE,
21553 PTA_HASWELL, PTA_BROADWELL, PTA_SKYLAKE, PTA_SKYLAKE_AVX512,
21554 PTA_CASCADELAKE, PTA_COOPERLAKE, PTA_CANNONLAKE, PTA_ICELAKE_CLIENT,
21555 PTA_ICELAKE_SERVER, PTA_TIGERLAKE, PTA_SAPPHIRERAPIDS, PTA_ALDERLAKE,
21556 PTA_KNL, PTA_BONNELL, PTA_SILVERMONT, PTA_GOLDMONT, PTA_GOLDMONT_PLUS,
21557 PTA_TREMONT, PTA_KNM): Use constexpr instead of const.
21559 2021-02-18 Jakub Jelinek <jakub@redhat.com>
21561 PR middle-end/99109
21562 * gimple-array-bounds.cc (build_zero_elt_array_type): Rename to ...
21563 (build_printable_array_type): ... this. Add nelts argument. For
21564 overaligned eltype, use TYPE_MAIN_VARIANT (eltype) instead. If
21565 nelts, call build_array_type_nelts.
21566 (array_bounds_checker::check_mem_ref): Use build_printable_array_type
21567 instead of build_zero_elt_array_type and build_array_type_nelts.
21569 2021-02-18 Jakub Jelinek <jakub@redhat.com>
21572 * config/i386/i386.c (distance_non_agu_define): Don't call
21573 extract_insn_cached here.
21574 (ix86_lea_outperforms): Save and restore recog_data around call
21575 to distance_non_agu_define and distance_agu_use.
21576 (ix86_ok_to_clobber_flags): Remove.
21577 (ix86_avoid_lea_for_add): Don't call ix86_ok_to_clobber_flags.
21578 (ix86_avoid_lea_for_addr): Likewise. Adjust function comment.
21579 * config/i386/i386.md (*lea<mode>): Change from define_insn_and_split
21580 into define_insn. Move the splitting to define_peephole2 and
21581 check there using peep2_regno_dead_p if FLAGS_REG is dead.
21583 2021-02-17 Julian Brown <julian@codesourcery.com>
21585 * gimplify.c (gimplify_scan_omp_clauses): Handle ATTACH_DETACH
21588 2021-02-17 Xi Ruoyao <xry111@mengyan1223.wang>
21591 * config/mips/mips.c (mips_symbol_insns): Do not use
21592 MSA_SUPPORTED_MODE_P if mode is MAX_MACHINE_MODE.
21594 2021-02-16 Vladimir N. Makarov <vmakarov@redhat.com>
21596 PR inline-asm/98096
21597 * stmt.c (resolve_operand_name_1): Take inout operands into account
21598 for access to labels by names.
21599 * doc/extend.texi: Describe counting operands for accessing labels.
21601 2021-02-16 Richard Biener <rguenther@suse.de>
21603 PR tree-optimization/38474
21604 * tree-ssa-structalias.c (variable_info::address_taken): New.
21605 (new_var_info): Initialize address_taken.
21606 (process_constraint): Set address_taken.
21607 (solve_constraints): Use the new address_taken flag rather
21608 than is_reg_var for sorting variables.
21609 (dump_constraint): Dump the variable number if the name
21612 2021-02-16 Jakub Jelinek <jakub@redhat.com>
21615 * tree-vect-stmts.c (vectorizable_simd_clone_call): For num_calls != 1
21616 multiply by 4096 and for inbranch by 8192.
21617 * config/i386/i386.c (ix86_simd_clone_usable): For TARGET_AVX512F,
21618 return 3, 2 or 1 for mangle letters 'b', 'c' or 'd'.
21620 2021-02-15 Maya Rashish <coypu@sdf.org>
21622 * config/aarch64/aarch64.c (aarch64_init_builtins):
21623 Call SUBTARGET_INIT_BUILTINS.
21625 2021-02-15 Peter Bergner <bergner@linux.ibm.com>
21627 PR rtl-optimization/98872
21628 * init-regs.c (initialize_uninitialized_regs): Skip initialization
21629 if CONST0_RTX is NULL.
21631 2021-02-15 Richard Sandiford <richard.sandiford@arm.com>
21633 PR rtl-optimization/98863
21634 * rtl-ssa/functions.h (function_info::bb_live_out_info): Delete.
21635 (function_info::build_info): Turn into a declaration, moving the
21636 definition to internals.h.
21637 (function_info::bb_walker): Declare.
21638 (function_info::create_reg_use): Likewise.
21639 (function_info::calculate_potential_phi_regs): Take a build_info
21641 (function_info::place_phis, function_info::create_ebbs): Declare.
21642 (function_info::calculate_ebb_live_in_for_debug): Likewise.
21643 (function_info::populate_backedge_phis): Delete.
21644 (function_info::start_block, function_info::end_block): Declare.
21645 (function_info::populate_phi_inputs): Delete.
21646 (function_info::m_potential_phi_regs): Move information to build_info.
21647 * rtl-ssa/internals.h: New file.
21648 (function_info::bb_phi_info): New class.
21649 (function_info::build_info): Moved from functions.h.
21650 Add a constructor and destructor.
21651 (function_info::build_info::ebb_use): Delete.
21652 (function_info::build_info::ebb_def): Likewise.
21653 (function_info::build_info::bb_live_out): Likewise.
21654 (function_info::build_info::tmp_ebb_live_in_for_debug): New variable.
21655 (function_info::build_info::potential_phi_regs): Likewise.
21656 (function_info::build_info::potential_phi_regs_for_debug): Likewise.
21657 (function_info::build_info::ebb_def_regs): Likewise.
21658 (function_info::build_info::bb_phis): Likewise.
21659 (function_info::build_info::bb_mem_live_out): Likewise.
21660 (function_info::build_info::bb_to_rpo): Likewise.
21661 (function_info::build_info::def_stack): Likewise.
21662 (function_info::build_info::old_def_stack_limit): Likewise.
21663 * rtl-ssa/internals.inl (function_info::build_info::record_reg_def):
21664 Remove the regno argument. Push the previous definition onto the
21665 definition stack where necessary.
21666 * rtl-ssa/accesses.cc: Include internals.h.
21667 * rtl-ssa/changes.cc: Likewise.
21668 * rtl-ssa/blocks.cc: Likewise.
21669 (function_info::build_info::build_info): Define.
21670 (function_info::build_info::~build_info): Likewise.
21671 (function_info::bb_walker): New class.
21672 (function_info::bb_walker::bb_walker): Define.
21673 (function_info::add_live_out_use): Convert a logarithmic-complexity
21674 test into a linear one. Allow the same definition to be passed
21676 (function_info::calculate_potential_phi_regs): Moved from
21677 functions.cc. Take a build_info parameter and store the
21678 information there instead.
21679 (function_info::place_phis): New function.
21680 (function_info::add_entry_block_defs): Update call to record_reg_def.
21681 (function_info::calculate_ebb_live_in_for_debug): New function.
21682 (function_info::add_phi_nodes): Use bb_phis to decide which
21683 registers need phi nodes and initialize ebb_def_regs accordingly.
21684 Do not add degenerate phis here.
21685 (function_info::add_artificial_accesses): Use create_reg_use.
21686 Assert that all definitions are listed in the DF LR sets.
21687 Update call to record_reg_def.
21688 (function_info::record_block_live_out): Record live-out register
21689 values in the phis of successor blocks. Use the live-out set
21690 when processing the last block in an EBB, instead of always
21691 using the live-in sets of successor blocks. AND the live sets
21692 with the set of registers that have been defined in the EBB,
21693 rather than with all potential phi registers. Cope correctly
21694 with branches back to the start of the current EBB.
21695 (function_info::start_block): New function.
21696 (function_info::end_block): Likewise.
21697 (function_info::populate_phi_inputs): Likewise.
21698 (function_info::create_ebbs): Likewise.
21699 (function_info::process_all_blocks): Rewrite into a multi-phase
21701 * rtl-ssa/functions.cc: Include internals.h.
21702 (function_info::calculate_potential_phi_regs): Move to blocks.cc.
21703 (function_info::init_function_data): Remove caller.
21704 * rtl-ssa/insns.cc: Include internals.h
21705 (function_info::create_reg_use): New function. Lazily any
21706 degenerate phis needed by the linear RPO view.
21707 (function_info::record_use): Use create_reg_use. When processing
21708 debug uses, use potential_phi_regs and test it before checking
21709 whether the register is live on entry to the current EBB. Lazily
21710 calculate ebb_live_in_for_debug.
21711 (function_info::record_call_clobbers): Update call to record_reg_def.
21712 (function_info::record_def): Likewise.
21714 2021-02-15 Martin Liska <mliska@suse.cz>
21716 * toplev.c (init_asm_output): Free output of
21717 gen_command_line_string function.
21718 (process_options): Likewise.
21720 2021-02-15 Martin Liska <mliska@suse.cz>
21722 * params.opt: Add 2 missing Param keywords.
21724 2021-02-15 Eric Botcazou <ebotcazou@adacore.com>
21726 * df-core.c (df_worklist_dataflow_doublequeue): Use proper cast.
21728 2021-02-15 Jakub Jelinek <jakub@redhat.com>
21730 PR tree-optimization/99079
21731 * match.pd (A % (pow2pcst << N) -> A & ((pow2pcst << N) - 1)): Remove
21732 useless tree_nop_conversion_p (type, TREE_TYPE (@3)) check. Instead
21733 require both type and TREE_TYPE (@1) to be integral types and either
21734 type having smaller or equal precision, or TREE_TYPE (@1) being
21735 unsigned type, or type being signed type. If TREE_TYPE (@1)
21736 doesn't have wrapping overflow, perform the subtraction of one in
21739 2021-02-14 Jan Hubicka <hubicka@ucw.cz>
21740 Richard Biener <rguether@suse.de>
21743 * ipa-reference.c (ipa_init): Only conditinally initialize
21744 reference_vars_to_consider.
21745 (propagate): Conditionally deninitialize reference_vars_to_consider.
21746 (ipa_reference_write_optimization_summary): Sanity check that
21747 reference_vars_to_consider is not allocated.
21749 2021-02-13 Levy Hsu <admin@levyhsu.com>
21752 * config/riscv/riscv-shorten-memrefs.c (pass_shorten_memrefs): Add
21753 extend parameter to get_si_mem_base_reg declaration.
21754 (get_si_mem_base_reg): Add extend parameter. Set it.
21755 (analyze): Pass extend arg to get_si_mem_base_reg.
21756 (transform): Likewise. Use it when rewriting mems.
21757 * config/riscv/riscv.c (riscv_legitimize_move): Check for subword
21758 loads and emit sign/zero extending load followed by subreg move.
21760 2021-02-13 Jim Wilson <jimw@sifive.com>
21763 * config/riscv/riscv.c (riscv_compressed_lw_address_p): Drop early
21764 exit when !reload_completed. Only perform check for compressed reg
21765 if reload_completed.
21766 (riscv_rtx_costs): In MEM case, when optimizing for size and
21767 shorten memrefs, if not compressible, then increase cost.
21769 2021-02-13 Jakub Jelinek <jakub@redhat.com>
21771 PR rtl-optimization/98439
21772 * recog.c (pass_split_before_regstack::gate): Enable even when
21773 pass_split_before_sched2 is enabled if -fselective-scheduling2 is
21776 2021-02-13 Jakub Jelinek <jakub@redhat.com>
21779 * config/i386/mmx.md (*mmx_pshufd_1): Add a combine splitter for
21780 swap of V2SImode elements in memory into DImode memory rotate by 32.
21782 2021-02-12 Martin Sebor <msebor@redhat.com>
21784 * tree-pretty-print.c (print_generic_expr_to_str): Update comment.
21786 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
21788 * rtl-ssa/accesses.cc (function_info::make_use_available): Use
21789 m_temp_obstack rather than m_obstack to allocate the temporary use.
21791 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
21793 * df-problems.c (df_lr_bb_local_compute): Treat partial definitions
21794 as read-modify operations.
21796 2021-02-12 Richard Biener <rguenther@suse.de>
21798 PR middle-end/38474
21799 * ipa-fnsummary.c (unmodified_parm_1): Only walk when
21800 fbi->aa_walk_budget is bigger than zero. Update
21801 fbi->aa_walk_budget.
21802 (param_change_prob): Likewise.
21803 * ipa-prop.c (detect_type_change_from_memory_writes):
21804 Properly account walk_aliased_vdefs.
21805 (parm_preserved_before_stmt_p): Canonicalize updates.
21806 (parm_ref_data_preserved_p): Likewise.
21807 (parm_ref_data_pass_through_p): Likewise.
21808 (determine_known_aggregate_parts): Account own alias queries.
21810 2021-02-12 Martin Liska <mliska@suse.cz>
21812 * opts-common.c (decode_cmdline_option): Release werror_arg.
21813 * opts.c (gen_producer_string): Release output of
21814 gen_command_line_string.
21816 2021-02-12 Richard Biener <rguenther@suse.de>
21818 PR tree-optimization/38474
21819 * params.opt (-param=max-store-chains-to-track=): New param.
21820 (-param=max-stores-to-track=): Likewise.
21821 * doc/invoke.texi (max-store-chains-to-track): Document.
21822 (max-stores-to-track): Likewise.
21823 * gimple-ssa-store-merging.c (pass_store_merging::m_n_chains):
21825 (pass_store_merging::m_n_stores): Likewise.
21826 (pass_store_merging::terminate_and_process_chain): Update
21827 m_n_stores and m_n_chains.
21828 (pass_store_merging::process_store): Likewise. Terminate
21829 oldest chains if the number of stores or chains get too large.
21830 (imm_store_chain_info::terminate_and_process_chain): Dump
21833 2021-02-11 Eric Botcazou <ebotcazou@adacore.com>
21835 * config/i386/winnt.c (i386_pe_seh_unwind_emit): When switching to
21836 the cold section, emit a nop before the directive if the previous
21837 active instruction can throw.
21839 2021-02-11 Peter Bergner <bergner@linux.ibm.com>
21842 * config/rs6000/predicates.md (mma_assemble_input_operand): Restrict
21843 memory addresses that are legal for quad word accesses.
21845 2021-02-11 Andrea Corallo <andrea.corallo@arm.com>
21848 * config/arm/thumb2.md (*doloop_end_internal): Generate
21849 alternative sequence to handle long range branches.
21851 2021-02-11 Joel Hutton <joel.hutton@arm.com>
21853 PR tree-optimization/98772
21854 * optabs-tree.c (supportable_half_widening_operation): New function
21855 to check for supportable V8QI->V8HI widening patterns.
21856 * optabs-tree.h (supportable_half_widening_operation): New function.
21857 * tree-vect-stmts.c (vect_create_half_widening_stmts): New function
21858 to create promotion stmts for V8QI->V8HI widening patterns.
21859 (vectorizable_conversion): Add case for V8QI->V8HI.
21861 2021-02-11 Richard Biener <rguenther@suse.de>
21863 * sparseset.h (SPARSESET_ELT_BITS): Remove.
21864 (SPARSESET_ELT_TYPE): Use unsigned int.
21865 * fwprop.c: Do not include sparseset.h.
21867 2021-02-10 Jakub Jelinek <jakub@redhat.com>
21870 * varasm.c (declare_weak): For -fsyntax-only, allow even
21871 TREE_ASM_WRITTEN function decls.
21873 2021-02-10 Jakub Jelinek <jakub@redhat.com>
21876 * config/i386/sse.md (fix<fixunssuffix>_truncv2sfv2di2,
21877 <insn>v8qiv8hi2, <insn>v8qiv8si2, <insn>v4qiv4si2, <insn>v4hiv4si2,
21878 <insn>v8qiv8di2, <insn>v4qiv4di2, <insn>v2qiv2di2, <insn>v4hiv4di2,
21879 <insn>v2hiv2di2, <insn>v2siv2di2): Force operands[1] into REG before
21880 calling simplify_gen_subreg on it.
21882 2021-02-10 Martin Liska <mliska@suse.cz>
21884 * config/nvptx/nvptx.c (nvptx_option_override): Use
21885 flag_patchable_function_entry instead of the removed
21886 function_entry_patch_area_size.
21888 2021-02-10 Martin Liska <mliska@suse.cz>
21890 PR tree-optimization/99002
21891 PR tree-optimization/99026
21892 * gimple-if-to-switch.cc (if_chain::is_beneficial): Fix memory
21893 leak when adjacent cases are merged.
21894 * tree-switch-conversion.c (switch_decision_tree::analyze_switch_statement): Use
21896 (make_pass_lower_switch): Remove trailing whitespace.
21897 * tree-switch-conversion.h (release_clusters): New.
21899 2021-02-10 Richard Biener <rguenther@suse.de>
21901 PR rtl-optimization/99054
21902 * cfgrtl.c (rtl-optimization/99054): Return an auto_vec.
21903 (fixup_partitions): Adjust.
21904 (rtl_verify_edges): Likewise.
21906 2021-02-10 Jakub Jelinek <jakub@redhat.com>
21908 PR middle-end/99007
21909 * gimplify.c (gimplify_scan_omp_clauses): For MEM_REF on reductions,
21910 temporarily disable gimplify_ctxp->into_ssa around gimplify_expr
21913 2021-02-10 Richard Biener <rguenther@suse.de>
21916 * ipa-pure-const.c (propagate_malloc): Use an auto_vec<>
21919 2021-02-10 Richard Biener <rguenther@suse.de>
21921 PR tree-optimization/99024
21922 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Only
21923 clear loop->aux if it is associated with the destroyed loop_vinfo.
21925 2021-02-10 Martin Liska <mliska@suse.cz>
21927 PR tree-optimization/99002
21928 * gimple-if-to-switch.cc (find_conditions): Fix memory leak
21931 2021-02-10 Martin Liska <mliska@suse.cz>
21934 * ipa-icf.c (sem_item::add_reference): Fix memory leak when
21935 a reference exists.
21937 2021-02-10 Jakub Jelinek <jakub@redhat.com>
21940 * dwarf2out.c (prune_unused_types_walk): Mark DW_TAG_variable DIEs
21941 at class scope for DWARF5+.
21943 2021-02-09 Eric Botcazou <ebotcazou@adacore.com>
21945 PR rtl-optimization/96015
21946 * reorg.c (skip_consecutive_labels): Minor comment tweaks.
21947 (relax_delay_slots): When deleting a jump to the next active
21948 instruction over a barrier, first delete the barrier if the
21949 jump is the only way to reach the target label.
21951 2021-02-09 Andre Vieira <andre.simoesdiasvieira@arm.com>
21953 * config/aarch64/aarch64-cost-tables.h: Add entries for vect.mul.
21954 * config/aarch64/aarch64.c (aarch64_rtx_mult_cost): Use vect.mul for
21955 vector multiplies and vect.alu for SSRA.
21956 * config/arm/aarch-common-protos.h (struct vector_cost_table): Define
21957 vect.mul cost field.
21958 * config/arm/aarch-cost-tables.h: Add entries for vect.mul.
21959 * config/arm/arm.c: Likewise.
21961 2021-02-09 Richard Biener <rguenther@suse.de>
21963 PR tree-optimization/98863
21964 * tree-ssa-sccvn.h (vn_avail::next_undo): Add.
21965 * tree-ssa-sccvn.c (last_pushed_avail): New global.
21966 (rpo_elim::eliminate_push_avail): Chain pushed avails.
21967 (unwind_state::avail_top): Add.
21968 (do_unwind): Rewrite unwinding of avail entries.
21969 (do_rpo_vn): Initialize last_pushed_avail and
21970 avail_top of the undo state.
21972 2021-02-09 Jakub Jelinek <jakub@redhat.com>
21974 PR middle-end/99004
21975 * calls.c (maybe_warn_rdwr_sizes): Change s0 and s1 type from
21976 const char * to char * and free those pointers after use.
21978 2021-02-09 Richard Biener <rguenther@suse.de>
21980 PR tree-optimization/99017
21981 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Allow
21982 zero vector cost entries.
21984 2021-02-08 Andre Vieira <andre.simoesdiasvieira@arm.com>
21986 PR middle-end/98974
21987 * tree-vect-stmts.c (vectorizable_condition): Remove shadow vec_num
21988 parameter in vectorizable_condition.
21990 2021-02-08 Richard Biener <rguenther@suse.de>
21993 * tree.c (walk_tree_1): Walk VECTOR_CST elements.
21995 2021-02-08 Martin Liska <mliska@suse.cz>
21998 * cfgexpand.c (pass_expand::execute): Parse per-function option
21999 flag_patchable_function_entry and use it.
22000 * common.opt: Remove function_entry_patch_area_size and
22001 function_entry_patch_area_start global variables.
22002 * opts.c (parse_and_check_patch_area): New function.
22003 (common_handle_option): Use it.
22004 * opts.h (parse_and_check_patch_area): New function.
22005 * toplev.c (process_options): Parse and use
22006 function_entry_patch_area_size.
22008 2021-02-08 Martin Sebor <msebor@redhat.com>
22010 * doc/extend.texi (attribute malloc): Correct typos.
22012 2021-02-05 Nathan Sidwell <nathan@acm.org>
22015 * gcc.c (driver::maybe_run_linker): Check for input file
22016 accessibility if not linking.
22018 2021-02-05 Richard Biener <rguenther@suse.de>
22020 PR tree-optimization/98855
22021 * tree-vectorizer.h (add_stmt_cost): New overload.
22022 * tree-vect-slp.c (li_cost_vec_cmp): New.
22023 (vect_bb_slp_scalar_cost): Cost individual loop regions
22024 separately. Account for the scalar instance root stmt.
22026 2021-02-05 Tom de Vries <tdevries@suse.de>
22029 * tree-switch-conversion.c (jump_table_cluster::emit): Add loc
22031 (bit_test_cluster::emit): Reuse location_t for newly created
22033 (switch_decision_tree::try_switch_expansion): Preserve
22035 * tree-switch-conversion.h: Change function signatures.
22037 2021-02-05 Jakub Jelinek <jakub@redhat.com>
22040 * config/i386/i386-options.c (m_NONE, m_ALL): Define.
22041 * config/i386/x86-tune.def (X86_TUNE_BRANCH_PREDICTION_HINTS,
22042 X86_TUNE_PROMOTE_QI_REGS): Use m_NONE instead of 0U.
22043 (X86_TUNE_QIMODE_MATH): Use m_ALL instead of ~0U.
22045 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22047 * config/aarch64/aarch64-simd-builtins.def (get_high): Define builtin.
22048 * config/aarch64/aarch64-simd.md (aarch64_get_high<mode>): Define.
22049 * config/aarch64/arm_neon.h (__GET_HIGH): Delete.
22050 (vget_high_f16): Reimplement using new builtin.
22051 (vget_high_f32): Likewise.
22052 (vget_high_f64): Likewise.
22053 (vget_high_p8): Likewise.
22054 (vget_high_p16): Likewise.
22055 (vget_high_p64): Likewise.
22056 (vget_high_s8): Likewise.
22057 (vget_high_s16): Likewise.
22058 (vget_high_s32): Likewise.
22059 (vget_high_s64): Likewise.
22060 (vget_high_u8): Likewise.
22061 (vget_high_u16): Likewise.
22062 (vget_high_u32): Likewise.
22063 (vget_high_u64): Likewise.
22065 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22067 * config/aarch64/aarch64-simd-builtins.def (get_low): Define builtin.
22068 * config/aarch64/aarch64-simd.md (aarch64_get_low<mode>): Define.
22069 * config/aarch64/arm_neon.h (__GET_LOW): Delete.
22070 (vget_low_f16): Reimplement using new builtin.
22071 (vget_low_f32): Likewise.
22072 (vget_low_f64): Likewise.
22073 (vget_low_p8): Likewise.
22074 (vget_low_p16): Likewise.
22075 (vget_low_p64): Likewise.
22076 (vget_low_s8): Likewise.
22077 (vget_low_s16): Likewise.
22078 (vget_low_s32): Likewise.
22079 (vget_low_s64): Likewise.
22080 (vget_low_u8): Likewise.
22081 (vget_low_u16): Likewise.
22082 (vget_low_u32): Likewise.
22083 (vget_low_u64): Likewise.
22085 2021-02-05 Kito Cheng <kito.cheng@sifive.com>
22087 * gcc.c (print_multilib_info): Check all required argument is provided
22090 2021-02-05 liuhongt <hongtao.liu@intel.com>
22093 * config/i386/i386-expand.c (ix86_expand_sse_cmp): Don't
22094 generate integer mask comparison for 128/256-bits vector when
22095 op_true/op_false is NULL_RTX or CONSTM1_RTX/CONST0_RTX. Also
22096 delete redundant !maskcmp condition.
22097 (ix86_expand_int_vec_cmp): Ditto but no redundant deletion
22099 (ix86_expand_sse_movcc): Delete definition of maskcmp, add the
22100 condition directly to if (maskcmp), add extra check for
22101 cmpmode, it should be MODE_INT.
22102 (ix86_expand_fp_vec_cmp): Pass NULL to ix86_expand_sse_cmp's
22103 parameters op_true/op_false.
22104 (ix86_use_mask_cmp_p): New.
22106 2021-02-05 liuhongt <hongtao.liu@intel.com>
22109 * config/i386/x86-tune.def (X86_TUNE_AVX256_UNALIGNED_LOAD_OPTIMAL):
22110 Remove m_GENERIC from ~list.
22111 (X86_TUNE_AVX256_UNALIGNED_STORE_OPTIMAL): Ditto.
22113 2021-02-04 David Malcolm <dmalcolm@redhat.com>
22116 * diagnostic-show-locus.c (compatible_locations_p): Require
22117 locations in the same macro map to be either both from the
22118 macro definition, or both from the macro arguments.
22120 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
22122 * config/aarch64/aarch64-simd-builtins.def: Add
22123 [su]mull_hi_lane[q] builtin generator macros.
22124 * config/aarch64/aarch64-simd.md
22125 (aarch64_<su>mull_hi_lane<mode>_insn): Define.
22126 (aarch64_<su>mull_hi_lane<mode>): Define.
22127 (aarch64_<su>mull_hi_laneq<mode>_insn): Define.
22128 (aarch64_<su>mull_hi_laneq<mode>): Define.
22129 * config/aarch64/arm_neon.h (vmull_high_lane_s16): Use RTL
22130 builtin instead of inline asm.
22131 (vmull_high_lane_s32): Likewise.
22132 (vmull_high_lane_u16): Likewise.
22133 (vmull_high_lane_u32): Likewise.
22134 (vmull_high_laneq_s16): Likewise.
22135 (vmull_high_laneq_s32): Likewise.
22136 (vmull_high_laneq_u16): Likewise.
22137 (vmull_high_laneq_u32): Liekwise.
22139 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
22141 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_hi_n
22142 builtin generator macros.
22143 * config/aarch64/aarch64-simd.md
22144 (aarch64_<su>mull_hi_n<mode>_insn): Define.
22145 (aarch64_<su>mull_hi_n<mode>): Define.
22146 * config/aarch64/arm_neon.h (vmull_high_n_s16): Use RTL builtin
22147 instead of inline asm.
22148 (vmull_high_n_s32): Likewise.
22149 (vmull_high_n_u16): Likewise.
22150 (vmull_high_n_u32): Likewise.
22152 2021-02-04 Richard Biener <rguenther@suse.de>
22154 PR tree-optimization/98855
22155 * tree-vect-loop.c (vectorizable_phi): Do not cost
22156 single-argument PHIs.
22157 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Likewise.
22158 * tree-vect-stmts.c (vectorizable_bswap): Also perform
22159 costing for SLP operation.
22161 2021-02-04 Martin Liska <mliska@suse.cz>
22163 * doc/extend.texi: Mention -mprefer-vector-width in target
22166 2021-02-03 Martin Sebor <msebor@redhat.com>
22168 PR tree-optimization/98937
22169 * tree-ssa-strlen.c (strlen_dom_walker::~strlen_dom_walker): Define.
22170 Flush pointer_query cache.
22172 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
22174 * config/rs6000/genfusion.pl (gen_2logical): Add missing
22175 fixes based on patch review.
22176 * config/rs6000/fusion.md: Regenerate file.
22178 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
22180 * config/rs6000/t-rs6000: Comment out auto generation of
22183 2021-02-03 Andrew Stubbs <ams@codesourcery.com>
22185 * config/gcn/gcn-opts.h (enum processor_type): Add PROCESSOR_GFX908.
22186 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Add gfx908.
22187 (output_file_start): Add gfx908.
22188 * config/gcn/gcn.opt (gpu_type): Add gfx908.
22189 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Add march=gfx908.
22190 (MULTILIB_DIRNAMES): Add gfx908.
22191 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): New define.
22192 (main): Recognize gfx908.
22193 * config/gcn/t-omp-device: Add gfx908.
22195 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
22197 * config/aarch64/aarch64-simd-builtins.def: Add
22198 [su]mlsl_hi_lane[q] builtin macro generators.
22199 * config/aarch64/aarch64-simd.md
22200 (aarch64_<su>mlsl_hi_lane<mode>_insn): Define.
22201 (aarch64_<su>mlsl_hi_lane<mode>): Define.
22202 (aarch64_<su>mlsl_hi_laneq<mode>_insn): Define.
22203 (aarch64_<su>mlsl_hi_laneq<mode>): Define.
22204 * config/aarch64/arm_neon.h (vmlsl_high_lane_s16): Use RTL
22205 builtin instead of inline asm.
22206 (vmlsl_high_lane_s32): Likewise.
22207 (vmlsl_high_lane_u16): Likewise.
22208 (vmlsl_high_lane_u32): Likewise.
22209 (vmlsl_high_laneq_s16): Likewise.
22210 (vmlsl_high_laneq_s32): Likewise.
22211 (vmlsl_high_laneq_u16): Likewise.
22212 (vmlsl_high_laneq_u32): Likewise.
22213 (vmlal_high_laneq_u32): Likewise.
22215 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
22217 * config/aarch64/aarch64-simd-builtins.def: Add
22218 [su]mlal_hi_lane[q] builtin generator macros.
22219 * config/aarch64/aarch64-simd.md
22220 (aarch64_<su>mlal_hi_lane<mode>_insn): Define.
22221 (aarch64_<su>mlal_hi_lane<mode>): Define.
22222 (aarch64_<su>mlal_hi_laneq<mode>_insn): Define.
22223 (aarch64_<su>mlal_hi_laneq<mode>): Define.
22224 * config/aarch64/arm_neon.h (vmlal_high_lane_s16): Use RTL
22225 builtin instead of inline asm.
22226 (vmlal_high_lane_s32): Likewise.
22227 (vmlal_high_lane_u16): Likewise.
22228 (vmlal_high_lane_u32): Likewise.
22229 (vmlal_high_laneq_s16): Likewise.
22230 (vmlal_high_laneq_s32): Likewise.
22231 (vmlal_high_laneq_u16): Likewise.
22232 (vmlal_high_laneq_u32): Likewise.
22234 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
22236 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_hi_n
22237 builtin generator macros.
22238 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_hi_n<mode>_insn):
22240 (aarch64_<su>mlsl_hi_n<mode>): Define.
22241 * config/aarch64/arm_neon.h (vmlsl_high_n_s16): Use RTL builtin
22242 instead of inline asm.
22243 (vmlsl_high_n_s32): Likewise.
22244 (vmlsl_high_n_u16): Likewise.
22245 (vmlsl_high_n_u32): Likewise.
22247 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
22249 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_hi_n
22250 builtin generator macros.
22251 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_hi_n<mode>_insn):
22253 (aarch64_<su>mlal_hi_n<mode>): Define.
22254 * config/aarch64/arm_neon.h (vmlal_high_n_s16): Use RTL builtin
22255 instead of inline asm.
22256 (vmlal_high_n_s32): Likewise.
22257 (vmlal_high_n_u16): Likewise.
22258 (vmlal_high_n_u32): Likewise.
22260 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
22262 * config/aarch64/aarch64-simd-builtins.def: Add RTL builtin
22264 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal_hi<mode>):
22266 (aarch64_<su>mlal_hi<mode>_insn): This.
22267 (aarch64_<su>mlal_hi<mode>): Define.
22268 * config/aarch64/arm_neon.h (vmlal_high_s8): Use RTL builtin
22269 instead of inline asm.
22270 (vmlal_high_s16): Likewise.
22271 (vmlal_high_s32): Likewise.
22272 (vmlal_high_u8): Likewise.
22273 (vmlal_high_u16): Likewise.
22274 (vmlal_high_u32): Likewise.
22276 2021-02-03 Ilya Leoshkevich <iii@linux.ibm.com>
22278 * lra-spills.c (remove_pseudos): Call lra_update_insn_recog_data()
22279 after calling alter_subreg() on a (mem).
22281 2021-02-03 Martin Liska <mliska@suse.cz>
22284 * lto-streamer-out.c (produce_lto_section): Fill up missing
22286 * lto-streamer.h (struct lto_section): Add _padding field.
22288 2021-02-03 Richard Biener <rguenther@suse.de>
22290 * lto-streamer.c (lto_get_section_name): Free temporary
22292 * tree-loop-distribution.c
22293 (loop_distribution::merge_dep_scc_partitions): Free edge data.
22295 2021-02-03 Jakub Jelinek <jakub@redhat.com>
22297 PR middle-end/97487
22298 * ifcvt.c (noce_can_force_operand): New function.
22299 (noce_emit_move_insn): Use it.
22300 (noce_try_sign_mask): Likewise. Formatting fix.
22302 2021-02-03 Jakub Jelinek <jakub@redhat.com>
22304 PR middle-end/97971
22305 * lra-constraints.c (process_alt_operands): For inline asm, don't call
22306 fatal_insn, but instead return false.
22308 2021-02-03 Jakub Jelinek <jakub@redhat.com>
22310 PR tree-optimization/98287
22311 * config/i386/mmx.md (<insn><mode>3): For shifts don't enable expander
22314 2021-02-03 Tamar Christina <tamar.christina@arm.com>
22316 PR tree-optimization/98928
22317 * tree-vect-loop.c (vect_analyze_loop_2): Change
22318 STMT_VINFO_SLP_VECT_ONLY to STMT_VINFO_SLP_VECT_ONLY_PATTERN.
22319 * tree-vect-slp-patterns.c (complex_pattern::build): Likewise.
22320 * tree-vectorizer.h (STMT_VINFO_SLP_VECT_ONLY_PATTERN): New.
22321 (class _stmt_vec_info): Add slp_vect_pattern_only_p.
22323 2021-02-02 Richard Biener <rguenther@suse.de>
22325 * gimple-loop-interchange.cc (prepare_data_references):
22327 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
22328 * tree-ssa-loop-im.c (hoist_memory_references): Likewise.
22329 * tree-vect-stmts.c (vectorizable_condition): Do not
22331 (vectorizable_comparison): Likewise.
22333 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22335 * config/aarch64/aarch64-simd-builtins.def (ursqrte): Define builtin.
22336 * config/aarch64/aarch64-simd.md (aarch64_ursqrte<mode>): New pattern.
22337 * config/aarch64/arm_neon.h (vrsqrte_u32): Reimplement using builtin.
22338 (vrsqrteq_u32): Likewise.
22340 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22342 * config/aarch64/aarch64-simd-builtins.def (sqxtun2): Define builtin.
22343 * config/aarch64/aarch64-simd.md (aarch64_sqxtun2<mode>_le): Define.
22344 (aarch64_sqxtun2<mode>_be): Likewise.
22345 (aarch64_sqxtun2<mode>): Likewise.
22346 * config/aarch64/arm_neon.h (vqmovun_high_s16): Reimplement using builtin.
22347 (vqmovun_high_s32): Likewise.
22348 (vqmovun_high_s64): Likewise.
22349 * config/aarch64/iterators.md (UNSPEC_SQXTUN2): Define.
22351 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22353 * config/aarch64/aarch64-simd-builtins.def (bfdot_lane, bfdot_laneq): Use
22355 (bfmlalb_lane, bfmlalt_lane, bfmlalb_lane_q, bfmlalt_lane_q): Use FP flags.
22357 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22359 * config/aarch64/aarch64-simd-builtins.def (fcmla_lane0, fcmla_lane90,
22360 fcmla_lane180, fcmla_lane270, fcmlaq_lane0, fcmlaq_lane90, fcmlaq_lane180,
22361 fcmlaq_lane270, scvtf, ucvtf, fcvtzs, fcvtzu, scvtfsi, scvtfdi, ucvtfsi,
22362 ucvtfdi, fcvtzshf, fcvtzuhf, fmlal_lane_low, fmlsl_lane_low,
22363 fmlal_laneq_low, fmlsl_laneq_low, fmlalq_lane_low, fmlslq_lane_low,
22364 fmlalq_laneq_low, fmlslq_laneq_low, fmlal_lane_high, fmlsl_lane_high,
22365 fmlal_laneq_high, fmlsl_laneq_high, fmlalq_lane_high, fmlslq_lane_high,
22366 fmlalq_laneq_high, fmlslq_laneq_high): Use FP flags.
22368 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22370 * config/aarch64/aarch64-builtins.c (FLAG_LOAD): Define.
22371 * config/aarch64/aarch64-simd-builtins.def (ld1x2, ld2, ld3, ld4, ld2r,
22372 ld3r, ld4r, ld1, ld1x3, ld1x4): Use LOAD flags.
22374 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22376 * config/aarch64/aarch64-simd-builtins.def (combine, zip1, zip2,
22377 uzp1, uzp2, trn1, trn2, simd_bsl): Use AUTO_FP flags.
22379 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22381 * config/aarch64/aarch64-simd-builtins.def (clrsb, clz, ctz, popcount,
22382 vec_smult_lane_, vec_smlal_lane_, vec_smult_laneq_, vec_smlal_laneq_,
22383 vec_umult_lane_, vec_umlal_lane_, vec_umult_laneq_, vec_umlal_laneq_,
22384 ashl, sshl, ushl, srshl, urshl, sdot_lane, udot_lane, sdot_laneq,
22385 udot_laneq, usdot_lane, usdot_laneq, sudot_lane, sudot_laneq, ashr,
22386 ashr_simd, lshr, lshr_simd, srshr_n, urshr_n, ssra_n, usra_n, srsra_n,
22387 ursra_n, sshll_n, ushll_n, sshll2_n, ushll2_n, ssri_n, usri_n, ssli_n,
22388 ssli_n, usli_n, bswap, rbit, simd_bsl, eor3q, rax1q, xarq, bcaxq): Use
22389 NONE builtin flags.
22391 2021-02-02 Jakub Jelinek <jakub@redhat.com>
22393 PR tree-optimization/98848
22394 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Punt if
22395 STMT_VINFO_DEF_TYPE (last_stmt_info) is vect_reduction_def.
22397 2021-02-02 Kito Cheng <kito.cheng@sifive.com>
22400 * expr.c: Check mode before calling store_expr.
22402 2021-02-02 Christophe Lyon <christophe.lyon@linaro.org>
22404 * config/arm/iterators.md (supf): Remove VORNQ_S and VORNQ_U.
22406 * config/arm/mve.md (mve_vornq_s<mode>): New entry for vorn
22407 instruction using expression ior.
22408 (mve_vornq_u<mode>): New expander.
22409 (mve_vornq_f<mode>): Use ior code instead of unspec.
22410 * config/arm/unspecs.md (VORNQ_S, VORNQ_U, VORNQ_F): Remove.
22412 2021-02-02 Alexandre Oliva <oliva@adacore.com>
22414 * tree-nested.c (convert_nonlocal_reference_op): Move
22415 current_function_decl restore after re-gimplification.
22416 (convert_local_reference_op): Likewise.
22418 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22420 * config/aarch64/aarch64-simd-builtins.def (rshrn, rshrn2):
22422 * config/aarch64/aarch64-simd.md (aarch64_rshrn<mode>_insn_le):
22424 (aarch64_rshrn<mode>_insn_be): Likewise.
22425 (aarch64_rshrn<mode>): Likewise.
22426 (aarch64_rshrn2<mode>_insn_le): Likewise.
22427 (aarch64_rshrn2<mode>_insn_be): Likewise.
22428 (aarch64_rshrn2<mode>): Likewise.
22429 * config/aarch64/aarch64.md (unspec): Add UNSPEC_RSHRN.
22430 * config/aarch64/arm_neon.h (vrshrn_high_n_s16): Reimplement
22432 (vrshrn_high_n_s32): Likewise.
22433 (vrshrn_high_n_s64): Likewise.
22434 (vrshrn_high_n_u16): Likewise.
22435 (vrshrn_high_n_u32): Likewise.
22436 (vrshrn_high_n_u64): Likewise.
22437 (vrshrn_n_s16): Likewise.
22438 (vrshrn_n_s32): Likewise.
22439 (vrshrn_n_s64): Likewise.
22440 (vrshrn_n_u16): Likewise.
22441 (vrshrn_n_u32): Likewise.
22442 (vrshrn_n_u64): Likewise.
22444 2021-02-01 Sergei Trofimovich <siarheit@google.com>
22446 PR tree-optimization/98499
22447 * ipa-modref.c (analyze_ssa_name_flags): treat RVO
22448 conservatively and assume all possible side-effects.
22450 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22452 * config/aarch64/aarch64-simd-builtins.def (vec_unpacks_hi,
22453 vec_unpacku_hi_): Define builtins.
22454 * config/aarch64/arm_neon.h (vmovl_high_s8): Reimplement using
22456 (vmovl_high_s16): Likewise.
22457 (vmovl_high_s32): Likewise.
22458 (vmovl_high_u8): Likewise.
22459 (vmovl_high_u16): Likewise.
22460 (vmovl_high_u32): Likewise.
22462 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22464 * config/aarch64/aarch64-simd-builtins.def (sabdl, uabdl):
22466 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl<mode>): New
22468 * config/aarch64/aarch64.md (unspec): Define UNSPEC_SABDL,
22470 * config/aarch64/arm_neon.h (vabdl_s8): Reimplemet using
22472 (vabdl_s16): Likewise.
22473 (vabdl_s32): Likewise.
22474 (vabdl_u8): Likewise.
22475 (vabdl_u16): Likewise.
22476 (vabdl_u32): Likewise.
22477 * config/aarch64/iterators.md (ABDL): New int iterator.
22478 (sur): Handle UNSPEC_SABDL, UNSPEC_UABDL.
22480 2021-02-01 Martin Sebor <msebor@redhat.com>
22482 * tree.h (BLOCK_VARS): Add comment.
22483 (BLOCK_SUBBLOCKS): Same.
22484 (BLOCK_SUPERCONTEXT): Same.
22485 (BLOCK_ABSTRACT_ORIGIN): Same.
22486 (inlined_function_outer_scope_p): Same.
22488 2021-02-01 Martin Sebor <msebor@redhat.com>
22490 PR middle-end/97172
22491 * attribs.c (attr_access::free_lang_data): Define new function.
22492 * attribs.h (attr_access::free_lang_data): Declare new function.
22494 2021-02-01 Richard Biener <rguenther@suse.de>
22496 * vec.h (auto_vec::auto_vec): Add memory stat parameters
22498 * bitmap.h (auto_bitmap::auto_bitmap): Likewise.
22500 2021-02-01 Tamar Christina <tamar.christina@arm.com>
22502 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>,
22503 aarch64_<su>mlsl<mode>, aarch64_<su>mlsl_n<mode>): Flip mult operands.
22505 2021-02-01 Richard Biener <rguenther@suse.de>
22507 PR rtl-optimization/98863
22508 * config/i386/i386-features.c (convert_scalars_to_vector):
22509 Set DF_RD_PRUNE_DEAD_DEFS.
22511 2021-01-31 Eric Botcazou <ebotcazou@adacore.com>
22513 * system.h (SIZE_MAX): Define if not already defined.
22515 2021-01-30 Aaron Sawdey <acsawdey@linux.ibm.com>
22517 * config/rs6000/genfusion.pl (gen_2logical): New function to
22518 generate patterns for logical-logical fusion.
22519 * config/rs6000/fusion.md: Regenerated patterns.
22520 * config/rs6000/rs6000-cpus.def: Add
22521 OPTION_MASK_P10_FUSION_2LOGICAL.
22522 * config/rs6000/rs6000.c (rs6000_option_override_internal):
22523 Enable logical-logical fusion for p10.
22524 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2logical.
22526 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
22528 * config/rs6000/rs6000.opt: Add periods to new AIX options.
22530 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
22532 * config/rs6000/rs6000.opt (mabi=vec-extabi): New.
22533 (mabi=vec-default): New.
22534 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
22535 __EXTABI__ for AIX Vector extended ABI.
22536 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Print AIX Vector
22538 (conditional_register_usage): If AIX vec_extabi enabled, vs20-vs31
22540 * doc/invoke.texi (PowerPC mabi): Add AIX vec-extabi and vec-default.
22542 2021-01-30 Jakub Jelinek <jakub@redhat.com>
22544 * config/i386/i386-features.c (remove_partial_avx_dependency): Clear
22545 DF_DEFER_INSN_RESCAN after calling df_process_deferred_rescans.
22547 2021-01-29 Vladimir N. Makarov <vmakarov@redhat.com>
22550 * lra-constraints.c (in_class_p): Don't narrow class only for REG
22553 2021-01-29 Will Schmidt <will_schmidt@vnet.ibm.com>
22555 * config/rs6000/rs6000-call.c (rs6000_expand_binup_builtin): Add
22556 clauses for CODE_FOR_vsx_xvcvuxddp_scale and
22557 CODE_FOR_vsx_xvcvsxddp_scale to the parameter checking code.
22559 2021-01-29 Andrew MacLeod <amacleod@redhat.com>
22561 PR tree-optimization/98866
22562 * gimple-range-gori.h (gori_compute:set_range_invariant): New.
22563 * gimple-range-gori.cc (gori_map::set_range_invariant): New.
22564 (gori_map::m_maybe_invariant): Rename from all_outgoing.
22565 (gori_map::gori_map): Rename all_outgoing to m_maybe_invariant.
22566 (gori_map::is_export_p): Ditto.
22567 (gori_map::calculate_gori): Ditto.
22568 (gori_compute::set_range_invariant): New.
22569 * gimple-range.cc (gimple_ranger::range_of_stmt): Set range
22570 invariant for pointers evaluating to [1, +INF].
22572 2021-01-29 Richard Biener <rguenther@suse.de>
22574 PR rtl-optimization/98863
22575 * config/i386/i386-features.c (remove_partial_avx_dependency):
22576 Do not perform DF analysis.
22577 (pass_data_remove_partial_avx_dependency): Remove
22580 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
22582 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_n
22583 builtin generator macros.
22584 * config/aarch64/aarch64-simd.md (aarch64_<su>mull_n<mode>):
22586 * config/aarch64/arm_neon.h (vmull_n_s16): Use RTL builtin
22587 instead of inline asm.
22588 (vmull_n_s32): Likewise.
22589 (vmull_n_u16): Likewise.
22590 (vmull_n_u32): Likewise.
22592 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22594 * config/aarch64/aarch64-simd-builtins.def (sabdl2, uabdl2):
22596 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
22598 (aarch64_<sur>abdl2<mode>): ... This.
22599 (<sur>sadv16qi): Adjust use of above.
22600 * config/aarch64/arm_neon.h (vabdl_high_s8): Reimplement using
22602 (vabdl_high_s16): Likewise.
22603 (vabdl_high_s32): Likewise.
22604 (vabdl_high_u8): Likewise.
22605 (vabdl_high_u16): Likewise.
22606 (vabdl_high_u32): Likewise.
22608 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22610 * config/aarch64/aarch64-simd-builtins.def (sabal2): Define
22612 (uabal2): Likewise.
22613 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal2<mode>): New
22615 * config/aarch64/aarch64.md (unspec): Add UNSPEC_SABAL2 and
22617 * config/aarch64/arm_neon.h (vabal_high_s8): Reimplement using
22619 (vabal_high_s16): Likewise.
22620 (vabal_high_s32): Likewise.
22621 (vabal_high_u8): Likewise.
22622 (vabal_high_u16): Likewise.
22623 (vabal_high_u32): Likewise.
22624 * config/aarch64/iterators.md (ABAL2): New mode iterator.
22625 (sur): Handle UNSPEC_SABAL2, UNSPEC_UABAL2.
22627 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22629 * config/aarch64/aarch64-simd-builtins.def (sabal): Define
22632 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal<mode>_4):
22634 (aarch64_<sur>abal<mode>): ... This
22635 (<sur>sadv16qi): Adust use of the above.
22636 * config/aarch64/arm_neon.h (vabal_s8): Reimplement using
22638 (vabal_s16): Likewise.
22639 (vabal_s32): Likewise.
22640 (vabal_u8): Likewise.
22641 (vabal_u16): Likewise.
22642 (vabal_u32): Likewise.
22644 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22646 * config/aarch64/aarch64-simd-builtins.def (saddlv, uaddlv):
22648 * config/aarch64/aarch64-simd.md (aarch64_<su>addlv<mode>):
22650 * config/aarch64/arm_neon.h (vaddlv_s8): Reimplement using
22652 (vaddlv_s16): Likewise.
22653 (vaddlv_u8): Likewise.
22654 (vaddlv_u16): Likewise.
22655 (vaddlvq_s8): Likewise.
22656 (vaddlvq_s16): Likewise.
22657 (vaddlvq_s32): Likewise.
22658 (vaddlvq_u8): Likewise.
22659 (vaddlvq_u16): Likewise.
22660 (vaddlvq_u32): Likewise.
22661 (vaddlv_s32): Likewise.
22662 (vaddlv_u32): Likewise.
22663 * config/aarch64/iterators.md (VDQV_L): New mode iterator.
22664 (unspec): Add UNSPEC_SADDLV, UNSPEC_UADDLV.
22665 (Vwstype): New mode attribute.
22667 (VWIDE_S): Likewise.
22668 (USADDLV): New int iterator.
22669 (su): Handle UNSPEC_SADDLV, UNSPEC_UADDLV.
22671 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
22673 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_lane[q]
22674 builtin generator macros.
22675 * config/aarch64/aarch64-simd.md (aarch64_vec_<su>mlsl_lane<Qlane>):
22677 * config/aarch64/arm_neon.h (vmlsl_lane_s16): Use RTL builtin
22678 instead of inline asm.
22679 (vmlsl_lane_s32): Likewise.
22680 (vmlsl_lane_u16): Likewise.
22681 (vmlsl_lane_u32): Likewise.
22682 (vmlsl_laneq_s16): Likewise.
22683 (vmlsl_laneq_s32): Likewise.
22684 (vmlsl_laneq_u16): Likewise.
22685 (vmlsl_laneq_u32): Likewise.
22687 2021-01-29 Richard Biener <rguenther@suse.de>
22689 * doc/invoke.texi (--param max-gcse-memory): Document unit
22691 * gcse.c (gcse_or_cprop_is_too_expensive): Adjust.
22692 * params.opt (--param max-gcse-memory): Adjust default and
22693 document unit of size.
22695 2021-01-29 Richard Biener <rguenther@suse.de>
22697 PR rtl-optimization/98863
22698 * gcse.c (gcse_or_cprop_is_too_expensive): Use unsigned
22699 HOST_WIDE_INT for the memory estimate.
22701 2021-01-29 Bin Cheng <bin.cheng@linux.alibaba.com>
22702 Richard Biener <rguenther@suse.de>
22704 PR tree-optimization/97627
22705 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
22706 Do not analyze fake edges.
22708 2021-01-29 Richard Biener <rguenther@suse.de>
22710 PR rtl-optimization/98144
22711 * df.h (df_mir_bb_info): Add con_visited member.
22712 * df-problems.c (df_mir_alloc): Initialize con_visited,
22713 do not fully populate IN and OUT.
22714 (df_mir_reset): Likewise.
22715 (df_mir_confluence_0): Set con_visited.
22716 (df_mir_confluence_n): Properly handle implicitely
22717 fully populated IN and OUT as designated by con_visited
22718 and update con_visited accordingly.
22720 2021-01-29 Jakub Jelinek <jakub@redhat.com>
22723 * config/arm/vec-common.md (mve_vshlq_<supf><mode>,
22724 vashl<mode>3, vashr<mode>3, vlshr<mode>3): Add
22725 && !TARGET_REALLY_IWMMXT to conditions.
22727 2021-01-29 Jakub Jelinek <jakub@redhat.com>
22730 * cfgbuild.c (find_bb_boundaries): Reset debug_insn when seeing
22733 2021-01-28 Marek Polacek <polacek@redhat.com>
22736 * stor-layout.c (finalize_type_size): If we reset TYPE_USER_ALIGN in
22737 the main variant, maybe reset it in its variants too.
22738 * tree.c (check_base_type): Return true only if TYPE_USER_ALIGN match.
22739 (check_aligned_type): Check if TYPE_USER_ALIGN match.
22741 2021-01-28 Christophe Lyon <christophe.lyon@linaro.org>
22744 * config/arm/arm.c (arm_rtx_costs_internal): Adjust cost of vector
22745 of constant zero for comparisons.
22747 2021-01-28 Michael Meissner <meissner@linux.ibm.com>
22749 * config/rs6000/rs6000.c (rs6000_mangle_decl_assembler_name): Add
22750 support for mapping built-in function names for long double
22751 built-in functions if long double is IEEE 128-bit.
22753 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
22755 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_n
22756 builtin generator macros.
22757 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_n<mode>):
22759 * config/aarch64/arm_neon.h (vmlsl_n_s16): Use RTL builtin
22760 instead of inline asm.
22761 (vmlsl_n_s32): Likewise.
22762 (vmlsl_n_u16): Likewise.
22763 (vmlsl_n_u32): Likewise.
22765 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
22767 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_n
22768 builtin generator macros.
22769 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>):
22771 * config/aarch64/arm_neon.h (vmlal_n_s16): Use RTL builtin
22772 instead of inline asm.
22773 (vmlal_n_s32): Likewise.
22774 (vmlal_n_u16): Likewise.
22775 (vmlal_n_u32): Likewise.
22777 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22779 * config/aarch64/aarch64-simd-builtins.def (shrn2): Define
22781 * config/aarch64/aarch64-simd.md (aarch64_shrn2<mode>_insn_le):
22783 (aarch64_shrn2<mode>_insn_be): Likewise.
22784 (aarch64_shrn2<mode>): Likewise.
22785 * config/aarch64/arm_neon.h (vshrn_high_n_s16): Reimlplement
22787 (vshrn_high_n_s32): Likewise.
22788 (vshrn_high_n_s64): Likewise.
22789 (vshrn_high_n_u16): Likewise.
22790 (vshrn_high_n_u32): Likewise.
22791 (vshrn_high_n_u64): Likewise.
22793 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22795 * config/aarch64/aarch64-simd-builtins.def (shrn): Define
22797 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le):
22799 (aarch64_shrn<mode>_insn_be): Likewise.
22800 (aarch64_shrn<mode>): Likewise.
22801 * config/aarch64/arm_neon.h (vshrn_n_s16): Reimplement using
22803 (vshrn_n_s32): Likewise.
22804 (vshrn_n_s64): Likewise.
22805 (vshrn_n_u16): Likewise.
22806 (vshrn_n_u32): Likewise.
22807 (vshrn_n_u64): Likewise.
22808 * config/aarch64/iterators.md (vn_mode): New mode attribute.
22810 2021-01-28 Richard Biener <rguenther@suse.de>
22812 PR rtl-optimization/80960
22813 * dse.c (check_mem_read_rtx): Call get_addr on the
22816 2021-01-28 Xionghu Luo <luoxhu@linux.ibm.com>
22817 David Edelsohn <dje.gcc@gmail.com>
22820 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
22821 Don't generate VIEW_CONVERT_EXPR for fcode ALTIVEC_BUILTIN_VEC_INSERT
22823 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
22825 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Remove the
22826 wrapper call rs6000_expand_vector_set_var for cleanup. Call
22827 rs6000_expand_vector_set_var_p9 and rs6000_expand_vector_set_var_p8
22829 (rs6000_expand_vector_set_var): Delete.
22830 (rs6000_expand_vector_set_var_p9): Make static.
22831 (rs6000_expand_vector_set_var_p8): Make static.
22833 2021-01-28 Xing GUO <higuoxing@gmail.com>
22835 * common/config/riscv/riscv-common.c
22836 (riscv_subset_list::parsing_subset_version): Fix -march option parsing
22837 when `p` extension exists.
22839 2021-01-27 Vladimir N. Makarov <vmakarov@redhat.com>
22841 PR rtl-optimization/97684
22842 * ira.c (ira): Call ira_set_pseudo_classes before
22843 update_equiv_regs when it is necessary.
22845 2021-01-27 Jakub Jelinek <jakub@redhat.com>
22848 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Use
22849 %w0, %w1 and %2 instead of %0, %1 and %2.
22851 2021-01-27 Aaron Sawdey <acsawdey@linux.ibm.com>
22853 * config/rs6000/genfusion.pl: New script to generate
22854 define_insn_and_split patterns so combine can arrange fused
22855 instructions next to each other.
22856 * config/rs6000/fusion.md: New file, generated fused instruction
22857 patterns for combine.
22858 * config/rs6000/predicates.md (const_m1_to_1_operand): New predicate.
22859 (non_update_memory_operand): New predicate.
22860 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_P10_FUSION and
22861 OPTION_MASK_P10_FUSION_LD_CMPI to ISA_3_1_MASKS_SERVER and
22863 * config/rs6000/rs6000-protos.h (address_is_non_pfx_d_or_x): Add
22865 * config/rs6000/rs6000.c (rs6000_option_override_internal):
22866 Automatically set OPTION_MASK_P10_FUSION and
22867 OPTION_MASK_P10_FUSION_LD_CMPI if target is power10.
22868 (rs600_opt_masks): Allow -mpower10-fusion
22869 in function attributes.
22870 (address_is_non_pfx_d_or_x): New function.
22871 * config/rs6000/rs6000.h: Add MASK_P10_FUSION.
22872 * config/rs6000/rs6000.md: Include fusion.md.
22873 * config/rs6000/rs6000.opt: Add -mpower10-fusion
22874 and -mpower10-fusion-ld-cmpi.
22875 * config/rs6000/t-rs6000: Add dependencies involving fusion.md.
22877 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
22879 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal
22880 builtin generator macros.
22881 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal<mode>):
22883 (aarch64_<su>mlal<mode>): This.
22884 * config/aarch64/arm_neon.h (vmlal_s8): Use RTL builtin
22885 instead of inline asm.
22886 (vmlal_s16): Likewise.
22887 (vmlal_s32): Likewise.
22888 (vmlal_u8): Likewise.
22889 (vmlal_u16): Likewise.
22890 (vmlal_u32): Likewise.
22892 2021-01-27 Richard Biener <rguenther@suse.de>
22894 PR tree-optimization/98854
22895 * tree-vect-slp.c (vect_build_slp_tree_2): Also build
22896 PHIs from scalars when the number of CTORs matches the
22897 number of children.
22899 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
22901 * config/aarch64/aarch64-simd-builtins.def: Add mls_n builtin
22903 * config/aarch64/aarch64-simd.md (*aarch64_mls_elt_merge<mode>):
22905 (aarch64_mls_n<mode>): This.
22906 * config/aarch64/arm_neon.h (vmls_n_s16): Use RTL builtin
22908 (vmls_n_s32): Likewise.
22909 (vmls_n_u16): Likewise.
22910 (vmls_n_u32): Likewise.
22911 (vmlsq_n_s16): Likewise.
22912 (vmlsq_n_s32): Likewise.
22913 (vmlsq_n_u16): Likewise.
22914 (vmlsq_n_u32): Likewise.
22916 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
22918 * config/aarch64/aarch64-simd-builtins.def: Add mls builtin
22920 * config/aarch64/arm_neon.h (vmls_s8): Use RTL builtin rather
22922 (vmls_s16): Likewise.
22923 (vmls_s32): Likewise.
22924 (vmls_u8): Likewise.
22925 (vmls_u16): Likewise.
22926 (vmls_u32): Likewise.
22927 (vmlsq_s8): Likewise.
22928 (vmlsq_s16): Likewise.
22929 (vmlsq_s32): Likewise.
22930 (vmlsq_u8): Likewise.
22931 (vmlsq_u16): Likewise.
22932 (vmlsq_u32): Likewise.
22934 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
22936 * config/aarch64/aarch64-simd-builtins.def: Add mla_n builtin
22938 * config/aarch64/aarch64-simd.md (*aarch64_mla_elt_merge<mode>):
22940 (aarch64_mla_n<mode>): This.
22941 * config/aarch64/arm_neon.h (vmla_n_s16): Use RTL builtin
22943 (vmla_n_s32): Likewise.
22944 (vmla_n_u16): Likewise.
22945 (vmla_n_u32): Likewise.
22946 (vmlaq_n_s16): Likewise.
22947 (vmlaq_n_s32): Likewise.
22948 (vmlaq_n_u16): Likewise.
22949 (vmlaq_n_u32): Likewise.
22951 2021-01-27 liuhongt <hongtao.liu@intel.com>
22954 * config/i386/sse.md (sse2_gt<mode>3): Drop !TARGET_XOP in condition.
22955 (*sse2_eq<mode>3): Ditto.
22957 2021-01-27 Jakub Jelinek <jakub@redhat.com>
22959 * tree-pass.h (PROP_trees): Rename to ...
22960 (PROP_gimple): ... this.
22961 * cfgexpand.c (pass_data_expand): Replace PROP_trees with PROP_gimple.
22962 * passes.c (execute_function_dump, execute_function_todo,
22963 execute_one_ipa_transform_pass, execute_one_pass): Likewise.
22964 * varpool.c (ctor_for_folding): Likewise.
22966 2021-01-27 Jakub Jelinek <jakub@redhat.com>
22968 PR tree-optimization/97260
22969 * varpool.c: Include tree-pass.h.
22970 (ctor_for_folding): In GENERIC return DECL_INITIAL for TREE_READONLY
22971 non-TREE_SIDE_EFFECTS automatic variables.
22973 2021-01-26 Paul Fee <paul.f.fee@gmail.com>
22975 * doc/cpp.texi (__cplusplus): Document value for -std=c++23
22977 * doc/invoke.texi: Document -std=c++23 and -std=gnu++23.
22978 * dwarf2out.c (highest_c_language): Recognise C++20 and C++23.
22979 (gen_compile_unit_die): Recognise C++23.
22981 2021-01-26 Jakub Jelinek <jakub@redhat.com>
22984 * dwarf2asm.c (dw2_assemble_integer): Cast DWARF2_ADDR_SIZE to int
22987 2021-01-26 Jakub Jelinek <jakub@redhat.com>
22990 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
22991 Use UINTVAL (shft_amnt) and UINTVAL (mask) instead of INTVAL (shft_amnt)
22992 and INTVAL (mask). Add && INTVAL (mask) > 0 condition.
22994 2021-01-26 Richard Biener <rguenther@suse.de>
22996 * gimple-pretty-print.c (dump_binary_rhs): Handle
22997 VEC_WIDEN_{PLUS,MINUS}_{LO,HI}_EXPR.
22999 2021-01-26 Richard Biener <rguenther@suse.de>
23001 PR middle-end/98726
23002 * tree.h (vector_cst_int_elt): Remove.
23003 * tree.c (vector_cst_int_elt): Use poly_wide_int for computations,
23006 2021-01-26 Andrew Stubbs <ams@codesourcery.com>
23008 * config/gcn/gcn.c (gcn_expand_reduc_scalar): Use move instructions
23009 for V64DFmode min/max reductions.
23011 2021-01-26 Jakub Jelinek <jakub@redhat.com>
23013 * dwarf2asm.c (dw2_assemble_integer): Handle size twice as large
23014 as DWARF2_ADDR_SIZE if x is not a scalar int by emitting it as
23015 two halves, one with x and the other with const0_rtx, ordered
23016 depending on endianity.
23018 2021-01-26 Alexandre Oliva <oliva@adacore.com>
23020 * gimplify.c (gimplify_decl_expr): Skip asan marking calls for
23021 temporaries not seen in binding block, and not about to be
23022 added as gimple variables.
23024 2021-01-25 Martin Sebor <msebor@redhat.com>
23027 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust warning text.
23029 2021-01-25 Martin Liska <mliska@suse.cz>
23031 * value-prof.c (get_nth_most_common_value): Use %s instead
23034 2021-01-25 Jakub Jelinek <jakub@redhat.com>
23037 * configure.ac (HAVE_AS_GDWARF_5_DEBUG_FLAG): Only define if
23038 readelf -wi is able to read the emitted .debug_info back.
23039 * configure: Regenerated.
23041 2021-01-25 Martin Liska <mliska@suse.cz>
23043 PR gcov-profile/98739
23044 * common.opt: Add missing sign symbol.
23045 * value-prof.c (get_nth_most_common_value): Restore handling
23046 of PROFILE_REPRODUCIBILITY_PARALLEL_RUNS and
23047 PROFILE_REPRODUCIBILITY_MULTITHREADED.
23049 2021-01-25 Richard Biener <rguenther@suse.de>
23051 PR middle-end/98807
23052 * tree.c (vector_element_bits): Always use precision of
23053 the element type for boolean vectors.
23055 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
23057 * config/rtems.h (STARTFILE_SPEC): Remove qnolinkcmds.
23058 (ENDFILE_SPEC): Evaluate qnolinkcmds.
23060 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
23062 * config/rtems.h (STARTFILE_SPEC): Remove nostdlib and
23063 nostartfiles handling since this is already done by
23064 LINK_COMMAND_SPEC. Evaluate qnolinkcmds.
23065 (ENDFILE_SPEC): Remove nostdlib and nostartfiles handling since this
23066 is already done by LINK_COMMAND_SPEC.
23067 (LIB_SPECS): Remove nostdlib and nodefaultlibs handling since
23068 this is already done by LINK_COMMAND_SPEC. Remove qnolinkcmds
23071 2021-01-25 Jakub Jelinek <jakub@redhat.com>
23074 * fold-const-call.c (host_size_t_cst_p): Renamed to ...
23075 (size_t_cst_p): ... this. Check and store unsigned HOST_WIDE_INT
23076 value rather than host size_t.
23077 (fold_const_call): Change type of s2 from size_t to
23078 unsigned HOST_WIDE_INT. Use size_t_cst_p instead of
23079 host_size_t_cst_p. For strncmp calls, pass MIN (s2, SIZE_MAX)
23080 instead of s2 as last argument.
23082 2021-01-25 Tamar Christina <tamar.christina@arm.com>
23084 * config/arm/iterators.md (rotsplit1, rotsplit2, conj_op, fcmac1,
23085 VCMLA_OP, VCMUL_OP): New.
23086 * config/arm/mve.md (mve_vcmlaq<mve_rot><mode>): Support vec_dup 0.
23087 * config/arm/neon.md (cmul<conj_op><mode>3): New.
23088 * config/arm/unspecs.md (UNSPEC_VCMLA_CONJ, UNSPEC_VCMLA180_CONJ,
23089 UNSPEC_VCMUL_CONJ): New.
23090 * config/arm/vec-common.md (cmul<conj_op><mode>3, arm_vcmla<rot><mode>,
23091 cml<fcmac1><conj_op><mode>4): New.
23093 2021-01-23 Jakub Jelinek <jakub@redhat.com>
23096 * config/rs6000/mmintrin.h (__m64): Add __may_alias__ attribute.
23098 2021-01-22 Jonathan Wright <jonathan.wright@arm.com>
23100 * config/aarch64/aarch64-simd-builtins.def: Add mla builtin
23102 * config/aarch64/arm_neon.h (vmla_s8): Use RTL builtin rather
23104 (vmla_s16): Likewise.
23105 (vmla_s32): Likewise.
23106 (vmla_u8): Likewise.
23107 (vmla_u16): Likewise.
23108 (vmla_u32): Likewise.
23109 (vmlaq_s8): Likewise.
23110 (vmlaq_s16): Likewise.
23111 (vmlaq_s32): Likewise.
23112 (vmlaq_u8): Likewise.
23113 (vmlaq_u16): Likewise.
23114 (vmlaq_u32): Likewise.
23116 2021-01-22 David Malcolm <dmalcolm@redhat.com>
23118 * doc/invoke.texi (GCC_EXTRA_DIAGNOSTIC_OUTPUT): Add @findex
23121 2021-01-22 Jakub Jelinek <jakub@redhat.com>
23124 * dwarf2out.c (output_file_names): For -gdwarf-5, if there are no
23125 filenames to emit, still emit the required 0 index directory and
23126 filename entries that match DW_AT_comp_dir and DW_AT_name of the
23129 2021-01-22 Marek Polacek <polacek@redhat.com>
23132 * doc/invoke.texi: Update C++ ABI Version 15 description.
23134 2021-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23136 PR tree-optimization/98766
23137 * tree-ssa-math-opts.c (convert_mult_to_fma): Use maybe_le when
23138 comparing against type size with param_avoid_fma_max_bits.
23140 2021-01-22 Richard Biener <rguenther@suse.de>
23142 PR middle-end/98793
23143 * tree.c (vector_element_bits): Key single-bit bool vector on
23144 integer mode rather than not vector mode.
23146 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
23149 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
23150 Generate ARRAY_REF(VIEW_CONVERT_EXPR) for P8 and later
23152 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): Update
23153 to call different path for P8 and P9.
23154 (rs6000_expand_vector_set_var_p9): New function.
23155 (rs6000_expand_vector_set_var_p8): New function.
23157 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
23161 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
23162 Ajdust variable index vec_insert from address dereference to
23163 ARRAY_REF(VIEW_CONVERT_EXPR) tree expression.
23164 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
23166 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): New function.
23168 2021-01-22 Martin Liska <mliska@suse.cz>
23170 PR gcov-profile/98739
23171 * profile.c (compute_value_histograms): Drop time profile for
23172 -fprofile-reproducible=multithreaded.
23174 2021-01-22 Nathan Sidwell <nathan@acm.org>
23176 * gcc.c (process_command): Don't check OPT_SPECIAL_input_file
23179 2021-01-22 Richard Biener <rguenther@suse.de>
23181 PR middle-end/98773
23182 * tree-data-ref.c (initalize_matrix_A): Revert previous
23183 change, retaining failing on HOST_WIDE_INT_MIN CHREC_RIGHT.
23185 2021-01-22 Jakub Jelinek <jakub@redhat.com>
23187 PR tree-optimization/90248
23188 * match.pd (X cmp 0.0 ? 1.0 : -1.0 -> copysign(1, +-X),
23189 X cmp 0.0 ? -1.0 : +1.0 -> copysign(1, -+X)): Remove
23191 (X * (X cmp 0.0 ? 1.0 : -1.0) -> +-abs(X),
23192 X * (X cmp 0.0 ? -1.0 : 1.0) -> +-abs(X)): New simplifications.
23194 2021-01-22 Jakub Jelinek <jakub@redhat.com>
23196 PR tree-optimization/98255
23197 * tree-dfa.c (get_ref_base_and_extent): For ARRAY_REFs, sign
23198 extend index - low_bound from sizetype's precision rather than index
23200 (get_addr_base_and_unit_offset_1): Likewise.
23201 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Likewise.
23202 * gimple-fold.c (fold_const_aggregate_ref_1): Likewise.
23204 2021-01-22 Richard Biener <rguenther@suse.de>
23206 PR tree-optimization/98786
23207 * tree-ssa-phiopt.c (factor_out_conditional_conversion): Avoid
23208 adding new uses of abnormals. Verify we deal with a conditional
23211 2021-01-22 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23214 * optc-save-gen.awk: Add arm_fp16_format to checked_options.
23216 2021-01-22 liuhongt <hongtao.liu@intel.com>
23220 * config/i386/sse.md (VI_128_256): New mode iterator.
23221 (*avx_cmp<mode>3_1, *avx_cmp<mode>3_2, *avx_cmp<mode>3_3,
23222 *avx_cmp<mode>3_4, *avx2_eq<mode>3, *avx2_pcmp<mode>3_1,
23223 *avx2_pcmp<mode>3_2, *avx2_gt<mode>3): New
23224 define_insn_and_split to lower avx512 vector comparison to avx
23225 version when dest is vector.
23226 (*<avx512>_cmp<mode>3,*<avx512>_cmp<mode>3,*<avx512>_ucmp<mode>3):
23227 define_insn_and_split for negating the comparison result.
23228 * config/i386/predicates.md (float_vector_all_ones_operand):
23230 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Use
23231 general NOT operator without UNSPEC_MASKOP.
23233 2021-01-21 Vladimir N. Makarov <vmakarov@redhat.com>
23235 PR rtl-optimization/98777
23236 * lra-int.h (lra_pmode_pseudo): New extern.
23237 * lra.c (lra_pmode_pseudo): New global.
23239 * lra-eliminations.c (eliminate_regs_in_insn): Use it.
23241 2021-01-21 Ilya Leoshkevich <iii@linux.ibm.com>
23243 * fwprop.c (fwprop_propagation::classify_result): Allow
23244 (subreg (mem)) simplifications.
23246 2021-01-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23248 * config/aarch64/aarch64-simd.md (aarch64_sqdml<SBINQOPS:as>l<mode>):
23250 (aarch64_sqdmlal<mode>): ... This...
23251 (aarch64_sqdmlsl<mode>): ... And this.
23252 (aarch64_sqdml<SBINQOPS:as>l_lane<mode>): Split into...
23253 (aarch64_sqdmlal_lane<mode>): ... This...
23254 (aarch64_sqdmlsl_lane<mode>): ... And this.
23255 (aarch64_sqdml<SBINQOPS:as>l_laneq<mode>): Split into...
23256 (aarch64_sqdmlsl_laneq<mode>): ... This...
23257 (aarch64_sqdmlal_laneq<mode>): ... And this.
23258 (aarch64_sqdml<SBINQOPS:as>l_n<mode>): Split into...
23259 (aarch64_sqdmlsl_n<mode>): ... This...
23260 (aarch64_sqdmlal_n<mode>): ... And this.
23261 (aarch64_sqdml<SBINQOPS:as>l2<mode>_internal): Split into...
23262 (aarch64_sqdmlal2<mode>_internal): ... This...
23263 (aarch64_sqdmlsl2<mode>_internal): ... And this.
23265 2021-01-21 Christophe Lyon <christophe.lyon@linaro.org>
23267 * config/arm/arm_mve.h (__arm_vcmpneq_s8): Fix return type.
23269 2021-01-21 Andrea Corallo <andrea.corallo@arm.com>
23272 * doc/sourcebuild.texi (arm_thumb2_no_arm_v8_1_lob): Document.
23274 2021-01-21 liuhongt <hongtao.liu@intel.com>
23276 PR rtl-optimization/98694
23277 * regcprop.c (copy_value): If SRC had been assigned a mode
23278 narrower than the copy, we can't link DEST into the chain even
23279 they have same hard_regno_nregs(i.e. HImode/SImode in i386
23282 2021-01-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23284 * config/aarch64/aarch64-simd.md (aarch64_get_lane<mode>):
23285 Convert to define_insn_and_split. Split into simple move when moving
23288 2021-01-20 Segher Boessenkool <segher@kernel.crashing.org>
23290 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_store): Change assert.
23291 Adjust comment. Simplify code.
23293 2021-01-20 Jakub Jelinek <jakub@redhat.com>
23296 * dwarf2out.c (reset_indirect_string): Also reset indirect strings
23297 with DW_FORM_line_strp form.
23298 (prune_unused_types_update_strings): Don't add into debug_str_hash
23299 indirect strings with DW_FORM_line_strp form.
23300 (adjust_name_comp_dir): New function.
23301 (dwarf2out_finish): Call it on CU DIEs after resetting
23302 debug_line_str_hash.
23304 2021-01-20 Vladimir N. Makarov <vmakarov@redhat.com>
23306 PR rtl-optimization/98722
23307 * lra-eliminations.c (eliminate_regs_in_insn): Check that target
23308 has no 3-op add insn to transform insns containing two pluses.
23310 2021-01-20 Richard Biener <rguenther@suse.de>
23312 * hwint.h (add_hwi): New function.
23313 (mul_hwi): Likewise.
23314 * tree-data-ref.c (initialize_matrix_A): Properly translate
23315 tree constants and avoid HOST_WIDE_INT_MIN.
23316 (lambda_matrix_row_add): Avoid undefined integer overflow
23317 and return true on such overflow.
23318 (lambda_matrix_right_hermite): Handle overflow from
23319 lambda_matrix_row_add gracefully. Simplify previous fix.
23320 (analyze_subscript_affine_affine): Likewise.
23322 2021-01-20 Eugene Rozenfeld <erozen@microsoft.com>
23324 PR tree-optimization/96674
23325 * match.pd: New patterns: x < y || y == XXX_MIN --> x <= y - 1
23326 x >= y && y != XXX_MIN --> x > y - 1
23328 2021-01-20 Richard Sandiford <richard.sandiford@arm.com>
23330 PR tree-optimization/98535
23331 * tree-vect-slp.c (duplicate_and_interleave): Use quick_grow_cleared.
23332 If the high and low permutes are the same, remove the high permutes
23333 from the working set and only continue with the low ones.
23335 2021-01-20 Jakub Jelinek <jakub@redhat.com>
23337 PR tree-optimization/98721
23338 * builtins.c (access_ref::inform_access): Don't assume
23339 SSA_NAME_IDENTIFIER must be non-NULL. Print messages about
23340 object whenever allocfn is NULL, rather than only when DECL_P
23341 is true. Use %qE instead of %qD for that. Formatting fixes.
23343 2021-01-20 Richard Biener <rguenther@suse.de>
23345 PR tree-optimization/98758
23346 * tree-data-ref.c (int_divides_p): Use lambda_int arguments.
23347 (lambda_matrix_right_hermite): Avoid undefinedness with
23348 signed integer abs and multiplication.
23349 (analyze_subscript_affine_affine): Use lambda_int.
23351 2021-01-20 David Malcolm <dmalcolm@redhat.com>
23354 * dwarf2out.c (output_line_info): Rename static variable
23355 "generation", moving it out of the function to...
23356 (output_line_info_generation): New.
23357 (init_sections_and_labels): Likewise, renaming the variable to...
23358 (init_sections_and_labels_generation): New.
23359 (dwarf2out_c_finalize): Reset the new variables.
23361 2021-01-19 Martin Sebor <msebor@redhat.com>
23363 PR middle-end/98664
23364 * tree-ssa-live.c (remove_unused_scope_block_p): Keep scopes for
23365 all functions, even if they're not declared artificial or inline.
23366 * tree.c (tree_inlined_location): Use macro expansion location
23367 only if scope traversal fails to expose one.
23369 2021-01-19 Richard Sandiford <richard.sandiford@arm.com>
23371 PR rtl-optimization/92294
23372 * alias.c (compare_base_symbol_refs): Take an extra parameter
23373 and add the distance between two symbols to it. Enshrine in
23374 comments that -1 means "either 0 or 1, but we can't tell
23375 which at compile time".
23376 (memrefs_conflict_p): Update call accordingly.
23377 (rtx_equal_for_memref_p): Likewise. Take the distance between symbols
23380 2021-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23382 * config/aarch64/aarch64-simd-builtins.def (sqshl, uqshl,
23383 sqrshl, uqrshl, sqadd, uqadd, sqsub, uqsub, suqadd, usqadd, sqmovn,
23384 uqmovn, sqxtn2, uqxtn2, sqabs, sqneg, sqdmlal, sqdmlsl, sqdmlal_lane,
23385 sqdmlsl_lane, sqdmlal_laneq, sqdmlsl_laneq, sqdmlal_n, sqdmlsl_n,
23386 sqdmlal2, sqdmlsl2, sqdmlal2_lane, sqdmlsl2_lane, sqdmlal2_laneq,
23387 sqdmlsl2_laneq, sqdmlal2_n, sqdmlsl2_n, sqdmull, sqdmull_lane,
23388 sqdmull_laneq, sqdmull_n, sqdmull2, sqdmull2_lane, sqdmull2_laneq,
23389 sqdmull2_n, sqdmulh, sqrdmulh, sqdmulh_lane, sqdmulh_laneq,
23390 sqrdmulh_lane, sqrdmulh_laneq, sqshrun_n, sqrshrun_n, sqshrn_n,
23391 uqshrn_n, sqrshrn_n, uqrshrn_n, sqshlu_n, sqshl_n, uqshl_n, sqrdmlah,
23392 sqrdmlsh, sqrdmlah_lane, sqrdmlsh_lane, sqrdmlah_laneq, sqrdmlsh_laneq,
23393 sqmovun): Use NONE flags.
23395 2021-01-19 Richard Biener <rguenther@suse.de>
23398 * ipa-modref.c (analyze_stmt): Only record a summary for a
23401 2021-01-19 Richard Biener <rguenther@suse.de>
23403 PR middle-end/98638
23404 * tree-ssanames.c (fini_ssanames): Zero SSA_NAME_DEF_STMT.
23406 2021-01-19 Daniel Hellstrom <daniel@gaisler.com>
23408 * config/sparc/rtemself.h (TARGET_OS_CPP_BUILTINS): Add
23409 built-in define __FIX_LEON3FT_TN0018.
23411 2021-01-19 Richard Biener <rguenther@suse.de>
23414 * tree-inline.c (tree_function_versioning): Set input_location
23415 to UNKNOWN_LOCATION throughout the function.
23417 2021-01-19 Tobias Burnus <tobias@codesourcery.com>
23420 * omp-low.c (lower_omp_target): Handle nonpointer is_device_ptr.
23422 2021-01-19 Martin Jambor <mjambor@suse.cz>
23425 * ipa-sra.c (ssa_name_only_returned_p): New parameter fun. Check
23426 whether non-call exceptions allow removal of a statement.
23427 (isra_analyze_call): Pass the appropriate function to
23428 ssa_name_only_returned_p.
23430 2021-01-19 Geng Qi <gengqi@linux.alibaba.com>
23432 * config/riscv/arch-canonicalize (longext_sort): New function for
23433 sorting 'multi-letter'.
23434 * config/riscv/multilib-generator: Adjusting the loop of 'alt' in
23435 'alts'. The 'arch' may not be the first of 'alts'.
23436 (_expand_combination): Add underline for the 'ext' without '*'.
23437 This is because, a single-letter extension can always be treated well
23438 with a '_' prefix, but it cannot be separated out if it is appended
23441 2021-01-18 Vladimir N. Makarov <vmakarov@redhat.com>
23444 * ira.c (ira): Skip abnormal critical edge splitting.
23446 2021-01-18 Jakub Jelinek <jakub@redhat.com>
23448 PR tree-optimization/98727
23449 * tree-ssa-math-opts.c (match_arith_overflow): Fix up computation of
23450 second .MUL_OVERFLOW operand for signed multiplication with overflow
23451 checking if the second operand of multiplication is not constant.
23453 2021-01-18 David Edelsohn <dje.gcc@gmail.com>
23455 * doc/invoke.texi (-gdwarf): TPF defaults to version 2 and AIX
23456 defaults to version 4.
23458 2021-01-18 David Malcolm <dmalcolm@redhat.com>
23460 * attribs.h (fndecl_dealloc_argno): New decl.
23461 * builtins.c (call_dealloc_argno): Split out second half of
23463 (fndecl_dealloc_argno): New.
23464 * doc/extend.texi (Common Function Attributes): Document the
23465 interaction between the analyzer and the malloc attribute.
23466 * doc/invoke.texi (Static Analyzer Options): Likewise.
23468 2021-01-17 David Edelsohn <dje.gcc@gmail.com>
23470 * config/rs6000/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Override
23471 dwarf_version to 4.
23472 * config/rs6000/aix72.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
23474 2021-01-17 Martin Jambor <mjambor@suse.cz>
23477 * cgraph.c (clone_of_p): Check also former_clone_of as we climb
23480 2021-01-17 Mark Wielaard <mark@klomp.org>
23482 * common.opt (gdwarf-): Init(5).
23483 * doc/invoke.texi (-gdwarf): Document default to 5.
23485 2021-01-16 Kwok Cheung Yeung <kcy@codesourcery.com>
23487 * builtin-types.def
23488 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT): Rename
23490 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR):
23491 ...this. Add extra argument.
23492 * gimplify.c (omp_default_clause): Ensure that event handle is
23493 firstprivate in a task region.
23494 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_DETACH.
23495 (gimplify_adjust_omp_clauses): Likewise.
23496 * omp-builtins.def (BUILT_IN_GOMP_TASK): Change function type to
23497 BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR.
23498 * omp-expand.c (expand_task_call): Add GOMP_TASK_FLAG_DETACH to flags
23499 if detach clause specified. Add detach argument when generating
23501 * omp-low.c (scan_sharing_clauses): Setup data environment for detach
23503 (finish_taskreg_scan): Move field for variable containing the event
23504 handle to the front of the struct.
23505 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_DETACH. Fix
23507 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
23508 OMP_CLAUSE_DETACH clause.
23509 (convert_local_omp_clauses): Handle OMP_CLAUSE_DETACH clause.
23510 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_DETACH.
23511 * tree.c (omp_clause_num_ops): Add entry for OMP_CLAUSE_DETACH.
23513 (omp_clause_code_name): Add entry for OMP_CLAUSE_DETACH. Fix
23515 (walk_tree_1): Handle OMP_CLAUSE_DETACH.
23517 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
23519 * config/nios2/t-rtems: Reset all MULTILIB_* variables. Shorten
23520 multilib directory names. Use MULTILIB_REQUIRED instead of
23521 MULTILIB_EXCEPTIONS. Add -mhw-mul -mhw-mulx -mhw-div
23522 -mcustom-fpu-cfg=fph2 multilib.
23524 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
23526 * config/nios2/nios2.c (NIOS2_FPU_CONFIG_NUM): Adjust value.
23527 (nios2_init_fpu_configs): Provide register values for new
23528 -mcustom-fpu-cfg=fph2 option variant.
23529 * doc/invoke.texi (-mcustom-fpu-cfg=fph2): Document new option
23532 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
23534 * config/nios2/nios2.c (nios2_custom_check_insns): Remove
23535 custom instruction warnings.
23537 2021-01-16 Jakub Jelinek <jakub@redhat.com>
23539 PR tree-optimization/96669
23540 * match.pd ((CST << x) & 1 -> x == 0): New simplification.
23542 2021-01-16 Jakub Jelinek <jakub@redhat.com>
23544 PR tree-optimization/96271
23545 * passes.def: Pass false argument to first two pass_cd_dce
23546 instances and true to last instance. Add comment that
23547 last instance rewrites no longer addressed locals.
23548 * tree-ssa-dce.c (pass_cd_dce): Add update_address_taken_p member and
23550 (pass_cd_dce::set_pass_param): New method.
23551 (pass_cd_dce::execute): Return TODO_update_address_taken from
23552 last cd_dce instance.
23554 2021-01-15 Carl Love <cel@us.ibm.com>
23556 * config/rs6000/altivec.h (vec_mulh, vec_div, vec_dive, vec_mod):
23558 * config/rs6000/altivec.md (VIlong): Move define to file vsx.md.
23559 * config/rs6000/rs6000-builtin.def (DIVES_V4SI, DIVES_V2DI,
23560 DIVEU_V4SI, DIVEU_V2DI, DIVS_V4SI, DIVS_V2DI, DIVU_V4SI,
23561 DIVU_V2DI, MODS_V2DI, MODS_V4SI, MODU_V2DI, MODU_V4SI,
23562 MULHS_V2DI, MULHS_V4SI, MULHU_V2DI, MULHU_V4SI, MULLD_V2DI):
23563 Add builtin define.
23564 (MULH, DIVE, MOD): Add new BU_P10_OVERLOAD_2 definitions.
23565 * config/rs6000/rs6000-call.c (VSX_BUILTIN_VEC_DIV,
23566 VSX_BUILTIN_VEC_DIVE, P10_BUILTIN_VEC_MOD, P10_BUILTIN_VEC_MULH):
23567 New overloaded definitions.
23568 (builtin_function_type) [P10V_BUILTIN_DIVEU_V4SI,
23569 P10V_BUILTIN_DIVEU_V2DI, P10V_BUILTIN_DIVU_V4SI,
23570 P10V_BUILTIN_DIVU_V2DI, P10V_BUILTIN_MODU_V2DI,
23571 P10V_BUILTIN_MODU_V4SI, P10V_BUILTIN_MULHU_V2DI,
23572 P10V_BUILTIN_MULHU_V4SI]: Add case
23573 statement for builtins.
23574 * config/rs6000/rs6000.md (bits): Add new attribute sizes V4SI, V2DI.
23575 * config/rs6000/vsx.md (VIlong): Moved from config/rs6000/altivec.md.
23576 (UNSPEC_VDIVES, UNSPEC_VDIVEU): New unspec definitions.
23577 (vsx_mul_v2di): Add if TARGET_POWER10 statement.
23578 (vsx_udiv_v2di): Add if TARGET_POWER10 statement.
23579 (dives_<mode>, diveu_<mode>, div<mode>3, uvdiv<mode>3,
23580 mods_<mode>, modu_<mode>, mulhs_<mode>, mulhu_<mode>, mulv2di3):
23581 Add define_insn, mode is VIlong.
23582 * doc/extend.texi (vec_mulh, vec_mul, vec_div, vec_dive, vec_mod):
23583 Add builtin descriptions.
23585 2021-01-15 Eric Botcazou <ebotcazou@adacore.com>
23587 * final.c (final_start_function_1): Reset force_source_line.
23589 2021-01-15 Jakub Jelinek <jakub@redhat.com>
23591 PR tree-optimization/96669
23592 * match.pd (((1 << A) & 1) != 0 -> A == 0,
23593 ((1 << A) & 1) == 0 -> A != 0): Generalize for 1s replaced by
23594 possibly different power of two constants and to right shift too.
23596 2021-01-15 Jakub Jelinek <jakub@redhat.com>
23598 PR tree-optimization/96681
23599 * match.pd ((x < 0) ^ (y < 0) to (x ^ y) < 0): New simplification.
23600 ((x >= 0) ^ (y >= 0) to (x ^ y) < 0): Likewise.
23601 ((x < 0) ^ (y >= 0) to (x ^ y) >= 0): Likewise.
23602 ((x >= 0) ^ (y < 0) to (x ^ y) >= 0): Likewise.
23604 2021-01-15 Alexandre Oliva <oliva@adacore.com>
23606 * opts.c (gen_command_line_string): Exclude -dumpbase-ext.
23608 2021-01-15 Tamar Christina <tamar.christina@arm.com>
23610 * config/aarch64/aarch64-simd.md (cml<fcmac1><conj_op><mode>4,
23611 cmul<conj_op><mode>3): New.
23612 * config/aarch64/iterators.md (UNSPEC_FCMUL,
23613 UNSPEC_FCMUL180, UNSPEC_FCMLA_CONJ, UNSPEC_FCMLA180_CONJ,
23614 UNSPEC_CMLA_CONJ, UNSPEC_CMLA180_CONJ, UNSPEC_CMUL, UNSPEC_CMUL180,
23615 FCMLA_OP, FCMUL_OP, conj_op, rotsplit1, rotsplit2, fcmac1, sve_rot1,
23616 sve_rot2, SVE2_INT_CMLA_OP, SVE2_INT_CMUL_OP, SVE2_INT_CADD_OP): New.
23617 (rot): Add UNSPEC_FCMUL, UNSPEC_FCMUL180.
23618 (rot_op): Renamed to conj_op.
23619 * config/aarch64/aarch64-sve.md (cml<fcmac1><conj_op><mode>4,
23620 cmul<conj_op><mode>3): New.
23621 * config/aarch64/aarch64-sve2.md (cml<fcmac1><conj_op><mode>4,
23622 cmul<conj_op><mode>3): New.
23624 2021-01-15 David Malcolm <dmalcolm@redhat.com>
23628 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
23629 Escape the tempfile name when constructing the expected output.
23631 2021-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23633 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlsl_hi<mode>):
23635 (aarch64_<su>mlsl_hi<mode>): ... This.
23636 (aarch64_<su>mlsl_hi<mode>): Define.
23637 (*aarch64_<su>mlsl<mode): Rename to...
23638 (aarch64_<su>mlsl<mode): ... This.
23639 * config/aarch64/aarch64-simd-builtins.def (smlsl, umlsl,
23640 smlsl_hi, umlsl_hi): Define builtins.
23641 * config/aarch64/arm_neon.h (vmlsl_high_s8, vmlsl_high_s16,
23642 vmlsl_high_s32, vmlsl_high_u8, vmlsl_high_u16, vmlsl_high_u32,
23643 vmlsl_s8, vmlsl_s16, vmlsl_s32, vmlsl_u8,
23644 vmlsl_u16, vmlsl_u32): Reimplement with builtins.
23646 2021-01-15 Uroš Bizjak <ubizjak@gmail.com>
23648 * config/i386/i386-c.c (ix86_target_macros):
23649 Use cpp_define_formatted for __SIZEOF_FLOAT80__ definition.
23651 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
23654 * config.gcc (aarch64*-*-*): Add aarch64-cc-fusion.o to extra_objs.
23655 * Makefile.in (RTL_SSA_H): New variable.
23656 * config/aarch64/t-aarch64 (aarch64-cc-fusion.o): New rule.
23657 * config/aarch64/aarch64-protos.h (make_pass_cc_fusion): Declare.
23658 * config/aarch64/aarch64-passes.def: Add pass_cc_fusion after
23660 * config/aarch64/aarch64-cc-fusion.cc: New file.
23662 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
23664 * recog.h (insn_change_watermark::~insn_change_watermark): Avoid
23665 calling cancel_changes for changes that no longer exist.
23667 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
23669 * rtl-ssa/functions.h (function_info::ref_defs): Rename to...
23670 (function_info::reg_defs): ...this.
23671 * rtl-ssa/member-fns.inl (function_info::ref_defs): Rename to...
23672 (function_info::reg_defs): ...this.
23674 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23677 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
23679 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23682 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23685 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
23687 2021-01-15 Richard Biener <rguenther@suse.de>
23689 PR tree-optimization/96376
23690 * tree-vect-stmts.c (get_load_store_type): Disregard alignment
23691 for VMAT_INVARIANT.
23693 2021-01-15 Martin Liska <mliska@suse.cz>
23695 * doc/install.texi: Document that some tests need pytest module.
23696 * doc/sourcebuild.texi: Likewise.
23698 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23701 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
23703 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23705 * config/arm/mve.md (mve_vshrq_n_s<mode>_imm): New entry.
23706 (mve_vshrq_n_u<mode>_imm): Likewise.
23707 * config/arm/neon.md (vashr<mode>3, vlshr<mode>3): Move to ...
23708 * config/arm/vec-common.md: ... here.
23710 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23712 * config/arm/mve.md (mve_vshlq_<supf><mode>): Move to
23714 * config/arm/neon.md (vashl<mode>3): Delete.
23715 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): New.
23716 (vasl<mode>3): New expander.
23718 2021-01-15 Richard Biener <rguenther@suse.de>
23720 PR tree-optimization/98685
23721 * tree-vect-slp.c (vect_schedule_slp_node): Refactor handling
23722 of vector extern defs.
23724 2021-01-14 David Malcolm <dmalcolm@redhat.com>
23727 * diagnostic.c (diagnostic_kind_text): Break out this array
23729 (diagnostic_build_prefix): ...here.
23730 (fancy_abort): Detect when diagnostic_initialize has not yet been
23731 called and fall back to a minimal implementation of printing the
23732 ICE, rather than segfaulting in internal_error.
23734 2021-01-14 David Malcolm <dmalcolm@redhat.com>
23736 * diagnostic.c (diagnostic_initialize): Eliminate
23737 parseable_fixits_p in favor of initializing extra_output_kind from
23738 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
23739 (convert_column_unit): New function, split out from...
23740 (diagnostic_converted_column): ...this.
23741 (print_parseable_fixits): Add "column_unit" and "tabstop" params.
23742 Use them to call convert_column_unit on the column values.
23743 (diagnostic_report_diagnostic): Eliminate conditional on
23744 parseable_fixits_p in favor of a switch statement on
23745 extra_output_kind, passing the appropriate values to the new
23746 params of print_parseable_fixits.
23747 (selftest::test_print_parseable_fixits_none): Update for new
23748 params of print_parseable_fixits.
23749 (selftest::test_print_parseable_fixits_insert): Likewise.
23750 (selftest::test_print_parseable_fixits_remove): Likewise.
23751 (selftest::test_print_parseable_fixits_replace): Likewise.
23752 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
23754 (selftest::diagnostic_c_tests): Call it.
23755 * diagnostic.h (enum diagnostics_extra_output_kind): New.
23756 (diagnostic_context::parseable_fixits_p): Delete field in favor
23758 (diagnostic_context::extra_output_kind): ...this new field.
23759 * doc/invoke.texi (Environment Variables): Add
23760 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
23761 * opts.c (common_handle_option): Update handling of
23762 OPT_fdiagnostics_parseable_fixits for change to diagnostic_context
23765 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23767 * tree-vect-slp-patterns.c (class complex_operations_pattern,
23768 complex_operations_pattern::matches,
23769 complex_operations_pattern::recognize,
23770 complex_operations_pattern::build): New.
23771 (slp_patterns): Use it.
23773 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23775 * internal-fn.def (COMPLEX_FMS, COMPLEX_FMS_CONJ): New.
23776 * optabs.def (cmls_optab, cmls_conj_optab): New.
23777 * doc/md.texi: Document them.
23778 * tree-vect-slp-patterns.c (class complex_fms_pattern,
23779 complex_fms_pattern::matches, complex_fms_pattern::recognize,
23780 complex_fms_pattern::build): New.
23782 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23784 * internal-fn.def (COMPLEX_FMA, COMPLEX_FMA_CONJ): New.
23785 * optabs.def (cmla_optab, cmla_conj_optab): New.
23786 * doc/md.texi: Document them.
23787 * tree-vect-slp-patterns.c (vect_match_call_p,
23788 class complex_fma_pattern, vect_slp_reset_pattern,
23789 complex_fma_pattern::matches, complex_fma_pattern::recognize,
23790 complex_fma_pattern::build): New.
23792 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23794 * internal-fn.def (COMPLEX_MUL, COMPLEX_MUL_CONJ): New.
23795 * optabs.def (cmul_optab, cmul_conj_optab): New.
23796 * doc/md.texi: Document them.
23797 * tree-vect-slp-patterns.c (vect_match_call_complex_mla,
23798 vect_normalize_conj_loc, is_eq_or_top, vect_validate_multiplication,
23799 vect_build_combine_node, class complex_mul_pattern,
23800 complex_mul_pattern::matches, complex_mul_pattern::recognize,
23801 complex_mul_pattern::build): New.
23803 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23805 * tree-vect-slp.c (optimize_load_redistribution_1): New.
23806 (optimize_load_redistribution, vect_is_slp_load_node): New.
23807 (vect_match_slp_patterns): Use it.
23809 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23811 * tree-vect-slp-patterns.c (complex_add_pattern::build):
23814 2021-01-14 Thomas Schwinge <thomas@codesourcery.com>
23816 * config/gcn/mkoffload.c (main): Create an offload image only in
23817 64-bit configurations.
23819 2021-01-14 H.J. Lu <hjl.tools@gmail.com>
23822 * config/i386/i386-options.c (ix86_option_override_internal):
23823 Issue an error for -fcf-protection with CF_BRANCH when compiling
23824 for 32-bit non-TARGET_CMOV targets.
23826 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
23829 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p):
23830 Remove declaration and initialization of shadow variable "ret".
23831 (ix86_option_override_internal): Remove delcaration of
23832 shadow variable "i". Redeclare shadowed variable to unsigned.
23833 * common/config/i386/i386-common.c (pta_size): Redeclare to unsigned.
23834 * config/i386/i386-builtins.c (get_builtin_code_for_version):
23835 Update for redeclaration.
23836 * config/i386/i386.h (pta_size): Ditto.
23838 2021-01-14 Richard Biener <rguenther@suse.de>
23840 PR tree-optimization/98674
23841 * tree-data-ref.c (base_supports_access_fn_components_p): New.
23842 (initialize_data_dependence_relation): For two bases without
23843 possible access fns resort to type size equality when determining
23844 shape compatibility.
23846 2021-01-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23849 * config/arm/arm_neon.h: Replace calls to __builtin_vcge* by
23850 <=, >= operators in vcle and vcge intrinsics respectively.
23851 * config/arm/arm_neon_builtins.def: Remove entry for
23854 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
23857 * config/i386/i386-options.c (ix86_function_specific_save):
23858 Remove redundant assignment to opts->x_ix86_branch_cost.
23859 * config/i386/i386.c (ix86_prefetch_sse):
23860 Rename from x86_prefetch_sse. Update all uses.
23861 * config/i386/i386.h: Update for rename.
23862 * config/i386/i386-options.h: Ditto.
23864 2021-01-14 Jakub Jelinek <jakub@redhat.com>
23867 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3,
23868 *sse4_1_zero_extendv4hiv4si2_3, *sse4_1_zero_extendv2siv2di2_3):
23869 Use Bm instead of m for non-avx. Add isa attribute.
23871 2021-01-14 Jakub Jelinek <jakub@redhat.com>
23873 PR tree-optimization/96688
23874 * match.pd (~(X >> Y) -> ~X >> Y): New simplification if
23875 ~X can be simplified.
23877 2021-01-14 Richard Sandiford <richard.sandiford@arm.com>
23879 * tree-vect-stmts.c (vect_model_load_cost): Account for unused
23880 IFN_LOAD_LANES results.
23882 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23884 * config/aarch64/aarch64-simd.md (aarch64_<su>xtl<mode>):
23886 (aarch64_xtn<mode>): Likewise.
23887 * config/aarch64/aarch64-simd-builtins.def (sxtl, uxtl, xtn):
23890 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
23892 (vmovl_s16): Likewise.
23893 (vmovl_s32): Likewise.
23894 (vmovl_u8): Likewise.
23895 (vmovl_u16): Likewise.
23896 (vmovl_u32): Likewise.
23897 (vmovn_s16): Likewise.
23898 (vmovn_s32): Likewise.
23899 (vmovn_s64): Likewise.
23900 (vmovn_u16): Likewise.
23901 (vmovn_u32): Likewise.
23902 (vmovn_u64): Likewise.
23904 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23906 * config/aarch64/aarch64-simd.md (aarch64_<su>qxtn2<mode>_le):
23908 (aarch64_<su>qxtn2<mode>_be): Likewise.
23909 (aarch64_<su>qxtn2<mode>): Likewise.
23910 * config/aarch64/aarch64-simd-builtins.def (sqxtn2, uqxtn2):
23912 * config/aarch64/iterators.md (SAT_TRUNC): Define code_iterator.
23913 (su): Handle ss_truncate and us_truncate.
23914 * config/aarch64/arm_neon.h (vqmovn_high_s16): Reimplement using
23916 (vqmovn_high_s32): Likewise.
23917 (vqmovn_high_s64): Likewise.
23918 (vqmovn_high_u16): Likewise.
23919 (vqmovn_high_u32): Likewise.
23920 (vqmovn_high_u64): Likewise.
23922 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23924 * config/aarch64/aarch64-simd.md (aarch64_xtn2<mode>_le):
23926 (aarch64_xtn2<mode>_be): Likewise.
23927 (aarch64_xtn2<mode>): Likewise.
23928 * config/aarch64/aarch64-simd-builtins.def (xtn2): Define
23930 * config/aarch64/arm_neon.h (vmovn_high_s16): Reimplement using
23932 (vmovn_high_s32): Likewise.
23933 (vmovn_high_s64): Likewise.
23934 (vmovn_high_u16): Likewise.
23935 (vmovn_high_u32): Likewise.
23936 (vmovn_high_u64): Likewise.
23938 2021-01-13 Stafford Horne <shorne@gmail.com>
23940 * config/or1k/or1k.h (ASM_PREFERRED_EH_DATA_FORMAT): New macro.
23942 2021-01-13 Stafford Horne <shorne@gmail.com>
23944 * config/or1k/linux.h (TARGET_ASM_FILE_END): Define macro.
23946 2021-01-13 Stafford Horne <shorne@gmail.com>
23948 * config/or1k/or1k.h (TARGET_CPU_CPP_BUILTINS): Add builtin
23949 define for __or1k_hard_float__.
23951 2021-01-13 Stafford Horne <shorne@gmail.com>
23953 * config/or1k/or1k.h (NO_PROFILE_COUNTERS): Define as 1.
23954 (PROFILE_HOOK): Define to call _mcount.
23955 (FUNCTION_PROFILER): Change from abort to no-op.
23957 2021-01-13 Jakub Jelinek <jakub@redhat.com>
23959 PR tree-optimization/96691
23960 * match.pd ((~X | C) ^ D -> (X | C) ^ (~D ^ C),
23961 (~X & C) ^ D -> (X & C) ^ (D ^ C)): New simplifications if
23962 (~D ^ C) or (D ^ C) can be simplified.
23964 2021-01-13 Richard Biener <rguenther@suse.de>
23966 PR tree-optimization/92645
23967 * match.pd (BIT_FIELD_REF to conversion): Delay canonicalization
23968 until after vector lowering.
23970 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
23972 * config/aarch64/aarch64-sve.md (fnma<mode>4): Extend from SVE_FULL_I
23974 (@aarch64_pred_fnma<mode>, cond_fnma<mode>, *cond_fnma<mode>_2)
23975 (*cond_fnma<mode>_4, *cond_fnma<mode>_any): Likewise.
23977 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
23979 * config/aarch64/aarch64-sve.md (fma<mode>4): Extend from SVE_FULL_I
23981 (@aarch64_pred_fma<mode>, cond_fma<mode>, *cond_fma<mode>_2)
23982 (*cond_fma<mode>_4, *cond_fma<mode>_any): Likewise.
23984 2021-01-13 Richard Biener <rguenther@suse.de>
23986 PR tree-optimization/92645
23987 * tree-vect-slp.c (vect_build_slp_tree_1): Relax supported
23988 BIT_FIELD_REF argument.
23989 (vect_build_slp_tree_2): Record the desired vector type
23990 on the external vector def.
23991 (vectorizable_slp_permutation): Handle required punning
23992 of existing vector defs.
23994 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
23996 * rtl-ssa/accesses.h (def_lookup): Fix order of comparison results.
23998 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
24000 * config/sh/sh.md (movsf_ie): Remove operands[2] test.
24002 2021-01-13 Samuel Thibault <samuel.thibault@ens-lyon.org>
24004 * config.gcc [$target == *-*-gnu*]: Enable
24005 'default_gnu_indirect_function'.
24007 2021-01-13 Jakub Jelinek <jakub@redhat.com>
24010 * optabs.c (expand_vec_perm_const): Don't force v0 and v1 into
24011 registers before calling targetm.vectorize.vec_perm_const, only after
24013 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const): Handle
24014 two argument permutation when one operand is zero vector and only
24015 after that force operands into registers.
24016 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_1): New
24017 define_insn_and_split pattern.
24018 (*avx512bw_zero_extendv32qiv32hi2_1): Likewise.
24019 (*avx512f_zero_extendv16hiv16si2_1): Likewise.
24020 (*avx2_zero_extendv8hiv8si2_1): Likewise.
24021 (*avx512f_zero_extendv8siv8di2_1): Likewise.
24022 (*avx2_zero_extendv4siv4di2_1): Likewise.
24023 * config/mips/mips.c (mips_vectorize_vec_perm_const): Force operands
24025 * config/arm/arm.c (arm_vectorize_vec_perm_const): Likewise.
24026 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Likewise.
24027 * config/ia64/ia64.c (ia64_vectorize_vec_perm_const): Likewise.
24028 * config/aarch64/aarch64.c (aarch64_vectorize_vec_perm_const): Likewise.
24029 * config/rs6000/rs6000.c (rs6000_vectorize_vec_perm_const): Likewise.
24030 * config/gcn/gcn.c (gcn_vectorize_vec_perm_const): Likewise. Use std::swap.
24032 2021-01-13 Martin Liska <mliska@suse.cz>
24034 PR tree-optimization/98455
24035 * gimple-if-to-switch.cc (condition_info::record_phi_mapping):
24036 Record also virtual PHIs.
24037 (pass_if_to_switch::execute): Return TODO_cleanup_cfg only
24040 2021-01-13 Jonathan Wakely <jwakely@redhat.com>
24042 * doc/invoke.texi (C++ Modules): Fix typos.
24044 2021-01-13 Richard Biener <rguenther@suse.de>
24046 PR tree-optimization/98640
24047 * tree-ssa-sccvn.c (visit_nary_op): Do not try to
24048 handle plus or minus from a truncated operand to be
24051 2021-01-13 Jakub Jelinek <jakub@redhat.com>
24054 * config/i386/i386.md (*btr<mode>_1, *btr<mode>_2): New
24055 define_insn_and_split patterns.
24056 (splitter after *btr<mode>_2): New splitter.
24058 2021-01-13 Martin Liska <mliska@suse.cz>
24061 * cgraphunit.c (analyze_functions): Remove dead code.
24063 2021-01-13 Qian Jianhua <qianjh@cn.fujitsu.com>
24065 * config/aarch64/aarch64-cost-tables.h (a64fx_extra_costs): New.
24066 * config/aarch64/aarch64.c (a64fx_addrcost_table): New.
24067 (a64fx_regmove_cost, a64fx_vector_cost): New.
24068 (a64fx_tunings): Use the new added cost tables.
24070 2021-01-13 Jakub Jelinek <jakub@redhat.com>
24073 * config/i386/predicates.md (pmovzx_parallel): New predicate.
24074 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): New
24075 define_insn_and_split pattern.
24076 (*sse4_1_zero_extendv4hiv4si2_3): Likewise.
24077 (*sse4_1_zero_extendv2siv2di2_3): Likewise.
24079 2021-01-13 Julian Brown <julian@codesourcery.com>
24081 * config/gcn/gcn.c (gcn_conditional_register_usage): Remove dead code
24082 to fix v0 register.
24084 2021-01-13 Julian Brown <julian@codesourcery.com>
24086 * config/gcn/gcn.c (gcn_md_reorg): Fix case where EXEC reg is live
24089 2021-01-13 Julian Brown <julian@codesourcery.com>
24091 * config/gcn/gcn-valu.md (recip<mode>2<exec>, recip<mode>2): Use unspec
24092 for reciprocal-approximation instructions.
24093 (div<mode>3): Use fused multiply-accumulate operations for reciprocal
24094 refinement and division result.
24095 * config/gcn/gcn.md (UNSPEC_RCP): New unspec constant.
24097 2021-01-13 Julian Brown <julian@codesourcery.com>
24099 * config/gcn/gcn-valu.md (subdf): Rename to...
24102 2021-01-12 Martin Liska <mliska@suse.cz>
24104 * gcov.c (source_info::debug): Fix printf format for 32-bit hosts.
24106 2021-01-12 Andrea Corallo <andrea.corallo@arm.com>
24108 * function-abi.h: Fix typo.
24110 2021-01-12 Christophe Lyon <christophe.lyon@linaro.org>
24114 * config/arm/arm.h (ARM_HAVE_NEON_V8QI_LDST): New macro.
24115 (ARM_HAVE_NEON_V16QI_LDST, ARM_HAVE_NEON_V4HI_LDST): Likewise.
24116 (ARM_HAVE_NEON_V8HI_LDST, ARM_HAVE_NEON_V2SI_LDST): Likewise.
24117 (ARM_HAVE_NEON_V4SI_LDST, ARM_HAVE_NEON_V4HF_LDST): Likewise.
24118 (ARM_HAVE_NEON_V8HF_LDST, ARM_HAVE_NEON_V4BF_LDST): Likewise.
24119 (ARM_HAVE_NEON_V8BF_LDST, ARM_HAVE_NEON_V2SF_LDST): Likewise.
24120 (ARM_HAVE_NEON_V4SF_LDST, ARM_HAVE_NEON_DI_LDST): Likewise.
24121 (ARM_HAVE_NEON_V2DI_LDST): Likewise.
24122 (ARM_HAVE_V8QI_LDST, ARM_HAVE_V16QI_LDST): Likewise.
24123 (ARM_HAVE_V4HI_LDST, ARM_HAVE_V8HI_LDST): Likewise.
24124 (ARM_HAVE_V2SI_LDST, ARM_HAVE_V4SI_LDST, ARM_HAVE_V4HF_LDST): Likewise.
24125 (ARM_HAVE_V8HF_LDST, ARM_HAVE_V4BF_LDST, ARM_HAVE_V8BF_LDST): Likewise.
24126 (ARM_HAVE_V2SF_LDST, ARM_HAVE_V4SF_LDST, ARM_HAVE_DI_LDST): Likewise.
24127 (ARM_HAVE_V2DI_LDST): Likewise.
24128 * config/arm/mve.md (*movmisalign<mode>_mve_store): New pattern.
24129 (*movmisalign<mode>_mve_load): New pattern.
24130 * config/arm/neon.md (movmisalign<mode>): Move to ...
24131 * config/arm/vec-common.md: ... here.
24133 2021-01-12 Vladimir N. Makarov <vmakarov@redhat.com>
24136 * lra-eliminations.c (eliminate_regs_in_insn): Add transformation
24137 of pattern 'plus (plus (hard reg, const), pseudo)'.
24139 2021-01-12 Richard Biener <rguenther@suse.de>
24141 PR tree-optimization/98550
24142 * tree-vect-slp.c (vect_record_max_nunits): Check whether
24143 the group size is a multiple of the vector element count.
24144 (vect_build_slp_tree_1): When we need to fail because
24145 the vector type choosen causes unrolling do so lazily
24146 without affecting matches only at the end to guide group splitting.
24148 2021-01-12 Martin Liska <mliska@suse.cz>
24151 * optc-save-gen.awk: Compare also n_target_save vars with
24154 2021-01-12 Martin Liska <mliska@suse.cz>
24156 * gcov.c (source_info::debug): New.
24157 (print_usage): Add --debug (-D) option.
24158 (process_args): Likewise.
24159 (generate_results): Call src->debug after
24160 accumulate_line_counts.
24161 (read_graph_file): Properly assign id for EXIT_BLOCK.
24162 * profile.c (branch_prob): Dump function body before it is
24165 2021-01-12 Jakub Jelinek <jakub@redhat.com>
24167 PR tree-optimization/98629
24168 * tree-ssa-math-opts.c (arith_overflow_check_p): Don't update use_stmt
24169 unless returning non-zero.
24171 2021-01-12 Jakub Jelinek <jakub@redhat.com>
24173 PR tree-optimization/95731
24174 * tree-ssa-reassoc.c (optimize_range_tests_cmp_bitwise): Also optimize
24175 x < 0 && y < 0 && z < 0 into (x | y | z) < 0 for signed x, y, z.
24176 (optimize_range_tests): Call optimize_range_tests_cmp_bitwise
24177 only after optimize_range_tests_var_bound.
24179 2021-01-12 Jakub Jelinek <jakub@redhat.com>
24181 * configure.ac: Ensure c/Make-lang.in comes first in @all_lang_makefrags@.
24182 * configure: Regenerated.
24184 2021-01-12 liuhongt <hongtao.liu@intel.com>
24187 * config/i386/i386-builtins.h (BUILTIN_DESC_SWAP_OPERANDS):
24189 * config/i386/i386-expand.c (ix86_expand_sse_comi): Delete
24192 2021-01-12 Alexandre Oliva <oliva@adacore.com>
24194 * ssa-iterators.h (end_imm_use_stmt_traverse): Forward
24196 (auto_end_imm_use_stmt_traverse): New struct.
24197 (FOR_EACH_IMM_USE_STMT): Use it.
24198 (BREAK_FROM_IMM_USE_STMT, RETURN_FROM_IMM_USE_STMT): Remove,
24200 * gimple-ssa-strength-reduction.c: ... here, ...
24201 * graphite-scop-detection.c: ... here, ...
24202 * ipa-modref.c, ipa-pure-const.c, ipa-sra.c: ... here, ...
24203 * tree-predcom.c, tree-ssa-ccp.c: ... here, ...
24204 * tree-ssa-dce.c, tree-ssa-dse.c: ... here, ...
24205 * tree-ssa-loop-ivopts.c, tree-ssa-math-opts.c: ... here, ...
24206 * tree-ssa-phiprop.c, tree-ssa.c: ... here, ...
24207 * tree-vect-slp.c: ... and here, ...
24208 * doc/tree-ssa.texi: ... and the example here.
24210 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24212 * config/aarch64/aarch64-sve.md (sdiv_pow2<mode>3): Extend from
24213 SVE_FULL_I to SVE_I. Generate an UNSPEC_PRED_X.
24214 (*sdiv_pow2<mode>3): New pattern.
24215 (@cond_<sve_int_op><mode>): Extend from SVE_FULL_I to SVE_I.
24216 Wrap the ASRD in an UNSPEC_PRED_X.
24217 (*cond_<sve_int_op><mode>_2): Likewise. Replace the UNSPEC_PRED_X
24218 predicate with a constant PTRUE, if it isn't already.
24219 (*cond_<sve_int_op><mode>_z): Replace with...
24220 (*cond_<sve_int_op><mode>_any): ...this new pattern.
24222 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24224 * config/aarch64/aarch64-sve.md (*cond_bic<mode>_2): Extend from
24225 SVE_FULL_I to SVE_I.
24226 (*cond_bic<mode>_any): Likewise.
24228 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24230 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart)
24231 (@aarch64_pred_<MUL_HIGHPART:optab><mode>): Extend from SVE_FULL_I
24234 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24236 * config/aarch64/aarch64-sve.md (<su>abd<mode>_3): Extend from
24237 SVE_FULL_I to SVE_I.
24238 (*aarch64_cond_<su>abd<mode>_2): Likewise.
24239 (*aarch64_cond_<su>abd<mode>_any): Likewise.
24240 (@aarch64_pred_<su>abd<mode>): Likewise. Use UNSPEC_PRED_X
24241 for the max and min but not for the minus.
24242 (*aarch64_cond_<su>abd<mode>_3): New pattern.
24244 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24246 * config/aarch64/iterators.md (SVE_24I): New iterator.
24247 * config/aarch64/aarch64-sve.md (*aarch64_adr<mode>_shift): Extend from
24248 SVE_FULL_SDI to SVE_24I. Use containers rather than elements.
24250 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24252 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_BINARY:optab><mode>)
24253 (*cond_<SVE_INT_BINARY:optab><mode>_2): Extend from SVE_FULL_I
24255 (*cond_<SVE_INT_BINARY:optab><mode>_3): Likewise.
24256 (*cond_<SVE_INT_BINARY:optab><mode>_any): Likewise.
24257 (*cond_<SVE_INT_BINARY:optab><mode>_2_const): Likewise.
24258 (*cond_<SVE_INT_BINARY:optab><mode>_any_const): Likewise.
24260 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24262 * config/aarch64/aarch64-sve.md (<SVE_INT_BINARY_IMM:optab><mode>3)
24263 (@aarch64_pred_<SVE_INT_BINARY_IMM:optab><mode>)
24264 (*post_ra_<SVE_INT_BINARY_IMM:optab><mode>3): Extend from SVE_FULL_I
24267 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24269 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3)
24270 (v<ASHIFT:optab><mode>3, @aarch64_pred_<optab><mode>)
24271 (*post_ra_v<ASHIFT:optab><mode>3): Extend from SVE_FULL_I to SVE_I.
24273 2021-01-11 Martin Liska <mliska@suse.cz>
24276 * symtab-clones.h (clone_info::release): Release
24277 symtab::m_clones with ggc_delete as it's a GGC memory.
24279 2021-01-11 Matthias Klose <doko@ubuntu.com>
24281 * Makefile.in (LINK_PROGRESS): Show the link target.
24283 2021-01-11 Richard Biener <rguenther@suse.de>
24285 PR tree-optimization/91403
24286 * tree-vect-data-refs.c (vect_analyze_group_access_1): Cap
24287 single-element interleaving group size at 4096 elements.
24289 2021-01-11 Richard Biener <rguenther@suse.de>
24291 PR tree-optimization/98526
24292 * tree-vect-loop.c (vect_model_reduction_cost): Remove costing
24293 of the actual reduction op for the regular case.
24294 (vectorizable_reduction): Cost the stmts
24295 vect_transform_reduction produces here.
24297 2021-01-11 Andreas Krebbel <krebbel@linux.ibm.com>
24299 * tree-ssa-forwprop.c (simplify_vector_constructor): For
24300 big-endian, use UNPACK[_FLOAT]_HI.
24302 2021-01-11 Tamar Christina <tamar.christina@arm.com>
24304 * tree-vect-slp-patterns.c (class complex_pattern,
24305 class complex_add_pattern): Add parameters to matches.
24306 (complex_add_pattern::build): Free memory.
24307 (complex_add_pattern::matches): Move validation end of match.
24308 (complex_add_pattern::recognize): Likewise.
24310 2021-01-11 Tamar Christina <tamar.christina@arm.com>
24312 * tree-vect-slp-patterns.c (linear_loads_p): Fix externals.
24314 2021-01-11 Tamar Christina <tamar.christina@arm.com>
24316 * tree-vect-slp-patterns.c (is_linear_load_p): Fix ambiguity.
24318 2021-01-11 Jakub Jelinek <jakub@redhat.com>
24320 PR tree-optimization/95867
24321 * tree-ssa-math-opts.h: New header.
24322 * tree-ssa-math-opts.c: Include tree-ssa-math-opts.h.
24323 (powi_as_mults): No longer static. Use build_one_cst instead of
24324 build_real. Formatting fix.
24325 * tree-ssa-reassoc.c: Include tree-ssa-math-opts.h.
24326 (attempt_builtin_powi): Handle multiplication reassociation without
24327 powi_fndecl using powi_as_mults.
24328 (reassociate_bb): For integral types don't require
24329 -funsafe-math-optimizations to call attempt_builtin_powi.
24331 2021-01-11 Jakub Jelinek <jakub@redhat.com>
24333 PR tree-optimization/95852
24334 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): Change
24335 mul_stmts parameter type to vec<gimple *> &. Before cond_stmt
24336 allow in the bb any of the stmts in that vector, div_stmt and
24337 up to 3 cast stmts.
24338 (arith_cast_equal_p): New function.
24339 (arith_overflow_check_p): Add cast_stmt argument, handle signed
24340 multiply overflow checks.
24341 (match_arith_overflow): Adjust caller. Handle signed multiply
24344 2021-01-11 Jakub Jelinek <jakub@redhat.com>
24346 PR tree-optimization/95852
24347 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): New function.
24348 (uaddsub_overflow_check_p): Renamed to ...
24349 (arith_overflow_check_p): ... this. Handle also multiplication
24350 with overflow check.
24351 (match_uaddsub_overflow): Renamed to ...
24352 (match_arith_overflow): ... this. Add cfg_changed argument. Handle
24353 also multiplication with overflow check. Adjust function comment.
24354 (math_opts_dom_walker::after_dom_children): Adjust callers. Call
24355 match_arith_overflow also for MULT_EXPR.
24357 2021-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24359 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
24360 __builtin_convertvector.
24361 (vmovl_s16): Likewise.
24362 (vmovl_s32): Likewise.
24363 (vmovl_u8): Likewise.
24364 (vmovl_u16): Likewise.
24365 (vmovl_u32): Likewise.
24366 (vmovn_s16): Likewise.
24367 (vmovn_s32): Likewise.
24368 (vmovn_s64): Likewise.
24369 (vmovn_u16): Likewise.
24370 (vmovn_u32): Likewise.
24371 (vmovn_u64): Likewise.
24373 2021-01-11 Martin Liska <mliska@suse.cz>
24375 * gimple-if-to-switch.cc (struct condition_info): Use auto_var.
24376 (if_chain::is_beneficial): Delete clusters
24377 (find_conditions): Make second argument of conditions_in_bbs a
24378 pointer so that we control over it's lifetime.
24379 (pass_if_to_switch::execute): Delete them.
24381 2021-01-11 Kewen Lin <linkw@linux.ibm.com>
24383 * ira.c (move_unallocated_pseudos): Check other_reg and skip if
24386 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
24388 * config/vax/vax.md (cc): Remove mode attribute.
24389 (subst_<cc>, subst_f<cc>): Rename to...
24390 (subst_<mode>, subst_f<VAXccnz:mode>): ... these respectively.
24391 (*cbranch<VAXint:mode>4_<VAXcc:mode>): Update for `cc' removal.
24392 (*cbranch<VAXfp:mode>4_<VAXccnz:mode>): Likewise.
24393 (*branch_<mode>, *branch_<mode>_reversed): Likewise.
24395 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
24397 * config/vax/vax.md (subst_f<cc>): Add mode to operands and
24398 `const_double_zero'.
24400 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
24402 * config/pdp11/pdp11.md (PDPfp): New mode iterator.
24403 (fcc_cc, fcc_ccnz): Use it. Add mode to `const_double_zero' and
24406 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
24408 * genemit.c (gen_exp) <CONST_DOUBLE>: Handle `const_double_zero'
24410 * read-rtl.c (rtx_reader::read_rtx_code): Handle machine mode
24411 with `const_double_zero'.
24412 * doc/rtl.texi (Constant Expression Types): Document it.
24414 2021-01-09 Jakub Jelinek <jakub@redhat.com>
24417 * tree-cfg.c (verify_gimple_assign_binary): Allow lhs of
24418 POINTER_DIFF_EXPR to be any integral type.
24420 2021-01-09 Jakub Jelinek <jakub@redhat.com>
24422 PR rtl-optimization/98603
24423 * function.c (instantiate_virtual_regs_in_insn): For asm goto
24424 with impossible constraints, drop all SETs, CLOBBERs, drop PARALLEL
24425 if any, set ASM_OPERANDS mode to VOIDmode and change
24426 ASM_OPERANDS_OUTPUT_CONSTRAINT and ASM_OPERANDS_OUTPUT_IDX.
24428 2021-01-09 Alexandre Oliva <oliva@gnu.org>
24431 * final.c (notice_source_line): Narrow down the condition to
24432 skip a line-0 marker.
24434 2021-01-08 Sergei Trofimovich <siarheit@google.com>
24436 * ipa-modref.c (merge_call_side_effects): Fix
24437 linebreak split by reordering two print calls.
24439 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
24441 * config/s390/vector.md (*tf_to_fprx2_0): Rename from
24442 "*mov_tf_to_fprx2_0" for consistency, fix constraint.
24443 (*tf_to_fprx2_1): Rename from "*mov_tf_to_fprx2_1" for
24444 consistency, fix constraint.
24446 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
24448 * config/s390/s390-c.c (s390_def_or_undef_macro): Accept
24449 callables instead of mask values.
24450 (struct target_flag_set_p): New predicate.
24451 (s390_cpu_cpp_builtins_internal): Define or undefine
24452 __LONG_DOUBLE_VX__ macro.
24454 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
24457 * config/i386/i386.c (x86_function_profiler): Use R10 and R11
24458 to call mcount in large model with PIC for NO_PROFILE_COUNTERS
24461 2021-01-08 Richard Biener <rguenther@suse.de>
24463 * tree-ssa-sccvn.c (pass_fre::execute): Reset the SCEV hash table.
24465 2021-01-08 Richard Biener <rguenther@suse.de>
24467 * tree-vect-slp.c (scalar_stmts_to_slp_tree_map_t): Fix.
24468 (vect_build_slp_tree): On cache hit release the matched
24469 scalar stmts vector.
24470 * tree-vect-stmts.c (vectorizable_store): Properly free
24471 vec_oprnds before possibly gathering them again.
24473 2021-01-08 Richard Biener <rguenther@suse.de>
24475 PR tree-optimization/98544
24476 * tree-vect-slp.c (vect_optimize_slp): Always materialize
24477 permutes at a permute node.
24479 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
24482 * config/i386/i386.c (x86_function_profiler): Use R10 to call
24483 mcount in large model. Sorry for large model with PIC.
24485 2021-01-08 Jakub Jelinek <jakub@redhat.com>
24488 * config/i386/i386.opt (ix86_cmodel, ix86_incoming_stack_boundary_arg,
24489 ix86_pmode, ix86_preferred_stack_boundary_arg, ix86_regparm,
24490 ix86_veclibabi_type): Remove x_ prefix, use TargetVariable instead of
24491 TargetSave and initialize for variables with enum types.
24492 (mfentry, mstack-protector-guard-reg=, mstack-protector-guard-offset=,
24493 mstack-protector-guard-symbol=): Add Save.
24494 * config/i386/i386-options.c (ix86_function_specific_save,
24495 ix86_function_specific_restore): Don't save or restore x_ix86_cmodel,
24496 x_ix86_incoming_stack_boundary_arg, x_ix86_pmode,
24497 x_ix86_preferred_stack_boundary_arg, x_ix86_regparm,
24498 x_ix86_veclibabi_type.
24500 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
24502 * config/aarch64/aarch64-sve.md (*cnot<mode>): Extend from
24503 SVE_FULL_I to SVE_I.
24504 (*cond_cnot<mode>_2, *cond_cnot<mode>_any): Likewise.
24506 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
24508 * config/aarch64/aarch64-sve.md (*cond_uxt<mode>_2): Extend from
24509 SVE_FULL_I to SVE_I.
24510 (*cond_uxt<mode>_any): Likewise.
24512 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24514 * config/aarch64/iterators.md (Vwhalf): New iterator.
24515 * config/aarch64/aarch64-simd.md (aarch64_<sur>adalp<mode>_3):
24517 (aarch64_<sur>adalp<mode>): ... This. Make more
24519 (<sur>sadv16qi): Adjust callsite of the above.
24520 * config/aarch64/aarch64-simd-builtins.def (sadalp, uadalp): New
24522 * config/aarch64/arm_neon.h (vpadal_s8): Reimplement using
24524 (vpadal_s16): Likewise.
24525 (vpadal_u8): Likewise.
24526 (vpadal_u16): Likewise.
24527 (vpadalq_s8): Likewise.
24528 (vpadalq_s16): Likewise.
24529 (vpadalq_s32): Likewise.
24530 (vpadalq_u8): Likewise.
24531 (vpadalq_u16): Likewise.
24532 (vpadalq_u32): Likewise.
24534 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24536 * config/aarch64/aarch64-simd.md (aarch64_<su>abd<mode>_3):
24538 (aarch64_<su>abd<mode>): ... This.
24539 (<sur>sadv16qi): Adjust callsite of the above.
24540 * config/aarch64/aarch64-simd-builtins.def (sabd, uabd): Define
24542 * config/aarch64/arm_neon.h (vabd_s8): Reimplement using
24544 (vabd_s16): Likewise.
24545 (vabd_s32): Likewise.
24546 (vabd_u8): Likewise.
24547 (vabd_u16): Likewise.
24548 (vabd_u32): Likewise.
24549 (vabdq_s8): Likewise.
24550 (vabdq_s16): Likewise.
24551 (vabdq_s32): Likewise.
24552 (vabdq_u8): Likewise.
24553 (vabdq_u16): Likewise.
24554 (vabdq_u32): Likewise.
24556 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24558 * config/aarch64/aarch64-simd-builtins.def (saba, uaba): Define
24560 * config/aarch64/arm_neon.h (vaba_s8): Implement using builtin.
24561 (vaba_s16): Likewise.
24562 (vaba_s32): Likewise.
24563 (vaba_u8): Likewise.
24564 (vaba_u16): Likewise.
24565 (vaba_u32): Likewise.
24566 (vabaq_s8): Likewise.
24567 (vabaq_s16): Likewise.
24568 (vabaq_s32): Likewise.
24569 (vabaq_u8): Likewise.
24570 (vabaq_u16): Likewise.
24571 (vabaq_u32): Likewise.
24573 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24575 * config/aarch64/aarch64-simd.md (aba<mode>_3): Rename to...
24576 (aarch64_<su>aba<mode>): ... This. Handle uaba as well.
24577 Change RTL pattern to match.
24579 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
24581 * common/config/riscv/riscv-common.c (riscv_current_subset_list): New.
24582 * config/riscv/riscv-c.c (riscv-subset.h): New.
24583 (INCLUDE_STRING): Define.
24584 (riscv_cpu_cpp_builtins): Add new style architecture extension
24586 * config/riscv/riscv-subset.h (riscv_subset_list::begin): New.
24587 (riscv_subset_list::end): New.
24588 (riscv_current_subset_list): New.
24590 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
24592 * common/config/riscv/riscv-common.c (RISCV_DONT_CARE_VERSION):
24593 Move to riscv-subset.h.
24594 (struct riscv_subset_t): Ditto.
24595 (class riscv_subset_list): Ditto.
24596 * config/riscv/riscv-subset.h (RISCV_DONT_CARE_VERSION): Move
24597 from riscv-common.c.
24598 (struct riscv_subset_t): Ditto.
24599 (class riscv_subset_list): Ditto.
24600 * config/riscv/t-riscv ($(common_out_file)): Add file
24603 2021-01-07 Jakub Jelinek <jakub@redhat.com>
24606 * config/i386/i386.md (*bmi_blsi_<mode>_cmp, *bmi_blsi_<mode>_ccno):
24607 New define_insn patterns.
24609 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
24611 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_UNARY:optab><mode>)
24612 (*cond_<SVE_INT_UNARY:optab><mode>_2): Extend from SVE_FULL_I to SVE_I.
24613 (*cond_<SVE_INT_UNARY:optab><mode>_any): Likewise.
24615 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
24617 PR tree-optimization/98560
24618 * internal-fn.def (IFN_VCONDU, IFN_VCONDEQ): Use type vec_cond.
24619 * internal-fn.c (vec_cond_mask_direct): Get the data mode from
24621 (vec_cond_direct): Likewise argument 2.
24622 (vec_condu_direct, vec_condeq_direct): Delete.
24623 (expand_vect_cond_optab_fn): Rename to...
24624 (expand_vec_cond_optab_fn): ...this, replacing old macro.
24625 (expand_vec_condu_optab_fn, expand_vec_condeq_optab_fn): Delete.
24626 (expand_vect_cond_mask_optab_fn): Rename to...
24627 (expand_vec_cond_mask_optab_fn): ...this, replacing old macro.
24628 (direct_vec_cond_mask_optab_supported_p): Treat the optab as a
24630 (direct_vec_cond_optab_supported_p): Likewise.
24631 (direct_vec_condu_optab_supported_p): Delete.
24632 (direct_vec_condeq_optab_supported_p): Delete.
24633 * gimple-isel.cc: Include internal-fn.h.
24634 (gimple_expand_vec_cond_expr): Check that IFN_VCONDEQ is supported
24637 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
24639 PR tree-optimization/98560
24640 * gimple-isel.cc (gimple_expand_vec_cond_expr): If we fail to use
24641 IFN_VCOND{,U,EQ}, fall back on IFN_VCOND_MASK.
24643 2021-01-07 Uroš Bizjak <ubizjak@gmail.com>
24645 * config/i386/i386.md (insn): Merge from plusminus_insn, shift_insn,
24646 rotate_insn and optab code attributes.
24647 Update all uses to merged code attribute.
24648 * config/i386/sse.md: Update all uses to merged code attribute.
24649 * config/i386/mmx.md: Update all uses to merged code attribute.
24651 2021-01-07 Jakub Jelinek <jakub@redhat.com>
24653 PR tree-optimization/98568
24654 * gimple-ssa-store-merging.c (bswap_view_convert): New function.
24655 (bswap_replace): Use it.
24657 2021-01-06 Vladimir N. Makarov <vmakarov@redhat.com>
24659 PR rtl-optimization/97978
24660 * lra-int.h (lra_hard_reg_split_p): New external.
24661 * lra.c (lra_hard_reg_split_p): New global.
24662 (lra): Set up lra_hard_reg_split_p after splitting a hard reg.
24663 * lra-assigns.c (lra_assign): Don't check allocation correctness
24664 after hard reg splitting.
24666 2021-01-06 Martin Sebor <msebor@redhat.com>
24669 * builtins.c (new_delete_mismatch_p): New overload.
24670 (new_delete_mismatch_p (tree, tree)): Call it.
24672 2021-01-06 Alexandre Oliva <oliva@adacore.com>
24674 * Makefile.in (T_GLIMITS_H): New.
24675 (stmp-int-hdrs): Depend on it, use it.
24676 * config/t-vxworks (T_GLIMITS_H): Override it.
24677 (vxw-glimits.h): New.
24679 2021-01-06 Richard Biener <rguenther@suse.de>
24681 PR tree-optimization/98513
24682 * value-range.cc (intersect_ranges): Compare the upper bounds
24683 for the expected relation.
24685 2021-01-06 Gerald Pfeifer <gerald@pfeifer.com>
24688 2020-12-28 Gerald Pfeifer <gerald@pfeifer.com>
24690 * doc/standards.texi (HSAIL): Remove section.
24692 2021-01-05 Samuel Thibault <samuel.thibault@ens-lyon.org>
24694 * configure: Re-generate.
24696 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24698 * doc/invoke.texi (-std=c++20): Adjust for the publication of
24699 ISO 14882:2020 standard.
24700 * doc/standards.texi: Likewise.
24702 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24704 PR tree-optimization/94802
24705 * expr.h (maybe_optimize_sub_cmp_0): Declare.
24706 * expr.c: Include tree-pretty-print.h and flags.h.
24707 (maybe_optimize_sub_cmp_0): New function.
24708 (do_store_flag): Use it.
24709 * cfgexpand.c (expand_gimple_cond): Likewise.
24711 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
24713 * mux-utils.h (pointer_mux::m_ptr): Tweak description of contents.
24714 * rtlanal.c (simple_regno_set): Tweak description to clarify the
24717 2021-01-05 Richard Biener <rguenther@suse.de>
24719 PR tree-optimization/98516
24720 * tree-vect-slp.c (vect_optimize_slp): Permute the incoming
24721 lanes when materializing on a VEC_PERM node.
24722 (vectorizable_slp_permutation): Dump the permute properly.
24724 2021-01-05 Richard Biener <rguenther@suse.de>
24726 * tree-vect-slp.c (vect_slp_region): Move debug counter
24727 to cover individual subgraphs.
24729 2021-01-05 Richard Biener <rguenther@suse.de>
24731 PR tree-optimization/98428
24732 * tree-vect-slp.c (vect_build_slp_tree_1): Properly reject
24733 vector lane extracts for loop vectorization.
24735 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24737 PR tree-optimization/98514
24738 * tree-ssa-reassoc.c (bb_rank): Change type from long * to
24740 (operand_rank): Change type from hash_map<tree, long> to
24741 hash_map<tree, int64_t>.
24742 (phi_rank): Change return type from long to int64_t.
24743 (loop_carried_phi): Change block_rank variable type from long to
24745 (propagate_rank): Change return type, rank parameter type and
24746 op_rank variable type from long to int64_t.
24747 (find_operand_rank): Change return type from long to int64_t
24748 and change slot variable type from long * to int64_t *.
24749 (insert_operand_rank): Change rank parameter type from long to
24751 (get_rank): Change return type and rank variable type from long to
24752 int64_t. Use PRId64 instead of ld to print the rank.
24753 (init_reassoc): Change rank variable type from long to int64_t
24754 and adjust correspondingly bb_rank and operand_rank initialization.
24756 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24758 PR tree-optimization/96928
24759 * tree-ssa-phiopt.c (xor_replacement): New function.
24760 (tree_ssa_phiopt_worker): Call it.
24762 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24764 PR tree-optimization/96930
24765 * match.pd ((A / (1 << B)) -> (A >> B)): If A is extended
24766 from narrower value which has the same type as 1 << B, perform
24767 the right shift on the narrower value followed by extension.
24769 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24771 PR tree-optimization/96239
24772 * gimple-ssa-store-merging.c (maybe_optimize_vector_constructor): New
24774 (get_status_for_store_merging): Don't return BB_INVALID for blocks
24775 with potential bswap optimizable CONSTRUCTORs.
24776 (pass_store_merging::execute): Optimize vector CONSTRUCTORs with bswap
24779 2021-01-05 Richard Biener <rguenther@suse.de>
24781 PR tree-optimization/98381
24782 * tree.c (vector_element_bits): Properly compute bool vector
24784 * tree-vect-loop.c (vectorizable_live_operation): Properly
24785 compute the last lane bit offset.
24787 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
24790 * config/i386/sse.md (sse_cvtps2pi): Redefine as define_insn_and_split.
24791 Clear the top 64 bytes of the input XMM register.
24792 (sse_cvttps2pi): Ditto.
24794 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
24797 * config/i386/xopintrin.h (_mm256_cmov_si256): New.
24799 2021-01-05 H.J. Lu <hjl.tools@gmail.com>
24802 * config/i386/xmmintrin.h (_mm_extract_pi16): Cast to unsigned
24805 2021-01-05 Claudiu Zissulescu <claziss@synopsys.com>
24807 * config/arc/arc.md (maddsidi4_split): Use ACC_REG_FIRST.
24808 (umaddsidi4_split): Likewise.
24810 2021-01-05 liuhongt <hongtao.liu@intel.com>
24813 * config/i386/sse.md (*sse2_pmovskb_zexthisi): New
24814 define_insn_and_split for zero_extend of subreg HI of pmovskb
24816 (*sse2_pmovskb_zexthisi): Add new combine splitters for
24817 zero_extend of not of subreg HI of pmovskb result.
24819 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
24822 * explow.c (convert_memory_address_addr_space_1): Handle UNSPECs
24824 * config/aarch64/aarch64.c (aarch64_expand_mov_immediate): Use
24825 convert_memory_address to convert symbolic immediates to ptr_mode
24826 before forcing them to memory.
24828 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
24830 PR rtl-optimization/97144
24831 * recog.c (constrain_operands): Initialize matching_operand
24832 for each alternative, rather than only doing it once.
24834 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
24836 PR rtl-optimization/98403
24837 * rtl-ssa/changes.cc (function_info::finalize_new_accesses): Explain
24838 why we don't remove call clobbers.
24839 (function_info::apply_changes_to_insn): Don't attempt to add
24840 call clobbers here.
24842 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
24844 PR tree-optimization/98371
24845 * tree-vect-loop.c (vect_reanalyze_as_main_loop): New function.
24846 (vect_analyze_loop): If an epilogue loop appears to be cheaper
24847 than the main loop, re-analyze it as a main loop before adopting
24850 2021-01-05 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
24853 * configure.ac (NETLIBS): Determine using AX_LIB_SOCKET_NSL.
24854 * aclocal.m4, configure: Regenerate.
24855 * Makefile.in (NETLIBS): Define.
24856 (BACKEND): Remove $(CODYLIB).
24858 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24860 PR rtl-optimization/98334
24861 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
24862 Optimize (X - 1) * Y + Y to X * Y or (X + 1) * Y - Y to X * Y.
24864 2021-01-05 Bernd Edlinger <bernd.edlinger@hotmail.de>
24866 * tree-inline.c (expand_call_inline): Restore input_location.
24867 Return result from recursive call.
24869 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
24871 PR tree-optimization/95401
24872 * config/aarch64/aarch64-sve-builtins.cc
24873 (gimple_folder::load_store_cookie): Use bits rather than bytes
24874 for the alignment argument to IFN_MASK_LOAD and IFN_MASK_STORE.
24875 * gimple-fold.c (gimple_fold_mask_load_store_mem_ref): Likewise.
24876 * tree-vect-stmts.c (vectorizable_store): Likewise.
24877 (vectorizable_load): Likewise.
24879 2021-01-04 Richard Biener <rguenther@suse.de>
24881 PR tree-optimization/98308
24882 * tree-vect-stmts.c (vectorizable_load): Set invariant mask
24885 2021-01-04 Jakub Jelinek <jakub@redhat.com>
24887 PR tree-optimization/95771
24888 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Handle types
24889 with precision smaller than int's precision and types with precision
24890 twice as large as long long. Formatting fixes.
24892 2021-01-04 Richard Biener <rguenther@suse.de>
24894 PR tree-optimization/98464
24895 * tree-ssa-sccvn.c (vn_valueize_for_srt): Rename from ...
24896 (vn_valueize_wrapper): ... this. Temporarily adjust vn_context_bb.
24897 (process_bb): Adjust.
24899 2021-01-04 Matthew Malcomson <matthew.malcomson@arm.com>
24902 * doc/invoke.texi (-fsanitize=address): Fix wording describing
24903 clash with -fsanitize=hwaddress.
24905 2021-01-04 Richard Biener <rguenther@suse.de>
24907 PR tree-optimization/98282
24908 * tree-ssa-sccvn.c (vn_get_stmt_kind): Classify tcc_reference on
24909 invariants as VN_NARY.
24911 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
24914 * config/aarch64/aarch64-simd.md (aarch64_combine<mode>): Accept
24915 aarch64_simd_reg_or_zero for operand 2. Use the combinez patterns
24916 to handle zero operands.
24918 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
24920 * config/aarch64/aarch64.c (offset_6bit_signed_scaled_p): New function.
24921 (offset_6bit_unsigned_scaled_p): Fix typo in comment.
24922 (aarch64_sve_prefetch_operand_p): Accept MUL VLs in the range
24925 2021-01-04 Richard Biener <rguenther@suse.de>
24927 PR tree-optimization/98393
24928 * tree-vect-slp.c (vect_build_slp_tree): Properly zero matches
24929 when hitting the limit.
24931 2021-01-04 Richard Biener <rguenther@suse.de>
24933 PR tree-optimization/98291
24934 * tree-vect-loop.c (vectorizable_reduction): Bypass
24935 associativity check for SLP reductions with VF 1.
24937 2021-01-04 Jakub Jelinek <jakub@redhat.com>
24939 PR tree-optimization/96782
24940 * match.pd (x == ~x -> false, x != ~x -> true): New simplifications.
24942 2021-01-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
24944 * collect-utils.c (collect_execute): Check dumppfx.
24945 * collect2.c (maybe_run_lto_and_relink, do_link): Pass atsuffix
24946 to collect_execute.
24947 (do_link): Add new parameter atsuffix.
24948 (main): Handle -dumpdir option. Skip one argument for
24949 -o, -isystem and -B options.
24950 * gcc.c (make_at_file): New helper function.
24951 (close_at_file): Use it.
24953 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
24955 * config/darwin.h (MIN_LD64_NO_COAL_SECTS): Adjust.
24956 Amend handling for LD64_VERSION fallback defaults.
24958 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
24960 * config.gcc: Compute default version information
24961 from the configured target. Likewise defaults for
24963 * config/darwin10.h: Removed.
24964 * config/darwin12.h: Removed.
24965 * config/darwin9.h: Removed.
24966 * config/rs6000/darwin8.h: Removed.
24968 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
24970 * config/darwin9.h (ASM_OUTPUT_ALIGNED_COMMON): Delete.
24972 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
24974 * config/darwin9.h (STACK_CHECK_STATIC_BUILTIN): Move from here..
24975 * config/darwin.h (STACK_CHECK_STATIC_BUILTIN): .. to here.
24977 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
24979 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move from
24981 * config/darwin.h (LINK_GCC_C_SEQUENCE_SPEC): ... to here.
24983 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
24985 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move the spec
24986 for the Darwin10 unwinder stub from here ...
24987 * config/darwin.h (LINK_COMMAND_SPEC_A): ... to here.
24989 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
24991 * config/darwin.h (DSYMUTIL_SPEC): Default to DWARF
24992 (ASM_DEBUG_SPEC):Only define if the assembler supports
24994 (PREFERRED_DEBUGGING_TYPE): Default to DWARF.
24995 (DARWIN_PREFER_DWARF): Define.
24996 * config/darwin9.h (PREFERRED_DEBUGGING_TYPE): Remove.
24997 (DARWIN_PREFER_DWARF): Likewise
24998 (DSYMUTIL_SPEC): Likewise.
24999 (COLLECT_RUN_DSYMUTIL): Likewise.
25000 (ASM_DEBUG_SPEC): Likewise.
25001 (ASM_DEBUG_OPTION_SPEC): Likewise.
25003 2021-01-02 Jan Hubicka <jh@suse.cz>
25005 * cfg.c (free_block): ggc_free bb.
25007 2021-01-01 Jakub Jelinek <jakub@redhat.com>
25009 * gcc.c (process_command): Update copyright notice dates.
25010 * gcov-dump.c (print_version): Ditto.
25011 * gcov.c (print_version): Ditto.
25012 * gcov-tool.c (print_version): Ditto.
25013 * gengtype.c (create_file): Ditto.
25014 * doc/cpp.texi: Bump @copying's copyright year.
25015 * doc/cppinternals.texi: Ditto.
25016 * doc/gcc.texi: Ditto.
25017 * doc/gccint.texi: Ditto.
25018 * doc/gcov.texi: Ditto.
25019 * doc/install.texi: Ditto.
25020 * doc/invoke.texi: Ditto.
25022 2021-01-01 Jakub Jelinek <jakub@redhat.com>
25024 * ChangeLog-2020: Rotate ChangeLog. New file.
25027 Copyright (C) 2021 Free Software Foundation, Inc.
25029 Copying and distribution of this file, with or without modification,
25030 are permitted in any medium without royalty provided the copyright
25031 notice and this notice are preserved.