1 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
3 PR tree-optimization/102546
4 * range-op.cc (operator_lshift::op1_range): Teach range-ops that
5 X << Y is non-zero implies X is also non-zero.
7 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
9 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
11 * config/aarch64/aarch64-tune.md: Regenerate.
12 * doc/invoke.texi: Update docs.
14 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
16 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
18 * config/aarch64/aarch64-tune.md: Regenerate.
19 * doc/invoke.texi: Update docs.
21 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
23 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
25 * config/aarch64/aarch64-tune.md: Regenerate.
26 * doc/invoke.texi: Update docs.
28 2021-10-01 Martin Sebor <msebor@redhat.com>
31 * doc/invoke.texi (-Waddress): Update.
32 * gengtype.c (write_types): Avoid -Waddress.
33 * poly-int.h (POLY_SET_COEFF): Avoid using null.
35 2021-10-01 John David Anglin <danglin@gcc.gnu.org>
38 * config/pa/pa.c (pa_option_override): Default to dwarf version 4
41 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
43 * config/aarch64/aarch64.h (AARCH64_FL_V9): Update value.
45 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
47 * gimple-range-path.cc (path_range_query::compute_ranges): Use
49 * gimple-range-path.h (class path_range_query): Remove shadowed
51 (path_range_query::get_path_oracle): New.
53 2021-10-01 Jakub Jelinek <jakub@redhat.com>
54 Richard Biener <rguenther@suse.de>
57 * doc/invoke.texi (-fsanitize=integer-divide-by-zero): Remove
58 INT_MIN / -1 division detection from here ...
59 (-fsanitize=signed-integer-overflow): ... and add it here.
61 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
63 * config/aarch64/aarch64-arches.def (AARCH64_ARCH): Added
65 * config/aarch64/aarch64.h (AARCH64_FL_V9): New.
66 (AARCH64_FL_FOR_ARCH9): New flags for Armv9-A.
67 (AARCH64_ISA_V9): New ISA flag.
68 * doc/invoke.texi: Update docs.
70 2021-10-01 Martin Liska <mliska@suse.cz>
72 * toplev.c (toplev::main): Save decoded optimization options.
73 * toplev.h (save_opt_decoded_options): New.
74 * doc/extend.texi: Be more clear about optimize and target
77 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
79 * explow.c: Include langhooks.h.
80 (set_stack_check_libfunc): Build a proper function type.
82 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
85 * config/i386/i386.c (legitimate_pic_address_disp_p): For PE-COFF do
86 not return true for external weak function symbols in medium model.
88 2021-10-01 Jakub Jelinek <jakub@redhat.com>
90 * tree.h (OMP_CLAUSE_ORDER_REPRODUCIBLE): Define.
91 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_ORDER>: Print
92 reproducible: for OMP_CLAUSE_ORDER_REPRODUCIBLE.
93 * omp-general.c (omp_extract_for_data): If OMP_CLAUSE_ORDER is seen
94 without OMP_CLAUSE_ORDER_UNCONSTRAINED, overwrite sched_kind to
95 OMP_CLAUSE_SCHEDULE_STATIC.
97 2021-10-01 Richard Biener <rguenther@suse.de>
100 * tree-inline.c (setup_one_parameter): Avoid substituting
101 an invariant into contexts where a GIMPLE register is not valid.
103 2021-09-30 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
105 * config/arm/arm-cpus.in: Add Cortex-R52+ CPU.
106 * config/arm/arm-tables.opt: Regenerate.
107 * config/arm/arm-tune.md: Regenerate.
108 * doc/invoke.texi: Update docs.
110 2021-09-30 Uroš Bizjak <ubizjak@gmail.com>
113 * config/i386/i386.md
114 (sign_extend:WIDE (any_logic:NARROW (memory, immediate)) splitters):
117 2021-09-30 Tobias Burnus <tobias@codesourcery.com>
119 * omp-low.c (omp_runtime_api_call): Add omp_aligned_{,c}alloc and
120 omp_{c,re}alloc, fix omp_alloc/omp_free.
122 2021-09-30 Martin Liska <mliska@suse.cz>
124 * defaults.h (ASM_OUTPUT_ASCII): Do not hide global variable
125 asm_out_file and stream directly to MYFILE.
127 2021-09-30 Richard Biener <rguenther@suse.de>
129 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
130 Restore and fix condition under which we apply npeel to
131 the DRs misalignment value.
133 2021-09-30 Richard Biener <rguenther@suse.de>
135 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
136 Fix npeel check for variable amount of peeling.
138 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
140 * lto-wrapper.c (run_gcc): Plug snprintf overflow.
142 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
144 * gimple-range.cc (gimple_ranger::debug): New.
145 * gimple-range.h (class gimple_ranger): Add debug.
147 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
150 * tree-vrp.c (hybrid_threader::~hybrid_threader): Free m_query.
152 2021-09-29 Indu Bhagat <indu.bhagat@oracle.com>
155 * btfout.c (GTY): Add GTY (()) albeit for cosmetic only purpose.
156 (btf_finalize): Empty the hash_map btf_var_ids.
158 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
160 * tree-vrp.c (thread_through_all_blocks): Return bool.
161 (execute_vrp_threader): Return TODO_* flags.
162 (pass_data_vrp_threader): Set todo_flags_finish to 0.
164 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
166 * timevar.def (TV_TREE_VRP_THREADER): New.
167 * tree-vrp.c: Use TV_TREE_VRP_THREADER for VRP threader pass.
169 2021-09-29 David Faust <david.faust@oracle.com>
171 * config.gcc (bpf-*-*): Do not overwrite extra_headers.
173 2021-09-29 Jonathan Wright <jonathan.wright@arm.com>
175 * config/aarch64/aarch64-builtins.c (TYPES_BINOP_PPU): Define
176 new type qualifier enum.
177 (TYPES_TERNOP_SSSU): Likewise.
178 (TYPES_TERNOP_PPPU): Likewise.
179 * config/aarch64/aarch64-simd-builtins.def: Define PPU, SSU,
180 PPPU and SSSU builtin generator macros for qtbl1 and qtbx1
182 * config/aarch64/arm_neon.h (vqtbl1_p8): Use type-qualified
183 builtin and remove casts.
184 (vqtbl1_s8): Likewise.
185 (vqtbl1q_p8): Likewise.
186 (vqtbl1q_s8): Likewise.
187 (vqtbx1_s8): Likewise.
188 (vqtbx1_p8): Likewise.
189 (vqtbx1q_s8): Likewise.
190 (vqtbx1q_p8): Likewise.
191 (vtbl1_p8): Likewise.
192 (vtbl2_p8): Likewise.
193 (vtbx2_p8): Likewise.
195 2021-09-29 Richard Biener <rguenther@suse.de>
197 * tree-vect-data-refs.c (vect_dr_misalign_for_aligned_access):
199 (vect_update_misalignment_for_peel): Use it to update
200 misaligned to the value necessary for an aligned access.
201 (vect_get_peeling_costs_all_drs): Likewise.
202 (vect_enhance_data_refs_alignment): Likewise.
204 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
206 * config/aarch64/aarch64.c (aarch64_expand_cpymem): Count number of
207 emitted operations and adjust heuristic for code size.
209 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
211 * config/aarch64/aarch64.c (aarch64_expand_setmem): Count number of
212 emitted operations and adjust heuristic for code size.
214 2021-09-29 Jakub Jelinek <jakub@redhat.com>
217 * gimplify.c (gimplify_scan_omp_clauses): Use omp_check_private even
218 in OMP_SCOPE clauses, not just on worksharing construct clauses.
220 2021-09-28 Geng Qi <gengqi@linux.alibaba.com>
222 * config/riscv/riscv.md (mulv<mode>4): Call gen_smul<mode>3_highpart.
223 (<u>mulditi3): Call <su>muldi3_highpart.
224 (<u>muldi3_highpart): Rename to <su>muldi3_highpart.
225 (<u>mulsidi3): Call <su>mulsi3_highpart.
226 (<u>mulsi3_highpart): Rename to <su>mulsi3_highpart.
228 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
230 * config/darwin.h (DSYMUTIL_SPEC): Recognize D sources.
232 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
234 * config/rs6000/darwin.h (FIXED_R13): Add for PPC64.
235 (FIRST_SAVED_GP_REGNO): Save from R13 even when it is one
238 2021-09-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
240 * config/aarch64/aarch64.h (AARCH64_FL_LS64): Define
241 (AARCH64_FL_V8_7): Likewise.
242 (AARCH64_FL_FOR_ARCH8_7): Likewise.
243 * config/aarch64/aarch64-arches.def (armv8.7-a): Define.
244 * config/aarch64/aarch64-option-extensions.def (ls64): Define.
245 * doc/invoke.texi: Document the above.
247 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
249 * dbgcnt.c (dbg_cnt_counter): New.
250 * dbgcnt.h (dbg_cnt_counter): New.
251 * dumpfile.c (dump_options): Add entry for TDF_THREADING.
252 * dumpfile.h (enum dump_flag): Add TDF_THREADING.
253 * gimple-range-path.cc (DEBUG_SOLVER): Use TDF_THREADING.
254 * tree-ssa-threadupdate.c (dump_jump_thread_path): Dump out
257 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
259 * cfgcleanup.c (pass_jump::execute): Check
260 flag_expensive_optimizations.
261 (pass_jump_after_combine::gate): Same.
262 * doc/invoke.texi (-fthread-jumps): Enable for -O1.
263 * opts.c (default_options_table): Enable -fthread-jumps at -O1.
264 * tree-ssa-threadupdate.c
265 (fwd_jt_path_registry::remove_jump_threads_including): Bail unless
268 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
270 * tree-ssa-reassoc.c (biased_names): New global.
271 (propagate_bias_p): New function.
272 (loop_carried_phi): Remove.
273 (propagate_rank): Propagate bias along single uses.
274 (get_rank): Update biased_names when needed.
276 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
278 * passes.def (pass_reassoc): Rename parameter to early_p.
279 * tree-ssa-reassoc.c (reassoc_bias_loop_carried_phi_ranks_p):
281 (phi_rank): Don't bias loop-carried phi ranks
282 before vectorization pass.
283 (execute_reassoc): Add bias_loop_carried_phi_ranks_p parameter.
284 (pass_reassoc::pass_reassoc): Add bias_loop_carried_phi_ranks_p
286 (pass_reassoc::set_param): Set bias_loop_carried_phi_ranks_p
288 (pass_reassoc::execute): Pass bias_loop_carried_phi_ranks_p to
290 (pass_reassoc::bias_loop_carried_phi_ranks_p): New member.
292 2021-09-28 Jakub Jelinek <jakub@redhat.com>
295 * config/i386/i386.c (standard_80387_constant_p): Don't recognize
296 special 80387 instruction XFmode constants if flag_rounding_math.
298 2021-09-28 Richard Biener <rguenther@suse.de>
300 PR tree-optimization/100112
301 * tree-ssa-sccvn.c (visit_reference_op_load): Record the
302 referece into the hashtable twice in case last_vuse is
303 different from the original vuse on the stmt.
305 2021-09-28 Jakub Jelinek <jakub@redhat.com>
308 * gimplify.c (gimplify_adjust_omp_clauses_1): Don't call the
309 omp_finish_clause langhook on implicitly added OMP_CLAUSE_PRIVATE
310 clauses on SIMD constructs.
312 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
314 PR tree-optimization/102511
315 * gimple-range-path.cc (path_range_query::range_on_path_entry):
316 Return VARYING when nothing found.
318 2021-09-28 Hongyu Wang <hongyu.wang@intel.com>
321 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): Add
323 (VALID_SSE2_REG_VHF_MODE): Add V4HFmode and V2HFmode.
324 (VALID_MMX_REG_MODE): Add V4HFmode.
325 (SSE_REG_MODE_P): Replace VALID_AVX512FP16_REG_MODE with
326 vector mode condition.
327 * config/i386/i386.c (classify_argument): Parse V4HF/V2HF
329 (function_arg_32): Add V4HFmode.
330 (function_arg_advance_32): Likewise.
331 * config/i386/i386.md (mode): Add V4HF/V2HF.
332 (MODE_SIZE): Likewise.
333 * config/i386/mmx.md (MMXMODE): Add V4HF mode.
334 (V_32): Add V2HF mode.
335 (VHF_32_64): New mode iterator.
336 (*mov<mode>_internal): Adjust sse alternatives to support
338 (*mov<mode>_internal): Adjust sse alternatives to support
340 (<insn><mode>3): New define_insn for add/sub/mul/div.
342 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
344 * tree-ssa-threadbackward.c (pass_thread_jumps::gate): Check
346 (pass_early_thread_jumps::gate): Same.
347 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
348 Return if !flag_thread_jumps.
349 * tree-ssa-threadupdate.c
350 (jt_path_registry::register_jump_thread): Assert that
351 flag_thread_jumps is true.
353 2021-09-28 liuhongt <hongtao.liu@intel.com>
356 (simplify_context::simplify_binary_operation_1): Relax
357 condition of simplifying (vec_concat:M (vec_select op0
358 index0)(vec_select op1 index1)) to allow different modes
359 between op0 and M, but have same inner mode.
361 2021-09-28 liuhongt <hongtao.liu@intel.com>
363 * config/i386/i386-expand.c (emit_reduc_half): Handle
364 V8HF/V16HF/V32HFmode.
365 * config/i386/sse.md (REDUC_SSE_PLUS_MODE): Add V8HF.
366 (REDUC_SSE_SMINMAX_MODE): Ditto.
367 (REDUC_PLUS_MODE): Add V16HF and V32HF.
368 (REDUC_SMINMAX_MODE): Ditto.
370 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
372 * gimple-range-path.cc
373 (path_range_query::precompute_ranges_in_block): Rename to...
374 (path_range_query::compute_ranges_in_block): ...this.
375 (path_range_query::precompute_ranges): Rename to...
376 (path_range_query::compute_ranges): ...this.
377 (path_range_query::precompute_relations): Rename to...
378 (path_range_query::compute_relations): ...this.
379 (path_range_query::precompute_phi_relations): Rename to...
380 (path_range_query::compute_phi_relations): ...this.
381 * gimple-range-path.h: Rename precompute* to compute*.
382 * tree-ssa-threadbackward.c
383 (back_threader::find_taken_edge_switch): Same.
384 (back_threader::find_taken_edge_cond): Same.
385 * tree-ssa-threadedge.c
386 (hybrid_jt_simplifier::compute_ranges_from_state): Same.
387 (hybrid_jt_state::register_equivs_stmt): Inline...
388 * tree-ssa-threadedge.h: ...here.
390 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
392 * tree-vrp.c (lhs_of_dominating_assert): Remove.
393 (class vrp_jt_state): Remove.
394 (class vrp_jt_simplifier): Remove.
395 (vrp_jt_simplifier::simplify): Remove.
396 (class vrp_jump_threader): Remove.
397 (vrp_jump_threader::vrp_jump_threader): Remove.
398 (vrp_jump_threader::~vrp_jump_threader): Remove.
399 (vrp_jump_threader::before_dom_children): Remove.
400 (vrp_jump_threader::after_dom_children): Remove.
402 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
404 * passes.def (pass_vrp_threader): New.
405 * tree-pass.h (make_pass_vrp_threader): Add make_pass_vrp_threader.
406 * tree-ssa-threadedge.c (hybrid_jt_state::register_equivs_stmt): New.
407 (hybrid_jt_simplifier::hybrid_jt_simplifier): New.
408 (hybrid_jt_simplifier::simplify): New.
409 (hybrid_jt_simplifier::compute_ranges_from_state): New.
410 * tree-ssa-threadedge.h (class hybrid_jt_state): New.
411 (class hybrid_jt_simplifier): New.
412 * tree-vrp.c (execute_vrp): Remove ASSERT_EXPR based jump
414 (class hybrid_threader): New.
415 (hybrid_threader::hybrid_threader): New.
416 (hybrid_threader::~hybrid_threader): New.
417 (hybrid_threader::before_dom_children): New.
418 (hybrid_threader::after_dom_children): New.
419 (execute_vrp_threader): New.
420 (class pass_vrp_threader): New.
421 (make_pass_vrp_threader): New.
423 2021-09-27 Martin Liska <mliska@suse.cz>
425 * output.h (enum section_flag): New.
426 (SECTION_FORGET): Remove.
427 (SECTION_ENTSIZE): Make it (1UL << 8) - 1.
428 (SECTION_STYLE_MASK): Define it based on other enum
430 * varasm.c (switch_to_section): Remove unused handling of
433 2021-09-27 Martin Liska <mliska@suse.cz>
435 * common.opt: Add new variable flag_default_complex_method.
436 * opts.c (finish_options): Handle flags related to
437 x_flag_complex_method.
438 * toplev.c (process_options): Remove option handling related
439 to flag_complex_method.
441 2021-09-27 Richard Biener <rguenther@suse.de>
444 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid using
445 type_for_size, instead use int_mode_for_size.
447 2021-09-27 Andrew Pinski <apinski@marvell.com>
450 * gimplify.c (gimplify_save_expr): Return early
451 if the type of val is error_mark_node.
453 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
455 * tree-ssanames.c (ssa_name_has_boolean_range): Use
458 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
460 * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Remove
462 * tree-ssa-dom.c (cprop_operand): Convert to range_query API.
463 (cprop_into_stmt): Same.
464 (dom_opt_dom_walker::optimize_stmt): Same.
466 2021-09-27 Richard Biener <rguenther@suse.de>
468 PR tree-optimization/97351
469 PR tree-optimization/97352
470 PR tree-optimization/82426
471 * tree-vectorizer.h (dr_misalignment): Add vector type
473 (aligned_access_p): Likewise.
474 (known_alignment_for_access_p): Likewise.
475 (vect_supportable_dr_alignment): Likewise.
476 (vect_known_alignment_in_bytes): Likewise. Refactor.
477 (DR_MISALIGNMENT): Remove.
478 (vect_update_shared_vectype): Likewise.
479 * tree-vect-data-refs.c (dr_misalignment): Refactor, handle
480 a vector type with larger alignment requirement and apply
481 the negative step adjustment here.
482 (vect_calculate_target_alignment): Remove.
483 (vect_compute_data_ref_alignment): Get explicit vector type
484 argument, do not apply a negative step alignment adjustment
486 (vect_slp_analyze_node_alignment): Re-analyze alignment
487 when we re-visit the DR with a bigger desired alignment but
488 keep more precise results from smaller alignments.
489 * tree-vect-slp.c (vect_update_shared_vectype): Remove.
490 (vect_slp_analyze_node_operations_1): Do not update the
491 shared vector type on stmts.
492 * tree-vect-stmts.c (vect_analyze_stmt): Push/pop the
493 vector type of an SLP node to the representative stmt-info.
494 (vect_transform_stmt): Likewise.
496 2021-09-27 liuhongt <hongtao.liu@intel.com>
499 2021-09-09 liuhongt <hongtao.liu@intel.com>
502 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
503 (reduc_plus_scal_v4sf): .. this, New define_expand.
504 (reduc_plus_scal_v2df): .. and this, New define_expand.
506 2021-09-26 liuhongt <hongtao.liu@intel.com>
508 * doc/extend.texi (Half-Precision): Remove storage only
509 description for _Float16 w/o avx512fp16.
511 2021-09-25 Dimitar Dimitrov <dimitar@dinux.eu>
513 * config/pru/constraints.md (Rrio): New constraint.
514 * config/pru/predicates.md (regio_operand): New predicate.
515 * config/pru/pru-pragma.c (pru_register_pragmas): Register
516 the __regio_symbol address space.
517 * config/pru/pru-protos.h (pru_symref2ioregno): Declaration.
518 * config/pru/pru.c (pru_symref2ioregno): New helper function.
519 (pru_legitimate_address_p): Remove.
520 (pru_addr_space_legitimate_address_p): Use the address space
522 (pru_nongeneric_pointer_addrspace): New helper function.
523 (pru_insert_attributes): New function to validate __regio_symbol
525 (TARGET_INSERT_ATTRIBUTES): New macro.
526 (TARGET_LEGITIMATE_ADDRESS_P): Remove.
527 (TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P): New macro.
528 * config/pru/pru.h (enum reg_class): Add REGIO_REGS class.
529 * config/pru/pru.md (*regio_readsi): New pattern to read I/O
531 (*regio_nozext_writesi): New pattern to write to I/O registers.
532 (*regio_zext_write_r30<EQS0:mode>): Ditto.
533 * doc/extend.texi: Document the new PRU Named Address Space.
535 2021-09-24 Patrick Palka <ppalka@redhat.com>
539 * real.c (encode_ieee_double): Avoid unwanted sign extension.
540 (encode_ieee_quad): Likewise.
542 2021-09-24 Vladimir Makarov <vmakarov@redhat.com>
544 PR rtl-optimization/102147
545 * ira-build.c (ira_conflict_vector_profitable_p): Make
546 profitability calculation independent of host compiler pointer and
549 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
551 * gimple-range-path.cc (path_range_query::path_range_query):
552 Move debugging header...
553 (path_range_query::precompute_ranges): ...here.
554 (path_range_query::internal_range_of_expr): Do not call
555 range_on_path_entry if NAME is defined in the current block.
557 2021-09-24 Richard Biener <rguenther@suse.de>
559 * cfghooks.c (verify_flow_info): Verify unallocated BB and
560 edge flags are not set.
562 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
564 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
566 (jt_path_registry::register_jump_thread): Call
567 cancel_invalid_paths.
568 * tree-ssa-threadupdate.h (class jt_path_registry): Add
569 cancel_invalid_paths.
571 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
573 PR tree-optimization/102400
574 * tree-ssa-sccvn.c (vn_reference_insert_pieces): Initialize
575 result_vdef to zero value.
577 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
579 PR tree-optimization/102451
580 * tree-ssa-dse.c (delete_dead_or_redundant_call): Record bb of stmt
583 2021-09-24 Hongyu Wang <hongyu.wang@intel.com>
585 * config/i386/sse.md (cond_<insn><mode>): Extend to support
587 (cond_mul<mode>): Likewise.
588 (cond_div<mode>): Likewise.
589 (cond_<code><mode>): Likewise.
590 (cond_fma<mode>): Likewise.
591 (cond_fms<mode>): Likewise.
592 (cond_fnma<mode>): Likewise.
593 (cond_fnms<mode>): Likewise.
595 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
597 PR tree-optimization/102463
598 * gimple-range-fold.cc (fold_using_range::relation_fold_and_or): If
599 there is no range-ops handler, don't look for a relation.
601 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
603 * gimple-range-cache.cc (ranger_cache::ranger_cache): Take
604 non-executable_edge flag as parameter.
605 * gimple-range-cache.h (ranger_cache): Adjust prototype.
606 * gimple-range-gori.cc (gori_compute::gori_compute): Take
607 non-executable_edge flag as parameter.
608 (gori_compute::outgoing_edge_range_p): Check new flag.
609 * gimple-range-gori.h (gori_compute): Adjust prototype.
610 * gimple-range.cc (gimple_ranger::gimple_ranger): Create new flag.
611 (gimple_ranger::range_on_edge): Check new flag.
612 * gimple-range.h (gimple_ranger::non_executable_edge_flag): New.
613 * gimple-ssa-evrp.c (rvrp_folder): Pass ranger flag to simplifer.
614 (hybrid_folder::hybrid_folder): Set ranger non-executable flag value.
615 (hybrid_folder::fold_stmt): Set flag value in the simplifer.
616 * vr-values.c (simplify_using_ranges::set_and_propagate_unexecutable):
617 Use not_executable flag if provided inmstead of EDGE_EXECUTABLE.
618 (simplify_using_ranges::simplify_switch_using_ranges): Clear
619 EDGE_EXECUTABLE like it originally did.
620 (simplify_using_ranges::cleanup_edges_and_switches): Clear any
621 NON_EXECUTABLE flags.
622 (simplify_using_ranges::simplify_using_ranges): Adjust.
623 * vr-values.h (class simplify_using_ranges): Adjust.
624 (simplify_using_ranges::set_range_query): Add non-executable flag param.
626 2021-09-23 Bill Schmidt <wschmidt@linux.ibm.com>
629 * config/rs6000/rs6000-call.c (rs6000_aggregate_candidate): Detect
630 zero-width bit fields and return indicator.
631 (rs6000_discover_homogeneous_aggregate): Diagnose when the
632 presence of a zero-width bit field changes parameter passing in
635 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
637 * gimple-range-fold.cc (fold_using_range::range_of_phi):
638 Remove dominator check.
640 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
642 * gimple-range-path.cc (path_range_query::precompute_relations):
643 Hoist edge calculations before using EDGE_SUCC.
645 2021-09-23 Jonathan Wakely <jwakely@redhat.com>
647 * configure.ac: Fix --with-multilib-list description.
648 * configure: Regenerate.
650 2021-09-23 Richard Biener <rguenther@suse.de>
652 PR tree-optimization/102448
653 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
654 Clear alignment info copied from DR_PTR_INFO.
656 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
658 * config/i386/i386-expand.c (ix86_use_mask_cmp_p): Enable
660 * config/i386/sse.md (sseintvecmodelower): Add HF vector modes.
661 (<avx512>_store<mode>_mask): Extend to support HF vector modes.
662 (vec_cmp<mode><avx512fmaskmodelower>): Likewise.
663 (vcond_mask_<mode><avx512fmaskmodelower>): Likewise.
664 (vcond<mode><mode>): New expander.
665 (vcond<mode><sseintvecmodelower>): Likewise.
666 (vcond<sseintvecmodelower><mode>): Likewise.
667 (vcondu<mode><sseintvecmodelower>): Likewise.
669 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
671 * config/i386/sse.md (extend<ssePHmodelower><mode>2):
673 (extendv4hf<mode>2): Likewise.
674 (extendv2hfv2df2): Likewise.
675 (trunc<mode><ssePHmodelower>2): Likewise.
676 (avx512fp16_vcvt<castmode>2ph_<mode>): Rename to ...
677 (trunc<mode>v4hf2): ... this, and drop constraints.
678 (avx512fp16_vcvtpd2ph_v2df): Rename to ...
679 (truncv2dfv2hf2): ... this, and likewise.
681 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
683 * config/i386/sse.md (float<floatunssuffix><mode><ssePHmodelower>2):
685 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>):
687 (float<floatunssuffix><mode>v4hf2): ... this, and drop constraints.
688 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Rename to ...
689 (float<floatunssuffix>v2div2hf2): ... this, and likewise.
691 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
693 * config/i386/i386.md (fix<fixunssuffix>_trunchf<mode>2): New expander.
694 (fixuns_trunchfhi2): Likewise.
695 (*fixuns_trunchfsi2zext): New define_insn.
696 * config/i386/sse.md (ssePHmodelower): New mode_attr.
697 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
698 New expander for same element vector fix_truncate.
699 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
700 Likewise for V4HF to V4SI/V4DI fix_truncate.
701 (fix<fixunssuffix>_truncv2hfv2di2):
702 Likeise for V2HF to V2DI fix_truncate.
704 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
706 * config/i386/i386.md (<code>hf3): New expander.
708 2021-09-23 liuhongt <hongtao.liu@intel.com>
710 * config/i386/sse.md (FMAMODEM): extend to handle FP16.
711 (VFH_SF_AVX512VL): Extend to handle HFmode.
712 (VF_SF_AVX512VL): Deleted.
714 2021-09-23 liuhongt <hongtao.liu@intel.com>
716 * config/i386/i386.md (rinthf2): New expander.
717 (nearbyinthf2): New expander.
719 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
721 * tree-ssa-dom.c (class dom_jump_threader_simplifier): Rename...
722 (class dom_jt_state): ...this and provide virtual overrides.
723 (dom_jt_state::register_equiv): New.
724 (class dom_jt_simplifier): Rename from
725 dom_jump_threader_simplifier.
726 (dom_jump_threader_simplifier::simplify): Rename...
727 (dom_jt_simplifier::simplify): ...to this.
728 (pass_dominator::execute): Use dom_jt_simplifier and
730 * tree-ssa-threadedge.c (jump_threader::jump_threader):
732 (jt_state::register_equivs_stmt): Abstract out...
733 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
735 (jump_threader::thread_around_empty_blocks): Update state.
736 (jump_threader::thread_through_normal_block): Same.
737 (jt_state::jt_state): Remove.
738 (jt_state::push): Remove pass specific bits. Keep block vector
740 (jt_state::append_path): New.
741 (jt_state::pop): Remove pass specific bits.
742 (jt_state::register_equiv): Same.
743 (jt_state::record_ranges_from_stmt): Same.
744 (jt_state::register_equivs_on_edge): Same. Rename...
745 (jt_state::register_equivs_edge): ...to this.
746 (jt_state::dump): New.
747 (jt_state::debug): New.
748 (jump_threader_simplifier::simplify): Remove.
749 (jt_state::get_path): New.
750 * tree-ssa-threadedge.h (class jt_simplifier): Make into a base
751 class. Expose common functionality as virtual methods.
752 (class jump_threader_simplifier): Same. Rename...
753 (class jt_simplifier): ...to this.
754 * tree-vrp.c (class vrp_jump_threader_simplifier): Rename...
755 (class vrp_jt_simplifier): ...to this. Provide pass specific
757 (class vrp_jt_state): New.
758 (vrp_jump_threader_simplifier::simplify): Rename...
759 (vrp_jt_simplifier::simplify): ...to this. Inline code from
760 what used to be the base class.
761 (vrp_jump_threader::vrp_jump_threader): Use vrp_jt_state and
764 2021-09-22 Tobias Burnus <tobias@codesourcery.com>
767 * doc/invoke.texi (-Wno-missing-include-dirs.): Document Fortran
770 2021-09-22 Roger Sayle <roger@nextmovesoftware.com>
771 Richard Biener <rguenther@suse.de>
773 * match.pd (negation simplifications): Implement some negation
774 folding transformations from fold-const.c's fold_negate_expr.
775 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Add a SIMPLIFY
776 argument, to control whether the op should be simplified prior
777 to looking up/assigning a value number.
778 (vn_nary_build_or_lookup): Update call to vn_nary_build_or_lookup_1.
779 (vn_nary_simplify): Likewise.
780 (visit_nary_op): Likewise, but when constructing a NEGATE_EXPR
781 now call vn_nary_build_or_lookup_1 disabling simplification.
783 2021-09-22 Jiufu Guo <guojiufu@linux.ibm.com>
785 PR tree-optimization/102087
786 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
787 Update bound/cmp/control for niter.
789 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
791 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
792 Move check for non-empty BB here.
793 (fur_source::register_outgoing_edges): ...from here.
795 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
797 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
798 Remove call to improve_range_with_equivs.
799 (path_range_query::improve_range_with_equivs): Remove
800 * gimple-range-path.h: Remove improve_range_with_equivs.
802 2021-09-22 dianhong xu <dianhong.xu@intel.com>
804 * config/i386/avx512fp16intrin.h:
805 (_mm512_mask_blend_ph): New intrinsic.
806 (_mm512_permutex2var_ph): Ditto.
807 (_mm512_permutexvar_ph): Ditto.
808 * config/i386/avx512fp16vlintrin.h:
809 (_mm256_mask_blend_ph): New intrinsic.
810 (_mm256_permutex2var_ph): Ditto.
811 (_mm256_permutexvar_ph): Ditto.
812 (_mm_mask_blend_ph): Ditto.
813 (_mm_permutex2var_ph): Ditto.
814 (_mm_permutexvar_ph): Ditto.
816 2021-09-22 dianhong xu <dianhong.xu@intel.com>
818 * config/i386/avx512fp16intrin.h: Add new intrinsics.
819 (_mm512_conj_pch): New intrinsic.
820 (_mm512_mask_conj_pch): Ditto.
821 (_mm512_maskz_conj_pch): Ditto.
822 * config/i386/avx512fp16vlintrin.h: Add new intrinsics.
823 (_mm256_conj_pch): New intrinsic.
824 (_mm256_mask_conj_pch): Ditto.
825 (_mm256_maskz_conj_pch): Ditto.
826 (_mm_conj_pch): Ditto.
827 (_mm_mask_conj_pch): Ditto.
828 (_mm_maskz_conj_pch): Ditto.
830 2021-09-22 dianhong xu <dianhong.xu@intel.com>
832 * config/i386/avx512fp16intrin.h (_MM512_REDUCE_OP): New macro
833 (_mm512_reduce_add_ph): New intrinsic.
834 (_mm512_reduce_mul_ph): Ditto.
835 (_mm512_reduce_min_ph): Ditto.
836 (_mm512_reduce_max_ph): Ditto.
837 * config/i386/avx512fp16vlintrin.h
838 (_MM256_REDUCE_OP/_MM_REDUCE_OP): New macro.
839 (_mm256_reduce_add_ph): New intrinsic.
840 (_mm256_reduce_mul_ph): Ditto.
841 (_mm256_reduce_min_ph): Ditto.
842 (_mm256_reduce_max_ph): Ditto.
843 (_mm_reduce_add_ph): Ditto.
844 (_mm_reduce_mul_ph): Ditto.
845 (_mm_reduce_min_ph): Ditto.
846 (_mm_reduce_max_ph): Ditto.
848 2021-09-22 dianhong xu <dianhong.xu@intel.com>
850 * config/i386/avx512fp16intrin.h (__m512h_u, __m256h_u,
851 __m128h_u): New typedef.
852 (_mm512_load_ph): New intrinsic.
853 (_mm256_load_ph): Ditto.
854 (_mm_load_ph): Ditto.
855 (_mm512_loadu_ph): Ditto.
856 (_mm256_loadu_ph): Ditto.
857 (_mm_loadu_ph): Ditto.
858 (_mm512_store_ph): Ditto.
859 (_mm256_store_ph): Ditto.
860 (_mm_store_ph): Ditto.
861 (_mm512_storeu_ph): Ditto.
862 (_mm256_storeu_ph): Ditto.
863 (_mm_storeu_ph): Ditto.
864 (_mm512_abs_ph): Ditto.
865 * config/i386/avx512fp16vlintrin.h
867 (_mm256_abs_ph): Ditto.
869 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
871 * config/s390/tpf.md (prologue_tpf, epilogue_tpf): Add cc clobber.
873 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
876 * config/s390/s390.c (s390_expand_insv): Emit a normal move if it
877 is actually a full copy of the source operand into the target.
878 Don't emit a strict low part move if source and target mode match.
880 2021-09-22 Jakub Jelinek <jakub@redhat.com>
883 * omp-expand.c (expand_omp_single): If region->exit is NULL,
884 assert region->entry is GIMPLE_OMP_SCOPE region and return.
886 2021-09-22 Jakub Jelinek <jakub@redhat.com>
888 * tree.h (OMP_CLAUSE_ALLOCATE_ALIGN): Define.
889 * tree.c (omp_clause_num_ops): Change number of OMP_CLAUSE_ALLOCATE
890 arguments from 2 to 3.
891 * tree-pretty-print.c (dump_omp_clause): Print allocator() around
892 allocate clause allocator and print align if present.
893 * omp-low.c (scan_sharing_clauses): Force allocate_map entry even
894 for omp_default_mem_alloc if align modifier is present. If align
895 modifier is present, use TREE_LIST to encode both allocator and
897 (lower_private_allocate, lower_rec_input_clauses, create_task_copyfn):
898 Handle align modifier on allocator clause if present.
900 2021-09-22 liuhongt <hongtao.liu@intel.com>
902 * config/i386/i386.md (define_attr "isa"): Add
904 (define_attr "enabled"): Correspond fma_or_avx512vl to
905 TARGET_FMA || TARGET_AVX512VL.
906 * config/i386/mmx.md (fmav2sf4): Extend to AVX512 fma.
911 2021-09-22 liuhongt <hongtao.liu@intel.com>
913 * config/i386/i386.md (cstorehf3): New define_expand.
915 2021-09-22 liuhongt <hongtao.liu@intel.com>
917 * config/i386/i386.md (<rounding_insn>hf2): New expander.
918 (sse4_1_round<mode>2): Extend from MODEF to MODEFH.
919 * config/i386/sse.md (*sse4_1_round<ssescalarmodesuffix>):
920 Extend from VF_128 to VFH_128.
922 2021-09-22 liuhongt <hongtao.liu@intel.com>
924 * config/i386/i386-features.c (i386-features.c): Handle
926 * config/i386/i386.md (sqrthf2): New expander.
927 (*sqrthf2): New define_insn.
929 (*<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
932 2021-09-22 liuhongt <hongtao.liu@intel.com>
934 * config/i386/avx512fp16intrin.h (_mm_mask_fcmadd_sch):
936 (_mm_mask3_fcmadd_sch): Likewise.
937 (_mm_maskz_fcmadd_sch): Likewise.
938 (_mm_fcmadd_sch): Likewise.
939 (_mm_mask_fmadd_sch): Likewise.
940 (_mm_mask3_fmadd_sch): Likewise.
941 (_mm_maskz_fmadd_sch): Likewise.
942 (_mm_fmadd_sch): Likewise.
943 (_mm_mask_fcmadd_round_sch): Likewise.
944 (_mm_mask3_fcmadd_round_sch): Likewise.
945 (_mm_maskz_fcmadd_round_sch): Likewise.
946 (_mm_fcmadd_round_sch): Likewise.
947 (_mm_mask_fmadd_round_sch): Likewise.
948 (_mm_mask3_fmadd_round_sch): Likewise.
949 (_mm_maskz_fmadd_round_sch): Likewise.
950 (_mm_fmadd_round_sch): Likewise.
951 (_mm_fcmul_sch): Likewise.
952 (_mm_mask_fcmul_sch): Likewise.
953 (_mm_maskz_fcmul_sch): Likewise.
954 (_mm_fmul_sch): Likewise.
955 (_mm_mask_fmul_sch): Likewise.
956 (_mm_maskz_fmul_sch): Likewise.
957 (_mm_fcmul_round_sch): Likewise.
958 (_mm_mask_fcmul_round_sch): Likewise.
959 (_mm_maskz_fcmul_round_sch): Likewise.
960 (_mm_fmul_round_sch): Likewise.
961 (_mm_mask_fmul_round_sch): Likewise.
962 (_mm_maskz_fmul_round_sch): Likewise.
963 * config/i386/i386-builtin.def: Add corresponding new builtins.
965 (avx512fp16_fmaddcsh_v8hf_maskz<round_expand_name>): New expander.
966 (avx512fp16_fcmaddcsh_v8hf_maskz<round_expand_name>): Ditto.
967 (avx512fp16_fma_<complexopname>sh_v8hf<mask_scalarcz_name><round_scalarcz_name>):
969 (avx512fp16_<complexopname>sh_v8hf_mask<round_name>): Ditto.
970 (avx512fp16_<complexopname>sh_v8hf<mask_scalarc_name><round_scalarcz_name>):
972 * config/i386/subst.md (mask_scalarcz_name): New.
973 (mask_scalarc_name): Ditto.
974 (mask_scalarc_operand3): Ditto.
975 (mask_scalarcz_operand4): Ditto.
976 (round_scalarcz_name): Ditto.
977 (round_scalarc_mask_operand3): Ditto.
978 (round_scalarcz_mask_operand4): Ditto.
979 (round_scalarc_mask_op3): Ditto.
980 (round_scalarcz_mask_op4): Ditto.
981 (round_scalarcz_constraint): Ditto.
982 (round_scalarcz_nimm_predicate): Ditto.
983 (mask_scalarcz): Ditto.
984 (mask_scalarc): Ditto.
985 (round_scalarcz): Ditto.
987 2021-09-22 liuhongt <hongtao.liu@intel.com>
989 * config/i386/avx512fp16intrin.h (_mm512_fcmadd_pch):
991 (_mm512_mask_fcmadd_pch): Likewise.
992 (_mm512_mask3_fcmadd_pch): Likewise.
993 (_mm512_maskz_fcmadd_pch): Likewise.
994 (_mm512_fmadd_pch): Likewise.
995 (_mm512_mask_fmadd_pch): Likewise.
996 (_mm512_mask3_fmadd_pch): Likewise.
997 (_mm512_maskz_fmadd_pch): Likewise.
998 (_mm512_fcmadd_round_pch): Likewise.
999 (_mm512_mask_fcmadd_round_pch): Likewise.
1000 (_mm512_mask3_fcmadd_round_pch): Likewise.
1001 (_mm512_maskz_fcmadd_round_pch): Likewise.
1002 (_mm512_fmadd_round_pch): Likewise.
1003 (_mm512_mask_fmadd_round_pch): Likewise.
1004 (_mm512_mask3_fmadd_round_pch): Likewise.
1005 (_mm512_maskz_fmadd_round_pch): Likewise.
1006 (_mm512_fcmul_pch): Likewise.
1007 (_mm512_mask_fcmul_pch): Likewise.
1008 (_mm512_maskz_fcmul_pch): Likewise.
1009 (_mm512_fmul_pch): Likewise.
1010 (_mm512_mask_fmul_pch): Likewise.
1011 (_mm512_maskz_fmul_pch): Likewise.
1012 (_mm512_fcmul_round_pch): Likewise.
1013 (_mm512_mask_fcmul_round_pch): Likewise.
1014 (_mm512_maskz_fcmul_round_pch): Likewise.
1015 (_mm512_fmul_round_pch): Likewise.
1016 (_mm512_mask_fmul_round_pch): Likewise.
1017 (_mm512_maskz_fmul_round_pch): Likewise.
1018 * config/i386/avx512fp16vlintrin.h (_mm_fmadd_pch):
1020 (_mm_mask_fmadd_pch): Likewise.
1021 (_mm_mask3_fmadd_pch): Likewise.
1022 (_mm_maskz_fmadd_pch): Likewise.
1023 (_mm256_fmadd_pch): Likewise.
1024 (_mm256_mask_fmadd_pch): Likewise.
1025 (_mm256_mask3_fmadd_pch): Likewise.
1026 (_mm256_maskz_fmadd_pch): Likewise.
1027 (_mm_fcmadd_pch): Likewise.
1028 (_mm_mask_fcmadd_pch): Likewise.
1029 (_mm_mask3_fcmadd_pch): Likewise.
1030 (_mm_maskz_fcmadd_pch): Likewise.
1031 (_mm256_fcmadd_pch): Likewise.
1032 (_mm256_mask_fcmadd_pch): Likewise.
1033 (_mm256_mask3_fcmadd_pch): Likewise.
1034 (_mm256_maskz_fcmadd_pch): Likewise.
1035 (_mm_fmul_pch): Likewise.
1036 (_mm_mask_fmul_pch): Likewise.
1037 (_mm_maskz_fmul_pch): Likewise.
1038 (_mm256_fmul_pch): Likewise.
1039 (_mm256_mask_fmul_pch): Likewise.
1040 (_mm256_maskz_fmul_pch): Likewise.
1041 (_mm_fcmul_pch): Likewise.
1042 (_mm_mask_fcmul_pch): Likewise.
1043 (_mm_maskz_fcmul_pch): Likewise.
1044 (_mm256_fcmul_pch): Likewise.
1045 (_mm256_mask_fcmul_pch): Likewise.
1046 (_mm256_maskz_fcmul_pch): Likewise.
1047 * config/i386/i386-builtin-types.def (V8HF_FTYPE_V8HF_V8HF_V8HF,
1048 V8HF_FTYPE_V16HF_V16HF_V16HF, V16HF_FTYPE_V16HF_V16HF_V16HF_UQI,
1049 V32HF_FTYPE_V32HF_V32HF_V32HF_INT,
1050 V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT): Add new builtin types.
1051 * config/i386/i386-builtin.def: Add new builtins.
1052 * config/i386/i386-expand.c: Handle new builtin types.
1053 * config/i386/subst.md (SUBST_CV): New.
1054 (maskc_name): Ditto.
1055 (maskc_operand3): Ditto.
1057 (sdc_maskz_name): Ditto.
1058 (sdc_mask_op4): Ditto.
1059 (sdc_mask_op5): Ditto.
1060 (sdc_mask_mode512bit_condition): Ditto.
1062 (round_maskc_operand3): Ditto.
1063 (round_sdc_mask_operand4): Ditto.
1064 (round_maskc_op3): Ditto.
1065 (round_sdc_mask_op4): Ditto.
1066 (round_saeonly_sdc_mask_operand5): Ditto.
1067 * config/i386/sse.md (unspec): Add complex fma unspecs.
1068 (avx512fmaskcmode): New.
1069 (UNSPEC_COMPLEX_F_C_MA): Ditto.
1070 (UNSPEC_COMPLEX_F_C_MUL): Ditto.
1071 (complexopname): Ditto.
1072 (<avx512>_fmaddc_<mode>_maskz<round_expand_name>): New expander.
1073 (<avx512>_fcmaddc_<mode>_maskz<round_expand_name>): Ditto.
1074 (fma_<complexopname>_<mode><sdc_maskz_name><round_name>): New
1076 (<avx512>_<complexopname>_<mode>_mask<round_name>): Ditto.
1077 (<avx512>_<complexopname>_<mode><maskc_name><round_name>): Ditto.
1079 2021-09-22 Kewen Lin <linkw@linux.ibm.com>
1081 * config/rs6000/rs6000.opt (rs6000-density-pct-threshold,
1082 rs6000-density-size-threshold, rs6000-density-penalty,
1083 rs6000-density-load-pct-threshold,
1084 rs6000-density-load-num-threshold): New parameter.
1085 * config/rs6000/rs6000.c (rs6000_density_test): Adjust with
1086 corresponding parameters.
1088 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1090 * gimple-range-path.cc (path_range_query::defined_outside_path):
1092 (path_range_query::range_on_path_entry): New.
1093 (path_range_query::internal_range_of_expr): Resolve unknowns
1095 (path_range_query::improve_range_with_equivs): New.
1096 (path_range_query::ssa_range_in_phi): Resolve unknowns with
1098 * gimple-range-path.h (class path_range_query): Add
1099 defined_outside_path, range_on_path_entry, and
1100 improve_range_with_equivs.
1102 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1104 * gimple-range-path.cc (path_range_query::add_to_imports): New.
1105 (path_range_query::add_copies_to_imports): New.
1106 (path_range_query::precompute_ranges): Call
1107 add_copies_to_imports.
1108 * gimple-range-path.h (class path_range_query): Add prototypes
1109 for add_copies_to_imports and add_to_imports.
1111 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1113 * gimple-range-path.cc (path_range_query::range_defined_in_block):
1114 Remove useless code.
1116 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1118 * gimple-range-fold.h (class fur_source): Make oracle protected.
1119 * gimple-range-path.cc (path_range_query::path_range_query): Add
1120 resolve argument. Initialize oracle.
1121 (path_range_query::~path_range_query): Delete oracle.
1122 (path_range_query::range_of_stmt): Adapt to use relations.
1123 (path_range_query::precompute_ranges): Pre-compute relations.
1124 (class jt_fur_source): New
1125 (jt_fur_source::jt_fur_source): New.
1126 (jt_fur_source::register_relation): New.
1127 (jt_fur_source::query_relation): New.
1128 (path_range_query::precompute_relations): New.
1129 (path_range_query::precompute_phi_relations): New.
1130 * gimple-range-path.h (path_range_query): Add resolve argument.
1131 Add oracle, precompute_relations, precompute_phi_relations.
1132 * tree-ssa-threadbackward.c (back_threader::back_threader): Pass
1133 resolve argument to solver.
1135 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1137 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
1138 Rename postfold_gcond_edges to register_outgoing_edges and
1140 (fold_using_range::postfold_gcond_edges): Rename...
1141 (fur_source::register_outgoing_edges): ...to this.
1142 * gimple-range-fold.h (postfold_gcond_edges): Rename to
1143 register_outgoing_edges and move to fur_source.
1145 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1147 * gimple-range-fold.cc (fold_using_range::range_of_phi): Check
1148 dom_info_available_p.
1150 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1152 * gimple-range-cache.cc (non_null_ref::non_null_ref): Use create
1153 and quick_grow_cleared instead of safe_grow_cleared.
1155 2021-09-21 Thomas Schwinge <thomas@codesourcery.com>
1158 * omp-oacc-neuter-broadcast.cc (oacc_do_neutering): Evaluate
1161 2021-09-21 Richard Earnshaw <rearnsha@arm.com>
1163 * configure.ac: Detect when the assembler supports new-style
1164 architecture extensions.
1165 * common/config/arm/arm-common.c (arm_rewrite_mcpu): Return
1166 the full CPU string if the assembler can grok it.
1167 (arm_rewrite_march): Likewise but for the architecture.
1168 * config.in: Regenerate.
1169 * configure: Regenerate.
1171 2021-09-21 Richard Biener <rguenther@suse.de>
1173 PR tree-optimization/102421
1174 * tree-vect-loop.c (vect_dissolve_slp_only_groups): Copy and
1175 adjust alignment info.
1177 2021-09-21 Kewen Lin <linkw@linux.ibm.com>
1179 * ipa-fnsummary.c (ipa_fn_summary_write): Remove inconsistent
1180 bitfield stream out.
1182 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
1184 * gimple-range-fold.cc (fold_using_range::range_of_phi): Ignore
1185 undefined edges, apply an equivalence if appropriate.
1186 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Return
1187 UNDEFINED if EDGE_EXECUTABLE is not set.
1188 * gimple-range.cc (gimple_ranger::gimple_ranger): Set all edges
1189 as EXECUTABLE upon startup.
1190 (gimple_ranger::range_on_edge): Return UNDEFINED for edges without
1191 EDGE_EXECUTABLE set.
1192 * vr-values.c (set_and_propagate_unexecutable): New.
1193 (simplify_using_ranges::fold_cond): Call set_and_propagate.
1194 (simplify_using_ranges::simplify_switch_using_ranges): Ditto.
1195 * vr-values.h: Add prototype.
1197 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
1199 * value-relation.cc (equiv_oracle::register_initial_def): New.
1200 (equiv_oracle::register_relation): Call register_initial_def.
1201 (equiv_oracle::add_equiv_to_block): New. Split register_relation.
1202 (relation_oracle::register_stmt): Check def block of PHI arguments.
1203 * value-relation.h (equiv_oracle): Add new prototypes.
1205 2021-09-20 Matthias Kretz <m.kretz@gsi.de>
1207 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
1208 Define __RECIPROCAL_MATH__, __NO_SIGNED_ZEROS__,
1209 __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__, and
1210 __ROUNDING_MATH__ according to their corresponding flags.
1211 * doc/cpp.texi: Document __RECIPROCAL_MATH__,
1212 __NO_SIGNED_ZEROS__, __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__,
1213 and __ROUNDING_MATH__.
1215 2021-09-20 Richard Biener <rguenther@suse.de>
1217 * tree-vect-stmts.c (vectorizable_load): Use the vectype
1220 2021-09-20 Richard Biener <rguenther@suse.de>
1222 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
1223 Do not compute alignment of the vectorized access here.
1225 2021-09-20 Richard Biener <rguenther@suse.de>
1227 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
1228 Store -1 for runtime alias peeling iterations.
1230 2021-09-20 Richard Biener <rguenther@suse.de>
1232 * config.gcc: Obsolete hppa[12]*-*-hpux10* and hppa[12]*-*-hpux11*.
1234 2021-09-20 Thomas Schwinge <thomas@codesourcery.com>
1236 * input.c (string_concat_db::record_string_concatenation)
1237 (string_concat_db::get_string_concatenation): Skip for
1238 'RESERVED_LOCATION_P'.
1240 2021-09-20 Richard Biener <rguenther@suse.de>
1242 PR tree-optimization/65206
1243 * tree-data-ref.h (struct data_reference): Add alt_indices,
1245 * tree-data-ref.c (free_data_ref): Release alt_indices.
1246 (dr_analyze_indices): Work on struct indices and get DR_REF as tree.
1247 (create_data_ref): Adjust.
1248 (initialize_data_dependence_relation): Split into head
1249 and tail. When the base objects fail to match up try
1250 again with pointer-based analysis of indices.
1251 * tree-vectorizer.c (vec_info_shared::check_datarefs): Do
1252 not compare the lazily computed alternate set of indices.
1254 2021-09-20 Iain Sandoe <iain@sandoe.co.uk>
1256 * gcc.c: Test for execute OK when we find the
1257 programs for assembler linker and dsymutil and those
1258 were specified at configure-time.
1260 2021-09-19 Martin Sebor <msebor@redhat.com>
1262 PR middle-end/102403
1263 * gimple-predicate-analysis.cc (predicate::init_from_control_deps):
1264 Correct a function pre/postcondition.
1266 2021-09-19 Martin Sebor <msebor@redhat.com>
1268 PR middle-end/102243
1269 * tree-ssa-strlen.c (get_range): Handle null cfun.
1271 2021-09-19 Iain Sandoe <iain@sandoe.co.uk>
1273 * config/darwin.h (LINK_COMMAND_SPEC_A): Use Darwin10
1274 unwinder shim as a convenience library.
1276 2021-09-19 Andrew Pinski <apinski@marvell.com>
1278 * doc/install.texi: Add note about
1279 binutils 2.35 is required for LTO usage.
1281 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
1283 * tree-ssa-threadbackward.c
1284 (back_threader_registry::register_path): Use push_edge.
1285 * tree-ssa-threadedge.c
1286 (jump_threader::thread_around_empty_blocks): Same.
1287 (jump_threader::thread_through_normal_block): Same.
1288 (jump_threader::thread_across_edge): Same. Also, use auto_bitmap.
1290 * tree-ssa-threadupdate.c
1291 (jt_path_registry::allocate_thread_edge): Remove.
1292 (jt_path_registry::push_edge): New.
1293 (dump_jump_thread_path): Make static.
1294 * tree-ssa-threadupdate.h (allocate_thread_edge): Remove.
1297 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
1299 * gimple-range-path.cc (path_range_query::path_range_query): Add
1301 (path_range_query::dump): Remove extern declaration of dump_ranger.
1302 * gimple-range-trace.cc (dump_ranger): Add DEBUG_FUNCTION marker.
1303 * gimple-range-trace.h (dump_ranger): Add prototype.
1305 2021-09-19 John Ericson <git@JohnEricson.me>
1307 * gcc.c (find_a_program): New function, factored out of...
1308 (find_a_file): Here.
1309 (execute): Use find_a_program when looking for programs rather
1312 2021-09-19 Matwey V. Kornilov <matwey.kornilov@gmail.com>
1314 * config/avr/avr-mcus.def: Add atmega324pb.
1315 * doc/avr-mmcu.texi: Corresponding changes.
1317 2021-09-19 Roger Sayle <roger@nextmovesoftware.com>
1320 * match.pd (cmp @0 REAL_CST@1): When @0 is also REAL_CST, apply
1321 the same transformations as to @1. For comparisons against NaN,
1322 don't check HONOR_SNANS but confirm that neither operand is a
1325 2021-09-19 Benjamin Peterson <benjamin@locrian.net>
1327 * attribs.c (make_unique_name): Delete.
1328 * attribs.h (make_unique_name): Delete.
1330 2021-09-19 Andrew Pinski <apinski@marvell.com>
1332 * lra-constraints.c (check_and_process_move): Assert
1333 that dclass and sclass are greater than or equal to NO_REGS.
1335 2021-09-18 Jakub Jelinek <jakub@redhat.com>
1337 * tree.h (OMP_CLAUSE_ORDER_UNCONSTRAINED): Define.
1338 * tree-pretty-print.c (dump_omp_clause): Print unconstrained:
1339 for OMP_CLAUSE_ORDER_UNCONSTRAINED.
1341 2021-09-18 liuhongt <hongtao.liu@intel.com>
1343 * config/i386/i386-features.c (remove_partial_avx_dependency):
1344 Restrict TARGET_USE_VECTOR_FP_CONVERTS and
1345 TARGET_USE_VECTOR_CONVERTS to conversion instructions only.
1347 2021-09-18 Jakub Jelinek <jakub@redhat.com>
1349 * gimplify.c (omp_default_clause): For C/C++ default({,first}private),
1350 if file/namespace scope variable doesn't have predetermined sharing,
1351 treat it as if there was default(none).
1353 2021-09-18 liuhongt <hongtao.liu@intel.com>
1355 * config/i386/avx512fp16intrin.h (_mm_fmadd_sh):
1357 (_mm_mask_fmadd_sh): Likewise.
1358 (_mm_mask3_fmadd_sh): Likewise.
1359 (_mm_maskz_fmadd_sh): Likewise.
1360 (_mm_fmadd_round_sh): Likewise.
1361 (_mm_mask_fmadd_round_sh): Likewise.
1362 (_mm_mask3_fmadd_round_sh): Likewise.
1363 (_mm_maskz_fmadd_round_sh): Likewise.
1364 (_mm_fnmadd_sh): Likewise.
1365 (_mm_mask_fnmadd_sh): Likewise.
1366 (_mm_mask3_fnmadd_sh): Likewise.
1367 (_mm_maskz_fnmadd_sh): Likewise.
1368 (_mm_fnmadd_round_sh): Likewise.
1369 (_mm_mask_fnmadd_round_sh): Likewise.
1370 (_mm_mask3_fnmadd_round_sh): Likewise.
1371 (_mm_maskz_fnmadd_round_sh): Likewise.
1372 (_mm_fmsub_sh): Likewise.
1373 (_mm_mask_fmsub_sh): Likewise.
1374 (_mm_mask3_fmsub_sh): Likewise.
1375 (_mm_maskz_fmsub_sh): Likewise.
1376 (_mm_fmsub_round_sh): Likewise.
1377 (_mm_mask_fmsub_round_sh): Likewise.
1378 (_mm_mask3_fmsub_round_sh): Likewise.
1379 (_mm_maskz_fmsub_round_sh): Likewise.
1380 (_mm_fnmsub_sh): Likewise.
1381 (_mm_mask_fnmsub_sh): Likewise.
1382 (_mm_mask3_fnmsub_sh): Likewise.
1383 (_mm_maskz_fnmsub_sh): Likewise.
1384 (_mm_fnmsub_round_sh): Likewise.
1385 (_mm_mask_fnmsub_round_sh): Likewise.
1386 (_mm_mask3_fnmsub_round_sh): Likewise.
1387 (_mm_maskz_fnmsub_round_sh): Likewise.
1388 * config/i386/i386-builtin-types.def
1389 (V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT): New builtin type.
1390 * config/i386/i386-builtin.def: Add new builtins.
1391 * config/i386/i386-expand.c: Handle new builtin type.
1392 * config/i386/sse.md (fmai_vmfmadd_<mode><round_name>):
1393 Ajdust to support FP16.
1394 (fmai_vmfmsub_<mode><round_name>): Ditto.
1395 (fmai_vmfnmadd_<mode><round_name>): Ditto.
1396 (fmai_vmfnmsub_<mode><round_name>): Ditto.
1397 (*fmai_fmadd_<mode>): Ditto.
1398 (*fmai_fmsub_<mode>): Ditto.
1399 (*fmai_fnmadd_<mode><round_name>): Ditto.
1400 (*fmai_fnmsub_<mode><round_name>): Ditto.
1401 (avx512f_vmfmadd_<mode>_mask<round_name>): Ditto.
1402 (avx512f_vmfmadd_<mode>_mask3<round_name>): Ditto.
1403 (avx512f_vmfmadd_<mode>_maskz<round_expand_name>): Ditto.
1404 (avx512f_vmfmadd_<mode>_maskz_1<round_name>): Ditto.
1405 (*avx512f_vmfmsub_<mode>_mask<round_name>): Ditto.
1406 (avx512f_vmfmsub_<mode>_mask3<round_name>): Ditto.
1407 (*avx512f_vmfmsub_<mode>_maskz_1<round_name>): Ditto.
1408 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
1409 (*avx512f_vmfnmsub_<mode>_mask3<round_name>): Ditto.
1410 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
1411 (*avx512f_vmfnmadd_<mode>_mask<round_name>): Renamed to ...
1412 (avx512f_vmfnmadd_<mode>_mask<round_name>) ... this, and
1413 adjust to support FP16.
1414 (avx512f_vmfnmadd_<mode>_mask3<round_name>): Ditto.
1415 (avx512f_vmfnmadd_<mode>_maskz_1<round_name>): Ditto.
1416 (avx512f_vmfnmadd_<mode>_maskz<round_expand_name>): New
1419 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
1421 * config/i386/sse.md (avx512fmaskmodelower): Extend to support
1423 (maskload<mode><avx512fmaskmodelower>): Ditto.
1424 (maskstore<mode><avx512fmaskmodelower>): Ditto.
1426 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
1428 * config/i386/i386-expand.c (ix86_expand_fp_absneg_operator):
1430 (ix86_expand_copysign): Ditto.
1431 (ix86_expand_xorsign): Ditto.
1432 * config/i386/i386.c (ix86_build_const_vector): Handle HF vector
1434 (ix86_build_signbit_mask): Ditto.
1435 (ix86_can_change_mode_class): Ditto.
1436 * config/i386/i386.md
1437 (SSEMODEF): Add HFmode.
1438 (ssevecmodef): Ditto.
1439 (<code>hf2): New define_expand.
1440 (*<code>hf2_1): New define_insn_and_split.
1441 (copysign<mode>): Extend to support HFmode under AVX512FP16.
1442 (xorsign<mode>): Ditto.
1443 * config/i386/sse.md (VFB): New mode iterator.
1444 (VFB_128_256): Ditto.
1446 (sseintvecmode2): Support HF vector mode.
1447 (<code><mode>2): Use new mode iterator.
1448 (*<code><mode>2): Ditto.
1449 (copysign<mode>3): Ditto.
1450 (xorsign<mode>3): Ditto.
1451 (<code><mode>3<mask_name>): Ditto.
1452 (<code><mode>3<mask_name>): Ditto.
1453 (<sse>_andnot<mode>3<mask_name>): Adjust for HF vector mode.
1454 (<sse>_andnot<mode>3<mask_name>): Ditto.
1455 (*<code><mode>3<mask_name>): Ditto.
1456 (*<code><mode>3<mask_name>): Ditto.
1458 2021-09-18 liuhongt <hongtao.liu@intel.com>
1460 * config/i386/avx512fp16intrin.h (_mm512_mask_fmadd_ph):
1462 (_mm512_mask3_fmadd_ph): Likewise.
1463 (_mm512_maskz_fmadd_ph): Likewise.
1464 (_mm512_fmadd_round_ph): Likewise.
1465 (_mm512_mask_fmadd_round_ph): Likewise.
1466 (_mm512_mask3_fmadd_round_ph): Likewise.
1467 (_mm512_maskz_fmadd_round_ph): Likewise.
1468 (_mm512_fnmadd_ph): Likewise.
1469 (_mm512_mask_fnmadd_ph): Likewise.
1470 (_mm512_mask3_fnmadd_ph): Likewise.
1471 (_mm512_maskz_fnmadd_ph): Likewise.
1472 (_mm512_fnmadd_round_ph): Likewise.
1473 (_mm512_mask_fnmadd_round_ph): Likewise.
1474 (_mm512_mask3_fnmadd_round_ph): Likewise.
1475 (_mm512_maskz_fnmadd_round_ph): Likewise.
1476 (_mm512_fmsub_ph): Likewise.
1477 (_mm512_mask_fmsub_ph): Likewise.
1478 (_mm512_mask3_fmsub_ph): Likewise.
1479 (_mm512_maskz_fmsub_ph): Likewise.
1480 (_mm512_fmsub_round_ph): Likewise.
1481 (_mm512_mask_fmsub_round_ph): Likewise.
1482 (_mm512_mask3_fmsub_round_ph): Likewise.
1483 (_mm512_maskz_fmsub_round_ph): Likewise.
1484 (_mm512_fnmsub_ph): Likewise.
1485 (_mm512_mask_fnmsub_ph): Likewise.
1486 (_mm512_mask3_fnmsub_ph): Likewise.
1487 (_mm512_maskz_fnmsub_ph): Likewise.
1488 (_mm512_fnmsub_round_ph): Likewise.
1489 (_mm512_mask_fnmsub_round_ph): Likewise.
1490 (_mm512_mask3_fnmsub_round_ph): Likewise.
1491 (_mm512_maskz_fnmsub_round_ph): Likewise.
1492 * config/i386/avx512fp16vlintrin.h (_mm256_fmadd_ph):
1494 (_mm256_mask_fmadd_ph): Likewise.
1495 (_mm256_mask3_fmadd_ph): Likewise.
1496 (_mm256_maskz_fmadd_ph): Likewise.
1497 (_mm_fmadd_ph): Likewise.
1498 (_mm_mask_fmadd_ph): Likewise.
1499 (_mm_mask3_fmadd_ph): Likewise.
1500 (_mm_maskz_fmadd_ph): Likewise.
1501 (_mm256_fnmadd_ph): Likewise.
1502 (_mm256_mask_fnmadd_ph): Likewise.
1503 (_mm256_mask3_fnmadd_ph): Likewise.
1504 (_mm256_maskz_fnmadd_ph): Likewise.
1505 (_mm_fnmadd_ph): Likewise.
1506 (_mm_mask_fnmadd_ph): Likewise.
1507 (_mm_mask3_fnmadd_ph): Likewise.
1508 (_mm_maskz_fnmadd_ph): Likewise.
1509 (_mm256_fmsub_ph): Likewise.
1510 (_mm256_mask_fmsub_ph): Likewise.
1511 (_mm256_mask3_fmsub_ph): Likewise.
1512 (_mm256_maskz_fmsub_ph): Likewise.
1513 (_mm_fmsub_ph): Likewise.
1514 (_mm_mask_fmsub_ph): Likewise.
1515 (_mm_mask3_fmsub_ph): Likewise.
1516 (_mm_maskz_fmsub_ph): Likewise.
1517 (_mm256_fnmsub_ph): Likewise.
1518 (_mm256_mask_fnmsub_ph): Likewise.
1519 (_mm256_mask3_fnmsub_ph): Likewise.
1520 (_mm256_maskz_fnmsub_ph): Likewise.
1521 (_mm_fnmsub_ph): Likewise.
1522 (_mm_mask_fnmsub_ph): Likewise.
1523 (_mm_mask3_fnmsub_ph): Likewise.
1524 (_mm_maskz_fnmsub_ph): Likewise.
1525 * config/i386/i386-builtin.def: Add corresponding new builtins.
1526 * config/i386/sse.md
1527 (<avx512>_fmadd_<mode>_maskz<round_expand_name>): Adjust to
1528 support HF vector modes.
1529 (<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name><round_name>):
1531 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
1532 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
1533 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
1534 (<avx512>_fmadd_<mode>_mask<round_name>): Ditto.
1535 (<avx512>_fmadd_<mode>_mask3<round_name>): Ditto.
1536 (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Ditto.
1537 (<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name><round_name>):
1539 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
1540 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
1541 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
1542 (<avx512>_fmsub_<mode>_mask<round_name>): Ditto.
1543 (<avx512>_fmsub_<mode>_mask3<round_name>): Ditto.
1544 (<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name><round_name>):
1546 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
1547 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
1548 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
1549 (<avx512>_fnmadd_<mode>_mask<round_name>): Ditto.
1550 (<avx512>_fnmadd_<mode>_mask3<round_name>): Ditto.
1551 (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Ditto.
1552 (<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name><round_name>):
1554 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
1555 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
1556 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
1557 (<avx512>_fnmsub_<mode>_mask<round_name>): Ditto.
1558 (<avx512>_fnmsub_<mode>_mask3<round_name>): Ditto.
1560 2021-09-18 liuhongt <hongtao.liu@intel.com>
1562 * config/i386/avx512fp16intrin.h (_mm512_fmaddsub_ph):
1564 (_mm512_mask_fmaddsub_ph): Likewise.
1565 (_mm512_mask3_fmaddsub_ph): Likewise.
1566 (_mm512_maskz_fmaddsub_ph): Likewise.
1567 (_mm512_fmaddsub_round_ph): Likewise.
1568 (_mm512_mask_fmaddsub_round_ph): Likewise.
1569 (_mm512_mask3_fmaddsub_round_ph): Likewise.
1570 (_mm512_maskz_fmaddsub_round_ph): Likewise.
1571 (_mm512_mask_fmsubadd_ph): Likewise.
1572 (_mm512_mask3_fmsubadd_ph): Likewise.
1573 (_mm512_maskz_fmsubadd_ph): Likewise.
1574 (_mm512_fmsubadd_round_ph): Likewise.
1575 (_mm512_mask_fmsubadd_round_ph): Likewise.
1576 (_mm512_mask3_fmsubadd_round_ph): Likewise.
1577 (_mm512_maskz_fmsubadd_round_ph): Likewise.
1578 * config/i386/avx512fp16vlintrin.h (_mm256_fmaddsub_ph):
1580 (_mm256_mask_fmaddsub_ph): Likewise.
1581 (_mm256_mask3_fmaddsub_ph): Likewise.
1582 (_mm256_maskz_fmaddsub_ph): Likewise.
1583 (_mm_fmaddsub_ph): Likewise.
1584 (_mm_mask_fmaddsub_ph): Likewise.
1585 (_mm_mask3_fmaddsub_ph): Likewise.
1586 (_mm_maskz_fmaddsub_ph): Likewise.
1587 (_mm256_fmsubadd_ph): Likewise.
1588 (_mm256_mask_fmsubadd_ph): Likewise.
1589 (_mm256_mask3_fmsubadd_ph): Likewise.
1590 (_mm256_maskz_fmsubadd_ph): Likewise.
1591 (_mm_fmsubadd_ph): Likewise.
1592 (_mm_mask_fmsubadd_ph): Likewise.
1593 (_mm_mask3_fmsubadd_ph): Likewise.
1594 (_mm_maskz_fmsubadd_ph): Likewise.
1595 * config/i386/i386-builtin.def: Add corresponding new builtins.
1596 * config/i386/sse.md (VFH_SF_AVX512VL): New mode iterator.
1597 * (<avx512>_fmsubadd_<mode>_maskz<round_expand_name>): New expander.
1598 * (<avx512>_fmaddsub_<mode>_maskz<round_expand_name>): Use
1600 * (<sd_mask_codefor>fma_fmaddsub_<mode><sd_maskz_name><round_name>):
1602 * (<avx512>_fmaddsub_<mode>_mask<round_name>): Ditto.
1603 * (<avx512>_fmaddsub_<mode>_mask3<round_name>): Ditto.
1604 * (<sd_mask_codefor>fma_fmsubadd_<mode><sd_maskz_name><round_name>):
1606 * (<avx512>_fmsubadd_<mode>_mask<round_name>): Ditto.
1607 * (<avx512>_fmsubadd_<mode>_mask3<round_name>): Ditto.
1609 2021-09-18 liuhongt <hongtao.liu@intel.com>
1612 * config/i386/i386.c (ix86_print_operand): Handle
1613 V8HF/V16HF/V32HFmode.
1614 * config/i386/i386.h (VALID_BCST_MODE_P): Add HFmode.
1615 * config/i386/sse.md (avx512bcst): Remove.
1617 2021-09-17 Martin Sebor <msebor@redhat.com>
1619 * Makefile.in (OBJS): Add gimple-predicate-analysis.o.
1620 * tree-ssa-uninit.c (max_phi_args): Move to gimple-predicate-analysis.
1621 (MASK_SET_BIT, MASK_TEST_BIT, MASK_EMPTY): Same.
1622 (check_defs): Add comment.
1623 (can_skip_redundant_opnd): Update comment.
1624 (compute_uninit_opnds_pos): Adjust to namespace change.
1625 (find_pdom): Move to gimple-predicate-analysis.cc.
1627 (struct uninit_undef_val_t): New.
1628 (is_non_loop_exit_postdominating): Move to gimple-predicate-analysis.cc.
1629 (find_control_equiv_block): Same.
1630 (MAX_NUM_CHAINS, MAX_CHAIN_LEN, MAX_POSTDOM_CHECK): Same.
1631 (MAX_SWITCH_CASES): Same.
1632 (compute_control_dep_chain): Same.
1633 (find_uninit_use): Use predicate analyzer.
1634 (struct pred_info): Move to gimple-predicate-analysis.
1635 (convert_control_dep_chain_into_preds): Same.
1636 (find_predicates): Same.
1637 (collect_phi_def_edges): Same.
1638 (warn_uninitialized_phi): Use predicate analyzer.
1639 (find_def_preds): Move to gimple-predicate-analysis.
1640 (dump_pred_info): Same.
1641 (dump_pred_chain): Same.
1642 (dump_predicates): Same.
1643 (destroy_predicate_vecs): Remove.
1644 (execute_late_warn_uninitialized): New.
1645 (get_cmp_code): Move to gimple-predicate-analysis.
1646 (is_value_included_in): Same.
1647 (value_sat_pred_p): Same.
1648 (find_matching_predicate_in_rest_chains): Same.
1649 (is_use_properly_guarded): Same.
1650 (prune_uninit_phi_opnds): Same.
1651 (find_var_cmp_const): Same.
1652 (use_pred_not_overlap_with_undef_path_pred): Same.
1653 (pred_equal_p): Same.
1654 (is_neq_relop_p): Same.
1655 (is_neq_zero_form_p): Same.
1656 (pred_expr_equal_p): Same.
1657 (is_pred_expr_subset_of): Same.
1658 (is_pred_chain_subset_of): Same.
1659 (is_included_in): Same.
1660 (is_superset_of): Same.
1662 (simplify_pred): Same.
1663 (simplify_preds_2): Same.
1664 (simplify_preds_3): Same.
1665 (simplify_preds_4): Same.
1666 (simplify_preds): Same.
1668 (push_to_worklist): Same.
1669 (get_pred_info_from_cmp): Same.
1670 (is_degenerated_phi): Same.
1671 (normalize_one_pred_1): Same.
1672 (normalize_one_pred): Same.
1673 (normalize_one_pred_chain): Same.
1674 (normalize_preds): Same.
1675 (can_one_predicate_be_invalidated_p): Same.
1676 (can_chain_union_be_invalidated_p): Same.
1677 (uninit_uses_cannot_happen): Same.
1678 (pass_late_warn_uninitialized::execute): Define.
1679 * gimple-predicate-analysis.cc: New file.
1680 * gimple-predicate-analysis.h: New file.
1682 2021-09-17 Julian Brown <julian@codesourcery.com>
1684 * config/gcn/gcn.c (gimple.h): Include.
1685 (gcn_fork_join): Emit barrier for worker-level joins.
1686 * omp-oacc-neuter-broadcast.cc (find_local_vars_to_propagate): Add
1687 writes_gang_private bitmap parameter. Set bit for blocks
1688 containing gang-private variable writes.
1689 (worker_single_simple): Don't emit barrier after predicated block.
1690 (worker_single_copy): Don't emit barrier if we're not broadcasting
1691 anything and the block contains no gang-private writes.
1692 (neuter_worker_single): Don't predicate blocks that only contain
1693 NOPs or internal marker functions. Pass has_gang_private_write
1694 argument to worker_single_copy.
1695 (oacc_do_neutering): Add writes_gang_private bitmap handling.
1697 2021-09-17 Julian Brown <julian@codesourcery.com>
1699 * config/gcn/gcn-protos.h
1700 (gcn_goacc_create_worker_broadcast_record): Update prototype.
1701 * config/gcn/gcn-tree.c (gcn_goacc_get_worker_red_decl): Use
1702 preallocated block of LDS memory. Do not cache/share decls for
1703 reduction temporaries between invocations.
1704 (gcn_goacc_reduction_teardown): Unshare VAR on second use.
1705 (gcn_goacc_create_worker_broadcast_record): Add OFFSET parameter
1706 and return temporary LDS space at that offset. Return pointer in
1708 * config/gcn/gcn.c (acc_lds_size, gang_private_hwm, lds_allocs):
1710 (ACC_LDS_SIZE): Define as acc_lds_size.
1711 (gcn_init_machine_status): Don't initialise lds_allocated,
1712 lds_allocs, reduc_decls fields of machine function struct.
1713 (gcn_option_override): Handle default size for gang-private
1714 variables and -mgang-private-size option.
1715 (gcn_expand_prologue): Use LDS_SIZE instead of LDS_SIZE-1 when
1716 initialising M0_REG.
1717 (gcn_shared_mem_layout): New function.
1718 (gcn_print_lds_decl): Update comment. Use global lds_allocs map and
1719 gang_private_hwm variable.
1720 (TARGET_GOACC_SHARED_MEM_LAYOUT): Define target hook.
1721 * config/gcn/gcn.h (machine_function): Remove lds_allocated,
1722 lds_allocs, reduc_decls. Add reduction_base, reduction_limit.
1723 * config/gcn/gcn.opt (gang_private_size_opt): New global.
1724 (mgang-private-size=): New option.
1725 * doc/tm.texi.in (TARGET_GOACC_SHARED_MEM_LAYOUT): Place
1727 * doc/tm.texi: Regenerate.
1728 * omp-oacc-neuter-broadcast.cc (targhooks.h, diagnostic-core.h):
1730 (build_sender_ref): Handle sender_decl being pointer.
1731 (worker_single_copy): Add PLACEMENT and ISOLATE_BROADCASTS
1732 parameters. Pass placement argument to
1733 create_worker_broadcast_record hook invocations. Handle
1734 sender_decl being pointer and isolate_broadcasts inserting extra
1736 (blk_offset_map_t): Add typedef.
1737 (neuter_worker_single): Add BLK_OFFSET_MAP parameter. Pass
1738 preallocated range to worker_single_copy call.
1739 (dfs_broadcast_reachable_1): New function.
1740 (idx_decl_pair_t, used_range_vec_t): New typedefs.
1741 (sort_size_descending): New function.
1742 (addr_range): New class.
1743 (splay_tree_compare_addr_range, splay_tree_free_key)
1744 (first_fit_range, merge_ranges_1, merge_ranges): New functions.
1745 (execute_omp_oacc_neuter_broadcast): Rename to...
1746 (oacc_do_neutering): ... this. Add BOUNDS_LO, BOUNDS_HI
1747 parameters. Arrange layout of shared memory for broadcast
1749 (execute_omp_oacc_neuter_broadcast): New function.
1750 (pass_omp_oacc_neuter_broadcast::gate): Remove num_workers==1
1751 handling from here. Enable pass for all OpenACC routines in order
1752 to call shared memory-layout hook.
1753 * target.def (create_worker_broadcast_record): Add OFFSET
1755 (shared_mem_layout): New hook.
1757 2021-09-17 Julian Brown <julian@codesourcery.com>
1758 Thomas Schwinge <thomas@codesourcery.com>
1760 * omp-oacc-neuter-broadcast.cc
1761 (pass_omp_oacc_neuter_broadcast::gate): Disable if num_workers is
1763 (execute_omp_oacc_neuter_broadcast): Adjust.
1765 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
1767 * value-relation.cc (class equiv_chain): Move to header file.
1768 (path_oracle::path_oracle): New.
1769 (path_oracle::~path_oracle): New.
1770 (path_oracle::register_relation): New.
1771 (path_oracle::query_relation): New.
1772 (path_oracle::reset_path): New.
1773 (path_oracle::dump): New.
1774 * value-relation.h (class equiv_chain): Move to here.
1775 (class path_oracle): New.
1777 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
1779 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a DOM
1781 * gimple-range-fold.cc (fur_depend::register_relation): Use
1782 register_stmt/edge routines.
1783 * value-relation.cc (equiv_chain::find): Relocate from equiv_oracle.
1784 (equiv_oracle::equiv_oracle): Create self equivalence cache.
1785 (equiv_oracle::~equiv_oracle): Release same.
1786 (equiv_oracle::equiv_set): Return entry from self equiv cache if there
1787 are no equivalences.
1788 (equiv_oracle::find_equiv_block): Move list find to equiv_chain.
1789 (equiv_oracle::register_relation): Rename from register_equiv.
1790 (relation_chain_head::find_relation): Relocate from dom_oracle.
1791 (relation_oracle::register_stmt): New.
1792 (relation_oracle::register_edge): New.
1793 (dom_oracle::*): Rename from relation_oracle.
1794 (dom_oracle::register_relation): Adjust to call equiv_oracle.
1795 (dom_oracle::set_one_relation): Split from register_relation.
1796 (dom_oracle::register_transitives): Consolidate 2 methods.
1797 (dom_oracle::find_relation_block): Move core to relation_chain.
1798 (dom_oracle::query_relation): Rename from find_relation_dom and adjust.
1799 * value-relation.h (class relation_oracle): New pure virtual base.
1800 (class equiv_oracle): Inherit from relation_oracle and adjust.
1801 (class dom_oracle): Rename from old relation_oracle and adjust.
1803 2021-09-17 Martin Sebor <msebor@redhat.com>
1805 PR middle-end/102200
1806 * pointer-query.cc (access_ref::inform_access): Handle MIN/MAX_EXPR.
1807 (handle_min_max_size): Change argument. Store original SSA_NAME for
1808 operands to potentially distinct (sub)objects.
1809 (compute_objsize_r): Adjust call to the above.
1811 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
1813 * config/rs6000/rs6000.c (rs6000-builtins.h): New include.
1814 (rs6000_new_builtin_vectorized_function): New function.
1815 (rs6000_new_builtin_md_vectorized_function): Likewise.
1816 (rs6000_builtin_vectorized_function): Call
1817 rs6000_new_builtin_vectorized_function.
1818 (rs6000_builtin_md_vectorized_function): Call
1819 rs6000_new_builtin_md_vectorized_function.
1821 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
1823 * config/rs6000/rs6000-builtin-new.def (ASSEMBLE_ACC): Add mmaint flag.
1824 (ASSEMBLE_PAIR): Likewise.
1825 (BUILD_ACC): Likewise.
1826 (DISASSEMBLE_ACC): Likewise.
1827 (DISASSEMBLE_PAIR): Likewise.
1828 (PMXVBF16GER2): Likewise.
1829 (PMXVBF16GER2NN): Likewise.
1830 (PMXVBF16GER2NP): Likewise.
1831 (PMXVBF16GER2PN): Likewise.
1832 (PMXVBF16GER2PP): Likewise.
1833 (PMXVF16GER2): Likewise.
1834 (PMXVF16GER2NN): Likewise.
1835 (PMXVF16GER2NP): Likewise.
1836 (PMXVF16GER2PN): Likewise.
1837 (PMXVF16GER2PP): Likewise.
1838 (PMXVF32GER): Likewise.
1839 (PMXVF32GERNN): Likewise.
1840 (PMXVF32GERNP): Likewise.
1841 (PMXVF32GERPN): Likewise.
1842 (PMXVF32GERPP): Likewise.
1843 (PMXVF64GER): Likewise.
1844 (PMXVF64GERNN): Likewise.
1845 (PMXVF64GERNP): Likewise.
1846 (PMXVF64GERPN): Likewise.
1847 (PMXVF64GERPP): Likewise.
1848 (PMXVI16GER2): Likewise.
1849 (PMXVI16GER2PP): Likewise.
1850 (PMXVI16GER2S): Likewise.
1851 (PMXVI16GER2SPP): Likewise.
1852 (PMXVI4GER8): Likewise.
1853 (PMXVI4GER8PP): Likewise.
1854 (PMXVI8GER4): Likewise.
1855 (PMXVI8GER4PP): Likewise.
1856 (PMXVI8GER4SPP): Likewise.
1857 (XVBF16GER2): Likewise.
1858 (XVBF16GER2NN): Likewise.
1859 (XVBF16GER2NP): Likewise.
1860 (XVBF16GER2PN): Likewise.
1861 (XVBF16GER2PP): Likewise.
1862 (XVF16GER2): Likewise.
1863 (XVF16GER2NN): Likewise.
1864 (XVF16GER2NP): Likewise.
1865 (XVF16GER2PN): Likewise.
1866 (XVF16GER2PP): Likewise.
1867 (XVF32GER): Likewise.
1868 (XVF32GERNN): Likewise.
1869 (XVF32GERNP): Likewise.
1870 (XVF32GERPN): Likewise.
1871 (XVF32GERPP): Likewise.
1872 (XVF64GER): Likewise.
1873 (XVF64GERNN): Likewise.
1874 (XVF64GERNP): Likewise.
1875 (XVF64GERPN): Likewise.
1876 (XVF64GERPP): Likewise.
1877 (XVI16GER2): Likewise.
1878 (XVI16GER2PP): Likewise.
1879 (XVI16GER2S): Likewise.
1880 (XVI16GER2SPP): Likewise.
1881 (XVI4GER8): Likewise.
1882 (XVI4GER8PP): Likewise.
1883 (XVI8GER4): Likewise.
1884 (XVI8GER4PP): Likewise.
1885 (XVI8GER4SPP): Likewise.
1886 (XXMFACC): Likewise.
1887 (XXMTACC): Likewise.
1888 (XXSETACCZ): Likewise.
1889 (ASSEMBLE_PAIR_V): Likewise.
1890 (BUILD_PAIR): Likewise.
1891 (DISASSEMBLE_PAIR_V): Likewise.
1894 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_mma_builtin):
1895 Handle RS6000_BIF_LXVP and RS6000_BIF_STXVP.
1896 * config/rs6000/rs6000-gen-builtins.c (attrinfo): Add ismmaint.
1897 (parse_bif_attrs): Handle ismmaint.
1898 (write_decls): Add bif_mmaint_bit and bif_is_mmaint.
1899 (write_bif_static_init): Handle ismmaint.
1901 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
1903 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_builtin): New
1905 (rs6000_gimple_fold_builtin): Call rs6000_gimple_fold_new_builtin.
1906 (rs6000_new_builtin_valid_without_lhs): New function.
1907 (rs6000_gimple_fold_new_mma_builtin): Likewise.
1908 (rs6000_gimple_fold_new_builtin): Likewise.
1910 2021-09-17 Thomas Schwinge <thomas@codesourcery.com>
1912 * hash-table.h (hash_table<Descriptor, Lazy, Allocator>::expand):
1913 Destruct stale Value objects.
1914 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
1917 2021-09-17 Roger Sayle <roger@nextmovesoftware.com>
1920 * match.pd (shift optimizations): Disable recent sign-changing
1921 optimization for shifts by zero, these will be folded later.
1923 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
1925 * config/rs6000/rs6000-builtin-new.def (__builtin_mffsl): Move from
1926 [power9] to [always].
1928 2021-09-17 Richard Biener <rguenther@suse.de>
1930 * tree-vect-stmts.c (vectorizable_load): Do not frob
1933 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
1935 * config/i386/i386-features.c (remove_partial_avx_dependency):
1936 Also check TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY and
1937 and TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY before generating
1939 * config/i386/i386.h (TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY):
1941 (TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
1942 * config/i386/i386.md (SSE FP to FP splitters): Replace
1943 TARGET_SSE_PARTIAL_REG_DEPENDENCY with
1944 TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY.
1945 (SSE INT to FP splitter): Replace TARGET_SSE_PARTIAL_REG_DEPENDENCY
1946 with TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY.
1947 * config/i386/x86-tune.def
1948 (X86_TUNE_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY): New.
1949 (X86_TUNE_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
1951 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
1954 * config/i386/i386-features.c (remove_partial_avx_dependency):
1955 Check TARGET_USE_VECTOR_FP_CONVERTS and TARGET_USE_VECTOR_CONVERTS
1956 before generating vxorps.
1958 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
1960 * config/i386/i386-options.c (processor_cost_table): Use
1961 tremont_cost for Tremont.
1962 * config/i386/x86-tune-costs.h (tremont_memcpy): New.
1963 (tremont_memset): Likewise.
1964 (tremont_cost): Likewise.
1965 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
1968 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
1970 * common/config/i386/i386-common.c: Use Haswell scheduling model
1972 * config/i386/i386.c (ix86_sched_init_global): Prepare for Tremont
1974 * config/i386/x86-tune-sched.c (ix86_issue_rate): Change Tremont
1976 (ix86_adjust_cost): Handle Tremont.
1977 * config/i386/x86-tune.def (X86_TUNE_SSE_PARTIAL_REG_DEPENDENCY):
1979 (X86_TUNE_USE_LEAVE): Likewise.
1980 (X86_TUNE_PUSH_MEMORY): Likewise.
1981 (X86_TUNE_MISALIGNED_MOVE_STRING_PRO_EPILOGUES): Likewise.
1982 (X86_TUNE_USE_CLTD): Likewise.
1983 (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI): Likewise.
1984 (X86_TUNE_AVOID_MFENCE): Likewise.
1985 (X86_TUNE_SSE_TYPELESS_STORES): Likewise.
1986 (X86_TUNE_SSE_LOAD0_BY_PXOR): Likewise.
1987 (X86_TUNE_ACCUMULATE_OUTGOING_ARGS): Disable for Tremont.
1988 (X86_TUNE_FOUR_JUMP_LIMIT): Likewise.
1989 (X86_TUNE_OPT_AGU): Likewise.
1990 (X86_TUNE_AVOID_LEA_FOR_ADDR): Likewise.
1991 (X86_TUNE_AVOID_MEM_OPND_FOR_CMOVE): Likewise.
1992 (X86_TUNE_EXPAND_ABS): Likewise.
1993 (X86_TUNE_SPLIT_MEM_OPND_FOR_FP_CONVERTS): Likewise.
1994 (X86_TUNE_SLOW_PSHUFB): Likewise.
1996 2021-09-17 Eric Botcazou <ebotcazou@adacore.com>
1998 PR rtl-optimization/102306
1999 * combine.c (try_combine): Abort the combination if we are about to
2000 duplicate volatile references.
2002 2021-09-17 liuhongt <hongtao.liu@intel.com>
2004 * config/i386/avx512fp16intrin.h (_mm_undefined_ph):
2006 (_mm256_undefined_ph): Likewise.
2007 (_mm512_undefined_ph): Likewise.
2008 (_mm_cvtsh_h): Likewise.
2009 (_mm256_cvtsh_h): Likewise.
2010 (_mm512_cvtsh_h): Likewise.
2011 (_mm512_castph_ps): Likewise.
2012 (_mm512_castph_pd): Likewise.
2013 (_mm512_castph_si512): Likewise.
2014 (_mm512_castph512_ph128): Likewise.
2015 (_mm512_castph512_ph256): Likewise.
2016 (_mm512_castph128_ph512): Likewise.
2017 (_mm512_castph256_ph512): Likewise.
2018 (_mm512_zextph128_ph512): Likewise.
2019 (_mm512_zextph256_ph512): Likewise.
2020 (_mm512_castps_ph): Likewise.
2021 (_mm512_castpd_ph): Likewise.
2022 (_mm512_castsi512_ph): Likewise.
2023 * config/i386/avx512fp16vlintrin.h (_mm_castph_ps):
2025 (_mm256_castph_ps): Likewise.
2026 (_mm_castph_pd): Likewise.
2027 (_mm256_castph_pd): Likewise.
2028 (_mm_castph_si128): Likewise.
2029 (_mm256_castph_si256): Likewise.
2030 (_mm_castps_ph): Likewise.
2031 (_mm256_castps_ph): Likewise.
2032 (_mm_castpd_ph): Likewise.
2033 (_mm256_castpd_ph): Likewise.
2034 (_mm_castsi128_ph): Likewise.
2035 (_mm256_castsi256_ph): Likewise.
2036 (_mm256_castph256_ph128): Likewise.
2037 (_mm256_castph128_ph256): Likewise.
2038 (_mm256_zextph128_ph256): Likewise.
2040 2021-09-17 liuhongt <hongtao.liu@intel.com>
2042 * config/i386/avx512fp16intrin.h (_mm_cvtsh_ss):
2044 (_mm_mask_cvtsh_ss): Likewise.
2045 (_mm_maskz_cvtsh_ss): Likewise.
2046 (_mm_cvtsh_sd): Likewise.
2047 (_mm_mask_cvtsh_sd): Likewise.
2048 (_mm_maskz_cvtsh_sd): Likewise.
2049 (_mm_cvt_roundsh_ss): Likewise.
2050 (_mm_mask_cvt_roundsh_ss): Likewise.
2051 (_mm_maskz_cvt_roundsh_ss): Likewise.
2052 (_mm_cvt_roundsh_sd): Likewise.
2053 (_mm_mask_cvt_roundsh_sd): Likewise.
2054 (_mm_maskz_cvt_roundsh_sd): Likewise.
2055 (_mm_cvtss_sh): Likewise.
2056 (_mm_mask_cvtss_sh): Likewise.
2057 (_mm_maskz_cvtss_sh): Likewise.
2058 (_mm_cvtsd_sh): Likewise.
2059 (_mm_mask_cvtsd_sh): Likewise.
2060 (_mm_maskz_cvtsd_sh): Likewise.
2061 (_mm_cvt_roundss_sh): Likewise.
2062 (_mm_mask_cvt_roundss_sh): Likewise.
2063 (_mm_maskz_cvt_roundss_sh): Likewise.
2064 (_mm_cvt_roundsd_sh): Likewise.
2065 (_mm_mask_cvt_roundsd_sh): Likewise.
2066 (_mm_maskz_cvt_roundsd_sh): Likewise.
2067 * config/i386/i386-builtin-types.def
2068 (V8HF_FTYPE_V2DF_V8HF_V8HF_UQI_INT,
2069 V8HF_FTYPE_V4SF_V8HF_V8HF_UQI_INT,
2070 V2DF_FTYPE_V8HF_V2DF_V2DF_UQI_INT,
2071 V4SF_FTYPE_V8HF_V4SF_V4SF_UQI_INT): Add new builtin types.
2072 * config/i386/i386-builtin.def: Add corrresponding new builtins.
2073 * config/i386/i386-expand.c: Handle new builtin types.
2074 * config/i386/sse.md (VF48_128): New mode iterator.
2075 (avx512fp16_vcvtsh2<ssescalarmodesuffix><mask_scalar_name><round_saeonly_scalar_name>):
2077 (avx512fp16_vcvt<ssescalarmodesuffix>2sh<mask_scalar_name><round_scalar_name>):
2080 2021-09-17 liuhongt <hongtao.liu@intel.com>
2082 * config/i386/avx512fp16intrin.h (_mm512_cvtph_pd):
2084 (_mm512_mask_cvtph_pd): Likewise.
2085 (_mm512_maskz_cvtph_pd): Likewise.
2086 (_mm512_cvt_roundph_pd): Likewise.
2087 (_mm512_mask_cvt_roundph_pd): Likewise.
2088 (_mm512_maskz_cvt_roundph_pd): Likewise.
2089 (_mm512_cvtxph_ps): Likewise.
2090 (_mm512_mask_cvtxph_ps): Likewise.
2091 (_mm512_maskz_cvtxph_ps): Likewise.
2092 (_mm512_cvtx_roundph_ps): Likewise.
2093 (_mm512_mask_cvtx_roundph_ps): Likewise.
2094 (_mm512_maskz_cvtx_roundph_ps): Likewise.
2095 (_mm512_cvtxps_ph): Likewise.
2096 (_mm512_mask_cvtxps_ph): Likewise.
2097 (_mm512_maskz_cvtxps_ph): Likewise.
2098 (_mm512_cvtx_roundps_ph): Likewise.
2099 (_mm512_mask_cvtx_roundps_ph): Likewise.
2100 (_mm512_maskz_cvtx_roundps_ph): Likewise.
2101 (_mm512_cvtpd_ph): Likewise.
2102 (_mm512_mask_cvtpd_ph): Likewise.
2103 (_mm512_maskz_cvtpd_ph): Likewise.
2104 (_mm512_cvt_roundpd_ph): Likewise.
2105 (_mm512_mask_cvt_roundpd_ph): Likewise.
2106 (_mm512_maskz_cvt_roundpd_ph): Likewise.
2107 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_pd):
2109 (_mm_mask_cvtph_pd): Likewise.
2110 (_mm_maskz_cvtph_pd): Likewise.
2111 (_mm256_cvtph_pd): Likewise.
2112 (_mm256_mask_cvtph_pd): Likewise.
2113 (_mm256_maskz_cvtph_pd): Likewise.
2114 (_mm_cvtxph_ps): Likewise.
2115 (_mm_mask_cvtxph_ps): Likewise.
2116 (_mm_maskz_cvtxph_ps): Likewise.
2117 (_mm256_cvtxph_ps): Likewise.
2118 (_mm256_mask_cvtxph_ps): Likewise.
2119 (_mm256_maskz_cvtxph_ps): Likewise.
2120 (_mm_cvtxps_ph): Likewise.
2121 (_mm_mask_cvtxps_ph): Likewise.
2122 (_mm_maskz_cvtxps_ph): Likewise.
2123 (_mm256_cvtxps_ph): Likewise.
2124 (_mm256_mask_cvtxps_ph): Likewise.
2125 (_mm256_maskz_cvtxps_ph): Likewise.
2126 (_mm_cvtpd_ph): Likewise.
2127 (_mm_mask_cvtpd_ph): Likewise.
2128 (_mm_maskz_cvtpd_ph): Likewise.
2129 (_mm256_cvtpd_ph): Likewise.
2130 (_mm256_mask_cvtpd_ph): Likewise.
2131 (_mm256_maskz_cvtpd_ph): Likewise.
2132 * config/i386/i386-builtin.def: Add corresponding new builtins.
2133 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2134 * config/i386/i386-expand.c: Handle new builtin types.
2135 * config/i386/sse.md
2136 (VF4_128_8_256): New.
2137 (VF48H_AVX512VL): Ditto.
2138 (ssePHmode): Add HF vector modes.
2139 (castmode): Add new convertable modes.
2142 (avx512fp16_vcvt<castmode>2ph_<mode><mask_name><round_name>): Ditto.
2143 (avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
2144 (*avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
2145 (avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
2146 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
2147 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask_1): Ditto.
2148 (avx512fp16_float_extend_ph<mode>2<mask_name><round_saeonly_name>):
2150 (avx512fp16_float_extend_ph<mode>2<mask_name>): Ditto.
2151 (*avx512fp16_float_extend_ph<mode>2_load<mask_name>): Ditto.
2152 (avx512fp16_float_extend_phv2df2<mask_name>): Ditto.
2153 (*avx512fp16_float_extend_phv2df2_load<mask_name>): Ditto.
2155 2021-09-17 liuhongt <hongtao.liu@intel.com>
2157 * config/i386/avx512fp16intrin.h (_mm_cvttsh_i32):
2159 (_mm_cvttsh_u32): Likewise.
2160 (_mm_cvtt_roundsh_i32): Likewise.
2161 (_mm_cvtt_roundsh_u32): Likewise.
2162 (_mm_cvttsh_i64): Likewise.
2163 (_mm_cvttsh_u64): Likewise.
2164 (_mm_cvtt_roundsh_i64): Likewise.
2165 (_mm_cvtt_roundsh_u64): Likewise.
2166 * config/i386/i386-builtin.def: Add corresponding new builtins.
2167 * config/i386/sse.md
2168 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<round_saeonly_name>):
2171 2021-09-17 liuhongt <hongtao.liu@intel.com>
2173 * config/i386/avx512fp16intrin.h (_mm512_cvttph_epi32):
2175 (_mm512_mask_cvttph_epi32): Likewise.
2176 (_mm512_maskz_cvttph_epi32): Likewise.
2177 (_mm512_cvtt_roundph_epi32): Likewise.
2178 (_mm512_mask_cvtt_roundph_epi32): Likewise.
2179 (_mm512_maskz_cvtt_roundph_epi32): Likewise.
2180 (_mm512_cvttph_epu32): Likewise.
2181 (_mm512_mask_cvttph_epu32): Likewise.
2182 (_mm512_maskz_cvttph_epu32): Likewise.
2183 (_mm512_cvtt_roundph_epu32): Likewise.
2184 (_mm512_mask_cvtt_roundph_epu32): Likewise.
2185 (_mm512_maskz_cvtt_roundph_epu32): Likewise.
2186 (_mm512_cvttph_epi64): Likewise.
2187 (_mm512_mask_cvttph_epi64): Likewise.
2188 (_mm512_maskz_cvttph_epi64): Likewise.
2189 (_mm512_cvtt_roundph_epi64): Likewise.
2190 (_mm512_mask_cvtt_roundph_epi64): Likewise.
2191 (_mm512_maskz_cvtt_roundph_epi64): Likewise.
2192 (_mm512_cvttph_epu64): Likewise.
2193 (_mm512_mask_cvttph_epu64): Likewise.
2194 (_mm512_maskz_cvttph_epu64): Likewise.
2195 (_mm512_cvtt_roundph_epu64): Likewise.
2196 (_mm512_mask_cvtt_roundph_epu64): Likewise.
2197 (_mm512_maskz_cvtt_roundph_epu64): Likewise.
2198 (_mm512_cvttph_epi16): Likewise.
2199 (_mm512_mask_cvttph_epi16): Likewise.
2200 (_mm512_maskz_cvttph_epi16): Likewise.
2201 (_mm512_cvtt_roundph_epi16): Likewise.
2202 (_mm512_mask_cvtt_roundph_epi16): Likewise.
2203 (_mm512_maskz_cvtt_roundph_epi16): Likewise.
2204 (_mm512_cvttph_epu16): Likewise.
2205 (_mm512_mask_cvttph_epu16): Likewise.
2206 (_mm512_maskz_cvttph_epu16): Likewise.
2207 (_mm512_cvtt_roundph_epu16): Likewise.
2208 (_mm512_mask_cvtt_roundph_epu16): Likewise.
2209 (_mm512_maskz_cvtt_roundph_epu16): Likewise.
2210 * config/i386/avx512fp16vlintrin.h (_mm_cvttph_epi32):
2212 (_mm_mask_cvttph_epi32): Likewise.
2213 (_mm_maskz_cvttph_epi32): Likewise.
2214 (_mm256_cvttph_epi32): Likewise.
2215 (_mm256_mask_cvttph_epi32): Likewise.
2216 (_mm256_maskz_cvttph_epi32): Likewise.
2217 (_mm_cvttph_epu32): Likewise.
2218 (_mm_mask_cvttph_epu32): Likewise.
2219 (_mm_maskz_cvttph_epu32): Likewise.
2220 (_mm256_cvttph_epu32): Likewise.
2221 (_mm256_mask_cvttph_epu32): Likewise.
2222 (_mm256_maskz_cvttph_epu32): Likewise.
2223 (_mm_cvttph_epi64): Likewise.
2224 (_mm_mask_cvttph_epi64): Likewise.
2225 (_mm_maskz_cvttph_epi64): Likewise.
2226 (_mm256_cvttph_epi64): Likewise.
2227 (_mm256_mask_cvttph_epi64): Likewise.
2228 (_mm256_maskz_cvttph_epi64): Likewise.
2229 (_mm_cvttph_epu64): Likewise.
2230 (_mm_mask_cvttph_epu64): Likewise.
2231 (_mm_maskz_cvttph_epu64): Likewise.
2232 (_mm256_cvttph_epu64): Likewise.
2233 (_mm256_mask_cvttph_epu64): Likewise.
2234 (_mm256_maskz_cvttph_epu64): Likewise.
2235 (_mm_cvttph_epi16): Likewise.
2236 (_mm_mask_cvttph_epi16): Likewise.
2237 (_mm_maskz_cvttph_epi16): Likewise.
2238 (_mm256_cvttph_epi16): Likewise.
2239 (_mm256_mask_cvttph_epi16): Likewise.
2240 (_mm256_maskz_cvttph_epi16): Likewise.
2241 (_mm_cvttph_epu16): Likewise.
2242 (_mm_mask_cvttph_epu16): Likewise.
2243 (_mm_maskz_cvttph_epu16): Likewise.
2244 (_mm256_cvttph_epu16): Likewise.
2245 (_mm256_mask_cvttph_epu16): Likewise.
2246 (_mm256_maskz_cvttph_epu16): Likewise.
2247 * config/i386/i386-builtin.def: Add new builtins.
2248 * config/i386/sse.md
2249 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name><round_saeonly_name>):
2251 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name>): Ditto.
2252 (*avx512fp16_fix<fixunssuffix>_trunc<mode>2_load<mask_name>): Ditto.
2253 (avx512fp16_fix<fixunssuffix>_truncv2di2<mask_name>): Ditto.
2254 (avx512fp16_fix<fixunssuffix>_truncv2di2_load<mask_name>): Ditto.
2256 2021-09-17 liuhongt <hongtao.liu@intel.com>
2258 * config/i386/avx512fp16intrin.h (_mm_cvtsh_i32): New intrinsic.
2259 (_mm_cvtsh_u32): Likewise.
2260 (_mm_cvt_roundsh_i32): Likewise.
2261 (_mm_cvt_roundsh_u32): Likewise.
2262 (_mm_cvtsh_i64): Likewise.
2263 (_mm_cvtsh_u64): Likewise.
2264 (_mm_cvt_roundsh_i64): Likewise.
2265 (_mm_cvt_roundsh_u64): Likewise.
2266 (_mm_cvti32_sh): Likewise.
2267 (_mm_cvtu32_sh): Likewise.
2268 (_mm_cvt_roundi32_sh): Likewise.
2269 (_mm_cvt_roundu32_sh): Likewise.
2270 (_mm_cvti64_sh): Likewise.
2271 (_mm_cvtu64_sh): Likewise.
2272 (_mm_cvt_roundi64_sh): Likewise.
2273 (_mm_cvt_roundu64_sh): Likewise.
2274 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2275 * config/i386/i386-builtin.def: Add corresponding new builtins.
2276 * config/i386/i386-expand.c (ix86_expand_round_builtin):
2277 Handle new builtin types.
2278 * config/i386/sse.md
2279 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix><round_name>):
2281 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix>_2): Likewise.
2282 (avx512fp16_vcvt<floatsuffix>si2sh<rex64namesuffix><round_name>): Likewise.
2284 2021-09-16 Bill Schmidt <wschmidt@linux.ibm.com>
2286 * config/rs6000/rs6000-c.c (rs6000-builtins.h): New include.
2287 (altivec_resolve_new_overloaded_builtin): New forward decl.
2288 (rs6000_new_builtin_type_compatible): New function.
2289 (altivec_resolve_overloaded_builtin): Call
2290 altivec_resolve_new_overloaded_builtin.
2291 (altivec_build_new_resolved_builtin): New function.
2292 (altivec_resolve_new_overloaded_builtin): Likewise.
2293 * config/rs6000/rs6000-call.c (rs6000_new_builtin_is_supported):
2295 * config/rs6000/rs6000-gen-builtins.c (write_decls): Remove _p from
2296 name of rs6000_new_builtin_is_supported.
2298 2021-09-16 Uroš Bizjak <ubizjak@gmail.com>
2300 * config/i386/i386-protos.h (ix86_decompose_address):
2301 Change return type to bool.
2302 * config/i386/i386.c (ix86_decompose_address): Ditto.
2304 2021-09-16 Tobias Burnus <tobias@codesourcery.com>
2307 * config/rs6000/t-rs6000 (build/rs6000-gen-builtins.o, build/rbtree.o):
2308 Added 'build/' to target, use build/%.o rule.
2309 (build/rs6000-gen-builtins$(build_exeext)): Add 'build/' and
2310 '$(build_exeext)' to target and 'build/' for the *.o files.
2311 (rs6000-builtins.c): Update for those changes; run rs6000-gen-builtins
2314 2021-09-16 Martin Jambor <mjambor@suse.cz>
2316 * cgraph.c (cgraph_node::dump): Do not check caller count sums if
2317 the body has been removed. Remove trailing whitespace.
2319 2021-09-16 Richard Biener <rguenther@suse.de>
2321 PR middle-end/102360
2322 * internal-fn.c (expand_DEFERRED_INIT): Make pattern-init
2323 of non-memory more robust.
2325 2021-09-16 Daniel Cederman <cederman@gaisler.com>
2327 * config/sparc/sparc-opts.h (enum sparc_processor_type): Add LEON5
2328 * config/sparc/sparc.c (struct processor_costs): Add LEON5 costs
2329 (leon5_adjust_cost): Increase cost of store with data dependency
2330 on ALU instruction and FPU anti-dependencies.
2331 (sparc_option_override): Add LEON5 costs
2332 (sparc_adjust_cost): Add LEON5 cost adjustments
2333 * config/sparc/sparc.h: Add LEON5
2334 * config/sparc/sparc.md: Include LEON5 scheduling information
2335 * config/sparc/sparc.opt: Add LEON5
2336 * doc/invoke.texi: Add LEON5
2337 * config/sparc/leon5.md: New file.
2339 2021-09-16 Daniel Cederman <cederman@gaisler.com>
2341 * config/sparc/sparc.md (stack_protect_set32): Add NOP to prevent
2342 sensitive sequence for B2BST errata workaround.
2344 2021-09-16 Daniel Cederman <cederman@gaisler.com>
2346 * config/sparc/sparc.c (sparc_do_work_around_errata): Do not begin
2347 functions with atomic instruction in the UT700 errata workaround.
2349 2021-09-16 Daniel Cederman <cederman@gaisler.com>
2351 * config/sparc/sparc.c (next_active_non_empty_insn): New function
2352 that returns next active non empty assembly instruction.
2353 (sparc_do_work_around_errata): Use new function.
2355 2021-09-16 Daniel Cederman <cederman@gaisler.com>
2357 * config/sparc/sparc.c (store_insn_p): Add predicate for store
2359 (load_insn_p): Add predicate for load attributes.
2360 (sparc_do_work_around_errata): Use new predicates.
2362 2021-09-16 Andreas Larsson <andreas@gaisler.com>
2364 * config/sparc/sparc.c (dump_target_flag_bits): Print bit names for
2367 2021-09-16 Martin Liska <mliska@suse.cz>
2369 * config/mips/netbsd.h: Fix typo in name of a macro.
2371 2021-09-16 liuhongt <hongtao.liu@intel.com>
2373 PR middle-end/102080
2374 * match.pd: Check mask type when doing cond_op related gimple
2376 * tree.c (is_truth_type_for): New function.
2377 * tree.h (is_truth_type_for): New declaration.
2379 2021-09-16 liuhongt <hongtao.liu@intel.com>
2381 * config/i386/avx512fp16intrin.h (_mm512_cvtepi32_ph): New
2383 (_mm512_mask_cvtepi32_ph): Likewise.
2384 (_mm512_maskz_cvtepi32_ph): Likewise.
2385 (_mm512_cvt_roundepi32_ph): Likewise.
2386 (_mm512_mask_cvt_roundepi32_ph): Likewise.
2387 (_mm512_maskz_cvt_roundepi32_ph): Likewise.
2388 (_mm512_cvtepu32_ph): Likewise.
2389 (_mm512_mask_cvtepu32_ph): Likewise.
2390 (_mm512_maskz_cvtepu32_ph): Likewise.
2391 (_mm512_cvt_roundepu32_ph): Likewise.
2392 (_mm512_mask_cvt_roundepu32_ph): Likewise.
2393 (_mm512_maskz_cvt_roundepu32_ph): Likewise.
2394 (_mm512_cvtepi64_ph): Likewise.
2395 (_mm512_mask_cvtepi64_ph): Likewise.
2396 (_mm512_maskz_cvtepi64_ph): Likewise.
2397 (_mm512_cvt_roundepi64_ph): Likewise.
2398 (_mm512_mask_cvt_roundepi64_ph): Likewise.
2399 (_mm512_maskz_cvt_roundepi64_ph): Likewise.
2400 (_mm512_cvtepu64_ph): Likewise.
2401 (_mm512_mask_cvtepu64_ph): Likewise.
2402 (_mm512_maskz_cvtepu64_ph): Likewise.
2403 (_mm512_cvt_roundepu64_ph): Likewise.
2404 (_mm512_mask_cvt_roundepu64_ph): Likewise.
2405 (_mm512_maskz_cvt_roundepu64_ph): Likewise.
2406 (_mm512_cvtepi16_ph): Likewise.
2407 (_mm512_mask_cvtepi16_ph): Likewise.
2408 (_mm512_maskz_cvtepi16_ph): Likewise.
2409 (_mm512_cvt_roundepi16_ph): Likewise.
2410 (_mm512_mask_cvt_roundepi16_ph): Likewise.
2411 (_mm512_maskz_cvt_roundepi16_ph): Likewise.
2412 (_mm512_cvtepu16_ph): Likewise.
2413 (_mm512_mask_cvtepu16_ph): Likewise.
2414 (_mm512_maskz_cvtepu16_ph): Likewise.
2415 (_mm512_cvt_roundepu16_ph): Likewise.
2416 (_mm512_mask_cvt_roundepu16_ph): Likewise.
2417 (_mm512_maskz_cvt_roundepu16_ph): Likewise.
2418 * config/i386/avx512fp16vlintrin.h (_mm_cvtepi32_ph): New
2420 (_mm_mask_cvtepi32_ph): Likewise.
2421 (_mm_maskz_cvtepi32_ph): Likewise.
2422 (_mm256_cvtepi32_ph): Likewise.
2423 (_mm256_mask_cvtepi32_ph): Likewise.
2424 (_mm256_maskz_cvtepi32_ph): Likewise.
2425 (_mm_cvtepu32_ph): Likewise.
2426 (_mm_mask_cvtepu32_ph): Likewise.
2427 (_mm_maskz_cvtepu32_ph): Likewise.
2428 (_mm256_cvtepu32_ph): Likewise.
2429 (_mm256_mask_cvtepu32_ph): Likewise.
2430 (_mm256_maskz_cvtepu32_ph): Likewise.
2431 (_mm_cvtepi64_ph): Likewise.
2432 (_mm_mask_cvtepi64_ph): Likewise.
2433 (_mm_maskz_cvtepi64_ph): Likewise.
2434 (_mm256_cvtepi64_ph): Likewise.
2435 (_mm256_mask_cvtepi64_ph): Likewise.
2436 (_mm256_maskz_cvtepi64_ph): Likewise.
2437 (_mm_cvtepu64_ph): Likewise.
2438 (_mm_mask_cvtepu64_ph): Likewise.
2439 (_mm_maskz_cvtepu64_ph): Likewise.
2440 (_mm256_cvtepu64_ph): Likewise.
2441 (_mm256_mask_cvtepu64_ph): Likewise.
2442 (_mm256_maskz_cvtepu64_ph): Likewise.
2443 (_mm_cvtepi16_ph): Likewise.
2444 (_mm_mask_cvtepi16_ph): Likewise.
2445 (_mm_maskz_cvtepi16_ph): Likewise.
2446 (_mm256_cvtepi16_ph): Likewise.
2447 (_mm256_mask_cvtepi16_ph): Likewise.
2448 (_mm256_maskz_cvtepi16_ph): Likewise.
2449 (_mm_cvtepu16_ph): Likewise.
2450 (_mm_mask_cvtepu16_ph): Likewise.
2451 (_mm_maskz_cvtepu16_ph): Likewise.
2452 (_mm256_cvtepu16_ph): Likewise.
2453 (_mm256_mask_cvtepu16_ph): Likewise.
2454 (_mm256_maskz_cvtepu16_ph): Likewise.
2455 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2456 * config/i386/i386-builtin.def: Add corresponding new builtins.
2457 * config/i386/i386-expand.c
2458 (ix86_expand_args_builtin): Handle new builtin types.
2459 (ix86_expand_round_builtin): Ditto.
2460 * config/i386/i386-modes.def: Declare V2HF and V6HF.
2461 * config/i386/sse.md (VI2H_AVX512VL): New.
2463 (sseintvecmode): Add HF vector modes.
2464 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode><mask_name><round_name>):
2466 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
2467 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
2468 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
2469 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
2470 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask_1): Ditto.
2471 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
2472 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
2473 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
2474 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
2475 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask_1): Ditto.
2476 * config/i386/subst.md (round_qq2phsuff): New subst_attr.
2478 2021-09-16 liuhongt <hongtao.liu@intel.com>
2480 * config/i386/avx512fp16intrin.h (_mm512_cvtph_epi32):
2482 (_mm512_mask_cvtph_epi32): Likewise.
2483 (_mm512_maskz_cvtph_epi32): Likewise.
2484 (_mm512_cvt_roundph_epi32): Likewise.
2485 (_mm512_mask_cvt_roundph_epi32): Likewise.
2486 (_mm512_maskz_cvt_roundph_epi32): Likewise.
2487 (_mm512_cvtph_epu32): Likewise.
2488 (_mm512_mask_cvtph_epu32): Likewise.
2489 (_mm512_maskz_cvtph_epu32): Likewise.
2490 (_mm512_cvt_roundph_epu32): Likewise.
2491 (_mm512_mask_cvt_roundph_epu32): Likewise.
2492 (_mm512_maskz_cvt_roundph_epu32): Likewise.
2493 (_mm512_cvtph_epi64): Likewise.
2494 (_mm512_mask_cvtph_epi64): Likewise.
2495 (_mm512_maskz_cvtph_epi64): Likewise.
2496 (_mm512_cvt_roundph_epi64): Likewise.
2497 (_mm512_mask_cvt_roundph_epi64): Likewise.
2498 (_mm512_maskz_cvt_roundph_epi64): Likewise.
2499 (_mm512_cvtph_epu64): Likewise.
2500 (_mm512_mask_cvtph_epu64): Likewise.
2501 (_mm512_maskz_cvtph_epu64): Likewise.
2502 (_mm512_cvt_roundph_epu64): Likewise.
2503 (_mm512_mask_cvt_roundph_epu64): Likewise.
2504 (_mm512_maskz_cvt_roundph_epu64): Likewise.
2505 (_mm512_cvtph_epi16): Likewise.
2506 (_mm512_mask_cvtph_epi16): Likewise.
2507 (_mm512_maskz_cvtph_epi16): Likewise.
2508 (_mm512_cvt_roundph_epi16): Likewise.
2509 (_mm512_mask_cvt_roundph_epi16): Likewise.
2510 (_mm512_maskz_cvt_roundph_epi16): Likewise.
2511 (_mm512_cvtph_epu16): Likewise.
2512 (_mm512_mask_cvtph_epu16): Likewise.
2513 (_mm512_maskz_cvtph_epu16): Likewise.
2514 (_mm512_cvt_roundph_epu16): Likewise.
2515 (_mm512_mask_cvt_roundph_epu16): Likewise.
2516 (_mm512_maskz_cvt_roundph_epu16): Likewise.
2517 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_epi32):
2519 (_mm_mask_cvtph_epi32): Likewise.
2520 (_mm_maskz_cvtph_epi32): Likewise.
2521 (_mm256_cvtph_epi32): Likewise.
2522 (_mm256_mask_cvtph_epi32): Likewise.
2523 (_mm256_maskz_cvtph_epi32): Likewise.
2524 (_mm_cvtph_epu32): Likewise.
2525 (_mm_mask_cvtph_epu32): Likewise.
2526 (_mm_maskz_cvtph_epu32): Likewise.
2527 (_mm256_cvtph_epu32): Likewise.
2528 (_mm256_mask_cvtph_epu32): Likewise.
2529 (_mm256_maskz_cvtph_epu32): Likewise.
2530 (_mm_cvtph_epi64): Likewise.
2531 (_mm_mask_cvtph_epi64): Likewise.
2532 (_mm_maskz_cvtph_epi64): Likewise.
2533 (_mm256_cvtph_epi64): Likewise.
2534 (_mm256_mask_cvtph_epi64): Likewise.
2535 (_mm256_maskz_cvtph_epi64): Likewise.
2536 (_mm_cvtph_epu64): Likewise.
2537 (_mm_mask_cvtph_epu64): Likewise.
2538 (_mm_maskz_cvtph_epu64): Likewise.
2539 (_mm256_cvtph_epu64): Likewise.
2540 (_mm256_mask_cvtph_epu64): Likewise.
2541 (_mm256_maskz_cvtph_epu64): Likewise.
2542 (_mm_cvtph_epi16): Likewise.
2543 (_mm_mask_cvtph_epi16): Likewise.
2544 (_mm_maskz_cvtph_epi16): Likewise.
2545 (_mm256_cvtph_epi16): Likewise.
2546 (_mm256_mask_cvtph_epi16): Likewise.
2547 (_mm256_maskz_cvtph_epi16): Likewise.
2548 (_mm_cvtph_epu16): Likewise.
2549 (_mm_mask_cvtph_epu16): Likewise.
2550 (_mm_maskz_cvtph_epu16): Likewise.
2551 (_mm256_cvtph_epu16): Likewise.
2552 (_mm256_mask_cvtph_epu16): Likewise.
2553 (_mm256_maskz_cvtph_epu16): Likewise.
2554 * config/i386/i386-builtin-types.def: Add new builtin types.
2555 * config/i386/i386-builtin.def: Add new builtins.
2556 * config/i386/i386-expand.c
2557 (ix86_expand_args_builtin): Handle new builtin types.
2558 (ix86_expand_round_builtin): Ditto.
2559 * config/i386/sse.md (sseintconvert): New.
2561 (UNSPEC_US_FIX_NOTRUNC): Ditto.
2562 (sseintconvertsignprefix): Ditto.
2563 (avx512fp16_vcvtph2<sseintconvertsignprefix><sseintconvert>_<mode><mask_name><round_name>):
2566 2021-09-16 liuhongt <hongtao.liu@intel.com>
2568 * config/i386/avx512fp16intrin.h: (_mm_cvtsi16_si128):
2570 (_mm_cvtsi128_si16): Likewise.
2571 (_mm_mask_load_sh): Likewise.
2572 (_mm_maskz_load_sh): Likewise.
2573 (_mm_mask_store_sh): Likewise.
2574 (_mm_move_sh): Likewise.
2575 (_mm_mask_move_sh): Likewise.
2576 (_mm_maskz_move_sh): Likewise.
2577 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2578 * config/i386/i386-builtin.def: Add corresponding new builtins.
2579 * config/i386/i386-expand.c
2580 (ix86_expand_special_args_builtin): Handle new builtin types.
2581 (ix86_expand_vector_init_one_nonzero): Adjust for FP16 target.
2582 * config/i386/sse.md (VI2F): New mode iterator.
2583 (vec_set<mode>_0): Use new mode iterator.
2584 (avx512f_mov<ssescalarmodelower>_mask): Adjust for HF vector mode.
2585 (avx512f_store<mode>_mask): Ditto.
2587 2021-09-16 Kewen Lin <linkw@linux.ibm.com>
2589 * config/rs6000/rs6000.opt (-mtoc-fusion): Remove.
2591 2021-09-15 David Edelsohn <dje.gcc@gmail.com>
2593 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
2594 Proceed if no symbol summary or the symbol alias flag is false.
2596 2021-09-15 Jakub Jelinek <jakub@redhat.com>
2600 * varasm.c (output_constructor_regular_field): Instead of assertion
2601 that array_size_for_constructor result is equal to size of
2602 TREE_TYPE (local->val) in bytes, assert that the type size is greater
2603 or equal to array_size_for_constructor result and use type size as
2606 2021-09-15 Martin Liska <mliska@suse.cz>
2609 * config/i386/vxworks.h: Use new macro TARGET_CPU_P.
2611 2021-09-15 Martin Liska <mliska@suse.cz>
2614 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
2615 Check that we have a symbol summary for a symbol.
2617 2021-09-15 Richard Biener <rguenther@suse.de>
2620 * config/rs6000/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
2621 to inherit from elfos.h
2623 2021-09-15 liuhongt <hongtao.liu@intel.com>
2626 * config/i386/i386-expand.c
2627 (ix86_expand_vector_init_interleave): Use puncklwd to pack 2
2629 (ix86_expand_vector_set): Use blendw instead of pinsrw.
2630 * config/i386/i386.c (ix86_can_change_mode_class): Adjust for
2631 AVX512FP16 which supports 16bit vector load.
2632 * config/i386/sse.md (avx512bw_interleave_highv32hi<mask_name>):
2634 (avx512bw_interleave_high<mode><mask_name>): .. this, and
2635 extend to V32HFmode.
2636 (avx2_interleave_highv16hi<mask_name>): Rename to ..
2637 (avx2_interleave_high<mode><mask_name>): .. this, and extend
2639 (vec_interleave_highv8hi<mask_name>): Rename to ..
2640 (vec_interleave_high<mode><mask_name>): .. this, and extend to V8HFmode.
2641 (<mask_codefor>avx512bw_interleave_lowv32hi<mask_name>):
2643 (<mask_codefor>avx512bw_interleave_low<mode><mask_name>):
2644 this, and extend to V32HFmode.
2645 (avx2_interleave_lowv16hi<mask_name>): Rename to ..
2646 (avx2_interleave_low<mode><mask_name>): .. this, and extend to V16HFmode.
2647 (vec_interleave_lowv8hi<mask_name>): Rename to ..
2648 (vec_interleave_low<mode><mask_name>): .. this, and extend to V8HFmode.
2649 (sse4_1_pblendw): Rename to ..
2650 (sse4_1_pblend<blendsuf>): .. this, and extend to V8HFmode.
2651 (avx2_pblendph): New define_expand.
2652 (<sse2p4_1>_pinsr<ssemodesuffix>): Refactor, use
2653 sseintmodesuffix instead of ssemodesuffix.
2654 (blendsuf): New mode attr.
2656 2021-09-15 Richard Biener <rguenther@suse.de>
2658 * tree-vectorizer.h (dr_misalignment): Move out of line.
2659 (dr_target_alignment): New.
2660 (DR_TARGET_ALIGNMENT): Wrap dr_target_alignment.
2661 (set_dr_target_alignment): New.
2662 (SET_DR_TARGET_ALIGNMENT): Wrap set_dr_target_alignment.
2663 * tree-vect-data-refs.c (dr_misalignment): Compute and
2664 return the group members misalignment.
2665 (vect_compute_data_ref_alignment): Use SET_DR_TARGET_ALIGNMENT.
2666 (vect_analyze_data_refs_alignment): Compute alignment only
2667 for the first element of a DR group.
2668 (vect_slp_analyze_node_alignment): Likewise.
2670 2021-09-15 Hongyu Wang <hongyu.wang@intel.com>
2672 * config/i386/avx512fp16intrin.h: Adjust all builtin calls.
2673 * config/i386/avx512fp16vlintrin.h: Likewise.
2674 * config/i386/i386-builtin.def: Adjust builtin name and
2675 enumeration to match AVX512F style.
2677 2021-09-15 Richard Biener <rguenther@suse.de>
2679 PR tree-optimization/102318
2680 * tree-vect-loop.c (vect_transform_cycle_phi): Revert
2681 previous change and do the mode conversion separately from
2682 the sign conversion.
2684 2021-09-15 Hongtao Liu <hongtao.liu@intel.com>
2685 Peter Cordes <peter@cordes.ca>
2688 * config/i386/sse.md (extract_suf): Add V8SF/V8SI/V4DF/V4DI.
2689 (*vec_extract<mode><ssescalarmodelower>_valign): Output
2690 vextract{i,f}{32x4,64x2} instruction when byte_offset % 16 ==
2693 2021-09-15 Richard Biener <rguenther@suse.de>
2695 * config.gcc: Remove vax-*-openbsd* configuration.
2697 2021-09-15 Richard Biener <rguenther@suse.de>
2699 * config.gcc: Remove m68k-openbsd.
2701 2021-09-15 Max Filippov <jcmvbkbc@gmail.com>
2704 * config/xtensa/t-xtensa (TM_H): Add include/xtensa-config.h.
2706 2021-09-14 Peter Bergner <bergner@linux.ibm.com>
2708 * config/rs6000/mma.md (unspec): Delete UNSPEC_MMA_XXSETACCZ.
2709 (unspecv): Add UNSPECV_MMA_XXSETACCZ.
2710 (*mma_xxsetaccz): Delete.
2711 (mma_xxsetaccz): Change to define_insn. Remove operand 1.
2712 Use UNSPECV_MMA_XXSETACCZ. Update comment.
2713 * config/rs6000/rs6000.c (rs6000_rtx_costs): Use UNSPECV_MMA_XXSETACCZ.
2715 2021-09-14 Iain Sandoe <iain@sandoe.co.uk>
2717 * Makefile.in: Remove variables related to applying no-PIE
2718 to the exes on $build.
2719 * configure: Regenerate.
2720 * configure.ac: Remove configuration related to applying
2721 no-PIE to the exes on $build.
2723 2021-09-14 Claudiu Zissulescu <claziss@synopsys.com>
2725 * config/arc/arc.md (doloop_end): Add missing mode.
2726 (loop_end): Likewise.
2728 2021-09-14 Jakub Jelinek <jakub@redhat.com>
2730 * gimplify.c (goa_stabilize_expr): Add depth argument, propagate
2731 it to recursive calls, for depth above 7 just gimplify or return.
2732 Perform a test even for MODIFY_EXPR, ADDR_EXPR, COMPOUND_EXPR with
2733 __builtin_clear_padding and TARGET_EXPR.
2734 (gimplify_omp_atomic): Adjust goa_stabilize_expr callers.
2736 2021-09-14 liuhongt <hongtao.liu@intel.com>
2738 * config/i386/avx512fp16intrin.h (_mm_fpclass_sh_mask):
2740 (_mm_mask_fpclass_sh_mask): Likewise.
2741 (_mm512_mask_fpclass_ph_mask): Likewise.
2742 (_mm512_fpclass_ph_mask): Likewise.
2743 (_mm_getexp_sh): Likewise.
2744 (_mm_mask_getexp_sh): Likewise.
2745 (_mm_maskz_getexp_sh): Likewise.
2746 (_mm512_getexp_ph): Likewise.
2747 (_mm512_mask_getexp_ph): Likewise.
2748 (_mm512_maskz_getexp_ph): Likewise.
2749 (_mm_getexp_round_sh): Likewise.
2750 (_mm_mask_getexp_round_sh): Likewise.
2751 (_mm_maskz_getexp_round_sh): Likewise.
2752 (_mm512_getexp_round_ph): Likewise.
2753 (_mm512_mask_getexp_round_ph): Likewise.
2754 (_mm512_maskz_getexp_round_ph): Likewise.
2755 (_mm_getmant_sh): Likewise.
2756 (_mm_mask_getmant_sh): Likewise.
2757 (_mm_maskz_getmant_sh): Likewise.
2758 (_mm512_getmant_ph): Likewise.
2759 (_mm512_mask_getmant_ph): Likewise.
2760 (_mm512_maskz_getmant_ph): Likewise.
2761 (_mm_getmant_round_sh): Likewise.
2762 (_mm_mask_getmant_round_sh): Likewise.
2763 (_mm_maskz_getmant_round_sh): Likewise.
2764 (_mm512_getmant_round_ph): Likewise.
2765 (_mm512_mask_getmant_round_ph): Likewise.
2766 (_mm512_maskz_getmant_round_ph): Likewise.
2767 * config/i386/avx512fp16vlintrin.h (_mm_mask_fpclass_ph_mask):
2769 (_mm_fpclass_ph_mask): Likewise.
2770 (_mm256_mask_fpclass_ph_mask): Likewise.
2771 (_mm256_fpclass_ph_mask): Likewise.
2772 (_mm256_getexp_ph): Likewise.
2773 (_mm256_mask_getexp_ph): Likewise.
2774 (_mm256_maskz_getexp_ph): Likewise.
2775 (_mm_getexp_ph): Likewise.
2776 (_mm_mask_getexp_ph): Likewise.
2777 (_mm_maskz_getexp_ph): Likewise.
2778 (_mm256_getmant_ph): Likewise.
2779 (_mm256_mask_getmant_ph): Likewise.
2780 (_mm256_maskz_getmant_ph): Likewise.
2781 (_mm_getmant_ph): Likewise.
2782 (_mm_mask_getmant_ph): Likewise.
2783 (_mm_maskz_getmant_ph): Likewise.
2784 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2785 * config/i386/i386-builtin.def: Add corresponding new builtins.
2786 * config/i386/i386-expand.c
2787 (ix86_expand_args_builtin): Handle new builtin types.
2788 (ix86_expand_round_builtin): Ditto.
2789 * config/i386/sse.md (vecmemsuffix): Add HF vector modes.
2790 (<avx512>_getexp<mode><mask_name><round_saeonly_name>): Adjust
2791 to support HF vector modes.
2792 (avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name):
2794 (avx512dq_fpclass<mode><mask_scalar_merge_name>): Ditto.
2795 (avx512dq_vmfpclass<mode><mask_scalar_merge_name>): Ditto.
2796 (<avx512>_getmant<mode><mask_name><round_saeonly_name>): Ditto.
2797 (avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2800 2021-09-14 liuhongt <hongtao.liu@intel.com>
2802 * config/i386/avx512fp16intrin.h (_mm512_reduce_ph):
2804 (_mm512_mask_reduce_ph): Likewise.
2805 (_mm512_maskz_reduce_ph): Likewise.
2806 (_mm512_reduce_round_ph): Likewise.
2807 (_mm512_mask_reduce_round_ph): Likewise.
2808 (_mm512_maskz_reduce_round_ph): Likewise.
2809 (_mm_reduce_sh): Likewise.
2810 (_mm_mask_reduce_sh): Likewise.
2811 (_mm_maskz_reduce_sh): Likewise.
2812 (_mm_reduce_round_sh): Likewise.
2813 (_mm_mask_reduce_round_sh): Likewise.
2814 (_mm_maskz_reduce_round_sh): Likewise.
2815 (_mm512_roundscale_ph): Likewise.
2816 (_mm512_mask_roundscale_ph): Likewise.
2817 (_mm512_maskz_roundscale_ph): Likewise.
2818 (_mm512_roundscale_round_ph): Likewise.
2819 (_mm512_mask_roundscale_round_ph): Likewise.
2820 (_mm512_maskz_roundscale_round_ph): Likewise.
2821 (_mm_roundscale_sh): Likewise.
2822 (_mm_mask_roundscale_sh): Likewise.
2823 (_mm_maskz_roundscale_sh): Likewise.
2824 (_mm_roundscale_round_sh): Likewise.
2825 (_mm_mask_roundscale_round_sh): Likewise.
2826 (_mm_maskz_roundscale_round_sh): Likewise.
2827 * config/i386/avx512fp16vlintrin.h: (_mm_reduce_ph):
2829 (_mm_mask_reduce_ph): Likewise.
2830 (_mm_maskz_reduce_ph): Likewise.
2831 (_mm256_reduce_ph): Likewise.
2832 (_mm256_mask_reduce_ph): Likewise.
2833 (_mm256_maskz_reduce_ph): Likewise.
2834 (_mm_roundscale_ph): Likewise.
2835 (_mm_mask_roundscale_ph): Likewise.
2836 (_mm_maskz_roundscale_ph): Likewise.
2837 (_mm256_roundscale_ph): Likewise.
2838 (_mm256_mask_roundscale_ph): Likewise.
2839 (_mm256_maskz_roundscale_ph): Likewise.
2840 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2841 * config/i386/i386-builtin.def: Add corresponding new builtins.
2842 * config/i386/i386-expand.c
2843 (ix86_expand_args_builtin): Handle new builtin types.
2844 (ix86_expand_round_builtin): Ditto.
2845 * config/i386/sse.md (<mask_codefor>reducep<mode><mask_name>):
2847 (<mask_codefor>reducep<mode><mask_name><round_saeonly_name>):
2848 ... this, and adjust for round operands.
2849 (reduces<mode><mask_scalar_name>): Likewise, with ...
2850 (reduces<mode><mask_scalar_name><round_saeonly_scalar_name):
2852 (<avx512>_rndscale<mode><mask_name><round_saeonly_name>):
2853 Adjust for HF vector modes.
2854 (avx512f_rndscale<mode><mask_scalar_name><round_saeonly_scalar_name>):
2856 (*avx512f_rndscale<mode><round_saeonly_name>): Ditto.
2858 2021-09-14 liuhongt <hongtao.liu@intel.com>
2860 * config/i386/avx512fp16intrin.h: (_mm512_rcp_ph):
2862 (_mm512_mask_rcp_ph): Likewise.
2863 (_mm512_maskz_rcp_ph): Likewise.
2864 (_mm_rcp_sh): Likewise.
2865 (_mm_mask_rcp_sh): Likewise.
2866 (_mm_maskz_rcp_sh): Likewise.
2867 (_mm512_scalef_ph): Likewise.
2868 (_mm512_mask_scalef_ph): Likewise.
2869 (_mm512_maskz_scalef_ph): Likewise.
2870 (_mm512_scalef_round_ph): Likewise.
2871 (_mm512_mask_scalef_round_ph): Likewise.
2872 (_mm512_maskz_scalef_round_ph): Likewise.
2873 (_mm_scalef_sh): Likewise.
2874 (_mm_mask_scalef_sh): Likewise.
2875 (_mm_maskz_scalef_sh): Likewise.
2876 (_mm_scalef_round_sh): Likewise.
2877 (_mm_mask_scalef_round_sh): Likewise.
2878 (_mm_maskz_scalef_round_sh): Likewise.
2879 * config/i386/avx512fp16vlintrin.h (_mm_rcp_ph):
2881 (_mm256_rcp_ph): Likewise.
2882 (_mm_mask_rcp_ph): Likewise.
2883 (_mm256_mask_rcp_ph): Likewise.
2884 (_mm_maskz_rcp_ph): Likewise.
2885 (_mm256_maskz_rcp_ph): Likewise.
2886 (_mm_scalef_ph): Likewise.
2887 (_mm256_scalef_ph): Likewise.
2888 (_mm_mask_scalef_ph): Likewise.
2889 (_mm256_mask_scalef_ph): Likewise.
2890 (_mm_maskz_scalef_ph): Likewise.
2891 (_mm256_maskz_scalef_ph): Likewise.
2892 * config/i386/i386-builtin.def: Add new builtins.
2893 * config/i386/sse.md (VFH_AVX512VL): New.
2894 (avx512fp16_rcp<mode>2<mask_name>): Ditto.
2895 (avx512fp16_vmrcpv8hf2<mask_scalar_name>): Ditto.
2896 (avx512f_vmscalef<mode><mask_scalar_name><round_scalar_name>):
2897 Adjust to support HF vector modes.
2898 (<avx512>_scalef<mode><mask_name><round_name>): Ditto.
2900 2021-09-14 liuhongt <hongtao.liu@intel.com>
2902 * config/i386/avx512fp16intrin.h: (_mm512_sqrt_ph):
2904 (_mm512_mask_sqrt_ph): Likewise.
2905 (_mm512_maskz_sqrt_ph): Likewise.
2906 (_mm512_sqrt_round_ph): Likewise.
2907 (_mm512_mask_sqrt_round_ph): Likewise.
2908 (_mm512_maskz_sqrt_round_ph): Likewise.
2909 (_mm512_rsqrt_ph): Likewise.
2910 (_mm512_mask_rsqrt_ph): Likewise.
2911 (_mm512_maskz_rsqrt_ph): Likewise.
2912 (_mm_rsqrt_sh): Likewise.
2913 (_mm_mask_rsqrt_sh): Likewise.
2914 (_mm_maskz_rsqrt_sh): Likewise.
2915 (_mm_sqrt_sh): Likewise.
2916 (_mm_mask_sqrt_sh): Likewise.
2917 (_mm_maskz_sqrt_sh): Likewise.
2918 (_mm_sqrt_round_sh): Likewise.
2919 (_mm_mask_sqrt_round_sh): Likewise.
2920 (_mm_maskz_sqrt_round_sh): Likewise.
2921 * config/i386/avx512fp16vlintrin.h (_mm_sqrt_ph): New intrinsic.
2922 (_mm256_sqrt_ph): Likewise.
2923 (_mm_mask_sqrt_ph): Likewise.
2924 (_mm256_mask_sqrt_ph): Likewise.
2925 (_mm_maskz_sqrt_ph): Likewise.
2926 (_mm256_maskz_sqrt_ph): Likewise.
2927 (_mm_rsqrt_ph): Likewise.
2928 (_mm256_rsqrt_ph): Likewise.
2929 (_mm_mask_rsqrt_ph): Likewise.
2930 (_mm256_mask_rsqrt_ph): Likewise.
2931 (_mm_maskz_rsqrt_ph): Likewise.
2932 (_mm256_maskz_rsqrt_ph): Likewise.
2933 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2934 * config/i386/i386-builtin.def: Add corresponding new builtins.
2935 * config/i386/i386-expand.c
2936 (ix86_expand_args_builtin): Handle new builtins.
2937 (ix86_expand_round_builtin): Ditto.
2938 * config/i386/sse.md (VF_AVX512FP16VL): New.
2939 (sqrt<mode>2): Adjust for HF vector modes.
2940 (<sse>_sqrt<mode>2<mask_name><round_name>): Likewise.
2941 (<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
2943 (<sse>_rsqrt<mode>2<mask_name>): New.
2944 (avx512fp16_vmrsqrtv8hf2<mask_scalar_name>): Likewise.
2946 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
2949 * diagnostic-spec.c (warning_suppressed_at, copy_warning): Handle
2950 'RESERVED_LOCATION_P' locations.
2951 * warning-control.cc (get_nowarn_spec, suppress_warning)
2952 (copy_warning): Likewise.
2954 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
2956 * diagnostic-spec.h (typedef xint_hash_t): Use 'location_t' instead of...
2957 (typedef key_type_t): ... this. Remove.
2958 (nowarn_map): Document.
2959 * diagnostic-spec.c (nowarn_map): Likewise.
2960 * warning-control.cc (convert_to_key): Evolve functions into...
2961 (get_location): ... these. Adjust all users.
2963 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
2965 * warning-control.cc (copy_warning): Remove 'nowarn_map' setup.
2967 2021-09-13 Jason Merrill <jason@redhat.com>
2969 * params.opt: Add destructive-interference-size and
2970 constructive-interference-size.
2971 * doc/invoke.texi: Document them.
2972 * config/aarch64/aarch64.c (aarch64_override_options_internal):
2974 * config/arm/arm.c (arm_option_override): Set them.
2975 * config/i386/i386-options.c (ix86_option_override_internal):
2978 2021-09-13 Martin Liska <mliska@suse.cz>
2979 H.J. Lu <hjl.tools@gmail.com>
2982 * common/config/i386/cpuinfo.h (cpu_indicator_init): Add support
2983 for x86-64 micro levels for __builtin_cpu_supports.
2984 * common/config/i386/i386-cpuinfo.h (enum feature_priority):
2985 Add priorities for the micro-arch levels.
2986 (enum processor_features): Add new features.
2987 * common/config/i386/i386-isas.h: Add micro-arch features.
2988 * config/i386/i386-builtins.c (get_builtin_code_for_version):
2989 Support the micro-arch levels by callsing
2990 __builtin_cpu_supports.
2991 * doc/extend.texi: Document that the levels are support by
2992 __builtin_cpu_supports.
2994 2021-09-13 Andrew Pinski <apinski@marvell.com>
2997 * config/aarch64/aarch64-builtins.c (aarch64_fold_builtin_lane_check):
2999 (aarch64_general_fold_builtin): Handle AARCH64_SIMD_BUILTIN_LANE_CHECK.
3000 (aarch64_general_gimple_fold_builtin): Likewise.
3002 2021-09-13 Andrew Pinski <apinski@marvell.com>
3004 * config.gcc: Add m32r-*-linux* and m32rle-*-linux*
3005 to the Unsupported targets list.
3006 Remove support for m32r-*-linux* and m32rle-*-linux*.
3007 * config/m32r/linux.h: Removed.
3008 * config/m32r/t-linux: Removed.
3010 2021-09-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3013 * config/aarch64/aarch64.c (aarch64_classify_address): Don't allow
3014 register index for SVE predicate modes.
3016 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
3018 * tree-ssa-threadbackward.c
3019 (back_threader_profitability::profitable_path_p): Remove FSM
3021 (back_threader_registry::register_path): Same.
3022 * tree-ssa-threadedge.c
3023 (jump_threader::simplify_control_stmt_condition): Same.
3024 * tree-ssa-threadupdate.c (jt_path_registry::jt_path_registry):
3025 Add backedge_threads argument.
3026 (fwd_jt_path_registry::fwd_jt_path_registry): Pass
3027 backedge_threads argument.
3028 (back_jt_path_registry::back_jt_path_registry): Same.
3029 (dump_jump_thread_path): Adjust for FSM removal.
3030 (back_jt_path_registry::rewire_first_differing_edge): Same.
3031 (back_jt_path_registry::adjust_paths_after_duplication): Same.
3032 (back_jt_path_registry::update_cfg): Same.
3033 (jt_path_registry::register_jump_thread): Same.
3034 * tree-ssa-threadupdate.h (enum jump_thread_edge_type): Remove
3036 (class back_jt_path_registry): Add backedge_threads to
3039 2021-09-13 Martin Liska <mliska@suse.cz>
3042 * asan.h (sanitize_coverage_p): Handle when fn == NULL.
3044 2021-09-13 H.J. Lu <hjl.tools@gmail.com>
3047 * config/i386/i386.h (TARGET_AVX256_MOVE_BY_PIECES): New.
3048 (TARGET_AVX256_STORE_BY_PIECES): Likewise.
3049 (MOVE_MAX): Check TARGET_AVX256_MOVE_BY_PIECES and
3050 TARGET_AVX256_STORE_BY_PIECES instead of
3051 TARGET_AVX256_SPLIT_UNALIGNED_LOAD and
3052 TARGET_AVX256_SPLIT_UNALIGNED_STORE.
3053 (STORE_MAX_PIECES): Check TARGET_AVX256_STORE_BY_PIECES instead
3054 of TARGET_AVX256_SPLIT_UNALIGNED_STORE.
3055 * config/i386/x86-tune.def (X86_TUNE_AVX256_MOVE_BY_PIECES): New.
3056 (X86_TUNE_AVX256_STORE_BY_PIECES): Likewise.
3058 2021-09-13 liuhongt <hongtao.liu@intel.com>
3061 * expmed.c (extract_bit_field_using_extv): Use
3062 gen_lowpart_if_possible instead of gen_lowpart to avoid ICE.
3064 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
3066 * Makefile.in (OBJS): Add value-pointer-equiv.o.
3067 * gimple-ssa-evrp.c (class ssa_equiv_stack): Move to
3068 value-pointer-equiv.*.
3069 (ssa_equiv_stack::ssa_equiv_stack): Same.
3070 (ssa_equiv_stack::enter): Same.
3071 (ssa_equiv_stack::leave): Same.
3072 (ssa_equiv_stack::push_replacement): Same.
3073 (ssa_equiv_stack::get_replacement): Same.
3074 (is_pointer_ssa): Same.
3075 (class pointer_equiv_analyzer): Same.
3076 (pointer_equiv_analyzer::pointer_equiv_analyzer): Same.
3077 (pointer_equiv_analyzer::~pointer_equiv_analyzer): Same.
3078 (pointer_equiv_analyzer::set_global_equiv): Same.
3079 (pointer_equiv_analyzer::set_cond_equiv): Same.
3080 (pointer_equiv_analyzer::get_equiv): Same.
3081 (pointer_equiv_analyzer::enter): Same.
3082 (pointer_equiv_analyzer::leave): Same.
3083 (pointer_equiv_analyzer::get_equiv_expr): Same.
3084 (pta_valueize): Same.
3085 (pointer_equiv_analyzer::visit_stmt): Same.
3086 (pointer_equiv_analyzer::visit_edge): Same.
3087 (hybrid_folder::value_of_expr): Same.
3088 (hybrid_folder::value_on_edge): Same.
3089 * value-pointer-equiv.cc: New file.
3090 * value-pointer-equiv.h: New file.
3092 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
3095 * gimple-fold.c (gimple_fold_builtin_memory_op): Allow folding
3096 memcpy if the size is not more than MOVE_MAX * MOVE_RATIO.
3098 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
3101 * config/arm/arm.md (movmisaligndi): New define_expand.
3102 * config/arm/vec-common.md (movmisalign<mode>): Iterate over VDQ mode.
3104 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
3107 * emit-rtl.c (gen_highpart): Use adjust_address to handle
3108 MEM rather than calling simplify_gen_subreg.
3110 2021-09-13 Jan-Benedict Glaw <jbglaw@ług-owl.de>
3112 * config/alpha/vms.h (INIT_CUMULATIVE_ARGS): Wrap multi-statment
3113 define into a block.
3115 2021-09-13 Richard Biener <rguenther@suse.de>
3117 * config/darwin.h (DARWIN_PREFER_DWARF): Do not define.
3118 * config/i386/darwin.h (PREFERRED_DEBUGGING_TYPE): Do not
3119 change based on DARWIN_PREFER_DWARF not being defined.
3121 2021-09-13 Richard Biener <rguenther@suse.de>
3123 * config/i386/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
3124 to inherit from elfos.h
3126 2021-09-13 Richard Biener <rguenther@suse.de>
3128 * config.gcc: Add cr16-*-* to the list of obsoleted targets.
3130 2021-09-13 Richard Biener <rguenther@suse.de>
3132 * config/avr/elf.h (PREFERRED_DEBUGGING_TYPE): Remove
3133 override, pick up DWARF2_DEBUG define from elfos.h
3135 2021-09-13 Richard Biener <rguenther@suse.de>
3137 * config/rx/rx.h (PREFERRED_DEBUGGING_TYPE): Always define to
3140 2021-09-13 Richard Biener <rguenther@suse.de>
3142 * config/alpha/vms.h (PREFERRED_DEBUGGING_TYPE): Define to
3145 2021-09-13 Richard Biener <rguenther@suse.de>
3147 * config/i386/cygming.h: Always default to DWARF2 debugging.
3148 Do not define DBX_DEBUGGING_INFO, that's done via dbxcoff.h
3150 * doc/install.texi: Document binutils 2.16 as minimum
3151 requirement for mingw.
3153 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
3155 * config/rs6000/rs6000.c (struct rs6000_cost_data): New members
3156 nstmts, nloads and extra_ctor_cost.
3157 (rs6000_density_test): Add load density related heuristics. Do
3158 extra costing on vector construction statements if need.
3159 (rs6000_init_cost): Init new members.
3160 (rs6000_update_target_cost_per_stmt): New function.
3161 (rs6000_add_stmt_cost): Factor vect_nonmem hunk out to function
3162 rs6000_update_target_cost_per_stmt and call it.
3164 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
3166 * config/rs6000/rs6000.c (struct rs6000_cost_data): Remove typedef.
3167 (rs6000_init_cost): Adjust.
3169 2021-09-13 liuhongt <hongtao.liu@intel.com>
3171 * config/i386/i386.md: (UNSPEC_COPYSIGN): Remove.
3172 (UNSPEC_XORSIGN): Ditto.
3174 2021-09-12 Roger Sayle <roger@nextmovesoftware.com>
3176 * expr.c (convert_move): Preserve SUBREG_PROMOTED_VAR_P when
3177 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
3180 2021-09-11 Aldy Hernandez <aldyh@redhat.com>
3182 * tree-ssa-threadbackward.c (class back_threader_registry): Use
3183 back_jt_path_registry.
3184 * tree-ssa-threadedge.c (jump_threader::jump_threader): Use
3185 fwd_jt_path_registry.
3186 * tree-ssa-threadedge.h (class jump_threader): Same..
3187 * tree-ssa-threadupdate.c
3188 (jump_thread_path_registry::jump_thread_path_registry): Rename...
3189 (jt_path_registry::jt_path_registry): ...to this.
3190 (jump_thread_path_registry::~jump_thread_path_registry): Rename...
3191 (jt_path_registry::~jt_path_registry): ...this.
3192 (fwd_jt_path_registry::fwd_jt_path_registry): New.
3193 (fwd_jt_path_registry::~fwd_jt_path_registry): New.
3194 (jump_thread_path_registry::allocate_thread_edge): Rename...
3195 (jt_path_registry::allocate_thread_edge): ...to this.
3196 (jump_thread_path_registry::allocate_thread_path): Rename...
3197 (jt_path_registry::allocate_thread_path): ...to this.
3198 (jump_thread_path_registry::lookup_redirection_data): Rename...
3199 (fwd_jt_path_registry::lookup_redirection_data): ...to this.
3200 (jump_thread_path_registry::thread_block_1): Rename...
3201 (fwd_jt_path_registry::thread_block_1): ...to this.
3202 (jump_thread_path_registry::thread_block): Rename...
3203 (fwd_jt_path_registry::thread_block): ...to this.
3204 (jt_path_registry::thread_through_loop_header): Rename...
3205 (fwd_jt_path_registry::thread_through_loop_header): ...to this.
3206 (jump_thread_path_registry::mark_threaded_blocks): Rename...
3207 (fwd_jt_path_registry::mark_threaded_blocks): ...to this.
3208 (jump_thread_path_registry::debug_path): Rename...
3209 (jt_path_registry::debug_path): ...to this.
3210 (jump_thread_path_registry::dump): Rename...
3211 (jt_path_registry::debug): ...to this.
3212 (jump_thread_path_registry::rewire_first_differing_edge): Rename...
3213 (back_jt_path_registry::rewire_first_differing_edge): ...to this.
3214 (jump_thread_path_registry::adjust_paths_after_duplication): Rename...
3215 (back_jt_path_registry::adjust_paths_after_duplication): ...to this.
3216 (jump_thread_path_registry::duplicate_thread_path): Rename...
3217 (back_jt_path_registry::duplicate_thread_path): ...to this. Also,
3218 drop ill-formed candidates.
3219 (jump_thread_path_registry::remove_jump_threads_including): Rename...
3220 (fwd_jt_path_registry::remove_jump_threads_including): ...to this.
3221 (jt_path_registry::thread_through_all_blocks): New.
3222 (back_jt_path_registry::update_cfg): New.
3223 (fwd_jt_path_registry::update_cfg): New.
3224 (jump_thread_path_registry::register_jump_thread): Rename...
3225 (jt_path_registry::register_jump_thread): ...to this.
3226 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
3228 (class jt_path_registry): ...here.
3229 (class fwd_jt_path_registry): New.
3230 (class back_jt_path_registry): New.
3232 2021-09-10 liuhongt <hongtao.liu@intel.com>
3235 2021-09-01 liuhongt <hongtao.liu@intel.com>
3237 * emit-rtl.c (validate_subreg): Get rid of all float-int
3240 2021-09-10 Jakub Jelinek <jakub@redhat.com>
3242 * tree-core.h (enum omp_memory_order): Add OMP_MEMORY_ORDER_MASK,
3243 OMP_FAIL_MEMORY_ORDER_UNSPECIFIED, OMP_FAIL_MEMORY_ORDER_RELAXED,
3244 OMP_FAIL_MEMORY_ORDER_ACQUIRE, OMP_FAIL_MEMORY_ORDER_RELEASE,
3245 OMP_FAIL_MEMORY_ORDER_ACQ_REL, OMP_FAIL_MEMORY_ORDER_SEQ_CST and
3246 OMP_FAIL_MEMORY_ORDER_MASK enumerators.
3247 (OMP_FAIL_MEMORY_ORDER_SHIFT): Define.
3248 * gimple-pretty-print.c (dump_gimple_omp_atomic_load,
3249 dump_gimple_omp_atomic_store): Print [weak] for weak atomic
3251 * gimple.h (enum gf_mask): Change GF_OMP_ATOMIC_MEMORY_ORDER
3252 to 6-bit mask, adjust GF_OMP_ATOMIC_NEED_VALUE value and add
3254 (gimple_omp_atomic_weak_p, gimple_omp_atomic_set_weak): New inline
3256 * tree.h (OMP_ATOMIC_WEAK): Define.
3257 * tree-pretty-print.c (dump_omp_atomic_memory_order): Adjust for
3258 fail memory order being encoded in the same enum and also print
3259 fail clause if present.
3260 (dump_generic_node): Print weak clause if OMP_ATOMIC_WEAK.
3261 * gimplify.c (goa_stabilize_expr): Add target_expr and rhs arguments,
3262 handle pre_p == NULL case as a test mode that only returns value
3263 but doesn't change gimplify nor change anything otherwise, adjust
3264 recursive calls, add MODIFY_EXPR, ADDR_EXPR, COND_EXPR, TARGET_EXPR
3265 and CALL_EXPR handling, adjust COMPOUND_EXPR handling for
3266 __builtin_clear_padding calls, for !rhs gimplify as lvalue rather
3268 (gimplify_omp_atomic): Adjust goa_stabilize_expr caller. Handle
3269 COND_EXPR rhs. Set weak flag on gimple load/store for
3271 * omp-expand.c (omp_memory_order_to_fail_memmodel): New function.
3272 (omp_memory_order_to_memmodel): Adjust for fail clause encoded
3274 (expand_omp_atomic_cas): New function.
3275 (expand_omp_atomic_pipeline): Use omp_memory_order_to_fail_memmodel
3277 (expand_omp_atomic): Attempt to optimize atomic compare and exchange
3278 using expand_omp_atomic_cas.
3280 2021-09-10 Aldy Hernandez <aldyh@redhat.com>
3281 Michael Matz <matz@suse.de>
3283 * tree-pass.h (PROP_loop_opts_done): New.
3284 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
3285 Intersect with global range.
3286 * tree-ssa-loop.c (tree_ssa_loop_done): Set PROP_loop_opts_done.
3287 * tree-ssa-threadbackward.c
3288 (back_threader_profitability::profitable_path_p): Disable
3289 threading through latches until after loop optimizations have run.
3291 2021-09-10 David Faust <david.faust@oracle.com>
3293 * doc/invoke.texi: Document BPF -mcpu, -mjmpext, -mjmp32 and -malu32
3296 2021-09-10 David Faust <david.faust@oracle.com>
3298 * config/bpf/bpf-opts.h (bpf_isa_version): New enum.
3299 * config/bpf/bpf-protos.h (bpf_expand_cbranch): New.
3300 * config/bpf/bpf.c (bpf_option_override): Handle -mcpu option.
3301 (bpf_expand_cbranch): New function.
3302 * config/bpf/bpf.md (AM mode iterator): Conditionalize support for SI
3304 (zero_extendsidi2): Only use mov32 instruction if it is available.
3305 (SIM mode iterator): Conditionalize support for SI mode.
3306 (JM mode iterator): New.
3307 (cbranchdi4): Update name, use new JM iterator. Use bpf_expand_cbranch.
3308 (*branch_on_di): Update name, use new JM iterator.
3309 * config/bpf/bpf.opt: (mjmpext): New option.
3313 (bpf_isa): New enum.
3315 2021-09-10 David Faust <david.faust@oracle.com>
3317 * config/bpf/bpf.md (zero_extendhidi2): Add new output template
3318 for register-to-register extensions.
3319 (zero_extendqidi2): Likewise.
3321 2021-09-10 Richard Biener <rguenther@suse.de>
3323 PR middle-end/102273
3324 * internal-fn.c (expand_DEFERRED_INIT): Always expand non-SSA vars.
3326 2021-09-10 Richard Biener <rguenther@suse.de>
3328 PR middle-end/102269
3329 * gimplify.c (is_var_need_auto_init): Empty types do not need
3332 2021-09-10 Richard Biener <rguenther@suse.de>
3334 * configure.ac (--with-stabs): Remove.
3335 * configure: Regenerate.
3336 * doc/install.texi: Remove --with-stabs documentation.
3338 2021-09-10 liuhongt <hongtao.liu@intel.com>
3340 * config/i386/avx512fp16intrin.h: (_mm512_cmp_ph_mask):
3342 (_mm512_mask_cmp_ph_mask): Likewise.
3343 (_mm512_cmp_round_ph_mask): Likewise.
3344 (_mm512_mask_cmp_round_ph_mask): Likewise.
3345 (_mm_cmp_sh_mask): Likewise.
3346 (_mm_mask_cmp_sh_mask): Likewise.
3347 (_mm_cmp_round_sh_mask): Likewise.
3348 (_mm_mask_cmp_round_sh_mask): Likewise.
3349 (_mm_comieq_sh): Likewise.
3350 (_mm_comilt_sh): Likewise.
3351 (_mm_comile_sh): Likewise.
3352 (_mm_comigt_sh): Likewise.
3353 (_mm_comige_sh): Likewise.
3354 (_mm_comineq_sh): Likewise.
3355 (_mm_ucomieq_sh): Likewise.
3356 (_mm_ucomilt_sh): Likewise.
3357 (_mm_ucomile_sh): Likewise.
3358 (_mm_ucomigt_sh): Likewise.
3359 (_mm_ucomige_sh): Likewise.
3360 (_mm_ucomineq_sh): Likewise.
3361 (_mm_comi_round_sh): Likewise.
3362 (_mm_comi_sh): Likewise.
3363 * config/i386/avx512fp16vlintrin.h (_mm_cmp_ph_mask): New intrinsic.
3364 (_mm_mask_cmp_ph_mask): Likewise.
3365 (_mm256_cmp_ph_mask): Likewise.
3366 (_mm256_mask_cmp_ph_mask): Likewise.
3367 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3368 * config/i386/i386-builtin.def: Add corresponding new builtins.
3369 * config/i386/i386-expand.c
3370 (ix86_expand_args_builtin): Handle new builtin types.
3371 (ix86_expand_round_builtin): Ditto.
3372 * config/i386/i386.md (ssevecmode): Add HF mode.
3373 (MODEFH): New mode iterator.
3374 * config/i386/sse.md
3375 (V48H_AVX512VL): New mode iterator to support HF vector modes.
3376 Ajdust corresponding description.
3377 (ssecmpintprefix): New.
3378 (VI12_AVX512VL): Adjust to support HF vector modes.
3379 (cmp_imm_predicate): Likewise.
3380 (<avx512>_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
3382 (avx512f_vmcmp<mode>3<round_saeonly_name>): Likewise.
3383 (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Likewise.
3384 (<sse>_<unord>comi<round_saeonly_name>): Likewise.
3386 2021-09-10 liuhongt <hongtao.liu@intel.com>
3388 * config/i386/avx512fp16intrin.h: (_mm512_max_ph): New intrinsic.
3389 (_mm512_mask_max_ph): Likewise.
3390 (_mm512_maskz_max_ph): Likewise.
3391 (_mm512_min_ph): Likewise.
3392 (_mm512_mask_min_ph): Likewise.
3393 (_mm512_maskz_min_ph): Likewise.
3394 (_mm512_max_round_ph): Likewise.
3395 (_mm512_mask_max_round_ph): Likewise.
3396 (_mm512_maskz_max_round_ph): Likewise.
3397 (_mm512_min_round_ph): Likewise.
3398 (_mm512_mask_min_round_ph): Likewise.
3399 (_mm512_maskz_min_round_ph): Likewise.
3400 (_mm_max_sh): Likewise.
3401 (_mm_mask_max_sh): Likewise.
3402 (_mm_maskz_max_sh): Likewise.
3403 (_mm_min_sh): Likewise.
3404 (_mm_mask_min_sh): Likewise.
3405 (_mm_maskz_min_sh): Likewise.
3406 (_mm_max_round_sh): Likewise.
3407 (_mm_mask_max_round_sh): Likewise.
3408 (_mm_maskz_max_round_sh): Likewise.
3409 (_mm_min_round_sh): Likewise.
3410 (_mm_mask_min_round_sh): Likewise.
3411 (_mm_maskz_min_round_sh): Likewise.
3412 * config/i386/avx512fp16vlintrin.h (_mm_max_ph): New intrinsic.
3413 (_mm256_max_ph): Likewise.
3414 (_mm_mask_max_ph): Likewise.
3415 (_mm256_mask_max_ph): Likewise.
3416 (_mm_maskz_max_ph): Likewise.
3417 (_mm256_maskz_max_ph): Likewise.
3418 (_mm_min_ph): Likewise.
3419 (_mm256_min_ph): Likewise.
3420 (_mm_mask_min_ph): Likewise.
3421 (_mm256_mask_min_ph): Likewise.
3422 (_mm_maskz_min_ph): Likewise.
3423 (_mm256_maskz_min_ph): Likewise.
3424 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3425 * config/i386/i386-builtin.def: Add corresponding new builtins.
3426 * config/i386/i386-expand.c
3427 (ix86_expand_args_builtin): Handle new builtin types.
3428 * config/i386/sse.md
3429 (<code><mode>3<mask_name><round_saeonly_name>): Adjust to
3430 support HF vector modes.
3431 (*<code><mode>3<mask_name><round_saeonly_name>): Likewise.
3432 (ieee_<ieee_maxmin><mode>3<mask_name><round_saeonly_name>):
3434 (<sse>_vm<code><mode>3<mask_scalar_name><round_saeonly_scalar_name>):
3436 * config/i386/subst.md (round_saeonly_mode512bit_condition):
3437 Adjust for HF vector modes.
3439 2021-09-10 Liu, Hongtao <hongtao.liu@intel.com>
3441 * config/i386/avx512fp16intrin.h (_mm_add_sh): New intrinsic.
3442 (_mm_mask_add_sh): Likewise.
3443 (_mm_maskz_add_sh): Likewise.
3444 (_mm_sub_sh): Likewise.
3445 (_mm_mask_sub_sh): Likewise.
3446 (_mm_maskz_sub_sh): Likewise.
3447 (_mm_mul_sh): Likewise.
3448 (_mm_mask_mul_sh): Likewise.
3449 (_mm_maskz_mul_sh): Likewise.
3450 (_mm_div_sh): Likewise.
3451 (_mm_mask_div_sh): Likewise.
3452 (_mm_maskz_div_sh): Likewise.
3453 (_mm_add_round_sh): Likewise.
3454 (_mm_mask_add_round_sh): Likewise.
3455 (_mm_maskz_add_round_sh): Likewise.
3456 (_mm_sub_round_sh): Likewise.
3457 (_mm_mask_sub_round_sh): Likewise.
3458 (_mm_maskz_sub_round_sh): Likewise.
3459 (_mm_mul_round_sh): Likewise.
3460 (_mm_mask_mul_round_sh): Likewise.
3461 (_mm_maskz_mul_round_sh): Likewise.
3462 (_mm_div_round_sh): Likewise.
3463 (_mm_mask_div_round_sh): Likewise.
3464 (_mm_maskz_div_round_sh): Likewise.
3465 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3466 * config/i386/i386-builtin.def: Add corresponding new builtins.
3467 * config/i386/i386-expand.c
3468 (ix86_expand_round_builtin): Handle new builtins.
3469 * config/i386/sse.md (VF_128): Change description.
3470 (<sse>_vm<plusminus_insn><mode>3<mask_scalar_name><round_scalar_name>):
3471 Adjust to support HF vector modes.
3472 (<sse>_vm<multdiv_mnemonic><mode>3<mask_scalar_name><round_scalar_name>):
3475 2021-09-10 H.J. Lu <hjl.tools@gmail.com>
3477 * config/i386/i386-expand.c
3478 (ix86_avx256_split_vector_move_misalign): Handle V16HF mode.
3479 * config/i386/i386.c
3480 (ix86_preferred_simd_mode): Handle HF mode.
3481 * config/i386/sse.md (V_256H): New mode iterator.
3482 (avx_vextractf128<mode>): Use it.
3483 (VEC_INIT_MODE): Align vector HFmode condition to vector
3484 HImodes since there're no real HF instruction used.
3485 (VEC_INIT_HALF_MODE): Ditto.
3487 (VIHF_AVX512BW): Ditto.
3488 (*vec_extracthf): Ditto.
3489 (VEC_EXTRACT_MODE): Ditto.
3491 2021-09-10 Richard Biener <rguenther@suse.de>
3494 * config/dbx.h: Remove.
3495 * config/dbxcoff.h: Do not define PREFERRED_DEBUGGING_TYPE.
3496 * config/lynx.h: Likewise.
3498 2021-09-10 liuhongt <hongtao.liu@intel.com>
3500 * config/i386/i386-expand.c (ix86_expand_copysign): Expand
3501 right into ANDNOT + AND + IOR, using paradoxical subregs.
3502 (ix86_split_copysign_const): Remove.
3503 (ix86_split_copysign_var): Ditto.
3504 * config/i386/i386-protos.h (ix86_split_copysign_const): Dotto.
3505 (ix86_split_copysign_var): Ditto.
3506 * config/i386/i386.md (@copysign<mode>3_const): Ditto.
3507 (@copysign<mode>3_var): Ditto.
3509 2021-09-09 qing zhao <qing.zhao@oracle.com>
3511 * builtins.c (expand_builtin_memset): Make external visible.
3512 * builtins.h (expand_builtin_memset): Declare extern.
3513 * common.opt (ftrivial-auto-var-init=): New option.
3514 * doc/extend.texi: Document the uninitialized attribute.
3515 * doc/invoke.texi: Document -ftrivial-auto-var-init.
3516 * flag-types.h (enum auto_init_type): New enumerated type
3518 * gimple-fold.c (clear_padding_type): Add one new parameter.
3519 (clear_padding_union): Likewise.
3520 (clear_padding_emit_loop): Likewise.
3521 (clear_type_padding_in_mask): Likewise.
3522 (gimple_fold_builtin_clear_padding): Handle this new parameter.
3523 * gimplify.c (gimple_add_init_for_auto_var): New function.
3524 (gimple_add_padding_init_for_auto_var): New function.
3525 (is_var_need_auto_init): New function.
3526 (gimplify_decl_expr): Add initialization to automatic variables per
3528 (gimplify_call_expr): Add one new parameter for call to
3529 __builtin_clear_padding.
3530 (gimplify_init_constructor): Add padding initialization in the end.
3531 * internal-fn.c (INIT_PATTERN_VALUE): New macro.
3532 (expand_DEFERRED_INIT): New function.
3533 * internal-fn.def (DEFERRED_INIT): New internal function.
3534 * tree-cfg.c (verify_gimple_call): Verify calls to .DEFERRED_INIT.
3535 * tree-sra.c (generate_subtree_deferred_init): New function.
3536 (scan_function): Avoid setting cannot_scalarize_away_bitmap for
3537 calls to .DEFERRED_INIT.
3538 (sra_modify_deferred_init): New function.
3539 (sra_modify_function_body): Handle calls to DEFERRED_INIT specially.
3540 * tree-ssa-structalias.c (find_func_aliases_for_call): Likewise.
3541 * tree-ssa-uninit.c (warn_uninit): Handle calls to DEFERRED_INIT
3543 (check_defs): Likewise.
3544 (warn_uninitialized_vars): Likewise.
3545 * tree-ssa.c (ssa_undefined_value_p): Likewise.
3546 * tree.c (build_common_builtin_nodes): Build tree node for
3547 BUILT_IN_CLEAR_PADDING when needed.
3549 2021-09-09 Richard Biener <rguenther@suse.de>
3551 * tree-ssa-loop-im.c (fill_always_executed_in_1): Walk
3554 2021-09-09 Richard Biener <rguenther@suse.de>
3556 * tree-ssa-loop-im.c (fill_always_executed_in_1): Integrate
3557 DOM walk from get_loop_body_in_dom_order using a worklist
3560 2021-09-09 liuhongt <hongtao.liu@intel.com>
3562 * config.gcc: Add avx512fp16vlintrin.h.
3563 * config/i386/avx512fp16intrin.h: (_mm512_add_ph): New intrinsic.
3564 (_mm512_mask_add_ph): Likewise.
3565 (_mm512_maskz_add_ph): Likewise.
3566 (_mm512_sub_ph): Likewise.
3567 (_mm512_mask_sub_ph): Likewise.
3568 (_mm512_maskz_sub_ph): Likewise.
3569 (_mm512_mul_ph): Likewise.
3570 (_mm512_mask_mul_ph): Likewise.
3571 (_mm512_maskz_mul_ph): Likewise.
3572 (_mm512_div_ph): Likewise.
3573 (_mm512_mask_div_ph): Likewise.
3574 (_mm512_maskz_div_ph): Likewise.
3575 (_mm512_add_round_ph): Likewise.
3576 (_mm512_mask_add_round_ph): Likewise.
3577 (_mm512_maskz_add_round_ph): Likewise.
3578 (_mm512_sub_round_ph): Likewise.
3579 (_mm512_mask_sub_round_ph): Likewise.
3580 (_mm512_maskz_sub_round_ph): Likewise.
3581 (_mm512_mul_round_ph): Likewise.
3582 (_mm512_mask_mul_round_ph): Likewise.
3583 (_mm512_maskz_mul_round_ph): Likewise.
3584 (_mm512_div_round_ph): Likewise.
3585 (_mm512_mask_div_round_ph): Likewise.
3586 (_mm512_maskz_div_round_ph): Likewise.
3587 * config/i386/avx512fp16vlintrin.h: New header.
3588 * config/i386/i386-builtin-types.def (V16HF, V8HF, V32HF):
3589 Add new builtin types.
3590 * config/i386/i386-builtin.def: Add corresponding builtins.
3591 * config/i386/i386-expand.c
3592 (ix86_expand_args_builtin): Handle new builtin types.
3593 (ix86_expand_round_builtin): Likewise.
3594 * config/i386/immintrin.h: Include avx512fp16vlintrin.h
3595 * config/i386/sse.md (VFH): New mode_iterator.
3597 (avx512fmaskmode): Add HF vector modes.
3598 (avx512fmaskhalfmode): Likewise.
3599 (<plusminus_insn><mode>3<mask_name><round_name>): Adjust to for
3601 (*<plusminus_insn><mode>3<mask_name><round_name>): Likewise.
3602 (mul<mode>3<mask_name><round_name>): Likewise.
3603 (*mul<mode>3<mask_name><round_name>): Likewise.
3604 (div<mode>3): Likewise.
3605 (<sse>_div<mode>3<mask_name><round_name>): Likewise.
3606 * config/i386/subst.md (SUBST_V): Add HF vector modes.
3607 (SUBST_A): Likewise.
3608 (round_mode512bit_condition): Adjust for V32HFmode.
3610 2021-09-09 liuhongt <hongtao.liu@intel.com>
3613 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
3614 (reduc_plus_scal_v4sf): .. this, New define_expand.
3615 (reduc_plus_scal_v2df): .. and this, New define_expand.
3617 2021-09-09 liuhongt <hongtao.liu@intel.com>
3620 * config/i386/sse.md (*vec_extract<mode><ssescalarmodelower>_valign):
3623 2021-09-08 Jonathan Wakely <jwakely@redhat.com>
3626 * doc/trouble.texi (Copy Assignment): Fix description of
3627 behaviour and fix code in example.
3629 2021-09-08 Segher Boessenkool <segher@kernel.crashing.org>
3632 * config/rs6000/rs6000-logue.c (rs6000_emit_epilogue): For ELFv2 use
3633 r11 instead of r12 for restoring CR.
3635 2021-09-08 Jakub Jelinek <jakub@redhat.com>
3636 liuhongt <hongtao.liu@intel.com>
3639 * config/i386/i386.md (@xorsign<mode>3_1): Remove.
3640 * config/i386/i386-expand.c (ix86_expand_xorsign): Expand right away
3641 into AND with mask and XOR, using paradoxical subregs.
3642 (ix86_split_xorsign): Remove.
3643 * config/i386/i386-protos.h (ix86_split_xorsign): Remove.
3645 2021-09-08 Di Zhao <dizhao@os.amperecomputing.com>
3647 * tree-ssa-sccvn.c (vn_nary_op_insert_into): fix result compare
3649 2021-09-08 Jakub Jelinek <jakub@redhat.com>
3652 * config/i386/i386.md (xorsign<mode>3): If operands[1] is equal to
3653 operands[2], emit abs<mode>2 instead.
3654 (@xorsign<mode>3_1): Add early-clobbers for output operand, enable
3655 first alternative even for avx, add another alternative with
3656 =&Yv <- 0, Yv, Yvm constraints.
3657 * config/i386/i386-expand.c (ix86_split_xorsign): If op0 is equal
3658 to op1, emit vpandn instead.
3660 2021-09-08 liuhongt <hongtao.liu@intel.com>
3662 * config/i386/avx512fp16intrin.h (_mm_set_ph): New intrinsic.
3663 (_mm256_set_ph): Likewise.
3664 (_mm512_set_ph): Likewise.
3665 (_mm_setr_ph): Likewise.
3666 (_mm256_setr_ph): Likewise.
3667 (_mm512_setr_ph): Likewise.
3668 (_mm_set1_ph): Likewise.
3669 (_mm256_set1_ph): Likewise.
3670 (_mm512_set1_ph): Likewise.
3671 (_mm_setzero_ph): Likewise.
3672 (_mm256_setzero_ph): Likewise.
3673 (_mm512_setzero_ph): Likewise.
3674 (_mm_set_sh): Likewise.
3675 (_mm_load_sh): Likewise.
3676 (_mm_store_sh): Likewise.
3677 * config/i386/i386-builtin-types.def (V8HF): New type.
3678 (DEF_FUNCTION_TYPE (V8HF, V8HI)): New builtin function type
3679 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
3680 Support vector HFmodes.
3681 (ix86_expand_vector_init_one_nonzero): Likewise.
3682 (ix86_expand_vector_init_one_var): Likewise.
3683 (ix86_expand_vector_init_interleave): Likewise.
3684 (ix86_expand_vector_init_general): Likewise.
3685 (ix86_expand_vector_set): Likewise.
3686 (ix86_expand_vector_extract): Likewise.
3687 (ix86_expand_vector_init_concat): Likewise.
3688 (ix86_expand_sse_movcc): Handle vector HFmodes.
3689 (ix86_expand_vector_set_var): Ditto.
3690 * config/i386/i386-modes.def: Add HF vector modes in comment.
3691 * config/i386/i386.c (classify_argument): Add HF vector modes.
3692 (ix86_hard_regno_mode_ok): Allow HF vector modes for AVX512FP16.
3693 (ix86_vector_mode_supported_p): Likewise.
3694 (ix86_set_reg_reg_cost): Handle vector HFmode.
3695 (ix86_get_ssemov): Handle vector HFmode.
3696 (function_arg_advance_64): Pass unamed V16HFmode and V32HFmode
3698 (function_arg_advance_32): Pass V8HF/V16HF/V32HF by sse reg for 32bit
3700 (function_arg_advance_32): Ditto.
3701 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): New.
3702 (VALID_AVX256_REG_OR_OI_MODE): Rename to ..
3703 (VALID_AVX256_REG_OR_OI_VHF_MODE): .. this, and add V16HF.
3704 (VALID_SSE2_REG_VHF_MODE): New.
3705 (VALID_AVX512VL_128_REG_MODE): Add V8HF and TImode.
3706 (SSE_REG_MODE_P): Add vector HFmode.
3707 * config/i386/i386.md (mode): Add HF vector modes.
3708 (MODE_SIZE): Likewise.
3709 (ssemodesuffix): Add ph suffix for HF vector modes.
3710 * config/i386/sse.md (VFH_128): New mode iterator.
3711 (VMOVE): Adjust for HF vector modes.
3713 (V_256_512): Likewise.
3715 (avx512fmaskmode): Likewise.
3716 (shuffletype): Likewise.
3717 (sseinsnmode): Likewise.
3718 (ssedoublevecmode): Likewise.
3719 (ssehalfvecmode): Likewise.
3720 (ssehalfvecmodelower): Likewise.
3721 (ssePScmode): Likewise.
3722 (ssescalarmode): Likewise.
3723 (ssescalarmodelower): Likewise.
3724 (sseintprefix): Likewise.
3726 (bcstscalarsuff): Likewise.
3727 (xtg_mode): Likewise.
3728 (VI12HF_AVX512VL): New mode_iterator.
3729 (VF_AVX512FP16): Likewise.
3731 (VIHF_256): Likewise.
3732 (VIHF_AVX512BW): Likewise.
3733 (V16_256): Likewise.
3734 (V32_512): Likewise.
3735 (sseintmodesuffix): New mode_attr.
3736 (sse): Add scalar and vector HFmodes.
3737 (ssescalarmode): Add vector HFmode mapping.
3738 (ssescalarmodesuffix): Add sh suffix for HFmode.
3739 (*<sse>_vm<insn><mode>3): Use VFH_128.
3740 (*<sse>_vm<multdiv_mnemonic><mode>3): Likewise.
3741 (*ieee_<ieee_maxmin><mode>3): Likewise.
3742 (<avx512>_blendm<mode>): New define_insn.
3743 (vec_setv8hf): New define_expand.
3744 (vec_set<mode>_0): New define_insn for HF vector set.
3745 (*avx512fp16_movsh): Likewise.
3746 (avx512fp16_movsh): Likewise.
3747 (vec_extract_lo_v32hi): Rename to ...
3748 (vec_extract_lo_<mode>): ... this, and adjust to allow HF
3750 (vec_extract_hi_v32hi): Likewise.
3751 (vec_extract_hi_<mode>): Likewise.
3752 (vec_extract_lo_v16hi): Likewise.
3753 (vec_extract_lo_<mode>): Likewise.
3754 (vec_extract_hi_v16hi): Likewise.
3755 (vec_extract_hi_<mode>): Likewise.
3756 (vec_set_hi_v16hi): Likewise.
3757 (vec_set_hi_<mode>): Likewise.
3758 (vec_set_lo_v16hi): Likewise.
3759 (vec_set_lo_<mode>): Likewise.
3760 (*vec_extract<mode>_0): New define_insn_and_split for HF
3762 (*vec_extracthf): New define_insn.
3763 (VEC_EXTRACT_MODE): Add HF vector modes.
3764 (PINSR_MODE): Add V8HF.
3765 (sse2p4_1): Likewise.
3766 (pinsr_evex_isa): Likewise.
3767 (<sse2p4_1>_pinsr<ssemodesuffix>): Adjust to support
3768 insert for V8HFmode.
3769 (pbroadcast_evex_isa): Add HF vector modes.
3770 (AVX2_VEC_DUP_MODE): Likewise.
3771 (VEC_INIT_MODE): Likewise.
3772 (VEC_INIT_HALF_MODE): Likewise.
3773 (avx2_pbroadcast<mode>): Adjust to support HF vector mode
3775 (avx2_pbroadcast<mode>_1): Likewise.
3776 (<avx512>_vec_dup<mode>_1): Likewise.
3777 (<avx512>_vec_dup<mode><mask_name>): Likewise.
3778 (<mask_codefor><avx512>_vec_dup_gpr<mode><mask_name>):
3781 2021-09-08 Guo, Xuepeng <xuepeng.guo@intel.com>
3782 H.J. Lu <hongjiu.lu@intel.com>
3783 Liu Hongtao <hongtao.liu@intel.com>
3784 Wang Hongyu <hongyu.wang@intel.com>
3785 Xu Dianhong <dianhong.xu@intel.com>
3787 * common/config/i386/cpuinfo.h (get_available_features):
3788 Detect FEATURE_AVX512FP16.
3789 * common/config/i386/i386-common.c
3790 (OPTION_MASK_ISA_AVX512FP16_SET,
3791 OPTION_MASK_ISA_AVX512FP16_UNSET,
3792 OPTION_MASK_ISA2_AVX512FP16_SET,
3793 OPTION_MASK_ISA2_AVX512FP16_UNSET): New.
3794 (OPTION_MASK_ISA2_AVX512BW_UNSET,
3795 OPTION_MASK_ISA2_AVX512BF16_UNSET): Add AVX512FP16.
3796 (ix86_handle_option): Handle -mavx512fp16.
3797 * common/config/i386/i386-cpuinfo.h (enum processor_features):
3798 Add FEATURE_AVX512FP16.
3799 * common/config/i386/i386-isas.h: Add entry for AVX512FP16.
3800 * config.gcc: Add avx512fp16intrin.h.
3801 * config/i386/avx512fp16intrin.h: New intrinsic header.
3802 * config/i386/cpuid.h: Add bit_AVX512FP16.
3803 * config/i386/i386-builtin-types.def: (FLOAT16): New primitive type.
3804 * config/i386/i386-builtins.c: Support _Float16 type for i386
3806 (ix86_register_float16_builtin_type): New function.
3807 (ix86_float16_type_node): New.
3808 * config/i386/i386-c.c (ix86_target_macros_internal): Define
3810 * config/i386/i386-expand.c (ix86_expand_branch): Support
3812 (ix86_prepare_fp_compare_args): Adjust TARGET_SSE_MATH &&
3813 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
3814 (ix86_expand_fp_movcc): Ditto.
3815 * config/i386/i386-isa.def: Add PTA define for AVX512FP16.
3816 * config/i386/i386-options.c (isa2_opts): Add -mavx512fp16.
3817 (ix86_valid_target_attribute_inner_p): Add avx512fp16 attribute.
3818 * config/i386/i386.c (ix86_get_ssemov): Use
3819 vmovdqu16/vmovw/vmovsh for HFmode/HImode scalar or vector.
3820 (ix86_get_excess_precision): Use
3821 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when TARGET_AVX512FP16
3823 (sse_store_index): Use SFmode cost for HFmode cost.
3824 (inline_memory_move_cost): Add HFmode, and perfer SSE cost over
3825 GPR cost for HFmode.
3826 (ix86_hard_regno_mode_ok): Allow HImode in sse register.
3827 (ix86_mangle_type): Add manlging for _Float16 type.
3828 (inline_secondary_memory_needed): No memory is needed for
3829 16bit movement between gpr and sse reg under
3831 (ix86_multiplication_cost): Adjust TARGET_SSE_MATH &&
3832 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
3833 (ix86_division_cost): Ditto.
3834 (ix86_rtx_costs): Ditto.
3835 (ix86_add_stmt_cost): Ditto.
3836 (ix86_optab_supported_p): Ditto.
3837 * config/i386/i386.h (VALID_AVX512F_SCALAR_MODE): Add HFmode.
3838 (SSE_FLOAT_MODE_SSEMATH_OR_HF_P): Add HFmode.
3839 (PTA_SAPPHIRERAPIDS): Add PTA_AVX512FP16.
3840 * config/i386/i386.md (mode): Add HFmode.
3841 (MODE_SIZE): Add HFmode.
3842 (isa): Add avx512fp16.
3843 (enabled): Handle avx512fp16.
3844 (ssemodesuffix): Add sh suffix for HFmode.
3845 (comm): Add mult, div.
3846 (plusminusmultdiv): New code iterator.
3847 (insn): Add mult, div.
3848 (*movhf_internal): Adjust for avx512fp16 instruction.
3849 (*movhi_internal): Ditto.
3850 (*cmpi<unord>hf): New define_insn for HFmode.
3851 (*ieee_s<ieee_maxmin>hf3): Likewise.
3852 (extendhf<mode>2): Likewise.
3853 (trunc<mode>hf2): Likewise.
3854 (float<floatunssuffix><mode>hf2): Likewise.
3855 (*<insn>hf): Likewise.
3856 (cbranchhf4): New expander.
3857 (movhfcc): Likewise.
3858 (<insn>hf3): Likewise.
3861 * config/i386/i386.opt: Add mavx512fp16.
3862 * config/i386/immintrin.h: Include avx512fp16intrin.h.
3863 * doc/invoke.texi: Add mavx512fp16.
3864 * doc/extend.texi: Add avx512fp16 Usage Notes.
3866 2021-09-08 liuhongt <hongtao.liu@intel.com>
3868 * common.opt: Support -fexcess-precision=16.
3869 * config/aarch64/aarch64.c (aarch64_excess_precision): Return
3870 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when
3871 EXCESS_PRECISION_TYPE_FLOAT16.
3872 * config/arm/arm.c (arm_excess_precision): Ditto.
3873 * config/i386/i386.c (ix86_get_excess_precision): Ditto.
3874 * config/m68k/m68k.c (m68k_excess_precision): Issue an error
3875 when EXCESS_PRECISION_TYPE_FLOAT16.
3876 * config/s390/s390.c (s390_excess_precision): Ditto.
3877 * coretypes.h (enum excess_precision_type): Add
3878 EXCESS_PRECISION_TYPE_FLOAT16.
3879 * doc/tm.texi (TARGET_C_EXCESS_PRECISION): Update documents.
3880 * doc/tm.texi.in (TARGET_C_EXCESS_PRECISION): Ditto.
3881 * doc/extend.texi (Half-Precision): Document
3882 -fexcess-precision=16.
3883 * flag-types.h (enum excess_precision): Add
3884 EXCESS_PRECISION_FLOAT16.
3885 * target.def (excess_precision): Update document.
3886 * tree.c (excess_precision_type): Set excess_precision_type to
3887 EXCESS_PRECISION_FLOAT16 when -fexcess-precision=16.
3889 2021-09-08 liuhongt <hongtao.liu@intel.com>
3891 * doc/extend.texi: (@node Floating Types): Adjust the wording.
3892 (@node Half-Precision): Ditto.
3894 2021-09-07 Takayuki 'January June' Suwa <jjsuwa_sys3175@yahoo.co.jp>
3897 * config/xtensa/xtensa.c (xtensa_emit_move_sequence): Add
3898 'CONST_INT_P (src)' to the condition of the block that tries to
3899 eliminate literal when loading integer contant.
3901 2021-09-07 David Faust <david.faust@oracle.com>
3903 * doc/extend.texi (BPF Type Attributes) New node.
3904 Document new preserve_access_index attribute.
3905 Document new preserve_access_index builtin.
3906 * doc/invoke.texi: Document -mco-re and -mno-co-re options.
3908 2021-09-07 David Faust <david.faust@oracle.com>
3910 * config/bpf/bpf.c: Adjust includes.
3911 (bpf_handle_preserve_access_index_attribute): New function.
3912 (bpf_attribute_table): Use it here.
3913 (bpf_builtins): Add BPF_BUILTIN_PRESERVE_ACCESS_INDEX.
3914 (bpf_option_override): Handle "-mco-re" option.
3915 (bpf_asm_init_sections): New.
3916 (TARGET_ASM_INIT_SECTIONS): Redefine.
3917 (bpf_file_end): New.
3918 (TARGET_ASM_FILE_END): Redefine.
3919 (bpf_init_builtins): Add "__builtin_preserve_access_index".
3920 (bpf_core_compute, bpf_core_get_index): New.
3921 (is_attr_preserve_access): New.
3922 (bpf_expand_builtin): Handle new builtins.
3923 (bpf_core_newdecl, bpf_core_is_maybe_aggregate_access): New.
3924 (bpf_core_walk): New.
3925 (bpf_resolve_overloaded_builtin): New.
3926 (TARGET_RESOLVE_OVERLOADED_BUILTIN): Redefine.
3928 (pass_bpf_core_attr): New RTL pass.
3929 * config/bpf/bpf-passes.def: New file.
3930 * config/bpf/bpf-protos.h (make_pass_bpf_core_attr): New.
3931 * config/bpf/coreout.c: New file.
3932 * config/bpf/coreout.h: Likewise.
3933 * config/bpf/t-bpf (TM_H): Add $(srcdir)/config/bpf/coreout.h.
3934 (coreout.o): New rule.
3935 (PASSES_EXTRA): Add $(srcdir)/config/bpf/bpf-passes.def.
3936 * config.gcc (bpf): Add coreout.h to extra_headers.
3937 Add coreout.o to extra_objs.
3938 Add $(srcdir)/config/bpf/coreout.c to target_gtfiles.
3940 2021-09-07 David Faust <david.faust@oracle.com>
3942 * btfout.c (get_btf_id): Function is no longer static.
3943 * ctfc.h: Expose it here.
3945 2021-09-07 David Faust <david.faust@oracle.com>
3947 * ctfc.c (ctf_lookup_tree_type): New function.
3950 2021-09-07 David Faust <david.faust@oracle.com>
3952 * ctfc.c (ctf_dtd_lookup): Function is no longer static.
3953 * ctfc.h: Analogous change.
3955 2021-09-07 David Faust <david.faust@oracle.com>
3957 * dwarf2out.c (lookup_type_die): Function is no longer static.
3958 * dwarf2out.h: Expose it here.
3960 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
3962 * dwarf2ctf.c (ctf_debug_finalize): Make it static.
3963 (ctf_debug_early_finish): New definition.
3964 (ctf_debug_finish): Likewise.
3965 * dwarf2ctf.h (ctf_debug_finalize): Remove declaration.
3966 (ctf_debug_early_finish): New declaration.
3967 (ctf_debug_finish): Likewise.
3968 * dwarf2out.c (dwarf2out_finish): Invoke ctf_debug_finish.
3969 (dwarf2out_early_finish): Invoke ctf_debug_early_finish.
3971 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
3973 * config/bpf/bpf.c (bpf_option_override): For BPF backend, disable LTO
3974 support when compiling for CO-RE.
3975 * config/bpf/bpf.opt: Add new command line option -mco-re.
3977 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
3979 * flag-types.h (enum debug_info_type): Add new enum
3980 DINFO_TYPE_BTF_WITH_CORE.
3981 (BTF_WITH_CORE_DEBUG): New bitmask.
3982 * flags.h (btf_with_core_debuginfo_p): New declaration.
3983 * opts.c (btf_with_core_debuginfo_p): New definition.
3985 2021-09-07 Jason Merrill <jason@redhat.com>
3987 * tree.h (error_operand_p): Change to inline function.
3989 2021-09-07 Aldy Hernandez <aldyh@redhat.com>
3991 * tree-ssa-threadedge.c (forwarder_block_p): Rename to...
3992 (empty_block_with_phis_p): ...this.
3993 (potentially_threadable_block): Same.
3994 (jump_threader::thread_through_normal_block): Same.
3996 2021-09-07 Eric Botcazou <ebotcazou@adacore.com>
3999 * dwarf2out.c (mark_base_types): New overloaded function.
4000 (dwarf2out_early_finish): Invoke it on the COMDAT type list as well
4001 as the compilation unit, and call move_marked_base_types afterward.
4003 2021-09-07 H.J. Lu <hjl.tools@gmail.com>
4006 * config/i386/i386-expand.c (ix86_expand_convert_uns_sisf_sse):
4008 (ix86_expand_vector_convert_uns_vsivsf): Likewise.
4010 2021-09-07 Richard Biener <rguenther@suse.de>
4012 PR tree-optimization/102226
4013 * tree-vect-loop.c (vect_transform_cycle_phi): Record
4014 the converted value for the epilogue PHI use.
4016 2021-09-07 Martin Liska <mliska@suse.cz>
4018 PR gcov-profile/80223
4019 * ipa-inline.c (can_inline_edge_p): Similarly to sanitizer
4020 options, do not inline when no_profile_instrument_function
4021 attributes are different in early inliner. It's fine to inline
4022 it after PGO instrumentation.
4024 2021-09-07 Richard Biener <rguenther@suse.de>
4026 PR tree-optimization/101555
4027 * tree-ssa-pre.c (translate_vuse_through_block): Do not
4028 perform an alias walk to determine the validity of the
4029 mem at the start of the block which is already guaranteed
4030 by means of prune_clobbered_mems.
4031 (phi_translate_1): Pass edge to translate_vuse_through_block.
4033 2021-09-07 Xionghu Luo <luoxhu@linux.ibm.com>
4036 * config/rs6000/rs6000.md (fmod<mode>3): New define_expand.
4037 (remainder<mode>3): Likewise.
4039 2021-09-07 YunQiang Su <yunqiang.su@cipunited.com>
4041 * config/mips/mips.c (mips_file_start): add .module for
4044 2021-09-06 Roger Sayle <roger@nextmovesoftware.com>
4046 * wide-int.cc (wi::clz): Reorder tests to ensure the result
4047 is zero for all negative values.
4049 2021-09-06 Tobias Burnus <tobias@codesourcery.com>
4051 * doc/invoke.texi (-foffload-options): Fix @opindex.
4053 2021-09-06 H.J. Lu <hjl.tools@gmail.com>
4056 * config/i386/i386-expand.c (ix86_split_xorsign): Use operands[2].
4057 * config/i386/i386.md (@xorsign<mode>3_1): Add non-destructive
4058 source alternative for AVX.
4060 2021-09-06 liuhongt <hongtao.liu@intel.com>
4062 PR middle-end/102182
4063 * optabs.c (expand_fix): Add from1 to avoid from being
4066 2021-09-06 Eric Botcazou <ebotcazou@adacore.com>
4068 * dwarf2out.c (modified_type_die): Deal with all array types earlier
4069 and use local variable consistently throughout the function.
4071 2021-09-06 Jakub Jelinek <jakub@redhat.com>
4073 PR tree-optimization/102207
4074 * match.pd: Don't demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
4075 were promoted from signed to wider unsigned type.
4077 2021-09-06 Andrew Pinski <apinski@marvell.com>
4079 PR tree-optimization/63184
4080 * match.pd: Add simplification of pointer_diff of two pointer_plus
4081 with addr_expr in the first operand of each pointer_plus.
4082 Add simplificatoin of ne/eq of two pointer_plus with addr_expr
4083 in the first operand of each pointer_plus.
4085 2021-09-06 Richard Biener <rguenther@suse.de>
4087 PR tree-optimization/102176
4088 * tree-vect-slp.c (vect_slp_gather_vectorized_scalar_stmts):
4090 (vect_bb_slp_scalar_cost): Use the computed set of
4091 vectorized scalar stmts instead of relying on the out-of-date
4092 and not accurate PURE_SLP_STMT.
4093 (vect_bb_vectorization_profitable_p): Compute the set
4094 of vectorized scalar stmts.
4096 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
4098 * gimple-range-path.cc (path_range_query::range_of_stmt): Remove
4099 GIMPLE_COND special casing.
4100 (path_range_query::range_defined_in_block): Use range_of_stmt
4101 instead of calling fold_range directly.
4103 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
4105 * gimple-range-path.cc (path_range_query::range_of_expr): Set
4106 m_undefined_path when appropriate.
4107 (path_range_query::internal_range_of_expr): Copy from range_of_expr.
4108 (path_range_query::unreachable_path_p): New.
4109 (path_range_query::precompute_ranges): Set m_undefined_path.
4110 * gimple-range-path.h (path_range_query::unreachable_path_p): New.
4111 (path_range_query::internal_range_of_expr): New.
4112 * tree-ssa-threadbackward.c (back_threader::find_taken_edge_cond):
4113 Use unreachable_path_p.
4115 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
4117 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
4118 Remove argument and call find_taken_edge.
4119 (back_threader::resolve_phi): Do not calculate taken edge before
4120 calling maybe_register_path.
4121 (back_threader::find_paths_to_names): Same.
4123 2021-09-05 Jeff Law <jlaw@localhost.localdomain>
4125 * config/h8300/h8300.md (QHSI2 mode iterator): New mode iterator.
4126 * config/h8300/testcompare.md (store_c): Update name, use new
4128 (store_neg_c, store_shifted_c): New patterns.
4130 2021-09-03 Segher Boessenkool <segher@kernel.crashing.org>
4133 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): On ELFv2 use r11
4134 instead of r12 for CR save, in all cases.
4136 2021-09-03 Andrew Pinski <apinski@marvell.com>
4138 * config/aarch64/aarch64-sve-builtins.cc (register_vector_type):
4139 Handle error_mark_node as the type of the type_decl.
4141 2021-09-03 Andrew Pinski <apinski@marvell.com>
4143 * config/aarch64/aarch64-builtins.c (struct aarch64_simd_type_info):
4145 (aarch64_simd_types): Likewise.
4146 (aarch64_simd_intOI_type_node): Likewise.
4147 (aarch64_simd_intCI_type_node): Likewise.
4148 (aarch64_simd_intXI_type_node): Likewise.
4149 * config/aarch64/aarch64.h (aarch64_fp16_type_node): Likewise.
4150 (aarch64_fp16_ptr_type_node): Likewise.
4151 (aarch64_bf16_type_node): Likewise.
4152 (aarch64_bf16_ptr_type_node): Likewise.
4154 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4156 * range-op.cc (operator_minus::op1_op2_relation_effect): Abstract
4158 (minus_op1_op2_relation_effect): ...here.
4159 (class operator_pointer_diff): New.
4160 (operator_pointer_diff::op1_op2_relation_effect): Call
4161 minus_op1_op2_relation_effect.
4162 (integral_table::integral_table): Add entry for POINTER_DIFF_EXPR.
4164 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4166 * tree-ssa-threadbackward.c (back_threader::thread_through_all_blocks):
4167 Add may_peel_loop_headers.
4168 (back_threader_registry::thread_through_all_blocks): Same.
4169 (try_thread_blocks): Pass may_peel_loop_headers argument.
4170 (pass_early_thread_jumps::execute): Same.
4172 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4174 * tree-ssa-threadedge.c (has_phis_p): New.
4175 (forwarder_block_p): New.
4176 (potentially_threadable_block): Call forwarder_block_p.
4177 (jump_threader::thread_around_empty_blocks): Call has_phis_p.
4178 (jump_threader::thread_through_normal_block): Call
4181 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4183 * tree-ssa-threadbackward.c (back_threader::dump): New.
4184 (back_threader::debug): New.
4185 (back_threader_profitability::profitable_path_p): Dump blocks
4186 even if we are bailing early.
4188 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4190 * tree-ssa-threadupdate.c (cancel_thread): New.
4191 (jump_thread_path_registry::thread_block_1): Use cancel_thread.
4192 (jump_thread_path_registry::mark_threaded_blocks): Same.
4193 (jump_thread_path_registry::register_jump_thread): Same.
4195 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4197 * tree-ssa-threadedge.c (jt_state::push): Only call methods for
4198 which objects are available.
4199 (jt_state::pop): Same.
4200 (jt_state::register_equiv): Same.
4201 (jt_state::register_equivs_on_edge): Same.
4203 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4205 * tree-ssa-threadedge.c (jump_threader::thread_across_edge):
4206 Move pop until after a thread is registered.
4208 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4210 * tree-ssa-threadupdate.c (debug): New.
4212 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4214 * gimple-range-trace.cc (push_dump_file::push_dump_file): New.
4215 (push_dump_file::~push_dump_file): New.
4216 (dump_ranger): Change dump_file temporarily while dumping
4218 * gimple-range-trace.h (class push_dump_file): New.
4220 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4222 * gimple-range-trace.cc (debug_seed_ranger): Remove static.
4223 (dump_ranger): Dump function name.
4225 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4227 * gimple-range-path.cc (path_range_query::range_defined_in_block):
4228 Adjust for non-null.
4229 (path_range_query::adjust_for_non_null_uses): New.
4230 (path_range_query::precompute_ranges): Call
4231 adjust_for_non_null_uses.
4232 * gimple-range-path.h: Add m_non_null and
4233 adjust_for_non_null_uses.
4235 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4237 * gimple-range-path.cc (path_range_query::dump): Dump path
4239 (path_range_query::precompute_ranges): Dump entire path.
4241 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4243 * value-relation.cc (relation_oracle::debug): New.
4244 * value-relation.h (relation_oracle::debug): New.
4246 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4248 * tree-ssa-loop-ch.c: Remove unnecessary include file.
4250 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4252 * gimple-range-fold.cc (fold_using_range::postfold_gcond_edges):
4253 Skip statements with no defining BB.
4254 * gimple-range-path.cc (path_range_query::range_defined_in_block):
4255 Do not get confused by statements with no defining BB.
4257 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4259 * gimple-range-fold.cc (adjust_imagpart_expr): Move from
4260 gimple_range_adjustment. Add support for constants.
4261 (adjust_realpart_expr): New.
4262 (gimple_range_adjustment): Move IMAGPART_EXPR code to
4263 adjust_imagpart_expr.
4264 * range-op.cc (integral_table::integral_table): Add entry for
4267 2021-09-03 Jakub Jelinek <jakub@redhat.com>
4269 * omp-expand.c (expand_omp_atomic_pipeline): Use
4270 IFN_ATOMIC_COMPARE_EXCHANGE instead of
4271 BUILT_IN_SYNC_VAL_COMPARE_AND_SWAP_? so that memory order
4274 2021-09-03 Jakub Jelinek <jakub@redhat.com>
4277 * tree.h (DECL_FIELD_ABI_IGNORED): Changed into rvalue only macro
4278 that is false if DECL_BIT_FIELD.
4279 (SET_DECL_FIELD_ABI_IGNORED, DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD,
4280 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD): Define.
4281 * tree-streamer-out.c (pack_ts_decl_common_value_fields): For
4282 DECL_BIT_FIELD stream DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead
4283 of DECL_FIELD_ABI_IGNORED.
4284 * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Use
4285 SET_DECL_FIELD_ABI_IGNORED instead of writing to
4286 DECL_FIELD_ABI_IGNORED and for DECL_BIT_FIELD use
4287 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead.
4288 * lto-streamer-out.c (hash_tree): For DECL_BIT_FIELD hash
4289 DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead of DECL_FIELD_ABI_IGNORED.
4291 2021-09-03 liuhongt <hongtao.liu@intel.com>
4294 * config/i386/amxbf16intrin.h : Remove macro check for __AMX_BF16__.
4295 * config/i386/amxint8intrin.h : Remove macro check for __AMX_INT8__.
4296 * config/i386/amxtileintrin.h : Remove macro check for __AMX_TILE__.
4298 2021-09-02 Martin Sebor <msebor@redhat.com>
4300 PR tree-optimization/17506
4302 * tree-ssa-uninit.c (warn_uninit): Remove conditional guarding note.
4304 2021-09-02 Richard Biener <rguenther@suse.de>
4306 * tree-ssa-loop-im.c (fill_always_executed_in_1): Refine
4307 fix for PR78185 and continue processing when leaving
4310 2021-09-02 Jakub Jelinek <jakub@redhat.com>
4312 PR tree-optimization/99591
4313 * match.pd: Demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
4316 2021-09-02 Richard Biener <rguenther@suse.de>
4319 2021-09-02 Richard Biener <rguenther@suse.de>
4321 PR tree-optimization/102155
4322 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
4323 over a part of the RPO array and do not recurse here.
4324 Dump blocks marked as always executed.
4325 (fill_always_executed_in): Walk over the RPO array and
4326 process loops whose header we run into.
4327 (loop_invariant_motion_in_fun): Compute the first RPO
4328 using rev_post_order_and_mark_dfs_back_seme in iteration
4329 order and pass that to fill_always_executed_in.
4331 2021-09-02 liuhongt <hongtao.liu@intel.com>
4333 * config/i386/i386-modes.def (FLOAT_MODE): Define ieee HFmode.
4334 * config/i386/i386.c (enum x86_64_reg_class): Add
4336 (merge_classes): Handle X86_64_SSEHF_CLASS.
4337 (examine_argument): Ditto.
4338 (construct_container): Ditto.
4339 (classify_argument): Ditto, and set HFmode/HCmode to
4341 (function_value_32): Return _FLoat16/Complex Float16 by
4343 (function_value_64): Return _Float16/Complex Float16 by SSE
4345 (ix86_print_operand): Handle CONST_DOUBLE HFmode.
4346 (ix86_secondary_reload): Require gpr as intermediate register
4347 to store _Float16 from sse register when sse4 is not
4349 (ix86_libgcc_floating_mode_supported_p): Enable _FLoat16 under
4351 (ix86_scalar_mode_supported_p): Ditto.
4352 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Defined.
4353 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add HFmode.
4354 (VALID_INT_MODE_P): Add HFmode and HCmode.
4355 * config/i386/i386.md (*pushhf_rex64): New define_insn.
4357 (*movhf_internal): Ditto.
4358 * doc/extend.texi (Half-Precision Floating Point): Documemt
4361 2021-09-02 Richard Biener <rguenther@suse.de>
4363 PR tree-optimization/102155
4364 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
4365 over a part of the RPO array and do not recurse here.
4366 Dump blocks marked as always executed.
4367 (fill_always_executed_in): Walk over the RPO array and
4368 process loops whose header we run into.
4369 (loop_invariant_motion_in_fun): Compute the first RPO
4370 using rev_post_order_and_mark_dfs_back_seme in iteration
4371 order and pass that to fill_always_executed_in.
4373 2021-09-02 YunQiang Su <syq@debian.org>
4376 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
4378 * config/mips/mips.c (mips_module_isa_name): New.
4379 mips_file_start: add .module mipsREV to all asm output
4381 2021-09-01 Jeff Law <jlaw@localhost.localdomain>
4383 PR tree-optimization/102152
4384 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Reduce a vector
4385 comparison to a scalar comparison before calling
4386 update_stmt_if_modified.
4388 2021-09-01 Andrew Pinski <apinski@marvell.com>
4391 * config/aarch64/aarch64.c (aarch64_expand_setmem):
4392 Check STRICT_ALIGNMENT before creating an overlapping
4395 2021-09-01 Martin Sebor <msebor@redhat.com>
4397 * gimple-ssa-warn-access.cc (get_size_range): Add argument.
4398 (check_access): Pass additional argument.
4399 (check_memop_access): Remove template and make a member function.
4400 (maybe_check_dealloc_call): Make a pass_waccess member function.
4401 (class pass_waccess): Add, rename, and remove members.
4402 (pass_waccess::pass_waccess): Adjust to name change.
4403 (pass_waccess::~pass_waccess): Same.
4404 (check_alloca): Make a member function.
4405 (check_alloc_size_call): Same.
4406 (check_strcat): Same.
4407 (check_strncat): Same.
4408 (check_stxcpy): Same.
4409 (check_stxncpy): Same.
4410 (check_strncmp): Same.
4411 (maybe_warn_rdwr_sizes): Rename...
4412 (pass_waccess::maybe_check_access_sizes): ...to this.
4413 (pass_waccess::check_call): Adjust to name changes.
4414 (pass_waccess::maybe_check_dealloc_call): Make a pass_waccess member
4416 (pass_waccess::execute): Adjust to name changes.
4417 * gimple-ssa-warn-access.h (check_memop_access): Remove.
4418 * pointer-query.cc (access_ref::phi): Handle null pointer.
4419 (access_ref::inform_access): Same.
4420 (pointer_query::put_ref): Modify a cached value, not a copy of it.
4421 (pointer_query::dump): New function.
4422 (compute_objsize_r): Avoid overwriting access_ref::bndrng. Cache
4424 * pointer-query.h (pointer_query::dump): Declare.
4425 * tree-ssa-strlen.c (get_range): Simplify. Use function query.
4426 (dump_strlen_info): Use function query.
4427 (printf_strlen_execute): Factor code out into pointer_query::put_ref.
4429 2021-09-01 Thomas Schwinge <thomas@codesourcery.com>
4431 * tree.c (walk_tree_1) <OMP_CLAUSE>: Simplify.
4433 2021-09-01 Iain Sandoe <iain@sandoe.co.uk>
4435 * doc/extend.texi: Document unavailable attribute.
4436 * print-tree.c (print_node): Handle unavailable attribute.
4437 * tree-core.h (struct tree_base): Add a bit to carry unavailability.
4438 * tree.c (error_unavailable_use): New.
4439 * tree.h (TREE_UNAVAILABLE): New.
4440 (error_unavailable_use): New.
4442 2021-09-01 Jakub Jelinek <jakub@redhat.com>
4444 PR tree-optimization/102124
4445 * tree-vect-patterns.c (vect_recog_widen_op_pattern): For ORIG_CODE
4446 MINUS_EXPR, if itype is unsigned with smaller precision than type,
4447 add an extra cast to signed variant of itype to ensure sign-extension.
4449 2021-09-01 Martin Liska <mliska@suse.cz>
4451 * graph.c (draw_cfg_node_succ_edges): Do not color fallthru
4452 edges and rather use colors for TRUE and FALSE edges.
4454 2021-09-01 Richard Biener <rguenther@suse.de>
4456 PR tree-optimization/93491
4457 * tree-ssa-pre.c (compute_avail): Set BB_MAY_NOTRETURN
4458 after processing the stmt itself. Do not consider
4459 pure functions possibly not returning. Properly avoid
4460 adding possibly trapping calls to EXP_GEN when there's
4461 a preceeding possibly not returning call.
4462 * tree-ssa-sccvn.c (vn_reference_may_trap): Conservatively
4465 2021-09-01 Richard Biener <rguenther@suse.de>
4467 PR tree-optimization/102139
4468 * tree-vectorizer.h (vec_base_alignments): Adjust hash-map
4469 type to record a std::pair of the stmt-info and the innermost
4471 (dr_vec_info::group): New member.
4472 * tree-vect-data-refs.c (vect_record_base_alignment): Adjust.
4473 (vect_compute_data_ref_alignment): Verify the recorded
4474 base alignment can be used.
4475 (data_ref_pair): Remove.
4476 (dr_group_sort_cmp): Adjust.
4477 (vect_analyze_data_ref_accesses): Store the group-ID in the
4478 dr_vec_info and operate on a vector of dr_vec_infos.
4480 2021-09-01 YunQiang Su <yunqiang.su@cipunited.com>
4482 * read-md.c (md_reader::handle_enum): support value assignation.
4483 * doc/md.texi: record define_c_enum value assignation support.
4485 2021-09-01 Jakub Jelinek <jakub@redhat.com>
4487 PR tree-optimization/102141
4488 * gimple-ssa-store-merging.c (bswap_view_convert): Add BEFORE
4489 argument. If false, emit stmts after gsi instead of before, and
4491 (bswap_replace): Adjust callers. When converting output of bswap,
4492 emit VIEW_CONVERT prepratation stmts after a copy of gsi instead
4495 2021-09-01 liuhongt <hongtao.liu@intel.com>
4497 * emit-rtl.c (validate_subreg): Get rid of all float-int
4500 2021-09-01 liuhongt <hongtao.liu@intel.com>
4503 2021-08-30 liuhongt <hongtao.liu@intel.com>
4505 * expmed.c (extract_bit_field_1): Make sure we're playing with
4506 integral modes before call extract_integral_bit_field.
4507 (extract_integral_bit_field): Add a parameter of type
4508 scalar_int_mode which corresponds to of tmode.
4509 And call extract_and_convert_fixed_bit_field instead of
4510 extract_fixed_bit_field and convert_extracted_bit_field.
4511 (extract_and_convert_fixed_bit_field): New function, it's a
4512 combination of extract_fixed_bit_field and
4513 convert_extracted_bit_field.
4515 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
4517 * tree.c (walk_tree_1) <OMP_CLAUSE_TILE>: Handle three operands.
4519 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
4521 * omp-general.h (omp_is_reference): Rename to...
4522 (omp_privatize_by_reference): ... this. Adjust all users...
4523 * omp-general.c: ... here, ...
4524 * gimplify.c: ... here, ...
4525 * omp-expand.c: ... here, ...
4526 * omp-low.c: ... here.
4528 2021-08-31 Martin Sebor <msebor@redhat.com>
4530 * gimple-ssa-warn-access.cc (maybe_warn_alloc_args_overflow): Test
4531 pointer element for equality to zero, not that of the cotaining
4534 2021-08-31 Martin Sebor <msebor@redhat.com>
4536 * gcc-rich-location.h (gcc_rich_location): Make ctor explicit.
4538 2021-08-31 Martin Sebor <msebor@redhat.com>
4540 * function.h (function): Add comments.
4541 (get_range_query): Same. Add attribute returns nonnull.
4543 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
4545 * expr.c (convert_modes): Don't use subreg_promoted_mode on a
4546 SUBREG if it can't be guaranteed to a SUBREG_PROMOTED_VAR_P set.
4547 Instead use the standard (safer) is_a <scalar_int_mode> idiom.
4549 2021-08-31 Jeff Law <jlaw@localhost.localdomain>
4551 * config.gcc (cris-*-elf, cris-*-none): Remove dbxelf.h from
4553 (m32r-*-elf, m32rle-*-elf, m32r-*-linux): Likewise.
4554 (mn10300-*-*, am33_2.0-*-linux*): Likewise.
4555 (xtensa*-*-elf, xtensa*-*-linux, xtensa*-*-uclinux): Likewise.
4556 (m32c-*-elf*, m32c-*-rtems*): Likewise.
4557 * config/cris/cris.h (DBX_NO_XREFS): Remove.
4558 (DBX_CONTIN_LENGTH, DBX_CONTIN_CHAR): Likewise.
4559 * config/m32r/m32r.h (DBXOUT_SOURCE_LINE): Likewise.
4560 (DBX_DEBUGGING_INFO, DBX_CONTIN_LENGTH): Likewise.
4561 * config/mn10300/mn10300.h (DEFAULT_GDB_EXTENSIONS): Likewise.
4562 * config/mn10300/linux.h (DBX_REGISTER_NAMES): Likewise.
4564 2021-08-31 Marcel Vollweiler <marcel@codesourcery.com>
4566 * gimplify.c (gimplify_scan_omp_clauses): Error handling. 'ancestor' only
4567 allowed on target constructs and only with particular other clauses.
4568 * omp-expand.c (expand_omp_target): Output of 'sorry, not supported' if
4570 * omp-low.c (check_omp_nesting_restrictions): Error handling. No nested OpenMP
4571 structs when 'ancestor' is used.
4572 (scan_omp_1_stmt): No usage of OpenMP runtime routines in a target region when
4574 * tree-pretty-print.c (dump_omp_clause): Append 'ancestor'.
4575 * tree.h (OMP_CLAUSE_DEVICE_ANCESTOR): Define macro.
4577 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
4579 * expr.c (convert_modes): Preserve SUBREG_PROMOTED_VAR_P when
4580 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
4582 * simplify-rtx.c (simplify_unary_operation_1) [SIGN_EXTEND]:
4583 Likewise, preserve SUBREG_PROMOTED_VAR_P when creating a (wider)
4584 partial subreg from a SUBREG_PROMOTED_VAR_P subreg. Generate
4585 SIGN_EXTEND of the SUBREG_REG when a subreg would be paradoxical.
4586 [ZERO_EXTEND]: Likewise, preserve SUBREG_PROMOTED_VAR_P when
4587 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
4588 subreg. Generate ZERO_EXTEND of the SUBREG_REG when a subreg
4589 would be paradoxical.
4591 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
4593 * combine.c (combine_simplify_rtx): Avoid converting an explicit
4594 TRUNCATE into a lowpart SUBREG on !TRULY_NOOP_TRUNCATION targets.
4595 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
4597 2021-08-31 Richard Biener <rguenther@suse.de>
4599 PR tree-optimization/102142
4600 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Fix
4601 condition under which to unset the visited flag.
4603 2021-08-31 Richard Biener <rguenther@suse.de>
4605 PR middle-end/102129
4606 * tree-ssa-ter.c (find_replaceable_in_bb): Do not move
4607 possibly trapping expressions across calls.
4609 2021-08-31 Jakub Jelinek <jakub@redhat.com>
4611 PR tree-optimization/102134
4612 * tree-ssa-ccp.c (bit_value_binop) <case RSHIFT_EXPR>: If sgn is
4613 UNSIGNED and r1val | r1mask has MSB set, ensure lzcount doesn't
4616 2021-08-31 Andrew Pinski <apinski@marvell.com>
4619 * collect-utils.c (setup_signals): New declaration.
4620 * collect-utils.h (setup_signals): New function.
4621 * collect2.c (handler): Delete.
4622 (main): Instead of manually setting up the signals,
4623 just call setup_signals.
4624 * lto-wrapper.c (main): Likewise.
4626 2021-08-31 Andrew Pinski <apinski@marvell.com>
4629 * config/i386/i386-protos.h (x86_output_aligned_bss):
4630 Change align argument to unsigned type.
4631 (x86_elf_aligned_decl_common): Likewise.
4632 * config/i386/i386.c (x86_elf_aligned_decl_common): Likewise.
4633 (x86_output_aligned_bss): Likewise.
4635 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
4637 * config/mips/mips.c (mips_module_isa_name): New.
4638 mips_file_start: add .module mipsREV to all asm output
4640 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
4642 * config/mips/mips.h (struct mips_cpu_info): define enum mips_isa;
4643 use enum instead of int for 'isa' member.
4644 * config.gcc, config/mips/mips.c, config/mips/mips-cpus.def,
4645 config/mips/netbsd.h: replace hardcoded numbers with enum.
4647 2021-08-31 liuhongt <hongtao.liu@intel.com>
4649 * config/i386/sse.md (*<avx512>_ucmp<mode>3_1): Change from
4650 define_split to define_insn_and_split.
4651 (*avx2_eq<mode>3): Removed.
4652 (<avx512>_eq<mode>3<mask_scalar_merge_name>): Adjust pattern
4653 (<avx512>_eq<mode>3<mask_scalar_merge_name>_1): Rename to ..
4654 (*<avx512>_eq<mode>3<mask_scalar_merge_name>_1): .. this, and
4656 (*avx2_gt<mode>3): Removed.
4657 (<avx512>_gt<mode>3<mask_scalar_merge_name>): Change from
4658 define_insn to define_expand, and adjust pattern.
4659 (UNSPEC_MASKED_EQ, UNSPEC_MASKED_GT): Removed.
4661 2021-08-30 David Malcolm <dmalcolm@redhat.com>
4664 * Makefile.in (ANALYZER_OBJS): Add analyzer/call-info.o.
4666 2021-08-30 Jason Merrill <jason@redhat.com>
4668 * doc/invoke.texi: Document -Wmissing-requires.
4670 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
4672 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Remove
4673 TARGET_EXTRA_BUILTINS guard.
4675 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
4677 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Change
4678 initialization of V2DI_type_node and unsigned_V2DI_type_node.
4680 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
4682 * config/rs6000/darwin.h (SUBTARGET_INIT_BUILTINS): Use the new
4683 decl when new_builtins_are_live.
4684 * config/rs6000/rs6000-builtin-new.def (__builtin_cfstring): New
4687 2021-08-30 Pat Haugen <pthaugen@linux.ibm.com>
4689 * config/rs6000/rs6000-cpus.def (ISA_3_1_MASKS_SERVER): Add
4690 OPTION_MASK_P10_FUSION_2STORE.
4691 (POWERPC_MASKS): Likewise.
4692 * config/rs6000/rs6000.c (rs6000_option_override_internal): Enable
4693 store fusion for Power10.
4694 (is_fusable_store): New.
4695 (power10_sched_reorder): Likewise.
4696 (rs6000_sched_reorder): Do Power10 specific reordering.
4697 (rs6000_sched_reorder2): Likewise.
4698 * config/rs6000/rs6000.opt: Add new option.
4700 2021-08-30 Richard Biener <rguenther@suse.de>
4702 PR tree-optimization/102128
4703 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
4704 Move scanning for if-converted scalar code to the caller
4705 and instead delay clearing the visited flag for profitable
4707 (vect_slp_region): Cost all subgraphs before scheduling.
4708 For if-converted BB vectorization scan for scalar COND_EXPRs
4709 and do not vectorize if any found and the cost model is
4712 2021-08-30 Richard Biener <rguenther@suse.de>
4714 * common.opt (fexceptions): Mark
4715 EnabledBy(fnon-call-exceptions).
4716 * doc/invoke.texi (fnon-call-exceptions): Document this
4717 enables -fexceptions.
4719 2021-08-30 Sebastian Huber <sebastian.huber@embedded-brains.de>
4721 * tsystem.h (abort): Define abort() if inhibit_libc is defined and it
4722 is not already defined.
4724 2021-08-30 liuhongt <hongtao.liu@intel.com>
4726 * expmed.c (extract_bit_field_1): Make sure we're playing with
4727 integral modes before call extract_integral_bit_field.
4728 (extract_integral_bit_field): Add a parameter of type
4729 scalar_int_mode which corresponds to of tmode.
4730 And call extract_and_convert_fixed_bit_field instead of
4731 extract_fixed_bit_field and convert_extracted_bit_field.
4732 (extract_and_convert_fixed_bit_field): New function, it's a
4733 combination of extract_fixed_bit_field and
4734 convert_extracted_bit_field.
4736 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
4738 * config/darwin.c (darwin_libc_has_function): Do not run
4739 the checks for x86 or modern Darwin. Make sure that there
4740 is a value set for darwin_macosx_version_min before testing.
4742 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
4744 * config/i386/darwin.h (CLEAR_INSN_CACHE): New.
4746 2021-08-28 Jan Hubicka <hubicka@ucw.cz>
4748 * ipa-modref-tree.h (modref_access_node::merge): Break out
4749 logic combining offsets and logic merging ranges to ...
4750 (modref_access_node::combined_offsets): ... here
4751 (modref_access_node::update2): ... here
4752 (modref_access_node::closer_pair_p): New member function.
4753 (modref_access_node::forced_merge): New member function.
4754 (modre_ref_node::insert): Do merging when table is full.
4756 2021-08-28 YunQiang Su <yunqiang.su@cipunited.com>
4759 * config.gcc: MIPS: use N64 ABI by default if the triple end
4760 with -gnuabi64, which is used by Debian since 2013.
4762 2021-08-28 Alexandre Oliva <oliva@adacore.com>
4764 * ipa-modref.c (analyze_function): Skip debug stmts.
4765 * tree-inline.c (estimate_num_insn): Consider builtins even
4766 without a cgraph_node.
4768 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
4770 * config/h8300/bitfield.md (cstore<mode>4): Remove expander.
4771 * config/h8300/h8300.c (h8300_expand_branch): Remove function.
4772 * config/h8300/h8300-protos.h (h8300_expadn_branch): Remove prototype.
4773 * config/h8300/h8300.md (eqne): New code iterator.
4774 (geultu, geultu_to_c): Similarly.
4775 * config/h8300/testcompare.md (cstore<mode>4): Dummy expander.
4776 (store_c_<mode>, store_c_i_<mode>): New define_insn_and_splits
4777 (cmp<mode>_c): New pattern
4779 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
4781 * tree-ssa-dom.c (reduce_vector_comparison_to_scalar_comparison): New
4783 (dom_opt_dom_walker::optimize_stmt): Use it.
4785 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
4787 * config/darwin.c (finalize_ctors): Add a section-start linker-
4789 (finalize_dtors): Likewise.
4790 * config/darwin.h (MIN_LD64_INIT_TERM_START_LABELS): New.
4792 2021-08-27 Bill Schmidt <wschmidt@linux.ibm.com>
4794 * config/rs6000/rs6000-call.c (rs6000-builtins.h): New #include.
4795 (rs6000_init_builtins): Call rs6000_init_generated_builtins. Skip the
4796 old initialization logic when new builtins are enabled.
4797 * config/rs6000/rs6000-gen-builtins.c (write_decls): Rename
4798 rs6000_autoinit_builtins to rs6000_init_generated_builtins.
4799 (write_init_file): Likewise.
4801 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
4803 * configure.ac (darwin2[[0-9]]* | darwin19*): Alter use of
4804 gcc_GAS_CHECK_FEATURE to remove an extraneous parameter.
4805 (amdgcn-* | gcn-*) Likewise.
4807 2021-08-27 Anthony Sharp <anthonysharp15@gmail.com>
4809 * symbol-summary.h: Added missing template keyword.
4811 2021-08-27 Richard Biener <rguenther@suse.de>
4813 PR tree-optimization/45178
4814 * tree-ssa-dce.c (find_obviously_necessary_stmts): For
4815 infinite loops without exit do not mark control dependent
4816 edges of the latch necessary.
4818 2021-08-27 konglin1 <lingling.kong@intel.com>
4821 * config/i386/sse.md: (<avx512>scattersi<mode>): Add mask operand to
4823 (<avx512>scattersi<mode>): Likewise.
4824 (*avx512f_scattersi<VI48F:mode>): Merge mask operand to set_dest.
4825 (*avx512f_scatterdi<VI48F:mode>): Likewise
4827 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
4829 * config/rs6000/rs6000.c (rs6000_builtin_md_vectorized_function): Add
4830 support for built-in functions MISC_BUILTIN_DIVWE, MISC_BUILTIN_DIVWEU,
4831 MISC_BUILTIN_DIVDE, MISC_BUILTIN_DIVDEU, P10_BUILTIN_CFUGED,
4832 P10_BUILTIN_CNTLZDM, P10_BUILTIN_CNTTZDM, P10_BUILTIN_PDEPD and
4833 P10_BUILTIN_PEXTD on Power10.
4835 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
4837 * config/rs6000/rs6000-call.c (builtin_function_type): Add unsigned
4838 signedness for some Power10 bifs.
4840 2021-08-27 David Edelsohn <dje.gcc@gmail.com>
4843 * config/rs6000/rs6000.c (rs6000_adjust_field_align): Use
4844 computed alignment if the entire struct has attribute packed.
4846 2021-08-27 liuhongt <hongtao.liu@intel.com>
4850 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
4851 IX86_BUILTIN_SHUFPD512, IX86_BUILTIN_SHUFPS512,
4852 IX86_BUILTIN_SHUFPD256, IX86_BUILTIN_SHUFPS,
4853 IX86_BUILTIN_SHUFPS256.
4854 (ix86_masked_all_ones): New function.
4856 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
4858 * config/i386/i386.md (*btr<mode>_1): Call force_reg unconditionally.
4859 (conditional moves with memory inputs splitters): Ditto.
4860 * config/i386/sse.md (one_cmpl<mode>2): Simplify.
4862 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
4864 * ipa-modref-tree.h (modref_access_node::try_merge_with): Restart
4865 search after merging.
4867 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
4869 * config/rs6000/rs6000-overload.def: Add remaining overloads.
4871 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
4873 * config/rs6000/rs6000-builtin-new.def: Add cell stanza.
4875 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
4877 * config/rs6000/rs6000-builtin-new.def: Add ieee128-hw, dfp,
4878 crypto, and htm stanzas.
4880 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
4882 * config/rs6000/rs6000-builtin-new.def: Add mma stanza.
4884 2021-08-26 Martin Sebor <msebor@redhat.com>
4886 * tree-ssa-uninit.c (warn_uninit): Refactor and simplify.
4887 (warn_uninit_phi_uses): Remove argument from calls to warn_uninit.
4888 (warn_uninitialized_vars): Same. Reduce visibility of locals.
4889 (warn_uninitialized_phi): Same.
4891 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
4893 * tree-ssa-ccp.c (get_individual_bits): Helper function to
4894 extract the individual bits from a widest_int constant (mask).
4895 (gray_code_bit_flips): New read-only table for effiently
4896 enumerating permutations/combinations of bits.
4897 (bit_value_binop) [LROTATE_EXPR, RROTATE_EXPR]: Handle rotates
4898 by unknown counts that are guaranteed less than the target
4899 precision and four or fewer unknown bits by enumeration.
4900 [LSHIFT_EXPR, RSHIFT_EXPR]: Likewise, also handle shifts by
4901 enumeration under the same conditions. Handle remaining
4902 shifts as a mask based upon the minimum possible shift value.
4904 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
4905 Richard Biener <rguenther@suse.de>
4907 * match.pd (shift transformations): Remove a redundant
4908 !POINTER_TYPE_P check.
4910 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
4913 * config/i386/i386.md (cmove reg-reg move elimination peephole2s):
4914 Set all_regs to true in the call to replace_rtx.
4916 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
4918 * ipa-modref-tree.c (test_insert_search_collapse): Update test.
4919 * ipa-modref-tree.h (modref_base_node::insert): Be smarter when
4920 hiting --param modref-max-refs limit.
4921 (modref_tree:insert_base): Be smarter when hitting
4922 --param modref-max-bases limit. Add new parameter REF.
4923 (modref_tree:insert): Update.
4924 (modref_tree:merge): Update.
4925 * ipa-modref.c (read_modref_records): Update.
4927 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
4929 * params.opt: (modref-max-adjustments): Add full stop.
4931 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
4933 * ipa-modref-tree.h (modref_ref_node::verify): New member
4935 (modref_ref_node::insert): Use it.
4936 (modref_ref_node::try_mere_with): Fix off by one error.
4938 2021-08-26 Martin Liska <mliska@suse.cz>
4939 Stefan Kneifel <stefan.kneifel@bluewin.ch>
4941 * cgraph.h (create_version_clone_with_body): Add new parameter.
4942 * cgraphclones.c: Likewise.
4943 * multiple_target.c (create_dispatcher_calls): Do not use
4945 (create_target_clone): Likewise here.
4947 2021-08-26 Jonathan Yong <10walls@gmail.com>
4949 * doc/extend.texi: Add note about reserved priorities
4950 to the constructor attribute.
4952 2021-08-25 Martin Sebor <msebor@redhat.com>
4954 * gimple-range-cache.cc (ssa_global_cache::dump): Avoid printing
4955 range table header alone.
4956 * gimple-range.cc (gimple_ranger::export_global_ranges): Same.
4958 2021-08-25 Jan Hubicka <hubicka@ucw.cz>
4960 * doc/invoke.texi: Document --param modref-max-adjustments.
4961 * ipa-modref-tree.c (test_insert_search_collapse): Update.
4962 (test_merge): Update.
4963 * ipa-modref-tree.h (struct modref_access_node): Add adjustments;
4964 (modref_access_node::operator==): Fix handling of access ranges.
4965 (modref_access_node::contains): Constify parameter; handle also
4966 mismatched parm offsets.
4967 (modref_access_node::update): New function.
4968 (modref_access_node::merge): New function.
4969 (unspecified_modref_access_node): Update constructor.
4970 (modref_ref_node::insert_access): Add record_adjustments parameter;
4972 (modref_ref_node::try_merge_with): New private function.
4973 (modref_tree::insert): New record_adjustments parameter.
4974 (modref_tree::merge): New record_adjustments parameter.
4975 (modref_tree::copy_from): Update.
4976 * ipa-modref.c (dump_access): Dump adjustments field.
4977 (get_access): Update constructor.
4978 (record_access): Update call of insert.
4979 (record_access_lto): Update call of insert.
4980 (merge_call_side_effects): Add record_adjustments parameter.
4981 (get_access_for_fnspec): Update.
4982 (process_fnspec): Update.
4983 (analyze_call): Update.
4984 (analyze_function): Update.
4985 (read_modref_records): Update.
4986 (ipa_merge_modref_summary_after_inlining): Update.
4987 (propagate_unknown_call): Update.
4988 (modref_propagate_in_scc): Update.
4989 * params.opt (param-max-modref-adjustments=): New.
4991 2021-08-25 Michael Meissner <meissner@linux.ibm.com>
4993 * config/rs6000/vsx.md (UNSPEC_XXSPLTIDP): Rename from
4995 (xxspltiw_v4si): Use vecperm type attribute.
4996 (xxspltiw_v4si_inst): Use vecperm type attribute.
4997 (xxspltiw_v4sf_inst): Likewise.
4998 (xxspltidp_v2df): Use vecperm type attribute. Use
4999 UNSPEC_XXSPLTIDP instead of UNSPEC_XXSPLTID.
5000 (xxspltidp_v2df_inst): Likewise.
5001 (xxsplti32dx_v4si): Use vecperm type attribute.
5002 (xxsplti32dx_v4si_inst): Likewise.
5003 (xxsplti32dx_v4sf_inst): Likewise.
5004 (xxblend_<mode>): Likewise.
5005 (xxpermx): Likewise.
5006 (xxpermx_inst): Likewise.
5009 2021-08-25 Lewis Hyatt <lhyatt@gmail.com>
5012 * coretypes.h (typedef diagnostic_input_charset_callback): Declare.
5013 * diagnostic.c (diagnostic_initialize_input_context): New function.
5014 * diagnostic.h (diagnostic_initialize_input_context): Declare.
5015 * input.c (default_charset_callback): New function.
5016 (file_cache::initialize_input_context): New function.
5017 (file_cache_slot::create): Added ability to convert the input
5018 according to the input context.
5019 (file_cache::file_cache): Initialize the new input context.
5020 (class file_cache_slot): Added new m_alloc_offset member.
5021 (file_cache_slot::file_cache_slot): Initialize the new member.
5022 (file_cache_slot::~file_cache_slot): Handle potentially offset buffer.
5023 (file_cache_slot::maybe_grow): Likewise.
5024 (file_cache_slot::needs_read_p): Handle NULL fp, which is now possible.
5025 (file_cache_slot::get_next_line): Likewise.
5026 * input.h (class file_cache): Added input context member.
5028 2021-08-25 Richard Biener <rguenther@suse.de>
5030 PR tree-optimization/102046
5031 * tree-vect-slp.c (vect_build_slp_tree_2): Conservatively
5032 update ->any_pattern when swapping operands.
5034 2021-08-25 Hongyu Wang <hongyu.wang@intel.com>
5037 * config/i386/i386.c (ix86_live_on_entry): Adjust comment.
5038 (ix86_decompose_address): Remove retval check for ASHIFT,
5039 allow non-canonical zero extend if AND mask covers ASHIFT
5041 (ix86_legitimate_address_p): Adjust condition for decompose.
5042 (ix86_rtx_costs): Adjust cost for lea with non-canonical
5044 Co-Authored by: Uros Bizjak <ubizjak@gmail.com>
5046 2021-08-25 Jiufu Guo <guojiufu@linux.ibm.com>
5048 PR tree-optimization/101145
5049 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
5051 (number_of_iterations_lt): Invoke above function.
5052 (adjust_cond_for_loop_until_wrap):
5053 Merge to number_of_iterations_until_wrap.
5054 (number_of_iterations_cond): Update invokes for
5055 adjust_cond_for_loop_until_wrap and number_of_iterations_lt.
5057 2021-08-25 konglin1 <lingling.kong@intel.com>
5060 * config/i386/avx512dqintrin.h (_mm512_fpclass_ps_mask): Fix
5062 (_mm512_mask_fpclass_ps_mask): Ditto.
5064 2021-08-25 Kewen Lin <linkw@linux.ibm.com>
5066 * config/rs6000/altivec.md (vec_unpacku_hi_v16qi): Remove.
5067 (vec_unpacku_hi_v8hi): Likewise.
5068 (vec_unpacku_lo_v16qi): Likewise.
5069 (vec_unpacku_lo_v8hi): Likewise.
5070 (vec_unpacku_hi_<VP_small_lc>): New define_expand.
5071 (vec_unpacku_lo_<VP_small_lc>): Likewise.
5073 2021-08-24 David Edelsohn <dje.gcc@gmail.com>
5075 * config/rs6000/aix.h (SYSTEM_IMPLICIT_EXTERN_C): Delete.
5076 * config/rs6000/aix71.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
5077 * config/rs6000/aix72.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
5078 * config/rs6000/aix73.h (TARGET_AIX_VERSION): Increase to 73.
5080 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
5082 PR middle-end/102031
5083 * simplify-rtx.c (simplify_truncation): When comparing precisions
5084 use "subreg_prec" variable, not "subreg_mode".
5086 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
5088 * config/rs6000/rs6000-builtin-new.def: Add power10 and power10-64
5091 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
5093 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Initialize
5094 various pointer type nodes.
5095 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add enum
5096 values for various pointer types.
5097 (ptr_V16QI_type_node): New macro.
5098 (ptr_V1TI_type_node): New macro.
5099 (ptr_V2DI_type_node): New macro.
5100 (ptr_V2DF_type_node): New macro.
5101 (ptr_V4SI_type_node): New macro.
5102 (ptr_V4SF_type_node): New macro.
5103 (ptr_V8HI_type_node): New macro.
5104 (ptr_unsigned_V16QI_type_node): New macro.
5105 (ptr_unsigned_V1TI_type_node): New macro.
5106 (ptr_unsigned_V8HI_type_node): New macro.
5107 (ptr_unsigned_V4SI_type_node): New macro.
5108 (ptr_unsigned_V2DI_type_node): New macro.
5109 (ptr_bool_V16QI_type_node): New macro.
5110 (ptr_bool_V8HI_type_node): New macro.
5111 (ptr_bool_V4SI_type_node): New macro.
5112 (ptr_bool_V2DI_type_node): New macro.
5113 (ptr_bool_V1TI_type_node): New macro.
5114 (ptr_pixel_type_node): New macro.
5115 (ptr_intQI_type_node): New macro.
5116 (ptr_uintQI_type_node): New macro.
5117 (ptr_intHI_type_node): New macro.
5118 (ptr_uintHI_type_node): New macro.
5119 (ptr_intSI_type_node): New macro.
5120 (ptr_uintSI_type_node): New macro.
5121 (ptr_intDI_type_node): New macro.
5122 (ptr_uintDI_type_node): New macro.
5123 (ptr_intTI_type_node): New macro.
5124 (ptr_uintTI_type_node): New macro.
5125 (ptr_long_integer_type_node): New macro.
5126 (ptr_long_unsigned_type_node): New macro.
5127 (ptr_float_type_node): New macro.
5128 (ptr_double_type_node): New macro.
5129 (ptr_long_double_type_node): New macro.
5130 (ptr_dfloat64_type_node): New macro.
5131 (ptr_dfloat128_type_node): New macro.
5132 (ptr_ieee128_type_node): New macro.
5133 (ptr_ibm128_type_node): New macro.
5134 (ptr_vector_pair_type_node): New macro.
5135 (ptr_vector_quad_type_node): New macro.
5136 (ptr_long_long_integer_type_node): New macro.
5137 (ptr_long_long_unsigned_type_node): New macro.
5139 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
5141 * config/rs6000/rs6000-builtin-new.def: Add power9-vector, power9,
5142 and power9-64 stanzas.
5144 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
5145 Tom de Vries <tdevries@suse.de>
5147 * config.gcc (nvptx-*-*): Define {c,c++}_target_objs.
5148 * config/nvptx/nvptx-protos.h (nvptx_cpu_cpp_builtins): Prototype.
5149 * config/nvptx/nvptx.h (TARGET_CPU_CPP_BUILTINS): Implement with
5150 a call to the new nvptx_cpu_cpp_builtins function in nvptx-c.c.
5151 * config/nvptx/t-nvptx (nvptx-c.o): New rule.
5152 * config/nvptx/nvptx-c.c: New source file.
5153 (nvptx_cpu_cpp_builtins): Move implementation here.
5155 2021-08-24 Martin Sebor <msebor@redhat.com>
5157 PR middle-end/101600
5158 PR middle-end/101977
5159 * gimple-ssa-warn-access.cc (maybe_warn_for_bound): Tighten up
5160 the phrasing of a warning.
5161 (check_access): Use the remaining size after subtracting any offset
5162 rather than the whole object size.
5163 * pointer-query.cc (access_ref::get_ref): Clear BASE0 flag if it's
5164 clear for any nonnull PHI argument.
5165 (compute_objsize): Clear argument.
5167 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
5169 * config/rs6000/rs6000-builtin-new.def: Add power8-vector stanza.
5171 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
5173 * config/rs6000/rs6000-builtin-new.def: Add power7 and power7-64
5176 2021-08-24 Andrew MacLeod <amacleod@redhat.com>
5178 * value-relation.cc (rr_transitive_table): New.
5179 (relation_transitive): New.
5180 (value_relation::swap): Remove.
5181 (value_relation::apply_transitive): New.
5182 (relation_oracle::relation_oracle): Allocate a new tmp bitmap.
5183 (relation_oracle::register_relation): Call register_transitives.
5184 (relation_oracle::register_transitives): New.
5185 * value-relation.h (relation_oracle): Add new temporary bitmap and
5188 2021-08-24 H.J. Lu <hjl.tools@gmail.com>
5191 * config/i386/i386-expand.c (ix86_expand_vector_move): Broadcast
5192 from integer to a pseudo vector register.
5194 2021-08-24 Richard Biener <rguenther@suse.de>
5196 PR tree-optimization/100089
5197 * tree-vectorizer.h (vect_slp_bb): Rename to ...
5198 (vect_slp_if_converted_bb): ... this and get the original
5199 loop as new argument.
5200 * tree-vectorizer.c (try_vectorize_loop_1): Revert previous fix,
5201 pass original loop to vect_slp_if_converted_bb.
5202 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
5203 If orig_loop was passed scan the not vectorized stmts
5204 for COND_EXPRs and force not profitable if found.
5205 (vect_slp_region): Pass down all SLP instances to costing
5206 if orig_loop was specified.
5207 (vect_slp_bbs): Pass through orig_loop.
5208 (vect_slp_bb): Rename to ...
5209 (vect_slp_if_converted_bb): ... this and get the original
5210 loop as new argument.
5211 (vect_slp_function): Adjust.
5213 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
5216 * config/arm/arm.md (attribute arch): Add fix_vlldm.
5217 (arch_enabled): Use it.
5218 * config/arm/vfp.md (lazy_store_multiple_insn): Add alternative to
5219 use when erratum mitigation is needed.
5221 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
5224 * config/arm/arm.opt (mfix-cmse-cve-2021-35465): New option.
5225 * doc/invoke.texi (Arm Options): Document it.
5226 * config/arm/arm-cpus.in (quirk_vlldm): New feature bit.
5227 (ALL_QUIRKS): Add quirk_vlldm.
5228 (cortex-m33): Add quirk_vlldm.
5229 (cortex-m35p, cortex-m55): Likewise.
5230 * config/arm/arm.c (arm_option_override): Enable fix_vlldm if
5231 targetting an affected CPU and not explicitly controlled on
5234 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
5236 * config/arm/vfp.md (lazy_store_multiple_insn): Rewrite as valid RTL.
5237 (lazy_load_multiple_insn): Likewise.
5239 2021-08-24 liuhongt <hongtao.liu@intel.com>
5242 * config/i386/sse.md (<avx512>_vternlog<mode><sd_maskz_name>):
5243 Enable avx512 embedded broadcast.
5244 (*<avx512>_vternlog<mode>_all): Ditto.
5245 (<avx512>_vternlog<mode>_mask): Ditto.
5247 2021-08-24 liuhongt <hongtao.liu@intel.com>
5250 * config/i386/i386.c (ix86_rtx_costs): Define cost for
5252 * config/i386/i386.h (STRIP_UNARY): New macro.
5253 * config/i386/predicates.md (reg_or_notreg_operand): New
5255 * config/i386/sse.md (*<avx512>_vternlog<mode>_all): New define_insn.
5256 (*<avx512>_vternlog<mode>_1): New pre_reload
5257 define_insn_and_split.
5258 (*<avx512>_vternlog<mode>_2): Ditto.
5259 (*<avx512>_vternlog<mode>_3): Ditto.
5260 (any_logic1,any_logic2): New code iterator.
5261 (logic_op): New code attribute.
5262 (ternlogsuffix): Extend to VNxDF and VNxSF.
5264 2021-08-24 Richard Biener <rguenther@suse.de>
5266 * doc/invoke.texi (vect-inner-loop-cost-factor): Adjust.
5267 * params.opt (--param vect-inner-loop-cost-factor): Adjust
5269 * tree-vect-loop.c (vect_analyze_loop_form): Initialize
5270 inner_loop_cost_factor to the minimum of the estimated number
5271 of iterations of the inner loop and vect-inner-loop-cost-factor.
5273 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
5274 Richard Biener <rguenther@suse.de>
5276 * config/i386/i386-features.c (compute_convert_gain): Provide
5277 more accurate values for CONST_INT, when optimizing for size.
5278 * config/i386/i386.c (COSTS_N_BYTES): Move definition from here...
5279 * config/i386/i386.h (COSTS_N_BYTES): to here.
5281 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
5282 Jakub Jelinek <jakub@redhat.com>
5284 PR middle-end/102029
5285 * match.pd (shift transformations): Add an additional check for
5286 !POINTER_TYPE_P in the recently added left shift transformation.
5288 2021-08-24 liuhongt <hongtao.liu@intel.com>
5290 PR tree-optimization/100089
5291 * tree-vectorizer.c (try_vectorize_loop_1): Disable slp in
5292 loop vectorizer when cost model is very-cheap.
5294 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
5296 * config/rs6000/rs6000-gen-builtins.c (parse_bif_entry): Don't call
5297 asprintf, which is not available on AIX.
5299 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
5301 * config.gcc (target_gtfiles): Add ./rs6000-builtins.h.
5302 * config/rs6000/t-rs6000 (EXTRA_GTYPE_DEPS): Set.
5304 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
5306 * config.gcc (powerpc*-*-*): Add rs6000-builtins.o to extra_objs.
5307 * config/rs6000/rs6000-gen-builtins.c (main): Close init_file
5309 * config/rs6000/t-rs6000 (rs6000-gen-builtins.o): New target.
5310 (rbtree.o): Likewise.
5311 (rs6000-gen-builtins): Likewise.
5312 (rs6000-builtins.c): Likewise.
5313 (rs6000-builtins.h): Likewise.
5314 (rs6000.o): Add dependency.
5315 (EXTRA_HEADERS): Add rs6000-vecdefines.h.
5316 (rs6000-vecdefines.h): New target.
5317 (rs6000-builtins.o): Likewise.
5318 (rs6000-call.o): Add rs6000-builtins.h as a dependency.
5319 (rs6000-c.o): Likewise.
5321 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
5324 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace):
5325 Diagnose buffer overrun.
5326 (safe_inc_pos): Fix overrun detection.
5327 (match_identifier): Diagnose buffer overrun.
5328 (match_integer): Likewise.
5329 (match_to_right_bracket): Likewise.
5331 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
5333 * ipa-modref-tree.h (modref_access_node::range_info_useful_p):
5334 Improve range compare.
5335 (modref_access_node::contains): New member function.
5336 (modref_access_node::search): Remove.
5337 (modref_access_node::insert): Be smarter about subaccesses.
5339 2021-08-23 Thomas Schwinge <thomas@codesourcery.com>
5341 * config/i386/i386-options.c (ix86_omp_device_kind_arch_isa)
5342 <omp_device_arch> [ACCEL_COMPILER]: Match "intel_mic".
5343 * config/i386/t-omp-device (omp-device-properties-i386) <arch>:
5346 2021-08-23 Jeff Law <jlaw@localhost.localdomain>
5348 * config/h8300/h8300-protos.h (h8300_expand_epilogue): Add new
5350 * config/h8300/jumpcall.md (call, call_value): Restrict to
5351 !SIBLING_CALL_P cases.
5352 (subcall, sibcall_value): New patterns & expanders.
5353 * config/h8300/proepi.md (epilogue): Pass new argument to
5354 h8300_expand_epilogue.
5355 (sibcall_epilogue): New expander.
5356 * config/h8300/h8300.c (h8300_expand_epilogue): Handle sibcall
5358 (h8300_ok_for_sibcall_p): New function.
5359 (TARGET_FUNCTION_OK_FOR_SIBCALL): define.
5361 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
5363 * simplify-rtx.c (simplify_unary_operation_1): [TRUNCATE]:
5364 Handle case where the operand is already the desired mode.
5366 2021-08-23 Richard Biener <rguenther@suse.de>
5369 * tree-ssa-structalias.c (ipa_pta_execute): Check in_other_partition
5370 in addition to has_gimple_body.
5372 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
5374 PR middle-end/101949
5375 * ipa-modref.c (analyze_ssa_name_flags): Fix merging of
5378 2021-08-23 Martin Liska <mliska@suse.cz>
5380 * doc/invoke.texi: Put the option out of -mxl-mode-app-model
5383 2021-08-23 Richard Biener <rguenther@suse.de>
5385 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5386 Properly scale the inner loop cost only once.
5388 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
5390 * tree-ssa-ccp.c (bit_value_binop) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]:
5391 Provide bounds for unsigned (and signed with non-negative operands)
5392 division and modulus.
5394 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
5396 * simplify-rtx.c (simplify_truncation): Generalize simplification
5397 of (truncate:A (subreg:B X)).
5398 (simplify_unary_operation_1) [FLOAT_TRUNCATE, FLOAT_EXTEND,
5399 SIGN_EXTEND, ZERO_EXTEND]: Handle cases where the operand
5400 already has the desired machine mode.
5401 (test_scalar_int_ops): Add tests that useless extensions and
5402 truncations are optimized away.
5403 (test_scalar_int_ext_ops): New self-test function to confirm
5404 that truncations of extensions are correctly simplified.
5405 (test_scalar_int_ext_ops2): New self-test function to check
5406 truncations of truncations, extensions of extensions, and
5407 truncations of extensions.
5408 (test_scalar_ops): Call the above two functions with a
5409 representative sampling of integer machine modes.
5411 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
5413 * match.pd (shift transformations): Change the sign of an
5414 LSHIFT_EXPR if it reduces the number of explicit conversions.
5416 2021-08-23 Jakub Jelinek <jakub@redhat.com>
5418 PR tree-optimization/86723
5419 * gimple-ssa-store-merging.c (find_bswap_or_nop_finalize): Add
5420 cast64_to_32 argument, set *cast64_to_32 to false, unless n is
5421 non-memory permutation of 64-bit src which only has bytes of
5422 0 or [5..8] and n->range is 4.
5423 (find_bswap_or_nop): Add cast64_to_32 and mask arguments, adjust
5424 find_bswap_or_nop_finalize caller, support bswap with some bytes
5425 zeroed, as long as at least two bytes are not zeroed.
5426 (bswap_replace): Add mask argument and handle masking of bswap
5428 (maybe_optimize_vector_constructor): Adjust find_bswap_or_nop
5429 caller, punt if cast64_to_32 or mask is not all ones.
5430 (pass_optimize_bswap::execute): Adjust find_bswap_or_nop_finalize
5431 caller, for now punt if cast64_to_32.
5433 2021-08-23 Richard Biener <rguenther@suse.de>
5435 PR tree-optimization/79334
5436 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Record
5437 a type also for COMPONENT_REFs.
5438 (vn_reference_may_trap): Check ARRAY_REF with constant index
5439 against the array domain.
5441 2021-08-23 liuhongt <hongtao.liu@intel.com>
5444 * config/i386/sse.md (*avx512f_pshufb_truncv8hiv8qi_1): Add
5445 TARGET_AVX512BW to condition.
5447 2021-08-23 Jakub Jelinek <jakub@redhat.com>
5450 * dwarf2out.c (gen_variable_die): Add DW_AT_location for global
5451 register variables already during early_dwarf if possible.
5453 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
5455 * config/arm/arm_mve.h: Fix __arm_vctp16q return type.
5457 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
5460 * config/arm/arm.opt: Fix typo.
5461 * config/arm/t-rmprofile: Fix typo.
5463 2021-08-23 Jakub Jelinek <jakub@redhat.com>
5465 * tree.h (OMP_CLAUSE_GRAINSIZE_STRICT): Define.
5466 (OMP_CLAUSE_NUM_TASKS_STRICT): Define.
5467 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_GRAINSIZE,
5468 case OMP_CLAUSE_NUM_TASKS>: Print strict: modifier.
5469 * omp-expand.c (expand_task_call): Use GOMP_TASK_FLAG_STRICT in iflags
5470 if either grainsize or num_tasks clause has the strict modifier.
5472 2021-08-23 Martin Liska <mliska@suse.cz>
5474 * dbgcnt.def (DEBUG_COUNTER): New counter.
5475 * gimple.c (gimple_call_arg_flags): Use it in IPA PTA.
5477 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
5479 * ipa-modref.c (analyze_ssa_name_flags): Improve handling of return slot.
5481 2021-08-23 Xi Ruoyao <xry111@mengyan1223.wang>
5484 * config/mips/mips-protos.h (mips_msa_output_shift_immediate):
5486 * config/mips/mips.c (mips_msa_output_shift_immediate): New
5488 * config/mips/mips-msa.md (vashl<mode>3, vashr<mode>3,
5489 vlshr<mode>3): Call it.
5491 2021-08-22 Jan Hubicka <hubicka@ucw.cz>
5492 Martin Liska <mliska@suse.cz>
5494 PR middle-end/101949
5495 * ipa-modref.c (analyze_ssa_name_flags): Indirect call implies
5498 2021-08-21 Dragan Mladjenovic <OT_Dragan.Mladjenovic@mediatek.com>
5500 * config/mips/mips.c (mips_function_rodata_section,
5501 TARGET_ASM_FUNCTION_RODATA_SECTION): Removed.
5503 2021-08-21 John David Anglin <danglin@gcc.gnu.org>
5505 * config/pa/pa.c (pa_asm_output_aligned_common): Remove warning.
5507 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
5509 * configure.ac (thread-local storage support): Remove tls_first_major
5510 and tls_first_minor. Use "$conftest_s" to check support.
5511 * configure: Regenerate.
5513 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
5515 * configure.ac: Fixup formatting.
5517 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
5519 * acinclude.m4 (gcc_GAS_CHECK_FEATURE): Remove third argument and ...
5520 * configure.ac: ... update all callers.
5522 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
5525 * acinclude.m4 (_gcc_COMPUTE_GAS_VERSION, _gcc_GAS_VERSION_GTE_IFELSE)
5526 (gcc_GAS_VERSION_GTE_IFELSE): Remove.
5527 (gcc_GAS_CHECK_FEATURE): Do not handle in-tree case specially.
5528 * configure.ac: Remove gcc_cv_gas_major_version, gcc_cv_gas_minor_version.
5529 Remove remaining checks for in-tree assembler.
5530 * configure: Regenerate.
5532 2021-08-20 Jeff Law <jlaw@localhost.localdomain>
5534 * config/h8300/h8300.c (shift_alg_hi): Improve arithmetic shift right
5535 by 15 bits for H8/300H and H8/S. Improve logical shifts by 12
5537 (shift_alg_si): Improve arithmetic right shift by 28-30 bits for
5538 H8/300H. Improve arithmetic shift right by 15 bits for H8/S.
5539 Improve logical shifts by 27 bits for H8/S.
5540 (get_shift_alg): Corresponding changes.
5541 (h8300_option_override): Revert to loops for -Os when profitable.
5543 2021-08-20 Richard Biener <rguenther@suse.de>
5545 * tree-vect-data-refs.c (dr_group_sort_cmp): Do not compare
5547 (vect_analyze_data_ref_accesses): Likewise. Assign the BB
5548 index as group_id when dataref_groups were not computed.
5549 * tree-vect-slp.c (vect_slp_bbs): Bump current_group when
5550 we advace to the next BB.
5552 2021-08-20 Jakub Jelinek <jakub@redhat.com>
5554 * omp-builtins.def (BUILT_IN_GOMP_WARNING, BUILT_IN_GOMP_ERROR): New
5557 2021-08-20 Martin Liska <mliska@suse.cz>
5559 PR gcov-profile/89961
5560 * gcov.c (make_gcov_file_name): Rewrite using std::string.
5561 (mangle_name): Simplify, do not used the second argument.
5562 (strip_extention): New function.
5563 (get_md5sum): Likewise.
5564 (get_gcov_intermediate_filename): Handle properly -p and -x
5566 (output_gcov_file): Use string type.
5567 (generate_results): Likewise.
5568 (md5sum_to_hex): Remove.
5570 2021-08-20 Michael Meissner <meissner@linux.ibm.com>
5572 * config/rs6000/altivec.md (UNSPEC_XXEVAL): Move to vsx.md.
5573 (UNSPEC_XXSPLTIW): Move to vsx.md.
5574 (UNSPEC_XXSPLTID): Move to vsx.md.
5575 (UNSPEC_XXSPLTI32DX): Move to vsx.md.
5576 (UNSPEC_XXBLEND): Move to vsx.md.
5577 (UNSPEC_XXPERMX): Move to vsx.md.
5578 (VM3): Move to vsx.md.
5579 (VM3_char): Move to vsx.md.
5580 (xxspltiw_v4si): Move to vsx.md.
5581 (xxspltiw_v4sf): Move to vsx.md.
5582 (xxspltiw_v4sf_inst): Move to vsx.md.
5583 (xxspltidp_v2df): Move to vsx.md.
5584 (xxspltidp_v2df_inst): Move to vsx.md.
5585 (xxsplti32dx_v4si_inst): Move to vsx.md.
5586 (xxsplti32dx_v4sf): Move to vsx.md.
5587 (xxsplti32dx_v4sf_inst): Move to vsx.md.
5588 (xxblend_<mode>): Move to vsx.md.
5589 (xxpermx): Move to vsx.md.
5590 (xxpermx_inst): Move to vsx.md.
5591 * config/rs6000/vsx.md (UNSPEC_XXEVAL): Move from altivec.md.
5592 (UNSPEC_XXSPLTIW): Move from altivec.md.
5593 (UNSPEC_XXSPLTID): Move from altivec.md.
5594 (UNSPEC_XXSPLTI32DX): Move from altivec.md.
5595 (UNSPEC_XXBLEND): Move from altivec.md.
5596 (UNSPEC_XXPERMX): Move from altivec.md.
5597 (VM3): Move from altivec.md.
5598 (VM3_char): Move from altivec.md.
5599 (xxspltiw_v4si): Move from altivec.md.
5600 (xxspltiw_v4sf): Move from altivec.md.
5601 (xxspltiw_v4sf_inst): Move from altivec.md.
5602 (xxspltidp_v2df): Move from altivec.md.
5603 (xxspltidp_v2df_inst): Move from altivec.md.
5604 (xxsplti32dx_v4si_inst): Move from altivec.md.
5605 (xxsplti32dx_v4sf): Move from altivec.md.
5606 (xxsplti32dx_v4sf_inst): Move from altivec.md.
5607 (xxblend_<mode>): Move from altivec.md.
5608 (xxpermx): Move from altivec.md.
5609 (xxpermx_inst): Move from altivec.md.
5611 2021-08-19 Roger Sayle <roger@nextmovesoftware.com>
5613 * tree-vect-generic.c (expand_vector_operations_1): Use either
5614 gimplify_build1 or gimplify_build2 instead of gimple_build_assign
5615 when constructing scalar splat expressions.
5617 2021-08-19 Peter Bergner <bergner@linux.ibm.com>
5620 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Cast
5621 pointer to __vector_pair *.
5623 2021-08-19 Martin Sebor <msebor@redhat.com>
5625 * gimple-range.cc: Add comments.
5626 * gimple-range.h: Same.
5628 2021-08-19 Martin Sebor <msebor@redhat.com>
5630 PR middle-end/101984
5631 * gimple-ssa-warn-access.cc (pass_waccess::execute): Also call
5634 2021-08-19 Jeff Law <jlaw@localhost.localdomain>
5636 * config.gcc (h8300-*-elf*): Do not include dbxelf.h.
5637 (h8300-*-linux*, v850-*-rtems*, v850*-elf*): Likewise.
5638 * config/v850/v850.h (DEFAULT_GDB_EXTENSIONS): Remove.
5640 2021-08-19 Jakub Jelinek <jakub@redhat.com>
5642 PR middle-end/101950
5643 * optabs.c (expand_clrsb_using_clz): New function.
5644 (expand_unop): Use it as another clrsb expansion fallback.
5646 2021-08-19 liuhongt <hongtao.liu@intel.com>
5649 2021-07-28 liuhongt <hongtao.liu@intel.com>
5652 * config/i386/i386.h (processor_costs): Add new member
5654 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
5655 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
5656 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
5657 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
5658 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
5659 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
5660 generic_cost, core_cost): Initialize integer_to_sse same value
5662 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
5663 * config/i386/i386.c (ix86_builtin_vectorization_cost):
5664 Use integer_to_sse instead of sse_op to calculate the cost of
5667 2021-08-18 Iain Sandoe <iain@sandoe.co.uk>
5669 * config.gcc: Include rpath.opt for Darwin.
5670 * config/darwin.h (DRIVER_SELF_SPECS): Handle -rpath.
5672 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
5675 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
5678 2021-08-18 Jonathan Wright <jonathan.wright@arm.com>
5680 * config/aarch64/arm_neon.h (vld3_lane_f64): Use float RTL
5681 pattern and type cast.
5682 (vld4_lane_f32): Use float RTL pattern.
5683 (vld4q_lane_f64): Use float type cast.
5685 2021-08-18 Jan Hubicka <hubicka@ucw.cz>
5687 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Check also
5690 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
5692 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor): Extend.
5693 (test_map_of_type_with_ctor_and_dtor_expand): Add function.
5694 (hash_map_tests_c_tests): Call it.
5696 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
5698 * ggc.h (enum ggc_collect): New.
5699 (ggc_collect): Use it.
5700 * ggc-page.c: Adjust.
5701 * ggc-common.c: Likewise.
5702 * ggc-tests.c: Likewise.
5703 * read-rtl-function.c: Likewise.
5704 * selftest-run-tests.c: Likewise.
5705 * doc/gty.texi (Invoking the garbage collector): Likewise.
5707 2021-08-18 liuhongt <hongtao.liu@intel.com>
5710 * config/i386/i386.h (TARGET_V2DF_REDUCTION_PREFER_HADDPD):
5712 * config/i386/sse.md (*sse3_haddv2df3_low): Add
5713 TARGET_V2DF_REDUCTION_PREFER_HADDPD.
5714 (*sse3_hsubv2df3_low): Ditto.
5715 * config/i386/x86-tune.def
5716 (X86_TUNE_V2DF_REDUCTION_PREFER_HADDPD): New tune.
5718 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
5720 * gimple-range-gori.cc (gori_compute::gori_compute): Enable tracing.
5721 (gori_compute::compute_operand_range): Add tracing.
5722 (gori_compute::logical_combine): Ditto.
5723 (gori_compute::compute_logical_operands): Ditto.
5724 (gori_compute::compute_operand1_range): Ditto.
5725 (gori_compute::compute_operand2_range): Ditto.
5726 (gori_compute::outgoing_edge_range_p): Ditto.
5727 * gimple-range-gori.h (class gori_compute): Add range_tracer.
5729 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
5731 * flag-types.h (enum evrp_mode): Adjust evrp-mode values.
5732 * gimple-range-cache.cc (DEBUG_RANGE_CACHE): Relocate from.
5733 * gimple-range-trace.h (DEBUG_RANGE_CACHE): Here.
5734 * params.opt (--param=evrp-mode): Adjust options.
5736 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
5738 * Makefile.in (OBJS): Add gimple-range-trace.o.
5739 * gimple-range-cache.h (enable_new_values): Remove unused prototype.
5740 * gimple-range-fold.cc: Adjust headers.
5741 * gimple-range-trace.cc: New.
5742 * gimple-range-trace.h: New.
5743 * gimple-range.cc (gimple_ranger::gimple_ranger): Enable tracer.
5744 (gimple_ranger::range_of_expr): Add tracing.
5745 (gimple_ranger::range_on_entry): Ditto.
5746 (gimple_ranger::range_on_exit): Ditto.
5747 (gimple_ranger::range_on_edge): Ditto.
5748 (gimple_ranger::fold_range_internal): Ditto.
5749 (gimple_ranger::dump_bb): Do not calculate edge range twice.
5750 (trace_ranger::*): Remove.
5751 (enable_ranger): Never create a trace_ranger.
5752 (debug_seed_ranger): Move to gimple-range-trace.cc.
5753 (dump_ranger): Ditto.
5754 (debug_ranger): Ditto.
5755 * gimple-range.h: Include gimple-range-trace.h.
5756 (range_on_entry, range_on_exit): No longer virtual.
5757 (class trace_ranger): Remove.
5758 (DEBUG_RANGE_CACHE): Move to gimple-range-trace.h.
5760 2021-08-17 Martin Sebor <msebor@redhat.com>
5762 PR middle-end/101854
5763 * builtins.c (expand_builtin_alloca): Move warning code to check_alloca
5764 in gimple-ssa-warn-access.cc.
5765 * calls.c (alloc_max_size): Move code to check_alloca.
5766 (get_size_range): Move to pointer-query.cc.
5767 (maybe_warn_alloc_args_overflow): Move to gimple-ssa-warn-access.cc.
5768 (get_attr_nonstring_decl): Move to tree.c.
5769 (fntype_argno_type): Move to gimple-ssa-warn-access.cc.
5770 (append_attrname): Same.
5771 (maybe_warn_rdwr_sizes): Same.
5772 (initialize_argument_information): Move code to
5773 gimple-ssa-warn-access.cc.
5774 * calls.h (maybe_warn_alloc_args_overflow): Move to
5775 gimple-ssa-warn-access.h.
5776 (get_attr_nonstring_decl): Move to tree.h.
5777 (maybe_warn_nonstring_arg): Move to gimple-ssa-warn-access.h.
5778 (enum size_range_flags): Move to pointer-query.h.
5779 (get_size_range): Same.
5780 * gimple-ssa-warn-access.cc (has_location): Remove unused overload
5781 to avoid Clang -Wunused-function.
5782 (get_size_range): Declare static.
5783 (maybe_emit_free_warning): Rename...
5784 (maybe_check_dealloc_call): ...to this for consistency.
5785 (class pass_waccess): Add members.
5786 (pass_waccess::~pass_waccess): Defined.
5787 (alloc_max_size): Move here from calls.c.
5788 (maybe_warn_alloc_args_overflow): Same.
5789 (check_alloca): New function.
5790 (check_alloc_size_call): New function.
5791 (check_strncat): Handle another warning flag.
5792 (pass_waccess::check_builtin): Handle alloca.
5793 (fntype_argno_type): Move here from calls.c.
5794 (append_attrname): Same.
5795 (maybe_warn_rdwr_sizes): Same.
5796 (pass_waccess::check_call): Define.
5797 (check_nonstring_args): New function.
5798 (pass_waccess::check): Call new member functions.
5799 (pass_waccess::execute): Enable ranger.
5800 * gimple-ssa-warn-access.h (get_size_range): Move here from calls.h.
5801 (maybe_warn_nonstring_arg): Same.
5802 * gimple-ssa-warn-restrict.c: Remove #include.
5803 * pointer-query.cc (get_size_range): Move here from calls.c.
5804 * pointer-query.h (enum size_range_flags): Same.
5805 (get_size_range): Same.
5806 * tree.c (get_attr_nonstring_decl): Move here from calls.c.
5807 * tree.h (get_attr_nonstring_decl): Move here from calls.h.
5809 2021-08-17 Thomas Schwinge <thomas@codesourcery.com>
5811 * ggc.h (ggc_collect): Add 'force_collect' parameter.
5812 * ggc-page.c (ggc_collect): Use that one instead of global
5813 'ggc_force_collect'. Adjust all users.
5814 * doc/gty.texi (Invoking the garbage collector): Update.
5815 * ggc-internal.h (ggc_force_collect): Remove.
5816 * ggc-common.c (ggc_force_collect): Likewise.
5817 * selftest.h (forcibly_ggc_collect): Remove.
5818 * ggc-tests.c (selftest::forcibly_ggc_collect): Likewise.
5819 * read-rtl-function.c (test_loading_labels): Adjust.
5820 * selftest-run-tests.c (run_tests): Likewise.
5822 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
5824 * config/darwin.c (darwin_file_end): Reset and reclaim the
5825 section names table at the end of compile.
5827 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
5830 * config.in: Regenerate.
5831 * config/i386/darwin.h (EXTRA_ASM_OPTS): New
5832 (ASM_SPEC): Pass options to disable branch shortening where
5834 * configure: Regenerate.
5835 * configure.ac: Detect versions of 'as' that support the
5836 optimisation which has the bug.
5838 2021-08-17 Richard Biener <rguenther@suse.de>
5840 * optabs-query.c (supports_vec_gather_load_p): Also check
5842 (supports_vec_scatter_store_p): Likewise.
5843 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Fall
5844 back to masked variants if non-masked are not supported.
5845 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
5846 When we need to use masked gather/scatter but do not have
5847 a mask set up a constant true one.
5848 * tree-vect-stmts.c (vect_check_scalar_mask): Also allow
5851 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
5853 * tree-ssa-ccp.c (bit_value_binop) [MINUS_EXPR]: Use same
5854 algorithm as PLUS_EXPR to improve subtraction bit bounds.
5855 [POINTER_DIFF_EXPR]: Treat as synonymous with MINUS_EXPR.
5857 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
5859 * tree-ssa-ccp.c (bit_value_mult_const): New helper function to
5860 calculate the mask-value pair result of a multiplication by an
5862 (bit_value_binop) [MULT_EXPR]: Call it from here for
5863 multiplications by (sparse) non-negative constants.
5865 2021-08-17 Christophe Lyon <christophe.lyon@foss.st.com>
5868 * config.gcc (gcc_cv_initfini_array): Leave undefined for
5869 uclinuxfdpiceabi targets.
5871 2021-08-17 Alexandre Oliva <oliva@adacore.com>
5873 * tree-inline.c (maybe_move_debug_stmts_to_successors): Don't
5874 reverse debug stmts.
5876 2021-08-17 Alexandre Oliva <oliva@adacore.com>
5878 * tree-cfg.c (dump_function_to_file): Use fun, not cfun.
5880 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
5882 * config/aarch64/arm_neon.h (__LD4_LANE_FUNC): Delete.
5883 (__LD4Q_LANE_FUNC): Likewise.
5884 (vld4_lane_u8): Define without macro.
5885 (vld4_lane_u16): Likewise.
5886 (vld4_lane_u32): Likewise.
5887 (vld4_lane_u64): Likewise.
5888 (vld4_lane_s8): Likewise.
5889 (vld4_lane_s16): Likewise.
5890 (vld4_lane_s32): Likewise.
5891 (vld4_lane_s64): Likewise.
5892 (vld4_lane_f16): Likewise.
5893 (vld4_lane_f32): Likewise.
5894 (vld4_lane_f64): Likewise.
5895 (vld4_lane_p8): Likewise.
5896 (vld4_lane_p16): Likewise.
5897 (vld4_lane_p64): Likewise.
5898 (vld4q_lane_u8): Likewise.
5899 (vld4q_lane_u16): Likewise.
5900 (vld4q_lane_u32): Likewise.
5901 (vld4q_lane_u64): Likewise.
5902 (vld4q_lane_s8): Likewise.
5903 (vld4q_lane_s16): Likewise.
5904 (vld4q_lane_s32): Likewise.
5905 (vld4q_lane_s64): Likewise.
5906 (vld4q_lane_f16): Likewise.
5907 (vld4q_lane_f32): Likewise.
5908 (vld4q_lane_f64): Likewise.
5909 (vld4q_lane_p8): Likewise.
5910 (vld4q_lane_p16): Likewise.
5911 (vld4q_lane_p64): Likewise.
5912 (vld4_lane_bf16): Likewise.
5913 (vld4q_lane_bf16): Likewise.
5915 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
5917 * config/aarch64/arm_neon.h (__LD3_LANE_FUNC): Delete.
5918 (__LD3Q_LANE_FUNC): Delete.
5919 (vld3_lane_u8): Define without macro.
5920 (vld3_lane_u16): Likewise.
5921 (vld3_lane_u32): Likewise.
5922 (vld3_lane_u64): Likewise.
5923 (vld3_lane_s8): Likewise.
5924 (vld3_lane_s16): Likewise.
5925 (vld3_lane_s32): Likewise.
5926 (vld3_lane_s64): Likewise.
5927 (vld3_lane_f16): Likewise.
5928 (vld3_lane_f32): Likewise.
5929 (vld3_lane_f64): Likewise.
5930 (vld3_lane_p8): Likewise.
5931 (vld3_lane_p16): Likewise.
5932 (vld3_lane_p64): Likewise.
5933 (vld3q_lane_u8): Likewise.
5934 (vld3q_lane_u16): Likewise.
5935 (vld3q_lane_u32): Likewise.
5936 (vld3q_lane_u64): Likewise.
5937 (vld3q_lane_s8): Likewise.
5938 (vld3q_lane_s16): Likewise.
5939 (vld3q_lane_s32): Likewise.
5940 (vld3q_lane_s64): Likewise.
5941 (vld3q_lane_f16): Likewise.
5942 (vld3q_lane_f32): Likewise.
5943 (vld3q_lane_f64): Likewise.
5944 (vld3q_lane_p8): Likewise.
5945 (vld3q_lane_p16): Likewise.
5946 (vld3q_lane_p64): Likewise.
5947 (vld3_lane_bf16): Likewise.
5948 (vld3q_lane_bf16): Likewise.
5950 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
5952 * config/aarch64/arm_neon.h (__LD2_LANE_FUNC): Delete.
5953 (__LD2Q_LANE_FUNC): Likewise.
5954 (vld2_lane_u8): Define without macro.
5955 (vld2_lane_u16): Likewise.
5956 (vld2_lane_u32): Likewise.
5957 (vld2_lane_u64): Likewise.
5958 (vld2_lane_s8): Likewise.
5959 (vld2_lane_s16): Likewise.
5960 (vld2_lane_s32): Likewise.
5961 (vld2_lane_s64): Likewise.
5962 (vld2_lane_f16): Likewise.
5963 (vld2_lane_f32): Likewise.
5964 (vld2_lane_f64): Likewise.
5965 (vld2_lane_p8): Likewise.
5966 (vld2_lane_p16): Likewise.
5967 (vld2_lane_p64): Likewise.
5968 (vld2q_lane_u8): Likewise.
5969 (vld2q_lane_u16): Likewise.
5970 (vld2q_lane_u32): Likewise.
5971 (vld2q_lane_u64): Likewise.
5972 (vld2q_lane_s8): Likewise.
5973 (vld2q_lane_s16): Likewise.
5974 (vld2q_lane_s32): Likewise.
5975 (vld2q_lane_s64): Likewise.
5976 (vld2q_lane_f16): Likewise.
5977 (vld2q_lane_f32): Likewise.
5978 (vld2q_lane_f64): Likewise.
5979 (vld2q_lane_p8): Likewise.
5980 (vld2q_lane_p16): Likewise.
5981 (vld2q_lane_p64): Likewise.
5982 (vld2_lane_bf16): Likewise.
5983 (vld2q_lane_bf16): Likewise.
5985 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
5987 * haifa-sched.c (advance_one_cycle): Output more context-synchronization
5990 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
5992 * haifa-sched.c (enum rfs_decision, rfs_str): Add RFS_AUTOPREF.
5993 (rank_for_schedule): Use it.
5995 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
5997 PR rtl-optimization/91598
5998 * haifa-sched.c (autopref_rank_for_schedule): Prioritize "irrelevant"
5999 insns after memory reads and before memory writes.
6001 2021-08-17 Alistair_Lee <alistair.lee@arm.com>
6003 * rtl.h (CONST_VECTOR_P): New macro.
6004 * config/aarch64/aarch64.c (aarch64_get_sve_pred_bits): Use RTL
6005 code testing macros.
6006 (aarch64_ptrue_all_mode): Likewise.
6007 (aarch64_expand_mov_immediate): Likewise.
6008 (aarch64_const_vec_all_in_range_p): Likewise.
6009 (aarch64_rtx_costs): Likewise.
6010 (aarch64_legitimate_constant_p): Likewise.
6011 (aarch64_simd_valid_immediate): Likewise.
6012 (aarch64_simd_make_constant): Likewise.
6013 (aarch64_convert_mult_to_shift): Likewise.
6014 (aarch64_expand_sve_vec_perm): Likewise.
6015 (aarch64_vec_fpconst_pow_of_2): Likewise.
6017 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
6019 PR tree-optimization/101938
6020 * range-op.cc (operator_abs::op1_range): Special case
6021 -TYPE_MIN_VALUE for flag_wrapv.
6023 2021-08-17 Kewen Lin <linkw@linux.ibm.com>
6025 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Add the cost for
6028 2021-08-17 Jakub Jelinek <jakub@redhat.com>
6030 * tree.def (OMP_SCOPE): New tree code.
6031 * tree.h (OMP_SCOPE_BODY, OMP_SCOPE_CLAUSES): Define.
6032 * tree-nested.c (convert_nonlocal_reference_stmt,
6033 convert_local_reference_stmt, convert_gimple_call): Handle
6035 * tree-pretty-print.c (dump_generic_node): Handle OMP_SCOPE.
6036 * gimple.def (GIMPLE_OMP_SCOPE): New gimple code.
6037 * gimple.c (gimple_build_omp_scope): New function.
6038 (gimple_copy): Handle GIMPLE_OMP_SCOPE.
6039 * gimple.h (gimple_build_omp_scope): Declare.
6040 (gimple_has_substatements): Handle GIMPLE_OMP_SCOPE.
6041 (gimple_omp_scope_clauses, gimple_omp_scope_clauses_ptr,
6042 gimple_omp_scope_set_clauses): New inline functions.
6043 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_SCOPE.
6044 * gimple-pretty-print.c (dump_gimple_omp_scope): New function.
6045 (pp_gimple_stmt_1): Handle GIMPLE_OMP_SCOPE.
6046 * gimple-walk.c (walk_gimple_stmt): Likewise.
6047 * gimple-low.c (lower_stmt): Likewise.
6048 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
6049 (gimplify_scan_omp_clauses): For task reductions, handle OMP_SCOPE
6050 like ORT_WORKSHARE constructs. Adjust diagnostics for %<scope%>
6051 allowing task reductions. Reject inscan reductions on scope.
6052 (omp_find_stores_stmt): Handle GIMPLE_OMP_SCOPE.
6053 (gimplify_omp_workshare, gimplify_expr): Handle OMP_SCOPE.
6054 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_SCOPE.
6055 (estimate_num_insns): Likewise.
6056 * omp-low.c (build_outer_var_ref): Look through GIMPLE_OMP_SCOPE
6057 contexts if var isn't privatized there.
6058 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_SCOPE.
6059 (scan_omp_1_stmt): Likewise.
6060 (maybe_add_implicit_barrier_cancel): Look through outer
6062 (lower_omp_scope): New function.
6063 (lower_omp_task_reductions): Handle OMP_SCOPE.
6064 (lower_omp_1): Handle GIMPLE_OMP_SCOPE.
6065 (diagnose_sb_1, diagnose_sb_2): Likewise.
6066 * omp-expand.c (expand_omp_single): Support also GIMPLE_OMP_SCOPE.
6067 (expand_omp): Handle GIMPLE_OMP_SCOPE.
6068 (omp_make_gimple_edges): Likewise.
6069 * omp-builtins.def (BUILT_IN_GOMP_SCOPE_START): New built-in.
6071 2021-08-17 Richard Biener <rguenther@suse.de>
6073 PR tree-optimization/101925
6074 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Set
6075 reverse on COMPONENT_REF and ARRAY_REF according to
6076 what reverse_storage_order_for_component_p does.
6077 (vn_reference_eq): Compare reversed on reference ops.
6078 (reverse_storage_order_for_component_p): New overload.
6079 (vn_reference_lookup_3): Check reverse_storage_order_for_component_p
6080 on the reference looked up.
6082 2021-08-17 Jeff Law <jlaw@localhost.localdomain>
6084 * config/h8300/h8300.c (shift_alg_si): Avoid loops for most SImode
6086 (h8300_option_override): Use loops on H8/S more often when optimizing
6088 (get_shift_alg): Handle new "special" cases on H8/S. Simplify
6089 accordingly. Handle various arithmetic right shifts with special
6090 sequences that we couldn't handle before.
6092 2021-08-16 Jeff Law <jlaw@localhost.localdomain>
6094 * config.gcc (rl78-*-elf*): Do not include dbxelf.h.
6096 2021-08-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
6098 * config/sparc/rtemself.h (SPARC_GCOV_TYPE_SIZE): Define.
6099 * config/sparc/sparc.c (sparc_gcov_type_size): New.
6100 (TARGET_GCOV_TYPE_SIZE): Redefine if SPARC_GCOV_TYPE_SIZE is defined.
6101 * coverage.c (get_gcov_type): Use targetm.gcov_type_size().
6102 * doc/tm.texi (TARGET_GCOV_TYPE_SIZE): Add hook under "Misc".
6103 * doc/tm.texi.in: Regenerate.
6104 * target.def (gcov_type_size): New target hook.
6105 * targhooks.c (default_gcov_type_size): New.
6106 * targhooks.h (default_gcov_type_size): Declare.
6107 * tree-profile.c (gimple_gen_edge_profiler): Use precision of
6109 (gimple_gen_time_profiler): Likewise.
6111 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
6113 * dwarf2out.c (add_scalar_info): Deal with DW_AT_data_bit_offset.
6115 2021-08-16 Tobias Burnus <tobias@codesourcery.com>
6117 PR middle-end/101931
6118 * omp-low.c (omp_runtime_api_call): Update for routines
6119 added in the meanwhile.
6121 2021-08-16 Martin Liska <mliska@suse.cz>
6123 PR tree-optimization/100393
6124 * tree-switch-conversion.c (group_cluster::dump): Use
6125 get_comparison_count.
6126 (jump_table_cluster::find_jump_tables): Pre-compute number of
6127 comparisons and then decrement it. Cache also max_ratio.
6128 (jump_table_cluster::can_be_handled): Change signature.
6129 * tree-switch-conversion.h (get_comparison_count): New.
6131 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
6133 * dwarf2out.c (add_data_member_location_attribute): Use GNAT
6134 encodings only when -fgnat-encodings=all is specified.
6135 (add_bound_info): Likewise.
6136 (add_byte_size_attribute): Likewise.
6137 (gen_member_die): Likewise.
6139 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
6141 * omp-oacc-neuter-broadcast.cc
6142 (execute_omp_oacc_neuter_broadcast): Plug 'par' memory leak.
6144 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
6146 * omp-oacc-neuter-broadcast.cc
6147 (execute_omp_oacc_neuter_broadcast): Clarify memory management for
6150 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
6152 * omp-oacc-neuter-broadcast.cc (field_map): Move variable into...
6153 (execute_omp_oacc_neuter_broadcast): ... here.
6154 (install_var_field, build_receiver_ref, build_sender_ref): Take
6155 'field_map_t *' parameter. Adjust all users.
6156 (worker_single_copy, neuter_worker_single): Take a
6157 'record_field_map_t *' parameter. Adjust all users.
6159 2021-08-16 liuhongt <hongtao.liu@intel.com>
6162 * config/i386/i386.md (ldexp<mode>3): Force operands[1] to
6165 2021-08-16 Martin Liska <mliska@suse.cz>
6168 * multiple_target.c (create_dispatcher_calls): Make default
6169 function local only if it is a definition.
6171 2021-08-16 Martin Liska <mliska@suse.cz>
6174 * ipa-icf-gimple.c (func_checker::compare_ssa_name): Do not
6175 consider equal SSA_NAMEs when one is a param.
6177 2021-08-16 liuhongt <hongtao.liu@intel.com>
6180 * config/i386/i386-expand.c (ix86_expand_vec_perm_vpermt2):
6181 Support vpermi2b for V32QI/V16QImode.
6182 (ix86_extract_perm_from_pool_constant): New function.
6183 (ix86_expand_vec_one_operand_perm_avx512): Support
6184 vpermw/vpermb under TARGET_AVX512BW/TARGET_AVX512VBMI.
6185 (expand_vec_perm_1): Adjust comments for upper.
6186 * config/i386/i386-protos.h (ix86_extract_perm_from_pool_constant):
6188 * config/i386/predicates.md (permvar_truncate_operand): New predicate.
6189 (pshufb_truncv4siv4hi_operand): Ditto.
6190 (pshufb_truncv8hiv8qi_operand): Ditto.
6191 * config/i386/sse.md (*avx512bw_permvar_truncv16siv16hi_1):
6192 New pre_reload define_insn_and_split.
6193 (*avx512f_permvar_truncv8siv8hi_1): Ditto.
6194 (*avx512f_vpermvar_truncv8div8si_1): Ditto.
6195 (*avx512f_permvar_truncv32hiv32qi_1): Ditto.
6196 (*avx512f_permvar_truncv16hiv16qi_1): Ditto.
6197 (*avx512f_permvar_truncv4div4si_1): Ditto.
6198 (*avx512f_pshufb_truncv8hiv8qi_1): Ditto.
6199 (*avx512f_pshufb_truncv4siv4hi_1): Ditto.
6200 (*avx512f_pshufd_truncv2div2si_1): Ditto.
6202 2021-08-16 Kito Cheng <kito.cheng@sifive.com>
6204 * config/riscv/multilib-generator: Support code model option for
6206 * doc/install.texi: Add document of new option for
6207 --with-multilib-generator.
6209 2021-08-15 Clément Chigot <clement.chigot@atos.net>
6211 * config/rs6000/rs6000.c (xcoff_tls_exec_model_detected): New.
6212 (rs6000_legitimize_tls_address_aix): Use it.
6213 (rs6000_xcoff_file_end): Add ".ref __tls_get_addr" when
6214 xcoff_tls_exec_model_detected is true.
6216 2021-08-15 Jeff Law <jlaw@localhost.localdomain>
6218 * config/h8300/h8300.c (shift_alg_si): Retune H8/300H shifts
6219 to allow a bit more code growth, saving many dozens of cycles.
6220 (h8300_option_override): Adjus shift_alg_si if optimizing for
6222 (get_shift_alg): Use special + inline shifts for residuals
6225 2021-08-14 Stafford Horne <shorne@gmail.com>
6228 * config/or1k/or1k-opts.h: New file.
6229 * config/or1k/or1k.c (or1k_legitimize_address_1, print_reloc):
6230 Support generating gotha relocations if -mcmodel=large is
6232 * config/or1k/or1k.h (TARGET_CMODEL_SMALL, TARGET_CMODEL_LARGE):
6234 * config/or1k/or1k.opt (mcmodel=): New option.
6235 * doc/invoke.texi (OpenRISC Options): Document mcmodel.
6237 2021-08-14 Martin Sebor <msebor@redhat.com>
6239 PR middle-end/101791
6240 * gimple-ssa-warn-access.cc (new_delete_mismatch_p): Use new argument
6241 to valid_new_delete_pair_p.
6242 * tree.c (valid_new_delete_pair_p): Add argument.
6243 * tree.h (valid_new_delete_pair_p): Same.
6245 2021-08-14 Jakub Jelinek <jakub@redhat.com>
6248 * config/i386/i386-expand.c (expand_vec_perm_broadcast_1)
6249 <case E_V64QImode>: For this mode assert
6250 !TARGET_AVX512BW || d->perm[0] rather than !TARGET_AVX2 || d->perm[0].
6252 2021-08-13 Michael Meissner <meissner@linux.ibm.com>
6255 * config/rs6000/altivec.md (xxeval): Use register_predicate
6256 instead of altivec_register_predicate.
6258 2021-08-13 Martin Sebor <msebor@redhat.com>
6260 PR middle-end/101734
6261 * tree-ssa-uninit.c (maybe_warn_read_write_only): New function.
6262 (maybe_warn_operand): Call it.
6264 2021-08-13 Martin Liska <mliska@suse.cz>
6267 * attribs.c (decl_attributes): Make naked functions "noipa"
6270 2021-08-13 Martin Liska <mliska@suse.cz>
6273 * symtab.c (symtab_node::noninterposable_alias): Do not create
6274 local aliases for target_clone functions as the clonning pass
6277 2021-08-13 Martin Liska <mliska@suse.cz>
6279 * opts.c (LIVE_PATCHING_OPTION): Define.
6280 (control_options_for_live_patching): Use it in error messages.
6282 2021-08-13 Jan Hubicka <hubicka@ucw.cz>
6284 * ipa-modref.c (dump_eaf_flags): Dump EAF_NOREAD.
6285 (implicit_const_eaf_flags, implicit_pure_eaf_flags,
6286 ignore_stores_eaf_flags): New constants.
6287 (remove_useless_eaf_flags): New function.
6288 (eaf_flags_useful_p): Use it.
6289 (deref_flags): Add EAF_NOT_RETURNED if flag is unused;
6291 (modref_lattice::init): Add EAF_NOREAD.
6292 (modref_lattice::add_escape_point): Do not reacord escape point if
6294 (modref_lattice::merge): EAF_NOESCAPE implies EAF_NODIRECTESCAPE;
6295 use remove_useless_eaf_flags.
6296 (modref_lattice::merge_deref): Use ignore_stores_eaf_flags.
6297 (modref_lattice::merge_direct_load): Add EAF_NOREAD
6298 (analyze_ssa_name_flags): Fix handling EAF_NOT_RETURNED
6299 (analyze_parms): Use remove_useless_eaf_flags.
6300 (ipa_merge_modref_summary_after_inlining): Use ignore_stores_eaf_flags.
6301 (modref_merge_call_site_flags): Add caller and ecf_flags parameter;
6302 use remove_useless_eaf_flags.
6303 (modref_propagate_flags_in_scc): Update.
6304 * ipa-modref.h: Turn eaf_flags_t back to char.
6305 * tree-core.h (EAF_NOT_RETURNED): Fix.
6306 (EAF_NOREAD): New constant
6307 * tree-ssa-alias.c: (ref_maybe_used_by_call_p_1): Check for
6309 * tree-ssa-structalias.c (handle_rhs_call): Handle new flags.
6310 (handle_pure_call): Likewise.
6312 2021-08-12 Jakub Jelinek <jakub@redhat.com>
6314 * tree.def (OMP_MASKED): New tree code.
6315 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_FILTER.
6316 * tree.h (OMP_MASKED_BODY, OMP_MASKED_CLAUSES, OMP_MASKED_COMBINED,
6317 OMP_CLAUSE_FILTER_EXPR): Define.
6318 * tree.c (omp_clause_num_ops): Add OMP_CLAUSE_FILTER entry.
6319 (omp_clause_code_name): Likewise.
6320 (walk_tree_1): Handle OMP_CLAUSE_FILTER.
6321 * tree-nested.c (convert_nonlocal_omp_clauses,
6322 convert_local_omp_clauses): Handle OMP_CLAUSE_FILTER.
6323 (convert_nonlocal_reference_stmt, convert_local_reference_stmt,
6324 convert_gimple_call): Handle GIMPLE_OMP_MASTER.
6325 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_FILTER.
6326 (dump_generic_node): Handle OMP_MASTER.
6327 * gimple.def (GIMPLE_OMP_MASKED): New gimple code.
6328 * gimple.c (gimple_build_omp_masked): New function.
6329 (gimple_copy): Handle GIMPLE_OMP_MASKED.
6330 * gimple.h (gimple_build_omp_masked): Declare.
6331 (gimple_has_substatements): Handle GIMPLE_OMP_MASKED.
6332 (gimple_omp_masked_clauses, gimple_omp_masked_clauses_ptr,
6333 gimple_omp_masked_set_clauses): New inline functions.
6334 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_MASKED.
6335 * gimple-pretty-print.c (dump_gimple_omp_masked): New function.
6336 (pp_gimple_stmt_1): Handle GIMPLE_OMP_MASKED.
6337 * gimple-walk.c (walk_gimple_stmt): Likewise.
6338 * gimple-low.c (lower_stmt): Likewise.
6339 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
6340 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_FILTER. For clauses
6341 that take one expression rather than decl or constant, force
6342 gimplification of that into a SSA_NAME or temporary unless min
6344 (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_FILTER.
6345 (gimplify_expr): Handle OMP_MASKED.
6346 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_MASKED.
6347 (estimate_num_insns): Likewise.
6348 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_FILTER.
6349 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_MASKED. Adjust
6350 diagnostics for existence of masked construct.
6351 (scan_omp_1_stmt, lower_omp_master, lower_omp_1, diagnose_sb_1,
6352 diagnose_sb_2): Handle GIMPLE_OMP_MASKED.
6353 * omp-expand.c (expand_omp_synch, expand_omp, omp_make_gimple_edges):
6356 2021-08-12 Uroš Bizjak <ubizjak@gmail.com>
6359 * config/i386/i386.md (avx512f_scalef<mode>2): New insn pattern.
6360 (ldexp<mode>3): Use avx512f_scalef<mode>2.
6361 (UNSPEC_SCALEF): Move from ...
6362 * config/i386/sse.md (UNSPEC_SCALEF): ... here.
6364 2021-08-12 Jan Hubicka <hubicka@ucw.cz>
6366 * ipa-split.c (consider_split): Fix condition testing void functions.
6368 2021-08-12 Aldy Hernandez <aldyh@redhat.com>
6370 * doc/invoke.texi: Remove docs for threader-mode param.
6371 * flag-types.h (enum threader_mode): Remove.
6372 * params.opt: Remove threader-mode param.
6373 * tree-ssa-threadbackward.c (class back_threader): Remove
6374 path_is_unreachable_p.
6375 Make find_paths private.
6376 Add maybe_thread and thread_through_all_blocks.
6377 Remove reference marker for m_registry.
6378 Remove reference marker for m_profit.
6379 (back_threader::back_threader): Adjust for registry and profit not
6381 (dump_path): Move down.
6383 (class thread_jumps): Remove.
6384 (class back_threader_registry): Remove m_all_paths.
6386 (thread_jumps::thread_through_all_blocks): Move to back_threader
6388 (fsm_find_thread_path): Remove
6389 (back_threader::maybe_thread): New.
6390 (back_threader::thread_through_all_blocks): Move from
6392 (back_threader_registry::back_threader_registry): Remove
6394 (back_threader_registry::~back_threader_registry): Remove.
6395 (thread_jumps::find_taken_edge): Remove.
6396 (thread_jumps::check_subpath_and_update_thread_path): Remove.
6397 (thread_jumps::maybe_register_path): Remove.
6398 (thread_jumps::handle_phi): Remove.
6399 (handle_assignment_p): Remove.
6400 (thread_jumps::handle_assignment): Remove.
6401 (thread_jumps::fsm_find_control_statement_thread_paths): Remove.
6402 (thread_jumps::find_jump_threads_backwards): Remove.
6403 (thread_jumps::find_jump_threads_backwards_with_ranger): Remove.
6404 (try_thread_blocks): Rename find_jump_threads_backwards to
6406 (pass_early_thread_jumps::execute): Same.
6408 2021-08-12 Tobias Burnus <tobias@codesourcery.com>
6410 * tree-core.h (omp_clause_proc_bind_kind): Add
6411 OMP_CLAUSE_PROC_BIND_PRIMARY.
6412 * tree-pretty-print.c (dump_omp_clause): Add TODO comment to
6413 change 'master' to 'primary' in proc_bind for OpenMP 5.1.
6415 2021-08-12 Claudiu Zissulescu <claziss@synopsys.com>
6417 * common/config/arc/arc-common.c (arc_option_init_struct): Remove
6418 fno-common reference.
6419 * config/arc/arc.c (arc_override_options): Remove overriding of
6422 2021-08-12 Jakub Jelinek <jakub@redhat.com>
6425 * config/i386/i386-expand.c (ix86_expand_vec_one_operand_perm_avx512):
6426 If d->testing_p, return true after performing checks instead of
6427 actually expanding the insn.
6428 (expand_vec_perm_broadcast_1): Handle V32HImode - assert
6429 !TARGET_AVX512BW and return false.
6431 2021-08-12 Eric Botcazou <ebotcazou@gcc.gnu.org>
6433 * configure.ac (PE linker --disable-dynamicbase support): New check.
6434 * configure: Regenerate.
6435 * config.in: Likewise.
6436 * config/i386/mingw32.h (LINK_SPEC_DISABLE_DYNAMICBASE): New define.
6437 (LINK_SPEC): Use it.
6438 * config/i386/mingw-w64.h (LINK_SPEC_DISABLE_DYNAMICBASE): Likewise.
6439 (LINK_SPEC): Likewise.
6441 2021-08-12 liuhongt <hongtao.liu@intel.com>
6444 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_2): New
6445 post_reload define_insn_and_split.
6446 (*avx512bw_zero_extendv32qiv32hi2_2): Ditto.
6447 (*sse4_1_zero_extendv8qiv8hi2_4): Ditto.
6448 (*avx512f_zero_extendv16hiv16si2_2): Ditto.
6449 (*avx2_zero_extendv8hiv8si2_2): Ditto.
6450 (*sse4_1_zero_extendv4hiv4si2_4): Ditto.
6451 (*avx512f_zero_extendv8siv8di2_2): Ditto.
6452 (*avx2_zero_extendv4siv4di2_2): Ditto.
6453 (*sse4_1_zero_extendv2siv2di2_4): Ditto.
6454 (VI248_256, VI248_512, VI148_512, VI148_256, VI148_128): New
6457 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
6459 * config/rs6000/rs6000-builtin-new.def: Add always, power5, and
6462 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
6464 * config/rs6000/rs6000-builtin-new.def: Add vsx stanza.
6466 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
6468 * config/rs6000/rs6000-builtin-new.def: Finish altivec stanza.
6469 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Move
6470 initialization of pcvoid_type_node here...
6471 (altivec_init_builtins): ...from here.
6472 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add
6473 RS6000_BTI_const_ptr_void.
6474 (pcvoid_type_node): New macro.
6476 2021-08-11 Richard Biener <rguenther@suse.de>
6479 * tree-ssa-forwprop.c (pass_forwprop::execute): Do not decompose
6480 hard-register accesses.
6482 2021-08-11 Richard Biener <rguenther@suse.de>
6484 * tree-ssa-operands.c (operands_scanner::get_expr_operands):
6485 Do not look at COMPONENT_REF FIELD_DECLs TREE_THIS_VOLATILE
6486 to determine has_volatile_ops.
6488 2021-08-11 Eric Botcazou <ebotcazou@gcc.gnu.org>
6490 * cfgexpand.c (expand_used_vars): Reuse attribs local variable.
6492 2021-08-11 Jan Hubicka <hubicka@ucw.cz>
6493 Alexandre Oliva <oliva@adacore.com>
6495 * ipa-modref.c (modref_lattice::dump): Fix escape_point's min_flags
6497 (modref_lattice::merge_deref): Fix handling of indirect scape points.
6498 (update_escape_summary_1): Likewise.
6499 (update_escape_summary): Likewise.
6500 (ipa_merge_modref_summary_after_inlining): Likewise.
6502 2021-08-11 Richard Biener <rguenther@suse.de>
6504 PR middle-end/101858
6505 * fold-const.c (fold_binary_loc): Guard simplification
6506 of X < (cast) (1 << Y) to integer types.
6508 2021-08-11 Richard Biener <rguenther@suse.de>
6510 PR tree-optimization/101861
6511 * tree-vect-stmts.c (vectorizable_load): Fix error in
6512 previous change with regard to gather vectorization.
6514 2021-08-11 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
6517 * config/arm/arm_neon.h (vdup_n_s8): Replace call to builtin
6519 (vdup_n_s16): Likewise.
6520 (vdup_n_s32): Likewise.
6521 (vdup_n_s64): Likewise.
6522 (vdup_n_u8): Likewise.
6523 (vdup_n_u16): Likewise.
6524 (vdup_n_u32): Likewise.
6525 (vdup_n_u64): Likewise.
6526 (vdup_n_p8): Likewise.
6527 (vdup_n_p16): Likewise.
6528 (vdup_n_p64): Likewise.
6529 (vdup_n_f16): Likewise.
6530 (vdup_n_f32): Likewise.
6531 (vdupq_n_s8): Likewise.
6532 (vdupq_n_s16): Likewise.
6533 (vdupq_n_s32): Likewise.
6534 (vdupq_n_s64): Likewise.
6535 (vdupq_n_u8): Likewise.
6536 (vdupq_n_u16): Likewise.
6537 (vdupq_n_u32): Likewise.
6538 (vdupq_n_u64): Likewise.
6539 (vdupq_n_p8): Likewise.
6540 (vdupq_n_p16): Likewise.
6541 (vdupq_n_p64): Likewise.
6542 (vdupq_n_f16): Likewise.
6543 (vdupq_n_f32): Likewise.
6544 (vmov_n_s8): Replace call to builtin with call to corresponding
6546 (vmov_n_s16): Likewise.
6547 (vmov_n_s32): Likewise.
6548 (vmov_n_s64): Likewise.
6549 (vmov_n_u8): Likewise.
6550 (vmov_n_u16): Likewise.
6551 (vmov_n_u32): Likewise.
6552 (vmov_n_u64): Likewise.
6553 (vmov_n_p8): Likewise.
6554 (vmov_n_p16): Likewise.
6555 (vmov_n_f16): Likewise.
6556 (vmov_n_f32): Likewise.
6557 (vmovq_n_s8): Likewise.
6558 (vmovq_n_s16): Likewise.
6559 (vmovq_n_s32): Likewise.
6560 (vmovq_n_s64): Likewise.
6561 (vmovq_n_u8): Likewise.
6562 (vmovq_n_u16): Likewise.
6563 (vmovq_n_u32): Likewise.
6564 (vmovq_n_u64): Likewise.
6565 (vmovq_n_p8): Likewise.
6566 (vmovq_n_p16): Likewise.
6567 (vmovq_n_f16): Likewise.
6568 (vmovq_n_f32): Likewise.
6569 * config/arm/arm_neon_builtins.def: Remove entries for vdup_n.
6571 2021-08-11 liuhongt <hongtao.liu@intel.com>
6574 * config/i386/i386.md (ldexp<mode>3): Extend to vscalefs[sd]
6575 when TARGET_AVX512F and TARGET_SSE_MATH.
6577 2021-08-10 Jakub Jelinek <jakub@redhat.com>
6580 * config/i386/i386-expand.c (expand_vec_perm_even_odd): Return false
6581 for V32HImode if !TARGET_AVX512BW.
6582 (ix86_vectorize_vec_perm_const) <case E_V32HImode, case E_V64QImode>:
6583 If !TARGET_AVX512BW and TARGET_AVX512F and d.testing_p, don't fail
6584 early, but actually check the permutation.
6586 2021-08-10 Richard Biener <rguenther@suse.de>
6588 PR tree-optimization/101809
6589 * tree-vect-stmts.c (get_load_store_type): Allow emulated
6590 gathers with offset vector nunits being a constant multiple
6591 of the data vector nunits.
6592 (vect_get_gather_scatter_ops): Use the appropriate nunits
6593 for the offset vector defs.
6594 (vectorizable_store): Adjust call to
6595 vect_get_gather_scatter_ops.
6596 (vectorizable_load): Likewise. Handle the case of less
6597 offset vectors than data vectors.
6599 2021-08-10 Jakub Jelinek <jakub@redhat.com>
6602 * config/i386/sse.md (*avx512f_shuf_<shuffletype>64x2_1<mask_name>_1,
6603 *avx512f_shuf_<shuffletype>32x4_1<mask_name>_1): New define_insn
6606 2021-08-10 Richard Biener <rguenther@suse.de>
6608 PR tree-optimization/101801
6609 PR tree-optimization/101819
6610 * tree-vectorizer.h (vect_emulated_vector_p): Declare.
6611 * tree-vect-loop.c (vect_emulated_vector_p): New function.
6612 (vectorizable_reduction): Re-instantiate a check for emulated
6614 * tree-vect-stmts.c (vectorizable_shift): Likewise.
6615 (vectorizable_operation): Likewise. Cost emulated vector
6616 operations according to the scalar sequence synthesized by
6619 2021-08-10 Richard Biener <rguenther@suse.de>
6621 PR middle-end/101824
6622 * tree-nested.c (get_frame_field): Mark the COMPONENT_REF as
6623 volatile in case the variable was.
6625 2021-08-10 H.J. Lu <hjl.tools@gmail.com>
6628 * config/i386/constraints.md (BC): Document for integer SSE
6629 constant all bits set operand.
6630 (BF): New constraint for const floating-point all bits set
6632 * config/i386/i386.c (standard_sse_constant_p): Likewise.
6633 (standard_sse_constant_opcode): Likewise.
6634 * config/i386/sse.md (sseconstm1): New mode attribute.
6635 (mov<mode>_internal): Replace BC with <sseconstm1>.
6637 2021-08-10 liuhongt <hongtao.liu@intel.com>
6639 * config/i386/sse.md (cond_<insn><mode>): New expander.
6640 (VI248_AVX512VLBW): New mode iterator.
6641 * config/i386/predicates.md
6642 (nonimmediate_or_const_vec_dup_operand): New predicate.
6644 2021-08-09 Andrew MacLeod <amacleod@redhat.com>
6646 PR tree-optimization/101741
6647 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Check
6648 type of parameter for toupper/tolower.
6650 2021-08-09 Martin Jambor <mjambor@suse.cz>
6653 * ipa-prop.c (propagate_controlled_uses): Removed a spurious space.
6655 2021-08-09 Pat Haugen <pthaugen@linux.ibm.com>
6657 * config/rs6000/rs6000.c (is_load_insn1): Verify destination is a
6659 (is_store_insn1): Verify source is a register.
6661 2021-08-09 Uroš Bizjak <ubizjak@gmail.com>
6664 * config/i386/mmx.md (<any_logic:code>v2sf3):
6665 Rename from *mmx_<any_logic:code>v2sf3
6667 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6669 * config/nvptx/nvptx.c: Cross-reference parts adapted in
6670 'gcc/omp-oacc-neuter-broadcast.cc'.
6671 * omp-low.c: Likewise.
6672 * omp-oacc-neuter-broadcast.cc: Cross-reference parts adapted from
6675 2021-08-09 Julian Brown <julian@codesourcery.com>
6676 Kwok Cheung Yeung <kcy@codesourcery.com>
6677 Thomas Schwinge <thomas@codesourcery.com>
6679 * config/gcn/gcn.c (gcn_init_builtins): Override decls for
6680 BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START,
6681 BUILT_IN_GOACC_SINGLE_COPY_END and BUILT_IN_GOACC_BARRIER.
6682 (gcn_goacc_validate_dims): Turn on worker partitioning unconditionally.
6683 (gcn_fork_join): Update comment.
6684 * config/gcn/gcn.opt (flag_worker_partitioning): Remove.
6685 (macc_experimental_workers): Remove unused option.
6687 2021-08-09 Julian Brown <julian@codesourcery.com>
6688 Nathan Sidwell <nathan@codesourcery.com> (via 'gcc/config/nvptx/nvptx.c' master)
6689 Kwok Cheung Yeung <kcy@codesourcery.com>
6690 Thomas Schwinge <thomas@codesourcery.com>
6692 * Makefile.in (OBJS): Add omp-oacc-neuter-broadcast.o.
6693 * doc/tm.texi.in (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD):
6694 Add documentation hook.
6695 * doc/tm.texi: Regenerate.
6696 * omp-oacc-neuter-broadcast.cc: New file.
6697 * omp-builtins.def (BUILT_IN_GOACC_BARRIER)
6698 (BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START)
6699 (BUILT_IN_GOACC_SINGLE_COPY_END): New builtins.
6700 * passes.def (pass_omp_oacc_neuter_broadcast): Add pass.
6701 * target.def (goacc.create_worker_broadcast_record): Add target
6703 * tree-pass.h (make_pass_omp_oacc_neuter_broadcast): Add
6705 * config/gcn/gcn-protos.h (gcn_goacc_adjust_propagation_record):
6706 Rename prototype to...
6707 (gcn_goacc_create_worker_broadcast_record): ... this.
6708 * config/gcn/gcn-tree.c (gcn_goacc_adjust_propagation_record): Rename
6710 (gcn_goacc_create_worker_broadcast_record): ... this.
6711 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_PROPAGATION_RECORD):
6713 (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD): ... this.
6715 2021-08-09 Tejas Belagod <tejas.belagod@arm.com>
6718 * config/aarch64/aarch64-simd.md (vlshr<mode>3, vashr<mode>3): Use
6721 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6723 * Makefile.in (GTFILES): Remove '$(srcdir)/omp-offload.c'.
6725 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6727 * builtins.def (DEF_GOACC_BUILTIN, DEF_GOMP_BUILTIN): Don't
6728 consider '-foffload-abi'.
6729 * common.opt (-foffload-abi): Remove 'Var', 'Init'.
6730 * opts.c (common_handle_option) <-foffload-abi> [ACCEL_COMPILER]:
6733 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6735 * optc-gen.awk: Sanity check that 'Init' doesn't appear without
6738 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6740 * omp-builtins.def (BUILT_IN_ACC_GET_DEVICE_TYPE): Remove.
6742 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6744 * doc/gty.texi (Files): Update.
6746 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6748 * doc/gty.texi (Files): Fix GTY header file example.
6750 2021-08-09 Roger Sayle <roger@nextmovesoftware.com>
6752 * tree-ssa-ccp.c (value_mask_to_min_max): Helper function to
6753 determine the upper and lower bounds from a mask-value pair.
6754 (bit_value_unop) [ABS_EXPR, ABSU_EXPR]: Add support for
6755 absolute value and unsigned absolute value expressions.
6756 (bit_value_binop): Initialize *VAL's precision.
6757 [LT_EXPR, LE_EXPR]: Use value_mask_to_min_max to determine
6758 upper and lower bounds of operands. Add LE_EXPR/GE_EXPR
6759 support when the operands are unknown but potentially equal.
6760 [MIN_EXPR, MAX_EXPR]: Support minimum/maximum expressions.
6762 2021-08-09 Bin Cheng <bin.cheng@linux.alibaba.com>
6764 * config/aarch64/aarch64.md
6765 (*extend<SHORT:mode><GPI:mode>2_aarch64): Use %<GPI:w>0.
6767 2021-08-08 Sergei Trofimovich <siarheit@google.com>
6769 * lra-constraints.c: Fix s/otput/output/ typo.
6771 2021-08-06 Martin Sebor <msebor@redhat.com>
6773 * builtins.c (expand_builtin_memchr): Move to gimple-ssa-warn-access.cc.
6774 (expand_builtin_strcat): Same.
6775 (expand_builtin_stpncpy): Same.
6776 (expand_builtin_strncat): Same.
6777 (check_read_access): Same.
6778 (check_memop_access): Same.
6779 (expand_builtin_strlen): Move checks to gimple-ssa-warn-access.cc.
6780 (expand_builtin_strnlen): Same.
6781 (expand_builtin_memcpy): Same.
6782 (expand_builtin_memmove): Same.
6783 (expand_builtin_mempcpy): Same.
6784 (expand_builtin_strcpy): Same.
6785 (expand_builtin_strcpy_args): Same.
6786 (expand_builtin_stpcpy_1): Same.
6787 (expand_builtin_strncpy): Same.
6788 (expand_builtin_memset): Same.
6789 (expand_builtin_bzero): Same.
6790 (expand_builtin_strcmp): Same.
6791 (expand_builtin_strncmp): Same.
6792 (expand_builtin): Remove handlers.
6793 (fold_builtin_strlen): Add a comment.
6794 * builtins.h (check_access): Move to gimple-ssa-warn-access.cc.
6795 * calls.c (maybe_warn_nonstring_arg): Same.
6796 * diagnostic-spec.c (nowarn_spec_t::nowarn_spec_t): Add warning option.
6797 * gimple-fold.c (gimple_fold_builtin_strcpy): Pass argument to callee.
6798 (gimple_fold_builtin_stpcpy): Same.
6799 * gimple-ssa-warn-access.cc (has_location): New function.
6800 (get_location): Same.
6801 (get_callee_fndecl): Same.
6804 (warn_string_no_nul): Define.
6805 (unterminated_array): Same.
6806 (check_nul_terminated_array): Same.
6807 (maybe_warn_nonstring_arg): Same.
6808 (maybe_warn_for_bound): Same.
6809 (warn_for_access): Same.
6810 (check_access): Same.
6811 (check_memop_access): Same.
6812 (check_read_access): Same.
6813 (warn_dealloc_offset): Use helper functions.
6814 (maybe_emit_free_warning): Same.
6815 (class pass_waccess): Add members.
6816 (check_strcat): New function.
6817 (check_strncat): New function.
6818 (check_stxcpy): New function.
6819 (check_stxncpy): New function.
6820 (check_strncmp): New function.
6821 (pass_waccess::check_builtin): New function.
6822 (pass_waccess::check): Call it.
6823 * gimple-ssa-warn-access.h (warn_string_no_nul): Move here from
6825 (maybe_warn_for_bound): Same.
6826 (check_access): Same.
6827 (check_memop_access): Same.
6828 (check_read_access): Same.
6829 * pointer-query.h (struct access_data): Define a ctor overload.
6831 2021-08-06 Richard Biener <rguenther@suse.de>
6833 PR tree-optimization/101801
6834 * tree-vectorizer.h (vect_worthwhile_without_simd_p): Rename...
6835 (vect_can_vectorize_without_simd_p): ... to this.
6836 * tree-vect-loop.c (vect_worthwhile_without_simd_p): Rename...
6837 (vect_can_vectorize_without_simd_p): ... to this and fold
6838 in vect_min_worthwhile_factor.
6839 (vect_min_worthwhile_factor): Remove.
6840 (vectorizable_reduction): Adjust and remove the cost part.
6841 * tree-vect-stmts.c (vectorizable_shift): Likewise.
6842 (vectorizable_operation): Likewise.
6844 2021-08-06 Uroš Bizjak <ubizjak@gmail.com>
6847 * config/i386/i386.md (cmove reg-to-reg move elimination peephole2s):
6848 Add general_gr_operand predicate to operand 3.
6850 2021-08-06 Roger Sayle <roger@nextmovesoftware.com>
6852 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern): Use
6853 CFN_BUILT_IN_CLRSB* instead of BUILT_IN_CLRSB* for consistency.
6855 2021-08-06 Tamar Christina <tamar.christina@arm.com>
6857 * config/aarch64/aarch64-sve-builtins.cc (register_svpattern,
6858 register_svprfop): Pass vec<> by pointer.
6859 * langhooks-def.h (lhd_simulate_enum_decl): Likewise.
6860 * langhooks.c (lhd_simulate_enum_decl): Likewise.
6861 * langhooks.h (struct lang_hooks_for_types): Likewise.
6863 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
6865 * config/aarch64/arm_neon.h (vst1_bf16_x2): Use
6866 __builtin_memcpy instead of constructing an additional
6867 __builtin_aarch64_simd_oi one vector at a time.
6868 (vst1q_bf16_x2): Likewise.
6869 (vst1_bf16_x3): Use __builtin_memcpy instead of constructing
6870 an additional __builtin_aarch64_simd_ci one vector at a time.
6871 (vst1q_bf16_x3): Likewise.
6872 (vst1_bf16_x4): Use __builtin_memcpy instead of a union.
6873 (vst1q_bf16_x4): Likewise.
6874 (vst2_bf16): Use __builtin_memcpy instead of constructing an
6875 additional __builtin_aarch64_simd_oi one vector at a time.
6876 (vst2q_bf16): Likewise.
6877 (vst3_bf16): Use __builtin_memcpy instead of constructing an
6878 additional __builtin_aarch64_simd_ci mode one vector at a
6880 (vst3q_bf16): Likewise.
6881 (vst4_bf16): Use __builtin_memcpy instead of constructing an
6882 additional __builtin_aarch64_simd_xi one vector at a time.
6883 (vst4q_bf16): Likewise.
6885 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
6887 * config/aarch64/arm_neon.h (__ST2_LANE_FUNC): Delete.
6888 (__ST2Q_LANE_FUNC): Delete.
6889 (vst2_lane_f16): Use __builtin_memcpy to copy vector
6890 structure instead of constructing __builtin_aarch64_simd_oi
6891 one vector at a time.
6892 (vst2_lane_f32): Likewise.
6893 (vst2_lane_f64): Likewise.
6894 (vst2_lane_p8): Likewise.
6895 (vst2_lane_p16): Likewise.
6896 (vst2_lane_p64): Likewise.
6897 (vst2_lane_s8): Likewise.
6898 (vst2_lane_s16): Likewise.
6899 (vst2_lane_s32): Likewise.
6900 (vst2_lane_s64): Likewise.
6901 (vst2_lane_u8): Likewise.
6902 (vst2_lane_u16): Likewise.
6903 (vst2_lane_u32): Likewise.
6904 (vst2_lane_u64): Likewise.
6905 (vst2_lane_bf16): Likewise.
6906 (vst2q_lane_f16): Use __builtin_memcpy to copy vector
6907 structure instead of using a union.
6908 (vst2q_lane_f32): Likewise.
6909 (vst2q_lane_f64): Likewise.
6910 (vst2q_lane_p8): Likewise.
6911 (vst2q_lane_p16): Likewise.
6912 (vst2q_lane_p64): Likewise.
6913 (vst2q_lane_s8): Likewise.
6914 (vst2q_lane_s16): Likewise.
6915 (vst2q_lane_s32): Likewise.
6916 (vst2q_lane_s64): Likewise.
6917 (vst2q_lane_u8): Likewise.
6918 (vst2q_lane_u16): Likewise.
6919 (vst2q_lane_u32): Likewise.
6920 (vst2q_lane_u64): Likewise.
6921 (vst2q_lane_bf16): Likewise.
6923 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
6925 * config/aarch64/arm_neon.h (__ST3_LANE_FUNC): Delete.
6926 (__ST3Q_LANE_FUNC): Delete.
6927 (vst3_lane_f16): Use __builtin_memcpy to copy vector
6928 structure instead of constructing __builtin_aarch64_simd_ci
6929 one vector at a time.
6930 (vst3_lane_f32): Likewise.
6931 (vst3_lane_f64): Likewise.
6932 (vst3_lane_p8): Likewise.
6933 (vst3_lane_p16): Likewise.
6934 (vst3_lane_p64): Likewise.
6935 (vst3_lane_s8): Likewise.
6936 (vst3_lane_s16): Likewise.
6937 (vst3_lane_s32): Likewise.
6938 (vst3_lane_s64): Likewise.
6939 (vst3_lane_u8): Likewise.
6940 (vst3_lane_u16): Likewise.
6941 (vst3_lane_u32): Likewise.
6942 (vst3_lane_u64): Likewise.
6943 (vst3_lane_bf16): Likewise.
6944 (vst3q_lane_f16): Use __builtin_memcpy to copy vector
6945 structure instead of using a union.
6946 (vst3q_lane_f32): Likewise.
6947 (vst3q_lane_f64): Likewise.
6948 (vst3q_lane_p8): Likewise.
6949 (vst3q_lane_p16): Likewise.
6950 (vst3q_lane_p64): Likewise.
6951 (vst3q_lane_s8): Likewise.
6952 (vst3q_lane_s16): Likewise.
6953 (vst3q_lane_s32): Likewise.
6954 (vst3q_lane_s64): Likewise.
6955 (vst3q_lane_u8): Likewise.
6956 (vst3q_lane_u16): Likewise.
6957 (vst3q_lane_u32): Likewise.
6958 (vst3q_lane_u64): Likewise.
6959 (vst3q_lane_bf16): Likewise.
6961 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
6963 * config/aarch64/arm_neon.h (__ST4_LANE_FUNC): Delete.
6964 (__ST4Q_LANE_FUNC): Delete.
6965 (vst4_lane_f16): Use __builtin_memcpy to copy vector
6966 structure instead of constructing __builtin_aarch64_simd_xi
6967 one vector at a time.
6968 (vst4_lane_f32): Likewise.
6969 (vst4_lane_f64): Likewise.
6970 (vst4_lane_p8): Likewise.
6971 (vst4_lane_p16): Likewise.
6972 (vst4_lane_p64): Likewise.
6973 (vst4_lane_s8): Likewise.
6974 (vst4_lane_s16): Likewise.
6975 (vst4_lane_s32): Likewise.
6976 (vst4_lane_s64): Likewise.
6977 (vst4_lane_u8): Likewise.
6978 (vst4_lane_u16): Likewise.
6979 (vst4_lane_u32): Likewise.
6980 (vst4_lane_u64): Likewise.
6981 (vst4_lane_bf16): Likewise.
6982 (vst4q_lane_f16): Use __builtin_memcpy to copy vector
6983 structure instead of using a union.
6984 (vst4q_lane_f32): Likewise.
6985 (vst4q_lane_f64): Likewise.
6986 (vst4q_lane_p8): Likewise.
6987 (vst4q_lane_p16): Likewise.
6988 (vst4q_lane_p64): Likewise.
6989 (vst4q_lane_s8): Likewise.
6990 (vst4q_lane_s16): Likewise.
6991 (vst4q_lane_s32): Likewise.
6992 (vst4q_lane_s64): Likewise.
6993 (vst4q_lane_u8): Likewise.
6994 (vst4q_lane_u16): Likewise.
6995 (vst4q_lane_u32): Likewise.
6996 (vst4q_lane_u64): Likewise.
6997 (vst4q_lane_bf16): Likewise.
6999 2021-08-06 Martin Liska <mliska@suse.cz>
7001 * config/rs6000/rs6000.c (rs6000_option_override_internal): When
7002 a target option is restored, it can have
7003 rs6000_long_double_type_size set to FLOAT_PRECISION_TFmode
7004 and error should not be emitted.
7006 2021-08-06 Sebastian Huber <sebastian.huber@embedded-brains.de>
7008 * gcov-io.h (gcov_write): Declare.
7009 * gcov-io.c (gcov_write): New.
7010 (gcov_write_counter): Remove.
7011 (gcov_write_tag_length): Likewise.
7012 (gcov_write_summary): Replace gcov_write_tag_length() with calls to
7013 gcov_write_unsigned().
7014 * doc/invoke.texi (fprofile-info-section): Mention
7015 __gcov_info_to_gdca().
7017 2021-08-06 Martin Sebor <msebor@redhat.com>
7019 * dominance.c (prune_bbs_to_update_dominators): Adjust by-value vec
7020 arguments to by-reference.
7021 (iterate_fix_dominators): Same.
7022 * dominance.h (iterate_fix_dominators): Same.
7023 * ipa-prop.h: Call auto_vec::to_vec_legacy.
7024 * tree-data-ref.c (dump_data_dependence_relation): Adjust by-value vec
7025 arguments to by-reference.
7026 (debug_data_dependence_relation): Same.
7027 (dump_data_dependence_relations): Same.
7028 * tree-data-ref.h (debug_data_dependence_relation): Same.
7029 (dump_data_dependence_relations): Same.
7030 * tree-predcom.c (dump_chains): Same.
7031 (initialize_root_vars_lm): Same.
7032 (determine_unroll_factor): Same.
7033 (replace_phis_by_defined_names): Same.
7034 (insert_init_seqs): Same.
7035 (pcom_worker::tree_predictive_commoning_loop): Call
7036 auto_vec::to_vec_legacy.
7037 * tree-ssa-pre.c (insert_into_preds_of_block): Adjust by-value vec
7038 arguments to by-reference.
7039 * tree-ssa-threadbackward.c (populate_worklist): Same.
7040 (back_threader::resolve_def): Same.
7041 * tree-vect-data-refs.c (vect_check_nonzero_value): Same.
7042 (vect_enhance_data_refs_alignment): Same.
7043 (vect_check_lower_bound): Same.
7044 (vect_prune_runtime_alias_test_list): Same.
7045 (vect_permute_store_chain): Same.
7046 * tree-vect-slp-patterns.c (vect_normalize_conj_loc): Same.
7047 * tree-vect-stmts.c (vect_create_vectorized_demotion_stmts): Same.
7048 * tree-vectorizer.h (vect_permute_store_chain): Same.
7049 * vec.c (test_init): New function.
7050 (vec_c_tests): Call new function.
7051 * vec.h (vec): Declare ctors, dtor, and assignment.
7052 (auto_vec::vec_to_legacy): New function.
7053 (vec::copy): Adjust initialization.
7055 2021-08-05 H.J. Lu <hjl.tools@gmail.com>
7058 * config/i386/i386.c (ix86_can_inline_p): Ignore MASK_80387 if
7059 callee only uses GPRs.
7060 * config/i386/ia32intrin.h: Revert commit 5463cee2770.
7061 * config/i386/serializeintrin.h: Revert commit 71958f740f1.
7062 * config/i386/x86gprintrin.h: Add
7063 #pragma GCC target("general-regs-only") and #pragma GCC pop_options
7064 to disable non-GPR ISAs.
7066 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
7068 PR middle-end/101787
7069 * doc/md.texi (cond_ashl, cond_ashr, cond_lshr): Document.
7071 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
7073 * tree-vectorizer.h (vect_is_store_elt_extraction, vect_is_reduction)
7074 (vect_reduc_type, vect_embedded_comparison_type, vect_comparison_type)
7075 (vect_is_extending_load, vect_is_integer_truncation): New functions,
7076 moved from aarch64.c but given different names.
7077 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction)
7078 (aarch64_is_reduction, aarch64_reduc_type)
7079 (aarch64_embedded_comparison_type, aarch64_comparison_type)
7080 (aarch64_extending_load_p, aarch64_integer_truncation_p): Delete
7081 in favor of the above. Update callers accordingly.
7083 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
7086 * config/arm/arm-cpus.in (generic-armv7-a): Add quirk to suppress
7087 writing .cpu directive in asm output.
7088 * config/arm/arm.c (arm_identify_fpu_from_isa): New variable.
7089 (arm_last_printed_arch_string): Delete.
7090 (arm_last-printed_fpu_string): Delete.
7091 (arm_configure_build_target): If use of floating-point/SIMD is
7092 disabled, remove all fp/simd related features from the target ISA.
7093 (last_arm_targ_options): New variable.
7094 (arm_print_asm_arch_directives): Add new parameters. Change order
7095 of emitted directives and handle all cases here.
7096 (arm_file_start): Always call arm_print_asm_arch_directives, move
7097 all generation of .arch/.arch_extension here.
7098 (arm_file_end): Call arm_print_asm_arch.
7099 (arm_declare_function_name): Call arm_print_asm_arch_directives
7100 instead of printing .arch/.fpu directives directly.
7102 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
7104 * config/arm/arm.c (arm_configure_build_target): Don't call
7105 arm_option_reconfigure_globals.
7106 (arm_option_restore): Call arm_option_reconfigure_globals after
7107 reconfiguring the target.
7108 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
7110 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
7112 * config/arm/arm.c (arm_configure_build_target): Ensure the target's
7113 arch_name is always set.
7115 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
7117 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
7118 of vec_select high-half from being added into Neon subtract
7121 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
7123 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
7124 of vec_select high-half from being added into Neon add cost.
7126 2021-08-05 Kewen Lin <linkw@linux.ibm.com>
7128 * cfgloop.h (loops_list::loops_list): Add one optional argument
7129 root and adjust accordingly, update loop tree walking and factor
7131 * cfgloop.c (loops_list::walk_loop_tree): ... this. New function.
7133 2021-08-05 Eric Botcazou <ebotcazou@gcc.gnu.org>
7135 PR tree-optimization/101626
7136 * tree-sra.c (propagate_subaccesses_from_rhs): Do not set the
7137 reverse scalar storage order on a pointer or vector component.
7139 2021-08-05 liuhongt <hongtao.liu@intel.com>
7141 * config/i386/sse.md (cond_<code><mode>): New expander.
7143 2021-08-05 liuhongt <hongtao.liu@intel.com>
7145 * config/i386/sse.md (cond_<code><mode>): New expander.
7147 2021-08-05 liuhongt <hongtao.liu@intel.com>
7149 * config/i386/sse.md (cond_<code><mode>): New expander.
7151 2021-08-04 David Malcolm <dmalcolm@redhat.com>
7154 * Makefile.in (ANALYZER_OBJS): Add analyzer/region-model-asm.o.
7156 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
7159 * config/i386/i386.h (STORE_MAX_PIECES): Allow 16/32/64 bytes
7160 only if TARGET_INTER_UNIT_MOVES_TO_VEC is true.
7162 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
7165 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
7166 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
7167 data with SSE register from one memory location to another.
7169 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
7171 * config/s390/s390.c (expand_perm_with_vpdi): New function.
7172 (vectorize_vec_perm_const_1): Call expand_perm_with_vpdi.
7173 * config/s390/vector.md (*vpdi1<mode>, @vpdi1<mode>): Enable a
7174 parameterized expander.
7175 (*vpdi4<mode>, @vpdi4<mode>): Likewise.
7177 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
7179 * config/s390/s390.c (MAX_VECT_LEN): Define macro.
7180 (struct expand_vec_perm_d): Define struct.
7181 (expand_perm_with_merge): New function.
7182 (vectorize_vec_perm_const_1): New function.
7183 (s390_vectorize_vec_perm_const): New function.
7184 (TARGET_VECTORIZE_VEC_PERM_CONST): Define target macro.
7186 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
7188 * config/s390/vector.md (V_HW_64): Remove mode iterator.
7189 (*vec_load_pair<mode>): Use V_HW_2 instead of V_HW_64.
7190 * config/s390/vx-builtins.md
7191 (vec_scatter_element<V_HW_2:mode>_SI): Use V_HW_2 instead of
7194 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
7196 * config/s390/s390.md (UNSPEC_VEC_PERMI): Remove constant
7198 * config/s390/vector.md (*vpdi1<mode>, *vpdi4<mode>): New pattern
7200 * config/s390/vx-builtins.md (*vec_permi<mode>): Emit generic rtx
7201 instead of an unspec.
7203 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
7205 * config/s390/s390-modes.def: Add more vector modes to support
7206 concatenation of two vectors.
7207 * config/s390/s390-protos.h (s390_expand_merge_perm_const): Add
7209 (s390_expand_merge): Likewise.
7210 * config/s390/s390.c (s390_expand_merge_perm_const): New function.
7211 (s390_expand_merge): New function.
7212 * config/s390/s390.md (UNSPEC_VEC_MERGEH, UNSPEC_VEC_MERGEL):
7213 Remove constant definitions.
7214 * config/s390/vector.md (V_HW_2): Add mode iterators.
7215 (VI_HW_4, V_HW_4): Rename VI_HW_4 to V_HW_4.
7216 (vec_2x_nelts, vec_2x_wide): New mode attributes.
7217 (*vmrhb, *vmrlb, *vmrhh, *vmrlh, *vmrhf, *vmrlf, *vmrhg, *vmrlg):
7218 New pattern definitions.
7219 (vec_widen_umult_lo_<mode>, vec_widen_umult_hi_<mode>)
7220 (vec_widen_smult_lo_<mode>, vec_widen_smult_hi_<mode>)
7221 (vec_unpacks_lo_v4sf, vec_unpacks_hi_v4sf, vec_unpacks_lo_v2df)
7222 (vec_unpacks_hi_v2df): Adjust expanders to emit non-unspec RTX for
7224 * config/s390/vx-builtins.md (V_HW_4): Remove mode iterator. Now
7226 (vec_mergeh<mode>, vec_mergel<mode>): Use s390_expand_merge to
7227 emit vec merge pattern.
7229 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
7231 * config/aarch64/aarch64.c (aarch64_strip_extend_vec_half):
7233 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent cost of
7234 vec_select high-half from being added into Neon multiply
7236 * rtlanal.c (vec_series_highpart_p): Define.
7237 * rtlanal.h (vec_series_highpart_p): Declare.
7239 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
7241 * config/aarch64/aarch64.c (aarch64_strip_duplicate_vec_elt):
7243 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent
7244 vec_select cost from being added into Neon multiply cost.
7246 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
7248 * tree-vect-loop.c (vect_better_loop_vinfo_p): Detect cases in
7249 which old_loop_vinfo is an epilogue loop that handles a constant
7250 number of iterations.
7252 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
7254 * tree-vect-loop.c (vect_analyze_loop): Print a dump message
7255 when a reanalyzed loop fails to be cheaper than the current
7258 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
7260 * config/aarch64/aarch64.c: Fix a typo.
7262 2021-08-04 Vincent Lefèvre <vincent-gcc@vinc17.net>
7264 PR gcov-profile/101773
7265 * gcov-io.c (gcov_close): Check return code of a fclose.
7267 2021-08-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
7270 * dwarf2out.c (dwarf2out_assembly_start): Emit a dummy
7271 .file statement when needed.
7273 2021-08-04 Richard Biener <rguenther@suse.de>
7275 * tree-vect-data-refs.c (vect_check_gather_scatter):
7276 Include widening conversions only when the result is
7277 still handed by native gather or the current offset
7278 size not already matches the data size.
7279 Also succeed analysis in case there's no native support,
7280 noted by a IFN_LAST ifn and a NULL decl.
7281 (vect_analyze_data_refs): Always consider gathers.
7282 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
7283 Test for no IFN gather rather than decl gather.
7284 * tree-vect-stmts.c (vect_model_load_cost): Pass in the
7285 gather-scatter info and cost emulated gathers accordingly.
7286 (vect_truncate_gather_scatter_offset): Properly test for
7288 (vect_use_strided_gather_scatters_p): Likewise.
7289 (get_load_store_type): Handle emulated gathers and its
7291 (vectorizable_load): Likewise. Emulate them by extracting
7292 scalar offsets, doing scalar loads and a vector construct.
7294 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
7297 * expr.c (op_by_pieces_d::op_by_pieces_d): Add a max_pieces
7298 argument to set m_max_size.
7299 (move_by_pieces_d): Pass MOVE_MAX_PIECES to op_by_pieces_d.
7300 (store_by_pieces_d): Pass STORE_MAX_PIECES to op_by_pieces_d.
7301 (compare_by_pieces_d): Pass COMPARE_MAX_PIECES to op_by_pieces_d.
7303 2021-08-04 Roger Sayle <roger@nextmovesoftware.com>
7304 Marc Glisse <marc.glisse@inria.fr>
7306 * match.pd (bit_ior, bit_xor): Canonicalize (X*C1)|(X*C2) and
7307 (X*C1)^(X*C2) as X*(C1+C2), and related variants, using
7308 tree_nonzero_bits to ensure that operands are bit-wise disjoint.
7310 2021-08-04 Richard Biener <rguenther@suse.de>
7312 * tree-ssa-forwprop.c (pass_forwprop::execute): Split
7313 out code to decompose vector loads ...
7314 (optimize_vector_load): ... here. Generalize it to
7315 handle intermediate widening and TARGET_MEM_REF loads
7316 and apply it to loads with a supported vector mode as well.
7318 2021-08-04 Richard Biener <rguenther@suse.de>
7320 PR tree-optimization/101756
7321 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Make sure
7322 the result of the reduction epilogue is compatible to the original
7325 2021-08-04 liuhongt <hongtao.liu@intel.com>
7328 * config/i386/i386.md (peephole2): Refine predicate from
7329 register_operand to general_reg_operand.
7331 2021-08-04 Aldy Hernandez <aldyh@redhat.com>
7333 * gimple-range-path.h (path_range_query::dump): Mark override.
7335 2021-08-04 Richard Biener <rguenther@suse.de>
7337 PR tree-optimization/101769
7338 * tree-tailcall.c (eliminate_tail_call): Add the created loop
7339 for the first recursion and return it via the new output parameter.
7340 (optimize_tail_call): Pass through new output param.
7341 (tree_optimize_tail_calls_1): After creating all latches,
7342 add the created loop to the loop tree. Do not mark loops for fixup.
7344 2021-08-04 Martin Liska <mliska@suse.cz>
7346 * doc/invoke.texi: Document threader-mode param.
7348 2021-08-04 liuhongt <hongtao.liu@intel.com>
7350 * config/i386/sse.md (cond_fma<mode>): New expander.
7351 (cond_fms<mode>): Ditto.
7352 (cond_fnma<mode>): Ditto.
7353 (cond_fnms<mode>): Ditto.
7355 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
7357 * config/rs6000/vsx.md (*vsx_le_perm_store_<mode>): Use && instead of &.
7359 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
7361 * config/rs6000/constraints.md: Remove "e" from the list of available
7362 constraint characters.
7364 2021-08-03 Eugene Rozenfeld <erozen@microsoft.com>
7366 PR gcov-profile/71672
7367 * auto-profile.c (afdo_indirect_call): Fix setup of the historgram value for indirect calls.
7369 2021-08-03 Paul A. Clarke <pc@us.ibm.com>
7371 * config/rs6000/smmintrin.h (_mm_minpos_epu16): New.
7373 2021-08-03 H.J. Lu <hjl.tools@gmail.com>
7375 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): In 64-bit mode,
7376 try XMM31 to avoid vzeroupper.
7378 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7380 * doc/invoke.texi: Document -mtune=neoverse-512tvb and
7381 -mcpu=neoverse-512tvb.
7382 * config/aarch64/aarch64-cores.def (neoverse-512tvb): New entry.
7383 * config/aarch64/aarch64-tune.md: Regenerate.
7384 * config/aarch64/aarch64.c (neoverse512tvb_sve_vector_cost)
7385 (neoverse512tvb_sve_issue_info, neoverse512tvb_vec_issue_info)
7386 (neoverse512tvb_vector_cost, neoverse512tvb_tunings): New structures.
7387 (aarch64_adjust_body_cost_sve): Handle -mtune=neoverse-512tvb.
7388 (aarch64_adjust_body_cost): Likewise.
7390 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7392 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Only
7393 record issue information for operations that occur in the
7396 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7398 * config/aarch64/aarch64.c (aarch64_multiply_add_p): Add a vec_flags
7399 parameter. Detect cases in which an Advanced SIMD MLA would almost
7400 certainly require a MOV.
7401 (aarch64_count_ops): Update accordingly.
7403 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7405 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction): New
7406 function, split out from...
7407 (aarch64_detect_vector_stmt_subtype): ...here.
7408 (aarch64_add_stmt_cost): Treat extracting element 0 as free.
7410 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7412 * config/aarch64/aarch64-protos.h (sve_vec_cost):
7413 Add gather_load_x32_cost and gather_load_x64_cost.
7414 * config/aarch64/aarch64.c (generic_sve_vector_cost)
7415 (a64fx_sve_vector_cost, neoversev1_sve_vector_cost): Update
7416 accordingly, using the values given by the scalar_load * number
7417 of elements calculation that we used previously.
7418 (aarch64_detect_vector_stmt_subtype): Use the new fields.
7420 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7422 * config/aarch64/aarch64.c (aarch64_adjust_body_cost_sve): New
7423 function, split out from...
7424 (aarch64_adjust_body_cost): ...here.
7426 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7428 * config/aarch64/fractional-cost.h: New file.
7429 * config/aarch64/aarch64.c: Include <algorithm> (indirectly)
7430 and cost_fraction.h.
7431 (vec_cost_fraction): New typedef.
7432 (aarch64_detect_scalar_stmt_subtype): Use it for statement costs.
7433 (aarch64_detect_vector_stmt_subtype): Likewise.
7434 (aarch64_sve_adjust_stmt_cost, aarch64_adjust_stmt_cost): Likewise.
7435 (aarch64_estimate_min_cycles_per_iter): Use vec_cost_fraction
7437 (aarch64_adjust_body_cost): Likewise.
7438 (aarch64_test_cost_fraction): New function.
7439 (aarch64_run_selftests): Call it.
7441 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7443 * config/aarch64/aarch64-protos.h (tune_params::sve_width): Turn
7445 * config/aarch64/aarch64.c (aarch64_cmp_autovec_modes): Update
7447 (aarch64_estimated_poly_value): Likewise. Use the least significant
7448 set bit for the minimum and likely values. Use the most significant
7449 set bit for the maximum value.
7451 2021-08-03 liuhongt <hongtao.liu@intel.com>
7453 * config/i386/sse.md (cond_<insn><mode>): New expander.
7454 (cond_mul<mode>): Ditto.
7456 2021-08-03 Kewen Lin <linkw@linux.ibm.com>
7458 * tree-cfg.c (move_sese_region_to_fn): Fix typos on dloop.
7460 2021-08-03 liuhongt <hongtao.liu@intel.com>
7462 * config/i386/sse.md (cond_<insn><mode>):New expander.
7463 (cond_mul<mode>): Ditto.
7464 (cond_div<mode>): Ditto.
7466 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
7468 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Also
7469 check stack_realign_needed for stack realignment.
7470 (ix86_legitimate_constant_p): Always allow CONST_WIDE_INT smaller
7471 than the largest integer supported by vector register.
7472 * config/i386/i386.h (MAX_MOVE_MAX): New. Set to 64.
7473 (MOVE_MAX): Set to bytes of the largest integer supported by
7475 (STORE_MAX_PIECES): New.
7477 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
7479 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
7480 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
7481 data from one memory location to another.
7483 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
7486 * config/i386/i386.c (TARGET_GEN_MEMSET_SCRATCH_RTX): New.
7488 2021-08-02 Aldy Hernandez <aldyh@redhat.com>
7490 PR tree-optimization/101724
7491 * params.opt: Remove --param=threader-iterative.
7492 * tree-ssa-threadbackward.c (pass_thread_jumps::execute): Remove
7495 2021-08-02 Tom de Vries <tdevries@suse.de>
7497 PR middle-end/101665
7498 * doc/extend.texi (nonnull attribute): Improve documentation.
7500 2021-08-02 Andrew Pinski <apinski@marvell.com>
7502 PR rtl-optimization/101683
7503 * rtlanal.c (may_trap_p_1): Handle UNSIGNED_FIX.
7505 2021-08-02 Roger Sayle <roger@nextmovesoftware.com>
7507 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern):
7508 Renamed from cond_removal_in_popcount_clz_ctz_pattern.
7509 Add support for BSWAP, FFS, PARITY and CLRSB builtins.
7510 (tree_ssa_phiop_worker): Update call to function above.
7512 2021-08-01 H.J. Lu <hjl.tools@gmail.com>
7515 * config/i386/i386.md (bsr_rex64_1_zext): New.
7516 (combine splitter for constant - clzll): Replace gen_bsr_rex64_1
7517 with gen_bsr_rex64_1_zext.
7519 2021-07-31 Jakub Jelinek <jakub@redhat.com>
7522 * config/i386/i386.md (bsr_rex64_1, bsr_1, bsr_zext_1): New
7523 define_insn patterns.
7524 (*bsr_rex64_2, *bsr_2): New define_insn_and_split patterns.
7525 Add combine splitters for constant - clz.
7526 (clz<mode>2): Use a temporary pseudo for bsr result.
7528 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
7530 * config/rs6000/smmintrin.h (_mm_floor_pd, _mm_floor_ps,
7531 _mm_floor_sd, _mm_floor_ss): New.
7533 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
7535 * config/rs6000/smmintrin.h (_mm_ceil_pd, _mm_ceil_ps,
7536 _mm_ceil_sd, _mm_ceil_ss): New.
7538 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
7540 * config/rs6000/smmintrin.h (_mm_blend_pd, _mm_blendv_pd,
7541 _mm_blend_ps, _mm_blendv_ps): New.
7543 2021-07-30 Roger Sayle <roger@nextmovesoftware.com>
7544 Uroš Bizjak <ubizjak@gmail.com>
7546 * config/i386/i386.md (*dec_cmov<mode>): New define_insn_and_split
7547 to generate a conditional move using the carry flag after sub $1.
7548 (peephole2): Eliminate a register-to-register move by inverting
7549 the condition of a conditional move.
7551 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
7553 * config/mmix/mmix.md ("call", "call_value", "*call_real")
7554 ("*call_value_real"): Don't generate rtx mentioning the generic
7555 operands 1 and 2 to "call", and similarly for "call_value".
7556 * config/mmix/mmix.c (mmix_print_operand_punct_valid_p)
7557 (mmix_print_operand): Use '!' instead of 'p'.
7559 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
7561 * doc/md.texi (call): Correct information about operand 2.
7562 * config/mmix/mmix.md ("call", "call_value"): Remove fixed FIXMEs.
7564 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
7566 * range-op.cc (operator_trunc_mod::wi_fold): Fold constants.
7568 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
7570 * range-op.cc (operator_div::wi_fold): Return UNDEFINED for [0, 0] divisor.
7572 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
7574 * gimple-range-cache.cc (*::set_bb_range): Change const basic_block to
7576 (*::get_bb_range): Ditto.
7577 (*::bb_range_p): Ditto.
7578 * gimple-range-cache.h: Change prototypes.
7580 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
7583 * builtins.c (builtin_memcpy_read_str): Change the mode argument
7584 from scalar_int_mode to fixed_size_mode.
7585 (builtin_strncpy_read_str): Likewise.
7586 (gen_memset_value_from_prev): New function.
7587 (builtin_memset_read_str): Change the mode argument from
7588 scalar_int_mode to fixed_size_mode. Use gen_memset_value_from_prev
7589 and support CONST_VECTOR.
7590 (builtin_memset_gen_str): Likewise.
7591 (try_store_by_multiple_pieces): Use by_pieces_constfn to declare
7593 * builtins.h (builtin_strncpy_read_str): Replace scalar_int_mode
7594 with fixed_size_mode.
7595 (builtin_memset_read_str): Likewise.
7596 * expr.c (widest_int_mode_for_size): Renamed to ...
7597 (widest_fixed_size_mode_for_size): Add a bool argument to
7598 indicate if QI vector mode can be used.
7599 (by_pieces_ninsns): Call widest_fixed_size_mode_for_size
7600 instead of widest_int_mode_for_size.
7601 (pieces_addr::adjust): Change the mode argument from
7602 scalar_int_mode to fixed_size_mode.
7603 (op_by_pieces_d): Make m_len read-only. Add a bool member,
7604 m_qi_vector_mode, to indicate that QI vector mode can be used.
7605 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
7606 initialize m_qi_vector_mode. Call widest_fixed_size_mode_for_size
7607 instead of widest_int_mode_for_size.
7608 (op_by_pieces_d::get_usable_mode): Change the mode argument from
7609 scalar_int_mode to fixed_size_mode. Call
7610 widest_fixed_size_mode_for_size instead of
7611 widest_int_mode_for_size.
7612 (op_by_pieces_d::smallest_fixed_size_mode_for_size): New member
7613 function to return the smallest integer or QI vector mode.
7614 (op_by_pieces_d::run): Call widest_fixed_size_mode_for_size
7615 instead of widest_int_mode_for_size. Call
7616 smallest_fixed_size_mode_for_size instead of
7617 smallest_int_mode_for_size.
7618 (store_by_pieces_d::store_by_pieces_d): Add a bool argument to
7619 indicate that QI vector mode can be used and pass it to
7620 op_by_pieces_d::op_by_pieces_d.
7621 (can_store_by_pieces): Call widest_fixed_size_mode_for_size
7622 instead of widest_int_mode_for_size. Pass memsetp to
7623 widest_fixed_size_mode_for_size to support QI vector mode.
7624 Allow all CONST_VECTORs for memset if vec_duplicate is supported.
7625 (store_by_pieces): Pass memsetp to
7626 store_by_pieces_d::store_by_pieces_d.
7627 (clear_by_pieces_1): Removed.
7628 (clear_by_pieces): Replace clear_by_pieces_1 with
7629 builtin_memset_read_str and pass true to store_by_pieces_d to
7630 support vector mode broadcast.
7631 (string_cst_read_str): Change the mode argument from
7632 scalar_int_mode to fixed_size_mode.
7633 * expr.h (by_pieces_constfn): Change scalar_int_mode to
7635 (by_pieces_prev): Likewise.
7636 * rtl.h (lowpart_subreg_regno): New.
7637 * rtlanal.c (lowpart_subreg_regno): New. A wrapper around
7638 simplify_subreg_regno.
7639 * target.def (gen_memset_scratch_rtx): New hook.
7640 * doc/tm.texi.in: Add TARGET_GEN_MEMSET_SCRATCH_RTX.
7641 * doc/tm.texi: Regenerated.
7643 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
7646 * config/mips/mips.c (mips_atomic_assign_expand_fenv): Use
7647 TARGET_EXPR instead of MODIFY_EXPR.
7649 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
7652 * config/mips/mips-protos.h (mips_expand_vec_cmp_expr): Declare.
7653 * config/mips/mips.c (mips_expand_vec_cmp_expr): New function.
7654 * config/mips/mips-msa.md (vec_cmp<MSA:mode><mode_i>): New
7656 (vec_cmpu<IMSA:mode><mode_i>): New expander.
7658 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
7661 * config/i386/i386-options.c (ix86_option_override_internal):
7662 Don't enable LZCNT/POPCNT if they have been disabled explicitly.
7664 2021-07-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
7667 * config/arm/arm_neon.h (vld1_p64): Replace call to builtin by
7668 explicitly dereferencing __a.
7669 (vld1_s64): Likewise.
7670 (vld1_u64): Likewise.
7671 * config/arm/arm_neon_builtins.def (vld1): Remove entry for di
7672 and change to VAR13.
7674 2021-07-30 Aldy Hernandez <aldyh@redhat.com>
7676 * gimple-loop-versioning.cc (lv_dom_walker::lv_dom_walker): Remove
7677 use of m_range_analyzer.
7678 (loop_versioning::lv_dom_walker::before_dom_children): Same.
7679 (loop_versioning::lv_dom_walker::after_dom_children): Remove.
7680 (loop_versioning::prune_loop_conditions): Replace vr_values use
7681 with range_query interface.
7682 (pass_loop_versioning::execute): Use ranger.
7684 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
7687 * ipa-devirt.c (ipa_odr_read_section): Compare the precision of
7688 enum values, and emit a warning if they mismatch.
7690 2021-07-30 Kewen Lin <linkw@linux.ibm.com>
7692 * cfgloop.h (as_const): New function.
7693 (class loop_iterator): Rename to ...
7694 (class loops_list): ... this.
7695 (loop_iterator::next): Rename to ...
7696 (loops_list::Iter::fill_curr_loop): ... this and adjust.
7697 (loop_iterator::loop_iterator): Rename to ...
7698 (loops_list::loops_list): ... this and adjust.
7699 (loops_list::Iter): New class.
7700 (loops_list::iterator): New type.
7701 (loops_list::const_iterator): New type.
7702 (loops_list::begin): New function.
7703 (loops_list::end): Likewise.
7704 (loops_list::begin const): Likewise.
7705 (loops_list::end const): Likewise.
7706 (FOR_EACH_LOOP): Remove.
7707 (FOR_EACH_LOOP_FN): Remove.
7708 * cfgloop.c (flow_loops_dump): Adjust FOR_EACH_LOOP* with range-based
7709 for loop with loops_list instance.
7710 (sort_sibling_loops): Likewise.
7711 (disambiguate_loops_with_multiple_latches): Likewise.
7712 (verify_loop_structure): Likewise.
7713 * cfgloopmanip.c (create_preheaders): Likewise.
7714 (force_single_succ_latches): Likewise.
7715 * config/aarch64/falkor-tag-collision-avoidance.c
7716 (execute_tag_collision_avoidance): Likewise.
7717 * config/mn10300/mn10300.c (mn10300_scan_for_setlb_lcc): Likewise.
7718 * config/s390/s390.c (s390_adjust_loops): Likewise.
7719 * doc/loop.texi: Likewise.
7720 * gimple-loop-interchange.cc (pass_linterchange::execute): Likewise.
7721 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
7722 * gimple-loop-versioning.cc (loop_versioning::analyze_blocks): Likewise.
7723 (loop_versioning::make_versioning_decisions): Likewise.
7724 * gimple-ssa-split-paths.c (split_paths): Likewise.
7725 * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
7726 * graphite.c (canonicalize_loop_form): Likewise.
7727 (graphite_transform_loops): Likewise.
7728 * ipa-fnsummary.c (analyze_function_body): Likewise.
7729 * ipa-pure-const.c (analyze_function): Likewise.
7730 * loop-doloop.c (doloop_optimize_loops): Likewise.
7731 * loop-init.c (loop_optimizer_finalize): Likewise.
7732 (fix_loop_structure): Likewise.
7733 * loop-invariant.c (calculate_loop_reg_pressure): Likewise.
7734 (move_loop_invariants): Likewise.
7735 * loop-unroll.c (decide_unrolling): Likewise.
7736 (unroll_loops): Likewise.
7737 * modulo-sched.c (sms_schedule): Likewise.
7738 * predict.c (predict_loops): Likewise.
7739 (pass_profile::execute): Likewise.
7740 * profile.c (branch_prob): Likewise.
7741 * sel-sched-ir.c (sel_finish_pipelining): Likewise.
7742 (sel_find_rgns): Likewise.
7743 * tree-cfg.c (replace_loop_annotate): Likewise.
7744 (replace_uses_by): Likewise.
7745 (move_sese_region_to_fn): Likewise.
7746 * tree-if-conv.c (pass_if_conversion::execute): Likewise.
7747 * tree-loop-distribution.c (loop_distribution::execute): Likewise.
7748 * tree-parloops.c (parallelize_loops): Likewise.
7749 * tree-predcom.c (tree_predictive_commoning): Likewise.
7750 * tree-scalar-evolution.c (scev_initialize): Likewise.
7751 (scev_reset): Likewise.
7752 * tree-ssa-dce.c (find_obviously_necessary_stmts): Likewise.
7753 * tree-ssa-live.c (remove_unused_locals): Likewise.
7754 * tree-ssa-loop-ch.c (ch_base::copy_headers): Likewise.
7755 * tree-ssa-loop-im.c (analyze_memory_references): Likewise.
7756 (tree_ssa_lim_initialize): Likewise.
7757 * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables): Likewise.
7758 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize): Likewise.
7759 * tree-ssa-loop-manip.c (get_loops_exits): Likewise.
7760 * tree-ssa-loop-niter.c (estimate_numbers_of_iterations): Likewise.
7761 (free_numbers_of_iterations_estimates): Likewise.
7762 * tree-ssa-loop-prefetch.c (tree_ssa_prefetch_arrays): Likewise.
7763 * tree-ssa-loop-split.c (tree_ssa_split_loops): Likewise.
7764 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Likewise.
7765 * tree-ssa-loop.c (gate_oacc_kernels): Likewise.
7766 (pass_scev_cprop::execute): Likewise.
7767 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
7768 * tree-ssa-sccvn.c (do_rpo_vn): Likewise.
7769 * tree-ssa-threadupdate.c
7770 (jump_thread_path_registry::thread_through_all_blocks): Likewise.
7771 * tree-vectorizer.c (vectorize_loops): Likewise.
7772 * tree-vrp.c (vrp_asserts::find_assert_locations): Likewise.
7774 2021-07-29 Hans-Peter Nilsson <hp@bitrange.com>
7776 * config/mmix/mmix.c (mmix_function_arg_1): Avoid
7777 generating a VOIDmode register for e.g the
7778 function_arg_info::end_marker.
7780 2021-07-29 Jeff Law <jeffreyalaw@gmail.com>
7782 * config/h8300/h8300-modes.def: Add CCZ, CCV and CCC, drop CCZNV.
7783 * config/h8300/h8300.md (H8cc mode iterator): Add CCZ.
7784 (cc mode_attr): Similarly.
7785 (ccz subst_attr): Similarly.
7786 * config/h8300/jumpcall.md: Add new patterns for branch-on-bit.
7787 * config/h8300/testcompare.md: Remove various cc0 based patterns
7788 that had been commented out. Add pattern to set CCZ from a bit
7791 2021-07-29 Thomas Schwinge <thomas@codesourcery.com>
7792 Julian Brown <julian@codesourcery.com>
7793 Kwok Cheung Yeung <kcy@codesourcery.com>
7795 * omp-offload.c (oacc_loop_xform_head_tail, oacc_loop_process):
7796 'update_stmt' after modification.
7797 (pass_oacc_loop_designation): New function, extracted out of...
7798 (pass_oacc_device_lower): ... this.
7799 (pass_data_oacc_loop_designation, pass_oacc_loop_designation)
7800 (make_pass_oacc_loop_designation): New
7801 * passes.def: Add it.
7802 * tree-parloops.c (create_parallel_loop): Adjust.
7803 * tree-pass.h (make_pass_oacc_loop_designation): New.
7805 2021-07-29 Aldy Hernandez <aldyh@redhat.com>
7807 * flag-types.h (enum threader_mode): New.
7808 * params.opt: Add entry for --param=threader-mode.
7809 * tree-ssa-threadbackward.c (THREADER_ITERATIVE_MODE): New.
7810 (class back_threader): New.
7811 (back_threader::back_threader): New.
7812 (back_threader::~back_threader): New.
7813 (back_threader::maybe_register_path): New.
7814 (back_threader::find_taken_edge): New.
7815 (back_threader::find_taken_edge_switch): New.
7816 (back_threader::find_taken_edge_cond): New.
7817 (back_threader::resolve_def): New.
7818 (back_threader::resolve_phi): New.
7819 (back_threader::find_paths_to_names): New.
7820 (back_threader::find_paths): New.
7823 (thread_jumps::find_jump_threads_backwards): Call ranger threader.
7824 (thread_jumps::find_jump_threads_backwards_with_ranger): New.
7825 (pass_thread_jumps::execute): Abstract out code...
7826 (try_thread_blocks): ...here.
7827 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
7828 Abstract out threading candidate code to...
7829 (single_succ_to_potentially_threadable_block): ...here.
7830 * tree-ssa-threadedge.h (single_succ_to_potentially_threadable_block):
7832 * tree-ssa-threadupdate.c (register_jump_thread): Return boolean.
7833 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
7834 Return bool from register_jump_thread.
7836 2021-07-29 Andreas Krebbel <krebbel@linux.ibm.com>
7838 * target.def: in0 and in1 do not need to be registers.
7839 * doc/tm.texi: Regenerate.
7841 2021-07-29 liuhongt <hongtao.liu@intel.com>
7844 * config/i386/i386.c (ix86_widen_mult_cost): New function.
7845 (ix86_add_stmt_cost): Use ix86_widen_mult_cost for
7848 2021-07-29 Jiufu Guo <guojiufu@linux.ibm.com>
7851 * config/rs6000/rs6000.c (TARGET_PREFERRED_DOLOOP_MODE): New hook.
7852 (rs6000_preferred_doloop_mode): New hook.
7853 * doc/tm.texi: Regenerate.
7854 * doc/tm.texi.in: Add hook preferred_doloop_mode.
7855 * target.def (preferred_doloop_mode): New hook.
7856 * targhooks.c (default_preferred_doloop_mode): New hook.
7857 * targhooks.h (default_preferred_doloop_mode): New hook.
7858 * tree-ssa-loop-ivopts.c (compute_doloop_base_on_mode): New function.
7859 (add_iv_candidate_for_doloop): Call targetm.preferred_doloop_mode
7860 and compute_doloop_base_on_mode.
7862 2021-07-28 Martin Sebor <msebor@redhat.com>
7864 PR middle-end/101494
7865 * tree-ssa-uninit.c (maybe_warn_operand): Correct object offset
7866 and size computation.
7868 2021-07-28 Martin Sebor <msebor@redhat.com>
7870 PR middle-end/101601
7871 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Remove
7873 Handle pointers to functions.
7875 2021-07-28 Martin Sebor <msebor@redhat.com>
7877 * Makefile.in (OBJS): Add gimple-ssa-warn-access.o and pointer-query.o.
7878 * attribs.h (fndecl_dealloc_argno): Move fndecl_dealloc_argno to tree.h.
7879 * builtins.c (compute_objsize_r): Move to pointer-query.cc.
7880 (access_ref::access_ref): Same.
7881 (access_ref::phi): Same.
7882 (access_ref::get_ref): Same.
7883 (access_ref::size_remaining): Same.
7884 (access_ref::offset_in_range): Same.
7885 (access_ref::add_offset): Same.
7886 (access_ref::inform_access): Same.
7887 (ssa_name_limit_t::visit_phi): Same.
7888 (ssa_name_limit_t::leave_phi): Same.
7889 (ssa_name_limit_t::next): Same.
7890 (ssa_name_limit_t::next_phi): Same.
7891 (ssa_name_limit_t::~ssa_name_limit_t): Same.
7892 (pointer_query::pointer_query): Same.
7893 (pointer_query::get_ref): Same.
7894 (pointer_query::put_ref): Same.
7895 (pointer_query::flush_cache): Same.
7896 (warn_string_no_nul): Move to gimple-ssa-warn-access.cc.
7897 (check_nul_terminated_array): Same.
7898 (unterminated_array): Same.
7899 (maybe_warn_for_bound): Same.
7900 (check_read_access): Same.
7901 (warn_for_access): Same.
7902 (get_size_range): Same.
7903 (check_access): Same.
7904 (gimple_call_alloc_size): Move to tree.c.
7905 (gimple_parm_array_size): Move to pointer-query.cc.
7906 (get_offset_range): Same.
7907 (gimple_call_return_array): Same.
7908 (handle_min_max_size): Same.
7909 (handle_array_ref): Same.
7910 (handle_mem_ref): Same.
7911 (compute_objsize): Same.
7912 (gimple_call_alloc_p): Move to gimple-ssa-warn-access.cc.
7913 (call_dealloc_argno): Same.
7914 (fndecl_dealloc_argno): Same.
7915 (new_delete_mismatch_p): Same.
7916 (matching_alloc_calls_p): Same.
7917 (warn_dealloc_offset): Same.
7918 (maybe_emit_free_warning): Same.
7919 * builtins.h (check_nul_terminated_array): Move to
7920 gimple-ssa-warn-access.h.
7921 (check_nul_terminated_array): Same.
7922 (warn_string_no_nul): Same.
7923 (unterminated_array): Same.
7924 (class ssa_name_limit_t): Same.
7925 (class pointer_query): Same.
7926 (struct access_ref): Same.
7927 (class range_query): Same.
7928 (struct access_data): Same.
7929 (gimple_call_alloc_size): Same.
7930 (gimple_parm_array_size): Same.
7931 (compute_objsize): Same.
7932 (class access_data): Same.
7933 (maybe_emit_free_warning): Same.
7934 * calls.c (initialize_argument_information): Remove call to
7935 maybe_emit_free_warning.
7936 * gimple-array-bounds.cc: Include new header..
7937 * gimple-fold.c: Same.
7938 * gimple-ssa-sprintf.c: Same.
7939 * gimple-ssa-warn-restrict.c: Same.
7940 * passes.def: Add pass_warn_access.
7941 * tree-pass.h (make_pass_warn_access): Declare.
7942 * tree-ssa-strlen.c: Include new headers.
7943 * tree.c (fndecl_dealloc_argno): Move here from builtins.c.
7944 * tree.h (fndecl_dealloc_argno): Move here from attribs.h.
7945 * gimple-ssa-warn-access.cc: New file.
7946 * gimple-ssa-warn-access.h: New file.
7947 * pointer-query.cc: New file.
7948 * pointer-query.h: New file.
7950 2021-07-28 Jakub Jelinek <jakub@redhat.com>
7952 PR middle-end/101624
7953 * ubsan.c (maybe_instrument_pointer_overflow,
7954 instrument_object_size): Only test DECL_REGISTER on VAR_DECLs,
7955 PARM_DECLs or RESULT_DECLs.
7956 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Likewise.
7958 2021-07-28 Jakub Jelinek <jakub@redhat.com>
7960 PR middle-end/101642
7961 * match.pd (bswap16 (x) == bswap16 (y)): Cast both operands
7962 to type of bswap16 for comparison.
7963 (bswap16 (x) == cst): Cast bswap16 operand to type of cst.
7965 2021-07-28 Richard Biener <rguenther@suse.de>
7967 PR tree-optimization/101615
7968 * tree-vect-slp.c (vect_optimize_slp): Materialize permutes
7969 at CTOR SLP graph entries.
7971 2021-07-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7973 * config/aarch64/aarch64.md (*extend<SHORT:mode><GPI:mode>2_aarch64):
7974 Add "r,w" alternative.
7976 2021-07-28 H.J. Lu <hjl.tools@gmail.com>
7979 * config/i386/i386.c (ix86_avx_u128_mode_needed): Don't set
7980 AVX_U128_DIRTY when all bits are zero.
7982 2021-07-28 Richard Biener <rguenther@suse.de>
7984 PR tree-optimization/101615
7985 * tree-vect-slp.c (vect_optimize_slp): Pre-existing vector
7986 external nodes cannot be permuted so make them perm_out 0.
7988 2021-07-28 Andrew Stubbs <ams@codesourcery.com>
7991 * config.in: Regenerate.
7992 * config/gcn/gcn-hsa.h (A_FIJI): New define.
7993 (A_900): New define.
7994 (A_906): New define.
7995 (A_908): New define.
7996 (ASM_SPEC): Use A_FIJI, A_900, A_906 and A_908.
7997 * config/gcn/gcn.c (output_file_start): Adjust attributes according
7998 to the assembler capabilities.
7999 * config/gcn/mkoffload.c (main): Likewise.
8000 * configure: Regenerate.
8001 * configure.ac: Add tests for LLVM assembler attribute features.
8003 2021-07-28 Andrew MacLeod <amacleod@redhat.com>
8005 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Check for
8006 cond_false and cond_true on branches.
8008 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
8010 * config/aarch64/aarch64.c (aarch64_gen_adjusted_ldpstp): use
8013 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
8015 * alias.c (init_alias_analysis): Don't skip prologue/epilogue.
8017 2021-07-28 Jakub Jelinek <jakub@redhat.com>
8020 * config/i386/sse.md (vashr<mode>3): Split into vashrv8di3 expander
8021 and vashrv4di3 expander, where the latter requires just TARGET_AVX2
8022 and has special !TARGET_AVX512VL expansion.
8023 (vashrv2di3<mask_name>): Rename to ...
8024 (vashrv2di3): ... this. Change condition to TARGET_XOP || TARGET_AVX2
8025 and add special !TARGET_XOP && !TARGET_AVX512VL expansion.
8027 2021-07-28 Martin Uecker <muecker@gwdg.de>
8029 * calls.c (maybe_warn_rdwr_sizes): Correct argument
8030 numbers in warning that were switched.
8032 2021-07-28 Kewen Lin <linkw@linux.ibm.com>
8034 PR tree-optimization/101596
8035 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Fix wrong check
8036 by using new_type's precision instead.
8038 2021-07-28 liuhongt <hongtao.liu@intel.com>
8041 * config/i386/i386.h (processor_costs): Add new member
8043 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
8044 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
8045 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
8046 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
8047 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
8048 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
8049 generic_cost, core_cost): Initialize integer_to_sse same value
8051 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
8052 * config/i386/i386.c (ix86_builtin_vectorization_cost):
8053 Use integer_to_sse instead of sse_op to calculate the cost of
8056 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
8058 * config/rs6000/rs6000-gen-builtins.c (write_ovld_static_init): New
8060 (write_init_file): Call write_ovld_static_init.
8062 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
8064 * config/rs6000/rs6000-gen-builtins.c (write_bif_static_init): New
8066 (write_init_file): Call write_bif_static_init.
8068 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
8070 * config/rs6000/rs6000-gen-builtins.c (typemap): New struct.
8071 (TYPE_MAP_SIZE): New macro.
8072 (type_map): New initialized variable.
8073 (typemap_cmp): New function.
8074 (write_type_node): Likewise.
8075 (write_fntype_init): Implement.
8077 2021-07-27 Martin Sebor <msebor@redhat.com>
8079 PR tree-optimization/101584
8080 * tree-ssa-uninit.c (builtin_call_nomodifying_p): New function.
8081 (check_defs): Call it.
8083 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
8085 * tree-ssa-dom.c (dom_jump_threader_simplifier):
8086 Put avail_exprs_stack in the class, instead of passing it to
8087 jump_threader_simplifier.
8088 (dom_jump_threader_simplifier::simplify): Add state argument.
8089 (dom_opt_dom_walker): Add state.
8090 (pass_dominator::execute): Pass state to threader.
8091 (dom_opt_dom_walker::before_dom_children): Use state.
8092 * tree-ssa-threadedge.c (jump_threader::jump_threader): Replace
8094 (jump_threader::record_temporary_equivalences_from_phis):
8095 Register equivalences through the state variable.
8096 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
8097 Record ranges in a statement through the state variable.
8098 (jump_threader::simplify_control_stmt_condition): Pass state to
8100 (jump_threader::simplify_control_stmt_condition_1): Same.
8101 (jump_threader::thread_around_empty_blocks): Remove obsolete
8103 (jump_threader::thread_through_normal_block): Record equivalences
8104 on edge through the state variable.
8105 (jump_threader::thread_across_edge): Abstract state pushing.
8106 (jt_state::jt_state): New.
8107 (jt_state::push): New.
8108 (jt_state::pop): New.
8109 (jt_state::register_equiv): New.
8110 (jt_state::record_ranges_from_stmt): New.
8111 (jt_state::register_equivs_on_edge): New.
8112 (jump_threader_simplifier::jump_threader_simplifier): Move from
8114 (jump_threader_simplifier::simplify): Add state argument.
8115 * tree-ssa-threadedge.h (class jt_state): New.
8116 (class jump_threader): Add state to constructor.
8117 (class jump_threader_simplifier): Add state to simplify. Remove
8118 avail_exprs_stack from class.
8119 * tree-vrp.c (vrp_jump_threader_simplifier::simplify): Add state
8121 (vrp_jump_threader::vrp_jump_threader): Add state.
8122 (vrp_jump_threader::~vrp_jump_threader): Cleanup state.
8124 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
8126 * Makefile.in (OBJS): Add gimple-range-path.o.
8127 * gimple-range-path.cc: New file.
8128 * gimple-range-path.h: New file.
8130 2021-07-27 Jonathan Wright <jonathan.wright@arm.com>
8132 * config/aarch64/aarch64-simd.md: Push sign/zero-extension
8133 inside vec_duplicate for all patterns.
8134 * simplify-rtx.c (simplify_context::simplify_unary_operation_1):
8135 Push sign/zero-extension inside vec_duplicate.
8137 2021-07-27 Richard Biener <rguenther@suse.de>
8139 PR tree-optimization/101573
8140 * tree-ssa-uninit.c (warn_uninit_phi_uses): New function
8141 looking at uninitialized PHI arg defs in some constrained cases.
8142 (warn_uninitialized_vars): Call it.
8143 (execute_early_warn_uninitialized): Calculate dominators.
8145 2021-07-27 Richard Biener <rguenther@suse.de>
8147 PR tree-optimization/39821
8148 * tree-vect-stmts.c (vect_model_promotion_demotion_cost): Use
8149 vector_stmt for widening arithmetic.
8150 (vectorizable_conversion): Adjust.
8152 2021-07-27 Martin Jambor <mjambor@suse.cz>
8154 * cgraph.h (ipa_replace_map): New field force_load_ref.
8155 * ipa-prop.h (ipa_param_descriptor): Reduce precision of move_cost,
8156 aded new flag load_dereferenced, adjusted comments.
8157 (ipa_get_param_dereferenced): New function.
8158 (ipa_set_param_dereferenced): Likewise.
8159 * cgraphclones.c (cgraph_node::create_virtual_clone): Follow it.
8160 * ipa-cp.c: Include gimple.h.
8161 (ipcp_discover_new_direct_edges): Take into account dereferenced flag.
8162 (get_replacement_map): New parameter force_load_ref, set the
8163 appropriate flag in ipa_replace_map if set.
8164 (struct symbol_and_index_together): New type.
8165 (adjust_refs_in_act_callers): New function.
8166 (adjust_references_in_caller): Likewise.
8167 (create_specialized_node): When appropriate, call
8168 adjust_references_in_caller and force only load references.
8169 * ipa-prop.c (load_from_dereferenced_name): New function.
8170 (ipa_analyze_controlled_uses): Also detect loads from a
8171 dereference, harden testing of call statements.
8172 (ipa_write_node_info): Stream the dereferenced flag.
8173 (ipa_read_node_info): Likewise.
8174 (ipa_set_jf_constant): Also create refdesc when jump function
8175 references a variable.
8176 (cgraph_node_for_jfunc): Rename to symtab_node_for_jfunc, work
8177 also on references of variables and return a symtab_node. Adjust
8179 (propagate_controlled_uses): Also remove references to VAR_DECLs.
8181 2021-07-27 Jakub Jelinek <jakub@redhat.com>
8183 PR middle-end/101586
8184 * gimple-fold.c (clear_padding_type): Ignore FIELD_DECLs with byte
8185 positions above or equal to sz except for diagnostics of flexible
8188 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
8190 PR tree-optimization/78888
8191 * gimple-range-fold.cc (get_letter_range): New.
8192 (fold_using_range::range_of_builtin_call): Call get_letter_range.
8194 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
8196 PR tree-optimization/78888
8197 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Add cases
8198 for CFN_BUILT_IN_TOUPPER and CFN_BUILT_IN_TOLOWER.
8200 2021-07-26 Roger Sayle <roger@nextmovesoftware.com>
8201 Marc Glisse <marc.glisse@inria.fr>
8203 * match.pd (rotate): Simplify equality/inequality of rotations.
8204 (bswap): Simplify equality/inequality tests of byte swapping.
8206 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
8208 * range-op.cc (operator_bitwise_xor::op1_op2_relation_effect):
8211 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
8213 * range-op.cc (operator_lshift::fold_range): Pass rel to
8214 base class fold_range.
8215 (operator_rshift::fold_range): Same.
8217 2021-07-26 Ashimida <ashimida@linux.alibaba.com>
8220 * toplev.h (min_align_loops_log): Remove declaration.
8221 (min_align_jumps_log, min_align_labels_log): Likewise.
8222 (min_align_functions_log): Likewise.
8224 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
8226 * tree-vrp.c (vrp_simplify_cond_using_ranges): Rename vr_values
8228 (execute_vrp): Abstract out simplification of conditionals...
8229 (simplify_casted_conds): ...here.
8231 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
8233 * gimple-array-bounds.cc (array_bounds_checker::get_value_range):
8234 Add gimple argument.
8235 (array_bounds_checker::check_array_ref): Same.
8236 (array_bounds_checker::check_addr_expr): Same.
8237 (array_bounds_checker::check_array_bounds): Pass statement to
8238 check_array_bounds and check_addr_expr.
8239 * gimple-array-bounds.h (check_array_bounds): Add gimple argument.
8240 (check_addr_expr): Same.
8241 (get_value_range): Same.
8243 2021-07-26 Tamar Christina <tamar.christina@arm.com>
8245 * config/aarch64/aarch64-simd-builtins.def (sdot, udot): Rename to..
8246 (sdot_prod, udot_prod): ... This.
8247 * config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>): Merged
8249 (<sur>dot_prod<vsi2qi>): ... this.
8250 (aarch64_<sur>dot_lane<vsi2qi>, aarch64_<sur>dot_laneq<vsi2qi>):
8251 Change operands order.
8252 (<sur>sadv16qi): Use new operands order.
8253 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32,
8254 vdotq_s32): Use new RTL ordering.
8256 2021-07-26 Tamar Christina <tamar.christina@arm.com>
8258 * config/aarch64/aarch64-builtins.c (TYPES_TERNOP_SUSS,
8259 aarch64_types_ternop_suss_qualifiers): New.
8260 * config/aarch64/aarch64-simd-builtins.def (usdot_prod): Use it.
8261 * config/aarch64/aarch64-simd.md (usdot_prod<vsi2qi>): Re-organize RTL.
8262 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Use it.
8264 2021-07-23 Jakub Jelinek <jakub@redhat.com>
8266 PR rtl-optimization/101562
8267 * expmed.c (store_integral_bit_field): Only use movstrict_optab
8268 if the operand isn't paradoxical.
8270 2021-07-23 Aldy Hernandez <aldyh@redhat.com>
8272 * gimple-array-bounds.h (class array_bounds_checker): Change
8273 ranges type to range_query.
8275 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8277 * config/aarch64/arm_neon.h (vst1_s64_x2): Use
8278 __builtin_memcpy instead of constructing
8279 __builtin_aarch64_simd_oi one vector at a time.
8280 (vst1_u64_x2): Likewise.
8281 (vst1_f64_x2): Likewise.
8282 (vst1_s8_x2): Likewise.
8283 (vst1_p8_x2): Likewise.
8284 (vst1_s16_x2): Likewise.
8285 (vst1_p16_x2): Likewise.
8286 (vst1_s32_x2): Likewise.
8287 (vst1_u8_x2): Likewise.
8288 (vst1_u16_x2): Likewise.
8289 (vst1_u32_x2): Likewise.
8290 (vst1_f16_x2): Likewise.
8291 (vst1_f32_x2): Likewise.
8292 (vst1_p64_x2): Likewise.
8293 (vst1q_s8_x2): Likewise.
8294 (vst1q_p8_x2): Likewise.
8295 (vst1q_s16_x2): Likewise.
8296 (vst1q_p16_x2): Likewise.
8297 (vst1q_s32_x2): Likewise.
8298 (vst1q_s64_x2): Likewise.
8299 (vst1q_u8_x2): Likewise.
8300 (vst1q_u16_x2): Likewise.
8301 (vst1q_u32_x2): Likewise.
8302 (vst1q_u64_x2): Likewise.
8303 (vst1q_f16_x2): Likewise.
8304 (vst1q_f32_x2): Likewise.
8305 (vst1q_f64_x2): Likewise.
8306 (vst1q_p64_x2): Likewise.
8308 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8310 * config/aarch64/arm_neon.h (vst1_s64_x3): Use
8311 __builtin_memcpy instead of constructing
8312 __builtin_aarch64_simd_ci one vector at a time.
8313 (vst1_u64_x3): Likewise.
8314 (vst1_f64_x3): Likewise.
8315 (vst1_s8_x3): Likewise.
8316 (vst1_p8_x3): Likewise.
8317 (vst1_s16_x3): Likewise.
8318 (vst1_p16_x3): Likewise.
8319 (vst1_s32_x3): Likewise.
8320 (vst1_u8_x3): Likewise.
8321 (vst1_u16_x3): Likewise.
8322 (vst1_u32_x3): Likewise.
8323 (vst1_f16_x3): Likewise.
8324 (vst1_f32_x3): Likewise.
8325 (vst1_p64_x3): Likewise.
8326 (vst1q_s8_x3): Likewise.
8327 (vst1q_p8_x3): Likewise.
8328 (vst1q_s16_x3): Likewise.
8329 (vst1q_p16_x3): Likewise.
8330 (vst1q_s32_x3): Likewise.
8331 (vst1q_s64_x3): Likewise.
8332 (vst1q_u8_x3): Likewise.
8333 (vst1q_u16_x3): Likewise.
8334 (vst1q_u32_x3): Likewise.
8335 (vst1q_u64_x3): Likewise.
8336 (vst1q_f16_x3): Likewise.
8337 (vst1q_f32_x3): Likewise.
8338 (vst1q_f64_x3): Likewise.
8339 (vst1q_p64_x3): Likewise.
8341 2021-07-23 H.J. Lu <hjl.tools@gmail.com>
8344 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): Don't return
8345 hard register when LRA is in progress.
8347 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8349 * config/aarch64/arm_neon.h (vst1_s8_x4): Use
8350 __builtin_memcpy instead of using a union.
8351 (vst1q_s8_x4): Likewise.
8352 (vst1_s16_x4): Likewise.
8353 (vst1q_s16_x4): Likewise.
8354 (vst1_s32_x4): Likewise.
8355 (vst1q_s32_x4): Likewise.
8356 (vst1_u8_x4): Likewise.
8357 (vst1q_u8_x4): Likewise.
8358 (vst1_u16_x4): Likewise.
8359 (vst1q_u16_x4): Likewise.
8360 (vst1_u32_x4): Likewise.
8361 (vst1q_u32_x4): Likewise.
8362 (vst1_f16_x4): Likewise.
8363 (vst1q_f16_x4): Likewise.
8364 (vst1_f32_x4): Likewise.
8365 (vst1q_f32_x4): Likewise.
8366 (vst1_p8_x4): Likewise.
8367 (vst1q_p8_x4): Likewise.
8368 (vst1_p16_x4): Likewise.
8369 (vst1q_p16_x4): Likewise.
8370 (vst1_s64_x4): Likewise.
8371 (vst1_u64_x4): Likewise.
8372 (vst1_p64_x4): Likewise.
8373 (vst1q_s64_x4): Likewise.
8374 (vst1q_u64_x4): Likewise.
8375 (vst1q_p64_x4): Likewise.
8376 (vst1_f64_x4): Likewise.
8377 (vst1q_f64_x4): Likewise.
8379 2021-07-23 Jonathan Wrightt <jonathan.wright@arm.com>
8381 * config/aarch64/arm_neon.h (vst2_s64): Use __builtin_memcpy
8382 instead of constructing __builtin_aarch64_simd_oi one vector
8384 (vst2_u64): Likewise.
8385 (vst2_f64): Likewise.
8386 (vst2_s8): Likewise.
8387 (vst2_p8): Likewise.
8388 (vst2_s16): Likewise.
8389 (vst2_p16): Likewise.
8390 (vst2_s32): Likewise.
8391 (vst2_u8): Likewise.
8392 (vst2_u16): Likewise.
8393 (vst2_u32): Likewise.
8394 (vst2_f16): Likewise.
8395 (vst2_f32): Likewise.
8396 (vst2_p64): Likewise.
8397 (vst2q_s8): Likewise.
8398 (vst2q_p8): Likewise.
8399 (vst2q_s16): Likewise.
8400 (vst2q_p16): Likewise.
8401 (vst2q_s32): Likewise.
8402 (vst2q_s64): Likewise.
8403 (vst2q_u8): Likewise.
8404 (vst2q_u16): Likewise.
8405 (vst2q_u32): Likewise.
8406 (vst2q_u64): Likewise.
8407 (vst2q_f16): Likewise.
8408 (vst2q_f32): Likewise.
8409 (vst2q_f64): Likewise.
8410 (vst2q_p64): Likewise.
8412 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8414 * config/aarch64/arm_neon.h (vst3_s64): Use __builtin_memcpy
8415 instead of constructing __builtin_aarch64_simd_ci one vector
8417 (vst3_u64): Likewise.
8418 (vst3_f64): Likewise.
8419 (vst3_s8): Likewise.
8420 (vst3_p8): Likewise.
8421 (vst3_s16): Likewise.
8422 (vst3_p16): Likewise.
8423 (vst3_s32): Likewise.
8424 (vst3_u8): Likewise.
8425 (vst3_u16): Likewise.
8426 (vst3_u32): Likewise.
8427 (vst3_f16): Likewise.
8428 (vst3_f32): Likewise.
8429 (vst3_p64): Likewise.
8430 (vst3q_s8): Likewise.
8431 (vst3q_p8): Likewise.
8432 (vst3q_s16): Likewise.
8433 (vst3q_p16): Likewise.
8434 (vst3q_s32): Likewise.
8435 (vst3q_s64): Likewise.
8436 (vst3q_u8): Likewise.
8437 (vst3q_u16): Likewise.
8438 (vst3q_u32): Likewise.
8439 (vst3q_u64): Likewise.
8440 (vst3q_f16): Likewise.
8441 (vst3q_f32): Likewise.
8442 (vst3q_f64): Likewise.
8443 (vst3q_p64): Likewise.
8445 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8447 * config/aarch64/arm_neon.h (vst4_s64): Use __builtin_memcpy
8448 instead of constructing __builtin_aarch64_simd_xi one vector
8450 (vst4_u64): Likewise.
8451 (vst4_f64): Likewise.
8452 (vst4_s8): Likewise.
8453 (vst4_p8): Likewise.
8454 (vst4_s16): Likewise.
8455 (vst4_p16): Likewise.
8456 (vst4_s32): Likewise.
8457 (vst4_u8): Likewise.
8458 (vst4_u16): Likewise.
8459 (vst4_u32): Likewise.
8460 (vst4_f16): Likewise.
8461 (vst4_f32): Likewise.
8462 (vst4_p64): Likewise.
8463 (vst4q_s8): Likewise.
8464 (vst4q_p8): Likewise.
8465 (vst4q_s16): Likewise.
8466 (vst4q_p16): Likewise.
8467 (vst4q_s32): Likewise.
8468 (vst4q_s64): Likewise.
8469 (vst4q_u8): Likewise.
8470 (vst4q_u16): Likewise.
8471 (vst4q_u32): Likewise.
8472 (vst4q_u64): Likewise.
8473 (vst4q_f16): Likewise.
8474 (vst4q_f32): Likewise.
8475 (vst4q_f64): Likewise.
8476 (vst4q_p64): Likewise.
8478 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8480 * config/aarch64/arm_neon.h (vtbx4_s8): Use __builtin_memcpy
8481 instead of constructing __builtin_aarch64_simd_oi one vector
8483 (vtbx4_u8): Likewise.
8484 (vtbx4_p8): Likewise.
8486 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8488 * config/aarch64/arm_neon.h (vtbl3_s8): Use __builtin_memcpy
8489 instead of constructing __builtin_aarch64_simd_oi one vector
8491 (vtbl3_u8): Likewise.
8492 (vtbl3_p8): Likewise.
8493 (vtbl4_s8): Likewise.
8494 (vtbl4_u8): Likewise.
8495 (vtbl4_p8): Likewise.
8497 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8499 * config/aarch64/arm_neon.h (vqtbx2_s8): Use __builtin_memcpy
8500 instead of constructing __builtin_aarch64_simd_oi one vector
8502 (vqtbx2_u8): Likewise.
8503 (vqtbx2_p8): Likewise.
8504 (vqtbx2q_s8): Likewise.
8505 (vqtbx2q_u8): Likewise.
8506 (vqtbx2q_p8): Likewise.
8507 (vqtbx3_s8): Use __builtin_memcpy instead of constructing
8508 __builtin_aarch64_simd_ci one vector at a time.
8509 (vqtbx3_u8): Likewise.
8510 (vqtbx3_p8): Likewise.
8511 (vqtbx3q_s8): Likewise.
8512 (vqtbx3q_u8): Likewise.
8513 (vqtbx3q_p8): Likewise.
8514 (vqtbx4_s8): Use __builtin_memcpy instead of constructing
8515 __builtin_aarch64_simd_xi one vector at a time.
8516 (vqtbx4_u8): Likewise.
8517 (vqtbx4_p8): Likewise.
8518 (vqtbx4q_s8): Likewise.
8519 (vqtbx4q_u8): Likewise.
8520 (vqtbx4q_p8): Likewise.
8522 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8524 * config/aarch64/arm_neon.h (vqtbl2_s8): Use __builtin_memcpy
8525 instead of constructing __builtin_aarch64_simd_oi one vector
8527 (vqtbl2_u8): Likewise.
8528 (vqtbl2_p8): Likewise.
8529 (vqtbl2q_s8): Likewise.
8530 (vqtbl2q_u8): Likewise.
8531 (vqtbl2q_p8): Likewise.
8532 (vqtbl3_s8): Use __builtin_memcpy instead of constructing
8533 __builtin_aarch64_simd_ci one vector at a time.
8534 (vqtbl3_u8): Likewise.
8535 (vqtbl3_p8): Likewise.
8536 (vqtbl3q_s8): Likewise.
8537 (vqtbl3q_u8): Likewise.
8538 (vqtbl3q_p8): Likewise.
8539 (vqtbl4_s8): Use __builtin_memcpy instead of constructing
8540 __builtin_aarch64_simd_xi one vector at a time.
8541 (vqtbl4_u8): Likewise.
8542 (vqtbl4_p8): Likewise.
8543 (vqtbl4q_s8): Likewise.
8544 (vqtbl4q_u8): Likewise.
8545 (vqtbl4q_p8): Likewise.
8547 2021-07-23 Haochen Gui <guihaoc@gcc.gnu.org>
8550 * config/rs6000/rs6000.md (cstore<mode>4): Fix wrong fall through.
8552 2021-07-22 Andrew Pinski <apinski@marvell.com>
8554 PR tree-optimization/10153
8555 * tree-tailcall.c (create_tailcall_accumulator):
8556 Don't call fold_convert as the type should be correct already.
8557 (tree_optimize_tail_calls_1): Use build_{one,zero}_cst instead
8558 of integer_{one,zero}_node for the call of create_tailcall_accumulator.
8560 2021-07-22 Aldy Hernandez <aldyh@redhat.com>
8562 * gimple-range-cache.cc (non_null_ref::adjust_range): Replace
8563 varying_p check for null/non-null check.
8565 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
8567 PR tree-optimization/101511
8568 * value-relation.cc (relation_oracle::query_relation): Check if ssa1
8569 is in ssa2's equiv set, and don't trap if so.
8571 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
8573 PR tree-optimization/101497
8574 * gimple-range-fold.cc (fold_using_range::range_of_cond_expr): Check
8577 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
8579 PR tree-optimization/101496
8580 * vr-values.c (simplify_using_ranges::fold_cond): Call range_of_stmt
8581 first, then vrp_visit_cond_Stmt.
8583 2021-07-22 liuhongt <hongtao.liu@intel.com>
8585 * config/i386/i386-expand.c
8586 (ix86_broadcast_from_integer_constant): Rename to ..
8587 (ix86_broadcast_from_constant): .. this, and extend it to
8589 (ix86_expand_vector_move): Extend to float mode.
8590 * config/i386/i386-features.c
8591 (replace_constant_pool_with_broadcast): Remove.
8592 (remove_partial_avx_dependency_gate): Ditto.
8593 (constant_pool_broadcast): Ditto.
8594 (class pass_constant_pool_broadcast): Ditto.
8595 (make_pass_constant_pool_broadcast): Ditto.
8596 (remove_partial_avx_dependency): Adjust gate.
8597 * config/i386/i386-passes.def: Remove pass_constant_pool_broadcast.
8598 * config/i386/i386-protos.h
8599 (make_pass_constant_pool_broadcast): Remove.
8601 2021-07-22 liuhongt <hongtao.liu@intel.com>
8603 * config/i386/constraints.md (Wb): New constraint.
8605 * config/i386/i386.md (*ashlhi3_1): Extend to avx512 mask
8607 (*ashlqi3_1): Ditto.
8608 (*<insn><mode>3_1): Split to ..
8609 (*ashr<mode>3_1): this, ...
8610 (*lshr<mode>3_1): and this, also extend this pattern to avx512
8612 (*<insn><mode>3_1): Split to ..
8613 (*ashr<mode>3_1): this, ...
8614 (*lshrqi3_1): and this, also extend this pattern to avx512
8616 (*lshrhi3_1): And this, also extend this pattern to avx512
8618 * config/i386/sse.md (k<code><mode>): New define_split after
8619 it to convert generic shift pattern to mask shift ones.
8621 2021-07-21 Thomas Schwinge <thomas@codesourcery.com>
8622 Joseph Myers <joseph@codesourcery.com>
8623 Cesar Philippidis <cesar@codesourcery.com>
8625 * tree-core.h (omp_clause_code): Add 'OMP_CLAUSE_NOHOST'.
8626 * tree.c (omp_clause_num_ops, omp_clause_code_name, walk_tree_1):
8628 * tree-pretty-print.c (dump_omp_clause): Likewise.
8629 * omp-general.c (oacc_verify_routine_clauses): Likewise.
8630 * gimplify.c (gimplify_scan_omp_clauses)
8631 (gimplify_adjust_omp_clauses): Likewise.
8632 * tree-nested.c (convert_nonlocal_omp_clauses)
8633 (convert_local_omp_clauses): Likewise.
8634 * omp-low.c (scan_sharing_clauses): Likewise.
8635 * omp-offload.c (execute_oacc_device_lower): Update.
8637 2021-07-21 Martin Sebor <msebor@redhat.com>
8639 * tree-ssa-alias.c (walk_aliased_vdefs_1): Fix typos in a comment.
8641 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8643 * config/rs6000/rs6000-gen-builtins.c (write_init_bif_table):
8646 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8648 * config/rs6000/rs6000-gen-builtins.c (write_fntype): New
8650 (write_fntype_init): New stub function.
8651 (write_init_bif_table): Likewise.
8652 (write_init_ovld_table): New function.
8653 (write_init_file): Implement.
8655 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8657 * config/rs6000/rs6000-gen-builtins.c
8658 (write_autogenerated_header): New function.
8659 (write_decls): Likewise.
8660 (write_extern_fntype): New callback function.
8661 (write_header_file): Implement.
8663 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8665 * config/rs6000/rs6000-gen-builtins.c (write_defines_file):
8668 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8670 * config/rs6000/rs6000-gen-builtins.c (complete_vector_type): New
8672 (complete_base_type): Likewise.
8673 (construct_fntype_id): Likewise.
8674 (parse_bif_entry): Call contruct_fntype_id.
8675 (parse_ovld_entry): Likewise.
8677 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8679 * config/rs6000/rs6000-gen-builtins.c (ovld_stanza): New struct.
8680 (MAXOVLDSTANZAS): New macro.
8681 (ovld_stanzas): New variable.
8682 (curr_ovld_stanza): Likewise.
8683 (MAXOVLDS): New macro.
8684 (ovlddata): New struct.
8685 (ovlds): New variable.
8686 (curr_ovld): Likewise.
8687 (max_ovld_args): Likewise.
8688 (parse_ovld_entry): New function.
8689 (parse_ovld_stanza): Likewise.
8690 (parse_ovld): Implement.
8692 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8694 * config/rs6000/rs6000-gen-builtins.c (parse_bif_attrs):
8697 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8699 * config/rs6000/rs6000-gen-builtins.c (parse_args): New function.
8700 (parse_prototype): Implement.
8702 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8704 * config/rs6000/rs6000-gen-builtins.c (bif_stanza): New enum.
8705 (curr_bif_stanza): New variable.
8706 (stanza_entry): New struct.
8707 (stanza_map): New initialized variable.
8708 (enable_string): Likewise.
8709 (fnkinds): New enum.
8710 (typelist): New struct.
8711 (attrinfo): Likewise.
8712 (MAXRESTROPNDS): New macro.
8713 (prototype): New struct.
8714 (MAXBIFS): New macro.
8715 (bifdata): New struct.
8716 (bifs): New variable.
8717 (curr_bif): Likewise.
8718 (bif_order): Likewise.
8719 (bif_index): Likewise.
8720 (fatal): New function.
8721 (stanza_name_to_stanza): Likewise.
8722 (parse_bif_attrs): New stub function.
8723 (parse_prototype): Likewise.
8724 (parse_bif_entry): New function.
8725 (parse_bif_stanza): Likewise.
8726 (parse_bif): Implement.
8727 (set_bif_order): New function.
8728 (create_bif_order): Implement.
8730 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8732 * config/rs6000/rs6000-gen-builtins.c (rbtree.h): New #include.
8733 (num_bifs): New variable.
8734 (num_ovld_stanzas): Likewise.
8735 (num_ovlds): Likewise.
8736 (parse_codes): New enum.
8737 (bif_rbt): New variable.
8738 (ovld_rbt): Likewise.
8739 (fntype_rbt): Likewise.
8740 (bifo_rbt): Likewise.
8741 (parse_bif): New stub function.
8742 (create_bif_order): Likewise.
8743 (parse_ovld): Likewise.
8744 (write_header_file): Likewise.
8745 (write_init_file): Likewise.
8746 (write_defines_file): Likewise.
8747 (delete_output_files): New function.
8750 2021-07-21 H.J. Lu <hjl.tools@gmail.com>
8753 * config/i386/i386-builtin.def: Remove OPTION_MASK_ISA_SSE4_2
8754 from CRC32 _builtin functions.
8756 2021-07-21 Sebastian Huber <sebastian.huber@embedded-brains.de>
8758 * coverage.c (build_gcov_info_var_registration): Mark the object placed
8759 in the linker set as referenced so that it does not get optimized away.
8761 2021-07-21 Kito Cheng <kito.cheng@sifive.com>
8764 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
8766 * config.gcc (riscv*-*-*): Detect which python is available.
8768 2021-07-21 Jakub Jelinek <jakub@redhat.com>
8770 PR middle-end/101535
8771 * gimplify.c (omp_check_private): Properly skip ORT_TARGET_DATA
8772 contexts in which decl isn't privatized and for ORT_TARGET return
8773 false if decl is mapped.
8775 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
8777 * gimple-loop-jam.c: Include tree-ssa-sccvn.h.
8778 (tree_loop_unroll_and_jam): Run value-numbering on a loop that
8779 has been successfully unrolled.
8781 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
8783 * tree-ssa-loop-manip.c (determine_exit_conditions): Return a null
8784 exit condition if no tail loop is needed, and if the original exit
8785 condition should therefore be kept as-is.
8786 (tree_transform_and_unroll_loop): Handle that case here too.
8788 2021-07-21 Kewen Lin <linkw@linux.ibm.com>
8790 * tree-data-ref.c (free_dependence_relations): Adjust to pass vec
8792 (free_data_refs): Likewise.
8793 * tree-data-ref.h (free_dependence_relations): Likewise.
8794 (free_data_refs): Likewise.
8795 * tree-predcom.c (struct chain): Use auto_vec instead of vec for
8797 (struct component): Likewise.
8798 (pcom_worker::pcom_worker): Adjust for auto_vec and renaming changes.
8799 (pcom_worker::~pcom_worker): Likewise.
8800 (pcom_worker::release_chain): Adjust as auto_vec changes.
8801 (pcom_worker::loop): Rename to ...
8802 (pcom_worker::m_loop): ... this.
8803 (pcom_worker::datarefs): Rename to ...
8804 (pcom_worker::m_datarefs): ... this. Use auto_vec instead of vec.
8805 (pcom_worker::dependences): Rename to ...
8806 (pcom_worker::m_dependences): ... this. Use auto_vec instead of vec.
8807 (pcom_worker::chains): Rename to ...
8808 (pcom_worker::m_chains): ... this. Use auto_vec instead of vec.
8809 (pcom_worker::looparound_phis): Rename to ...
8810 (pcom_worker::m_looparound_phis): ... this. Use auto_vec instead of
8812 (pcom_worker::cache): Rename to ...
8813 (pcom_worker::m_cache): ... this. Use auto_vec instead of vec.
8814 (pcom_worker::release_chain): Adjust for auto_vec changes.
8815 (pcom_worker::release_chains): Adjust for auto_vec and renaming
8817 (release_component): Remove.
8818 (release_components): Adjust for release_component removal.
8819 (component_of): Adjust to use vec.
8820 (merge_comps): Likewise.
8821 (pcom_worker::aff_combination_dr_offset): Adjust for renaming changes.
8822 (pcom_worker::determine_offset): Likewise.
8823 (class comp_ptrs): Remove.
8824 (pcom_worker::split_data_refs_to_components): Adjust for renaming
8825 changes, for comp_ptrs removal with auto_vec.
8826 (pcom_worker::suitable_component_p): Adjust for renaming changes.
8827 (pcom_worker::filter_suitable_components): Adjust for release_component
8829 (pcom_worker::valid_initializer_p): Adjust for renaming changes.
8830 (pcom_worker::find_looparound_phi): Likewise.
8831 (pcom_worker::add_looparound_copies): Likewise.
8832 (pcom_worker::determine_roots_comp): Likewise.
8833 (pcom_worker::single_nonlooparound_use): Likewise.
8834 (pcom_worker::execute_pred_commoning_chain): Likewise.
8835 (pcom_worker::execute_pred_commoning): Likewise.
8836 (pcom_worker::try_combine_chains): Likewise.
8837 (pcom_worker::prepare_initializers_chain): Likewise.
8838 (pcom_worker::prepare_initializers): Likewise.
8839 (pcom_worker::prepare_finalizers_chain): Likewise.
8840 (pcom_worker::prepare_finalizers): Likewise.
8841 (pcom_worker::tree_predictive_commoning_loop): Likewise.
8843 2021-07-20 Martin Sebor <msebor@redhat.com>
8845 PR middle-end/101397
8846 * builtins.c (gimple_call_return_array): Add argument. Correct
8847 offsets for memchr, mempcpy, stpcpy, and stpncpy.
8848 (compute_objsize_r): Adjust offset computation for argument returning
8851 2021-07-20 Martin Sebor <msebor@redhat.com>
8853 PR middle-end/101300
8854 * tree-ssa-uninit.c (check_defs): Handle UBSAN built-ins.
8856 2021-07-20 Jeff Law <jlaw@localhost.localdomain>
8858 * function.c (assign_parm_setup_block): Use adjust_address instead
8859 of change_address to preserve MEM_EXPR and friends.
8861 2021-07-20 Martin Sebor <msebor@redhat.com>
8863 * cfgloop.h (single_likely_exit): Adjust by-value argument to
8865 * cfgloopanal.c (single_likely_exit): Same.
8866 * cgraph.h (struct cgraph_node): Same.
8867 * cgraphclones.c (cgraph_node::create_virtual_clone): Same.
8868 * genautomata.c (merge_states): Same.
8869 * genextract.c (VEC_char_to_string): Same.
8870 * genmatch.c (dt_node::gen_kids_1): Same.
8871 (walk_captures): Adjust by-value argument to by-reference.
8872 * gimple-ssa-store-merging.c (check_no_overlap): Adjust by-value argument
8873 to by-const-reference.
8874 * gimple.c (gimple_build_call_vec): Same.
8875 (gimple_build_call_internal_vec): Same.
8876 (gimple_build_switch): Same.
8877 (sort_case_labels): Same.
8878 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
8880 * gimple.h (gimple_build_call_vec): Adjust by-value argument to
8882 (gimple_build_call_internal_vec): Same.
8883 (gimple_build_switch): Same.
8884 (sort_case_labels): Same.
8885 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
8887 * haifa-sched.c (calc_priorities): Adjust by-value argument to
8889 (sched_init_luids): Same.
8890 (haifa_init_h_i_d): Same.
8891 * ipa-cp.c (ipa_get_indirect_edge_target_1): Same.
8892 (adjust_callers_for_value_intersection): Adjust by-value argument to
8894 (find_more_scalar_values_for_callers_subset): Adjust by-value argument to
8896 (find_more_contexts_for_caller_subset): Same.
8897 (find_aggregate_values_for_callers_subset): Same.
8898 (copy_useful_known_contexts): Same.
8899 * ipa-fnsummary.c (remap_edge_summaries): Same.
8900 (remap_freqcounting_predicate): Same.
8901 * ipa-inline.c (add_new_edges_to_heap): Adjust by-value argument to
8903 * ipa-predicate.c (predicate::remap_after_inlining): Adjust by-value argument
8904 to by-const-reference.
8905 * ipa-predicate.h (predicate::remap_after_inlining): Same.
8906 * ipa-prop.c (ipa_find_agg_cst_for_param): Same.
8907 * ipa-prop.h (ipa_find_agg_cst_for_param): Same.
8908 * ira-build.c (ira_loop_tree_body_rev_postorder): Same.
8909 * read-rtl.c (add_overload_instance): Same.
8910 * rtl.h (native_decode_rtx): Same.
8911 (native_decode_vector_rtx): Same.
8912 * sched-int.h (sched_init_luids): Same.
8913 (haifa_init_h_i_d): Same.
8914 * simplify-rtx.c (native_decode_vector_rtx): Same.
8915 (native_decode_rtx): Same.
8916 * tree-call-cdce.c (gen_shrink_wrap_conditions): Same.
8917 (shrink_wrap_one_built_in_call_with_conds): Same.
8918 (shrink_wrap_conditional_dead_built_in_calls): Same.
8919 * tree-data-ref.c (create_runtime_alias_checks): Same.
8920 (compute_all_dependences): Same.
8921 * tree-data-ref.h (compute_all_dependences): Same.
8922 (create_runtime_alias_checks): Same.
8923 (index_in_loop_nest): Same.
8924 * tree-if-conv.c (mask_exists): Same.
8925 * tree-loop-distribution.c (class loop_distribution): Same.
8926 (loop_distribution::create_rdg_vertices): Same.
8927 (dump_rdg_partitions): Same.
8928 (debug_rdg_partitions): Same.
8929 (partition_contains_all_rw): Same.
8930 (loop_distribution::distribute_loop): Same.
8931 * tree-parloops.c (oacc_entry_exit_ok_1): Same.
8932 (oacc_entry_exit_single_gang): Same.
8933 * tree-ssa-loop-im.c (hoist_memory_references): Same.
8934 (loop_suitable_for_sm): Same.
8935 * tree-ssa-loop-niter.c (bound_index): Same.
8936 * tree-ssa-reassoc.c (update_ops): Same.
8937 (swap_ops_for_binary_stmt): Same.
8938 (rewrite_expr_tree): Same.
8939 (rewrite_expr_tree_parallel): Same.
8940 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Same.
8941 * tree-ssa-sccvn.h (ao_ref_init_from_vn_reference): Same.
8942 * tree-ssa-structalias.c (process_all_all_constraints): Same.
8943 (make_constraints_to): Same.
8944 (handle_lhs_call): Same.
8945 (find_func_aliases_for_builtin_call): Same.
8946 (sort_fieldstack): Same.
8947 (check_for_overlaps): Same.
8948 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Same.
8949 (vect_create_cond_for_unequal_addrs): Same.
8950 (vect_create_cond_for_lower_bounds): Same.
8951 (vect_create_cond_for_alias_checks): Same.
8952 * tree-vect-slp-patterns.c (vect_validate_multiplication): Same.
8953 * tree-vect-slp.c (vect_analyze_slp_instance): Same.
8954 (vect_make_slp_decision): Same.
8955 (vect_slp_bbs): Same.
8956 (duplicate_and_interleave): Same.
8957 (vect_transform_slp_perm_load): Same.
8958 (vect_schedule_slp): Same.
8959 * tree-vectorizer.h (vect_transform_slp_perm_load): Same.
8960 (vect_schedule_slp): Same.
8961 (duplicate_and_interleave): Same.
8962 * tree.c (build_vector_from_ctor): Same.
8963 (build_vector): Same.
8964 (check_vector_cst): Same.
8965 (check_vector_cst_duplicate): Same.
8966 (check_vector_cst_fill): Same.
8967 (check_vector_cst_stepped): Same.
8968 * tree.h (build_vector_from_ctor): Same.
8970 2021-07-20 Jakub Jelinek <jakub@redhat.com>
8973 * config/rs6000/rs6000-protos.h (easy_altivec_constant): Change return
8974 type from bool to int.
8975 * config/rs6000/rs6000.c (vspltis_constant): Fix up handling the
8976 EASY_VECTOR_MSB case if either step or copies is not 1.
8977 (vspltis_shifted): Fix comment typo.
8978 (easy_altivec_constant): Change return type from bool to int, instead
8979 of returning true return byte size of the element mode that should be
8980 used to synthetize the constant.
8981 * config/rs6000/predicates.md (easy_vector_constant_msb): Require
8982 that vspltis_shifted is 0, handle the case where easy_altivec_constant
8983 assumes using different vector mode from CONST_VECTOR's mode.
8984 * config/rs6000/altivec.md (easy_vector_constant_msb splitter): Use
8985 easy_altivec_constant to determine mode in which -1 >> -1 should be
8986 performed, use rs6000_expand_vector_init instead of gen_vec_initv4sisi.
8988 2021-07-20 Richard Biener <rguenther@suse.de>
8991 * dwarf2out.h (dwarf_file_data): Add key member.
8992 * dwarf2out.c (dwarf_file_hasher::equal): Compare key.
8993 (dwarf_file_hasher::hash): Hash key.
8994 (lookup_filename): Remap the filename and store it in the
8995 filename member of dwarf_file_data when creating a new
8997 (file_name_acquire): Do not remap the filename again.
8998 (maybe_emit_file): Likewise.
9000 2021-07-20 Jonathan Wright <jonathan.wright@arm.com>
9002 * config/aarch64/aarch64-simd-builtins.def: Use two variant
9003 generators for all TBL/TBX intrinsics and rename to
9004 consistent forms: qtbl[1234] or qtbx[1234].
9005 * config/aarch64/aarch64-simd.md (aarch64_tbl1<mode>):
9007 (aarch64_qtbl1<mode>): This.
9008 (aarch64_tbx1<mode>): Rename to...
9009 (aarch64_qtbx1<mode>): This.
9010 (aarch64_tbl2v16qi): Delete.
9011 (aarch64_tbl3<mode>): Rename to...
9012 (aarch64_qtbl2<mode>): This.
9013 (aarch64_tbx4<mode>): Rename to...
9014 (aarch64_qtbx2<mode>): This.
9015 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_1): Use
9016 renamed qtbl1 and qtbl2 RTL patterns.
9017 * config/aarch64/arm_neon.h (vqtbl1_p8): Use renamed qtbl1
9019 (vqtbl1_s8): Likewise.
9020 (vqtbl1_u8): Likewise.
9021 (vqtbl1q_p8): Likewise.
9022 (vqtbl1q_s8): Likewise.
9023 (vqtbl1q_u8): Likewise.
9024 (vqtbx1_s8): Use renamed qtbx1 RTL pattern.
9025 (vqtbx1_u8): Likewise.
9026 (vqtbx1_p8): Likewise.
9027 (vqtbx1q_s8): Likewise.
9028 (vqtbx1q_u8): Likewise.
9029 (vqtbx1q_p8): Likewise.
9030 (vtbl1_s8): Use renamed qtbl1 RTL pattern.
9031 (vtbl1_u8): Likewise.
9032 (vtbl1_p8): Likewise.
9033 (vtbl2_s8): Likewise
9034 (vtbl2_u8): Likewise.
9035 (vtbl2_p8): Likewise.
9036 (vtbl3_s8): Use renamed qtbl2 RTL pattern.
9037 (vtbl3_u8): Likewise.
9038 (vtbl3_p8): Likewise.
9039 (vtbl4_s8): Likewise.
9040 (vtbl4_u8): Likewise.
9041 (vtbl4_p8): Likewise.
9042 (vtbx2_s8): Use renamed qtbx2 RTL pattern.
9043 (vtbx2_u8): Likewise.
9044 (vtbx2_p8): Likewise.
9045 (vqtbl2_s8): Use renamed qtbl2 RTL pattern.
9046 (vqtbl2_u8): Likewise.
9047 (vqtbl2_p8): Likewise.
9048 (vqtbl2q_s8): Likewise.
9049 (vqtbl2q_u8): Likewise.
9050 (vqtbl2q_p8): Likewise.
9051 (vqtbx2_s8): Use renamed qtbx2 RTL pattern.
9052 (vqtbx2_u8): Likewise.
9053 (vqtbx2_p8): Likewise.
9054 (vqtbx2q_s8): Likewise.
9055 (vqtbx2q_u8): Likewise.
9056 (vqtbx2q_p8): Likewise.
9057 (vtbx4_s8): Likewise.
9058 (vtbx4_u8): Likewise.
9059 (vtbx4_p8): Likewise.
9061 2021-07-20 Uroš Bizjak <ubizjak@gmail.com>
9064 * config/i386/sync.md (define_peephole2 atomic_storedi_fpu):
9066 (define_peephole2 atomic_loaddi_fpu): Ditto.
9068 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
9070 * config.gcc (riscv*-*-*): Detect which python is available.
9072 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
9074 * config/rs6000/vsx.md (mulhs_<mode>): Rename to...
9075 (smul<mode>3_highpart): ... this.
9076 (mulhu_<mode>): Rename to...
9077 (umul<mode>3_highpart): ... this.
9078 * config/rs6000/rs6000-builtin.def (MULHS_V2DI, MULHS_V4SI,
9079 MULHU_V2DI, MULHU_V4SI): Adjust.
9081 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
9083 PR tree-optimization/100696
9084 * internal-fn.c (first_commutative_argument): Add info for IFN_MULH.
9085 * internal-fn.def (IFN_MULH): New internal function.
9086 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Add support to
9087 recog normal multiply highpart as IFN_MULH.
9088 * config/i386/i386.c (ix86_add_stmt_cost): Adjust for combined
9091 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
9093 * config/elfos.h (CTF_DEBUGGING_INFO): New definition.
9094 (BTF_DEBUGGING_INFO): Likewise.
9095 * doc/tm.texi.in: Document the new macros.
9096 * doc/tm.texi: Regenerated.
9097 * toplev.c: Guard initialization of debug hooks.
9099 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
9101 * flags.h (ctf_debuginfo_p): New function declaration.
9102 * opts.c (ctf_debuginfo_p): New function definition.
9104 2021-07-19 Andrew Stubbs <ams@codesourcery.com>
9107 * config/gcn/gcn-hsa.h (DRIVER_SELF_SPECS): New.
9108 (ASM_SPEC): Set -mattr for xnack and sram-ecc.
9109 * config/gcn/gcn-opts.h (enum sram_ecc_type): New.
9110 * config/gcn/gcn-valu.md: Add a warning comment.
9111 * config/gcn/gcn.c (gcn_option_override): Add "sorry" for -mxnack.
9112 (output_file_start): Add xnack and sram-ecc state to ".amdgcn_target".
9113 * config/gcn/gcn.md: Add a warning comment.
9114 * config/gcn/gcn.opt: Add -mxnack and -msram-ecc.
9115 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): Remove
9117 (EF_AMDGPU_XNACK): New.
9118 (EF_AMDGPU_SRAM_ECC): New.
9120 (copy_early_debug_info): Use elf_flags.
9121 (main): Handle -mxnack and -msram-ecc options.
9122 * doc/invoke.texi: Document -mxnack and -msram-ecc.
9124 2021-07-19 Andrew Pinski <apinski@marvell.com>
9127 * config/aarch64/aarch64.md (csneg3_uxtw_insn): Rename to ...
9128 (*cs<neg_not_cs>3_uxtw_insn4): and extend to NEG_NOT.
9130 2021-07-19 Richard Biener <rguenther@suse.de>
9132 PR tree-optimization/101505
9133 * tree-vect-patterns.c (vect_determine_precisions): Walk
9134 PHIs also for loop vectorization.
9136 2021-07-19 Richard Biener <rguenther@suse.de>
9138 * gimple.h (gimple_expr_type): Remove.
9139 * doc/gimple.texi: Remove gimple_expr_type documentation.
9141 2021-07-19 Richard Biener <rguenther@suse.de>
9143 * tree-ssa-sccvn.c (vn_reference_eq): Handle NULL vr->type.
9144 (ao_ref_init_from_vn_reference): Likewise.
9145 (fully_constant_reference): Likewise.
9146 (vn_reference_lookup_call): Do not set vr->type to random
9148 * tree-ssa-pre.c (compute_avail): Do not try to PRE calls
9150 * tree-vect-generic.c (expand_vector_piecewise): Pass in
9151 whether we expanded parallel.
9152 (expand_vector_parallel): Adjust.
9153 (expand_vector_addition): Likewise.
9154 (expand_vector_comparison): Likewise.
9155 (expand_vector_operation): Likewise.
9156 (expand_vector_scalar_condition): Likewise.
9157 (expand_vector_conversion): Likewise.
9159 2021-07-19 Richard Biener <rguenther@suse.de>
9161 * tree-vrp.c (register_edge_assert_for_2): Use the
9163 (vrp_folder::fold_predicate_in): Likewise.
9164 * vr-values.c (gimple_assign_nonzero_p): Likewise.
9165 (vr_values::extract_range_from_comparison): Likewise.
9166 (vr_values::extract_range_from_ubsan_builtin): Use the
9167 type of the first operand.
9168 (vr_values::extract_range_basic): Push down type
9169 computation, use the appropriate LHS.
9170 (vr_values::extract_range_from_assignment): Use the
9173 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
9176 * common/config/i386/i386-common.c (ix86_handle_option): For
9177 -mgeneral-regs-only, enable the GPR only instructions which are
9178 enabled implicitly by SSE ISAs unless they have been disabled
9181 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
9184 * config/i386/i386.c (ix86_check_avx_upper_stores): Moved before
9185 ix86_avx_u128_mode_needed.
9186 (ix86_avx_u128_mode_needed): Return AVX_U128_DIRTY if callee
9187 returns AVX register.
9189 2021-07-17 Jan Hubicka <hubicka@ucw.cz>
9191 * tree-ssa-structalias.c (handle_rhs_call): Support EAF_NOT_RETURNED.
9192 (handle_const_call): Liekise
9193 (handle_pure_call): Liekise
9195 2021-07-17 Andrew MacLeod <amacleod@redhat.com>
9197 PR tree-optimization/96542
9198 * range-op.cc (range_operator::wi_fold_in_parts): New.
9199 (range_operator::fold_range): Call wi_fold_in_parts.
9200 (operator_lshift::wi_fold): Fix broken lshift by [0,0].
9201 * range-op.h (wi_fold_in_parts): Add prototype.
9203 2021-07-16 David Malcolm <dmalcolm@redhat.com>
9205 * doc/analyzer.texi: Add __analyzer_dump_state.
9207 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9209 * config/rs6000/rbtree.c: New file.
9210 * config/rs6000/rbtree.h: New file.
9212 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9214 * config/rs6000/rs6000-gen-builtins.c (restriction): New enum.
9215 (typeinfo): Add restr field.
9216 (match_bracketed_pair): New function.
9217 (match_const_restriction): Implement.
9219 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9221 * config/rs6000/rs6000-gen-builtins.c (match_basetype): Implement.
9223 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9225 * config/rs6000/rs6000-gen-builtins.c (void_status): New enum.
9226 (basetype): Likewise.
9227 (typeinfo): Likewise.
9228 (handle_pointer): New function.
9229 (match_basetype): New stub function.
9230 (match_const_restriction): Likewise.
9231 (match_type): New function.
9233 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9235 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace): New
9237 (advance_line): Likewise.
9238 (safe_inc_pos): Likewise.
9239 (match_identifier): Likewise.
9240 (match_integer): Likewise.
9241 (match_to_right_bracket): Likewise.
9243 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9245 * config/rs6000/rs6000-gen-builtins.c (bif_file): New variable.
9246 (ovld_file): Likewise.
9247 (header_file): Likewise.
9248 (init_file): Likewise.
9249 (defines_file): Likewise.
9250 (pgm_path): Likewise.
9251 (bif_path): Likewise.
9252 (ovld_path): Likewise.
9253 (header_path): Likewise.
9254 (init_path): Likewise.
9255 (defines_path): Likewise.
9256 (LINELEN): New macro.
9257 (linebuf): New variable.
9261 (bif_diag): New function.
9262 (ovld_diag): Likewise.
9264 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9266 * config/rs6000/rs6000-builtin-new.def: New.
9267 * config/rs6000/rs6000-overload.def: New.
9269 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9271 * config/rs6000/rs6000-gen-builtins.c: New.
9273 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9275 * Makefile.in (EXTRA_GTYPE_DEPS): New variable.
9276 (s-gtype): Depend on EXTRA_GTYPE_DEPS.
9277 * gengtype-state.c (state_writer::write_state_file_list): Add a
9278 parameter to the fileslist expression for the number of build
9280 (read_state_files_list): Detect build headers and strip the
9281 initial "./" or ".\" from their names.
9282 * gengtype.c (build_headers): New global variable.
9283 (num_build_headers): Likewise.
9284 (open_base_files): Emit #include for each build header.
9285 (main): Detect and count build headers.
9286 * gengtype.h (build_headers): New extern variable.
9287 (num_build_headers): Likewise.
9289 2021-07-16 Richard Biener <rguenther@suse.de>
9291 * gimple-ssa-store-merging.c (verify_symbolic_number_p): Use
9292 the type of the LHS.
9293 (find_bswap_or_nop_1): Likewise.
9294 (find_bswap_or_nop): Likewise.
9295 * tree-vectorizer.h (vect_get_smallest_scalar_type): Adjust
9297 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
9298 Remove unused parameters, pass in the scalar type. Fix
9299 internal store function handling.
9300 * tree-vect-stmts.c (vect_analyze_stmt): Remove assert.
9301 (vect_get_vector_types_for_stmt): Move down check for
9302 existing vector stmt after we've determined a scalar type.
9303 Pass down the used scalar type to vect_get_smallest_scalar_type.
9304 * tree-vect-generic.c (expand_vector_condition): Use
9305 the type of the LHS.
9306 (expand_vector_scalar_condition): Likewise.
9307 (expand_vector_operations_1): Likewise.
9308 * tree-vect-patterns.c (vect_widened_op_tree): Likewise.
9309 (vect_recog_dot_prod_pattern): Likewise.
9310 (vect_recog_sad_pattern): Likewise.
9311 (vect_recog_widen_op_pattern): Likewise.
9312 (vect_recog_widen_sum_pattern): Likewise.
9313 (vect_recog_mixed_size_cond_pattern): Likewise.
9315 2021-07-16 Jan Hubicka <hubicka@ucw.cz>
9317 * ipa-modref.c (struct escape_entry): Use eaf_fleags_t.
9318 (dump_eaf_flags): Dump EAF_NOT_RETURNED
9319 (eaf_flags_useful_p): Use eaf_fleags_t; handle const functions
9320 and EAF_NOT_RETURNED.
9321 (modref_summary::useful_p): Likewise.
9322 (modref_summary_lto::useful_p): Likewise.
9323 (struct) modref_summary_lto: Use eaf_fleags_t.
9324 (deref_flags): Handle EAF_NOT_RETURNED.
9325 (struct escape_point): Use min_flags.
9326 (modref_lattice::init): Add EAF_NOT_RETURNED.
9327 (merge_call_lhs_flags): Ignore EAF_NOT_RETURNED functions
9328 (analyze_ssa_name_flags): Clear EAF_NOT_RETURNED on return;
9330 (analyze_parms): Also analyze const functions; update conition on
9332 (modref_write): Update streaming.
9333 (read_section): Update streaming.
9334 (remap_arg_flags): Use eaf_flags_t.
9335 (modref_merge_call_site_flags): Hanlde EAF_NOT_RETURNED.
9336 * ipa-modref.h: (eaf_flags_t): New typedef.
9337 (struct modref_summary): Use eaf_flags_t.
9338 * tree-core.h (EAF_NOT_RETURNED): New constant.
9340 2021-07-16 Richard Biener <rguenther@suse.de>
9342 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Use
9343 the type of the LHS.
9344 (gimple_assign_nonnegative_warnv_p): Likewise.
9345 (gimple_call_nonnegative_warnv_p): Likewise. Return false
9346 if the call has no LHS.
9347 * gimple.c (gimple_could_trap_p_1): Use the type of the LHS.
9348 * tree-eh.c (stmt_could_throw_1_p): Likewise.
9349 * tree-inline.c (insert_init_stmt): Likewise.
9350 * tree-ssa-loop-niter.c (get_val_for): Likewise.
9351 * tree-outof-ssa.c (ssa_is_replaceable_p): Use the type of
9353 * tree-ssa-sccvn.c (init_vn_nary_op_from_stmt): Take a
9354 gassign *. Use the type of the lhs.
9355 (vn_nary_op_lookup_stmt): Adjust.
9356 (vn_nary_op_insert_stmt): Likewise.
9358 2021-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
9360 * config/s390/predicates.md (bras_sym_operand): Accept all
9361 functions in 64-bit mode, use UNSPEC_PLT31.
9362 (larl_operand): Use UNSPEC_PLT31.
9363 * config/s390/s390.c (s390_loadrelative_operand_p): Likewise.
9364 (legitimize_pic_address): Likewise.
9365 (s390_emit_tls_call_insn): Mark __tls_get_offset as function,
9367 (s390_delegitimize_address): Use UNSPEC_PLT31.
9368 (s390_output_addr_const_extra): Likewise.
9369 (print_operand): Add @PLT to TLS calls, handle %K.
9370 (s390_function_profiler): Mark __fentry__/_mcount as function,
9371 use %K, use UNSPEC_PLT31.
9372 (s390_output_mi_thunk): Use only UNSPEC_GOT, use %K.
9373 (s390_emit_call): Use UNSPEC_PLT31.
9374 (s390_emit_tpf_eh_return): Mark __tpf_eh_return as function.
9375 * config/s390/s390.md (UNSPEC_PLT31): Rename from UNSPEC_PLT.
9376 (*movdi_64): Use %K.
9377 (reload_base_64): Likewise.
9378 (*sibcall_brc): Likewise.
9379 (*sibcall_brcl): Likewise.
9380 (*sibcall_value_brc): Likewise.
9381 (*sibcall_value_brcl): Likewise.
9384 (*bras_r): Likewise.
9385 (*brasl_r): Likewise.
9386 (*bras_tls): Likewise.
9387 (*brasl_tls): Likewise.
9388 (main_base_64): Likewise.
9389 (reload_base_64): Likewise.
9390 (@split_stack_call<mode>): Likewise.
9392 2021-07-16 Richard Biener <rguenther@suse.de>
9394 PR tree-optimization/101467
9395 * tree-vect-stmts.c (vect_gen_while): Properly guard
9396 make_temp_ssa_name usage.
9398 2021-07-16 Cooper Qu <cooper.qu@linux.alibaba.com>
9400 * config.gcc: Don't use forked print-sysroot-suffix.sh and
9401 t-sysroot-suffix for C-SKY.
9402 * config/csky/print-sysroot-suffix.sh: Delete.
9403 * config/csky/t-csky-linux: Delete.
9404 * config/csky/t-sysroot-suffix: Define MULTILIB_DIRNAMES
9405 instead of CSKY_MULTILIB_DIRNAMES.
9407 2021-07-16 Richard Biener <rguenther@suse.de>
9409 * tree-vect-loop.c (vect_transform_cycle_phi): Correct sign
9410 conversion issues with the partial reduction of the reused
9413 2021-07-16 Richard Biener <rguenther@suse.de>
9415 * config/i386/i386-options.c (ix86_option_override_internal): Set
9416 param_vect_partial_vector_usage to zero if not set.
9418 2021-07-15 Uroš Bizjak <ubizjak@gmail.com>
9421 * config/i386/i386.h (VALID_SSE_REG_MODE): Add TDmode.
9422 (VALID_INT_MODE_P): Add SDmode and DDmode.
9423 Add TDmode for TARGET_64BIT.
9424 (VALID_DFP_MODE_P): Remove.
9425 * config/i386/i386.c (ix86_hard_regno_mode_ok):
9426 Do not use VALID_DFP_MODE_P.
9428 2021-07-15 Andrew MacLeod <amacleod@redhat.com>
9430 * gimple-range-fold.cc (adjust_pointer_diff_expr): Use
9432 (fold_using_range::fold_stmt): Ditto.
9433 (fold_using_range::range_of_range_op): Ditto.
9434 (fold_using_range::range_of_phi): Ditto.
9435 (fold_using_range::range_of_call): Ditto.
9436 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
9437 (fold_using_range::range_of_builtin_call): Ditto.
9438 (fold_using_range::range_of_cond_expr): Ditto.
9439 * gimple-range-fold.h (gimple_range_type): New.
9441 2021-07-15 Martin Sebor <msebor@redhat.com>
9444 * tree-ssa-strlen.c (handle_assign): New function.
9445 (maybe_warn_overflow): Add argument.
9446 (nonzero_bytes_for_type): New function.
9447 (count_nonzero_bytes): Handle more tree types. Call
9448 nonzero_bytes_for_tye.
9449 (count_nonzero_bytes): Handle types.
9450 (handle_store): Handle stores from function calls.
9451 (strlen_check_and_optimize_call): Move code to handle_assign. Call
9452 it for assignments from function calls.
9454 2021-07-15 David Malcolm <dmalcolm@redhat.com>
9459 * doc/invoke.texi: Add -Wanalyzer-use-of-uninitialized-value.
9461 2021-07-15 David Malcolm <dmalcolm@redhat.com>
9463 * doc/invoke.texi (-fdump-analyzer-exploded-paths): New.
9465 2021-07-15 Martin Sebor <msebor@redhat.com>
9469 * fold-const.c (operand_compare::operand_equal_p): Handle OEP_DECL_NAME.
9470 (operand_compare::verify_hash_value): Same.
9471 * tree-core.h (OEP_DECL_NAME): New.
9473 2021-07-15 Martin Jambor <mjambor@suse.cz>
9475 * profile-count.h (profile_count::value): Change the return type to
9477 * gimple-pretty-print.c (dump_gimple_bb_header): Adjust print
9479 * tree-cfg.c (dump_function_to_file): Likewise.
9481 2021-07-15 Bill Schmidt <wschmidt@linux.ibm.com>
9484 * config/rs6000/rs6000-p8swap.c (has_part_mult): New.
9485 (rs6000_analyze_swaps): Insns containing a subreg of a mult are
9488 2021-07-15 Richard Biener <rguenther@suse.de>
9490 * tree-vectorizer.h (vect_gen_while): Match up with
9492 * tree-vect-stmts.c (vect_gen_while): Adjust API to that
9493 of vect_gen_while_not.
9494 (vect_gen_while_not): Adjust.
9495 * tree-vect-loop-manip.c (vect_set_loop_controls_directly): Likewise.
9497 2021-07-15 Aldy Hernandez <aldyh@redhat.com>
9499 * gimple-range-cache.cc (non_null_ref::adjust_range): New.
9500 (ranger_cache::range_of_def): Call adjust_range.
9501 (ranger_cache::entry_range): Same.
9502 * gimple-range-cache.h (non_null_ref::adjust_range): New.
9503 * gimple-range.cc (gimple_ranger::range_of_expr): Call
9505 (gimple_ranger::range_on_entry): Same.
9507 2021-07-15 Tamar Christina <tamar.christina@arm.com>
9510 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9512 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
9514 2021-07-15 Tamar Christina <tamar.christina@arm.com>
9517 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9519 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
9520 (sdot_prod, udot_prod): ...These.
9521 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
9522 (aarch64_<sur>dot<vsi2qi>): Rename to...
9523 (<sur>dot_prod<vsi2qi>): ...This.
9524 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
9527 2021-07-15 Jakub Jelinek <jakub@redhat.com>
9529 PR middle-end/101437
9530 * gimplify.c (gimplify_expr): Throw away volatile reads from empty
9531 types even if they have non-BLKmode TYPE_MODE.
9533 2021-07-15 Richard Biener <rguenther@suse.de>
9536 * gcc.c (process_command): Process -gtoggle like process_options
9537 would after parsing options.
9539 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
9541 * cfgexpand.c (expand_asm_loc): Adjust.
9542 (expand_asm_stmt): Likewise.
9543 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Likewise.
9544 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
9545 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
9546 * config/avr/avr.c (avr_md_asm_adjust): Likewise.
9547 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
9548 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
9549 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
9550 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
9551 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
9552 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
9553 * config/s390/s390.c (s390_md_asm_adjust): Likewise.
9554 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
9555 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
9556 * doc/tm.texi: Regenerate.
9557 * target.def: Add location argument to md_asm_adjust.
9559 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
9561 * tree-diagnostic.c (diagnostic_report_current_function): Use the
9562 diagnostic's location, not input_location.
9564 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
9566 * cfgexpand.c (tree_conflicts_with_clobbers_p): Pass location to
9568 (expand_asm_stmt): Likewise.
9570 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
9572 * config/rs6000/rs6000.c (adjacent_mem_locations): Return the lower
9573 addressed memory rtx, if any.
9574 (rs6000_split_multireg_move): Fix code formatting.
9575 Handle MMA build built-ins with operands in adjacent memory locations.
9577 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
9579 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Move to later
9582 2021-07-14 Jason Merrill <jason@redhat.com>
9584 * sel-sched-ir.h (get_all_loop_exits): Use auto_vec.
9586 2021-07-14 Jason Merrill <jason@redhat.com>
9588 * doc/invoke.texi: -fdelete-dead-exceptions is on by default for
9591 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9593 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
9594 Remove erroneous line.
9596 2021-07-14 Andrew MacLeod <amacleod@redhat.com>
9598 * params.opt (param_evrp_mode): Change default.
9600 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9602 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
9603 (sdot_prod, udot_prod): ...These.
9604 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
9605 (aarch64_<sur>dot<vsi2qi>): Rename to...
9606 (<sur>dot_prod<vsi2qi>): ...This.
9607 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
9610 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9612 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
9614 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9616 * doc/sourcebuild.texi (arm_v8_2a_i8mm_neon_hw): Document.
9618 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9620 * config/arm/neon.md (usdot_prod<vsi2qi>): New.
9622 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9624 * config/aarch64/aarch64-simd.md (aarch64_usdot<vsi2qi>): Rename to...
9625 (usdot_prod<vsi2qi>): ... This.
9626 * config/aarch64/aarch64-simd-builtins.def (usdot): Rename to...
9627 (usdot_prod): ...This.
9628 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Likewise.
9629 * config/aarch64/aarch64-sve.md (@aarch64_<sur>dot_prod<vsi2qi>):
9631 (@<sur>dot_prod<vsi2qi>): ...This.
9632 * config/aarch64/aarch64-sve-builtins-base.cc
9633 (svusdot_impl::expand): Use it.
9635 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9637 * optabs.def (usdot_prod_optab): New.
9638 * doc/md.texi: Document it and clarify other dot prod optabs.
9639 * optabs-tree.h (enum optab_subtype): Add optab_vector_mixed_sign.
9640 * optabs-tree.c (optab_for_tree_code): Support usdot_prod_optab.
9641 * optabs.c (expand_widen_pattern_expr): Likewise.
9642 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
9643 * tree-vect-loop.c (vectorizable_reduction): Query dot-product kind.
9644 * tree-vect-patterns.c (vect_supportable_direct_optab_p): Take optional
9646 (vect_widened_op_tree): Optionally ignore
9648 (vect_recog_dot_prod_pattern): Support usdot_prod_optab.
9650 2021-07-14 H.J. Lu <hjl.tools@gmail.com>
9653 * config/i386/driver-i386.c (host_detect_local_cpu): Check
9654 "arch [32|64]" and "tune [32|64]" for 32-bit and 64-bit codegen.
9655 Enable UINTR only for 64-bit codegen.
9656 * config/i386/i386-options.c
9657 (ix86_option_override_internal::DEF_PTA): Skip PTA_UINTR if not
9659 * config/i386/i386.h (ARCH_ARG): New.
9660 (CC1_CPU_SPEC): Pass "[arch|tune] 32" for 32-bit codegen and
9661 "[arch|tune] 64" for 64-bit codegen.
9663 2021-07-14 Richard Biener <rguenther@suse.de>
9665 PR tree-optimization/101445
9666 * tree-vect-stmts.c (vectorizable_load): Do the gap adjustment
9667 of the IV in the correct direction for negative stride
9670 2021-07-14 Jakub Jelinek <jakub@redhat.com>
9673 * godump.c (godump_str_hash): New type.
9674 (godump_container::pot_dummy_types): Use string_hash instead of
9675 ptr_hash in the hash_set.
9677 2021-07-14 Richard Biener <rguenther@suse.de>
9679 * tree-vect-loop.c (vect_find_reusable_accumulator): Handle
9680 vector types where the old vector type has a multiple of
9681 the new vector type elements.
9682 (vect_create_partial_epilog): New function, split out from...
9683 (vect_create_epilog_for_reduction): ... here.
9684 (vect_transform_cycle_phi): Reduce the re-used accumulator
9685 to the new vector type.
9687 2021-07-14 Alexandre Oliva <oliva@adacore.com>
9689 * tree-ssa-alias.c (attr_fnspec::verify): Fix index in
9690 non-'t'-sized arg check.
9692 2021-07-14 Alexandre Oliva <oliva@adacore.com>
9694 * tree-cfg.c (cleanup_dead_labels_eh): Update
9695 post_landing_pad label upon change of landing pad block's
9697 (cleanup_dead_labels): Check that a removed label is not that
9700 2021-07-13 Jonathan Wright <jonathan.wright@arm.com>
9702 * combine.c (combine_simplify_rtx): Add vec_select -> subreg
9704 * config/aarch64/aarch64.md (*zero_extend<SHORT:mode><GPI:mode>2_aarch64):
9705 Add Neon to general purpose register case for zero-extend
9707 * config/arm/vfp.md (*arm_movsi_vfp): Remove "*" from *t -> r
9708 case to prevent some cases opting to go through memory.
9709 * cse.c (fold_rtx): Add vec_select -> subreg simplification.
9710 * rtl.c (rtvec_series_p): Define predicate to determine
9711 whether a vector contains a linear series of integers.
9712 * rtl.h (rtvec_series_p): Define.
9713 * rtlanal.c (vec_series_lowpart_p): Define predicate to
9714 determine if a vector selection is equivalent to the low part
9716 * rtlanal.h (vec_series_lowpart_p): Define.
9717 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
9718 Add vec_select -> subreg simplification.
9720 2021-07-13 Paul A. Clarke <pc@us.ibm.com>
9722 * config/rs6000/smmintrin.h (_mm_testz_si128, _mm_testc_si128,
9723 _mm_testnzc_si128, _mm_test_all_ones, _mm_test_all_zeros,
9724 _mm_test_mix_ones_zeros): New.
9726 2021-07-13 Roger Sayle <roger@nextmovesoftware.com>
9727 Richard Biener <rguenther@suse.de>
9729 * gimple.c (gimple_could_trap_p_1): Make S argument a
9730 "const gimple*". Preserve constness in call to
9731 gimple_asm_volatile_p.
9732 (gimple_could_trap_p): Make S argument a "const gimple*".
9733 * gimple.h (gimple_could_trap_p_1, gimple_could_trap_p):
9734 Update function prototypes.
9736 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9738 * tree-vectorizer.h (vect_reusable_accumulator): New structure.
9739 (_loop_vec_info::main_loop_edge): New field.
9740 (_loop_vec_info::skip_main_loop_edge): Likewise.
9741 (_loop_vec_info::skip_this_loop_edge): Likewise.
9742 (_loop_vec_info::reusable_accumulators): Likewise.
9743 (_stmt_vec_info::reduc_scalar_results): Likewise.
9744 (_stmt_vec_info::reused_accumulator): Likewise.
9745 (vect_get_main_loop_result): Declare.
9746 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize
9747 reduc_scalar_inputs.
9748 (vec_info::free_stmt_vec_info): Free reduc_scalar_inputs.
9749 * tree-vect-loop-manip.c (vect_get_main_loop_result): New function.
9750 (vect_do_peeling): Fill an epilogue loop's main_loop_edge,
9751 skip_main_loop_edge and skip_this_loop_edge fields.
9752 * tree-vect-loop.c (INCLUDE_ALGORITHM): Define.
9753 (vect_emit_reduction_init_stmts): New function.
9754 (get_initial_def_for_reduction): Use it.
9755 (get_initial_defs_for_reduction): Likewise. Change the vinfo
9756 parameter to a loop_vec_info.
9757 (vect_create_epilog_for_reduction): Store the scalar results
9758 in the reduc_info. If an epilogue loop is reusing an accumulator
9759 from the main loop, and if the epilogue loop can also be skipped,
9760 try to place the reduction code in the join block. Record
9761 accumulators that could potentially be reused by epilogue loops.
9762 (vect_transform_cycle_phi): When vectorizing epilogue loops,
9763 try to reuse accumulators from the main loop. Record the initial
9764 value in reduc_info for non-SLP reductions too.
9766 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9768 * tree-vect-loop.c (get_initial_def_for_reduction): Remove
9769 adjustment handling. Take the neutral value as an argument,
9770 in place of the code argument.
9771 (vect_transform_cycle_phi): Update accordingly. Handle the
9772 initial values of cond reductions separately from code reductions.
9773 Choose the adjustment here rather than in
9774 get_initial_def_for_reduction. Sink the splat of vec_initial_def.
9776 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9778 * tree-vect-loop.c (neutral_op_for_slp_reduction): Replace with...
9779 (neutral_op_for_reduction): ...this, providing a more general
9781 (vect_create_epilog_for_reduction): Update accordingly.
9782 (vectorizable_reduction): Likewise.
9783 (vect_transform_cycle_phi): Likewise.
9785 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9787 * tree-vect-loop.c (get_initial_def_for_reduction): Take the
9788 reduc_info instead of the original stmt_vec_info.
9789 (vect_transform_cycle_phi): Update accordingly.
9791 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9793 * tree-vect-loop.c (get_initial_defs_for_reduction): Take the
9794 reduc_info as an additional parameter.
9795 (vect_transform_cycle_phi): Update accordingly.
9797 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9799 * tree-vectorizer.h: Include tree-ssa-operands.h.
9800 (vect_phi_initial_value): New function.
9801 * tree-vect-loop.c (neutral_op_for_slp_reduction): Use it.
9802 (get_initial_defs_for_reduction, info_for_reduction): Likewise.
9803 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
9804 (vect_transform_cycle_phi, vectorizable_induction): Likewise.
9806 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9808 * tree-vect-loop.c (vect_create_epilog_for_reduction): Convert
9809 the phi results to vectype after creating them. Remove later
9810 conversion code that thus becomes redundant.
9812 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9814 * tree-vect-loop.c (vect_create_epilog_for_reduction): Replace
9815 the new_phis vector with a reduc_inputs vector. Combine handling
9816 of reduction chains and ncopies > 1.
9818 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9820 * tree-vect-loop.c (vect_create_epilog_for_reduction): Truncate
9821 scalar_results to group_size elements after reducing down from
9822 N*group_size elements. Construct an array_slice of the live-out
9823 stmts and assert that there is one stmt per scalar result.
9825 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9827 * tree-vect-loop.c (vect_create_epilog_for_reduction): Remove
9828 nested_in_vect_loop and use double_reduc everywhere. Remove dead
9829 assignment to "loop".
9831 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9833 * internal-fn.c (vectorized_internal_fn_supported_p): Handle
9834 vector types first. For scalar types, consider both the preferred
9835 vector mode and the alternative vector modes.
9836 * optabs-query.c (can_vec_mask_load_store_p): Use the same
9837 structure as above, in particular using related_vector_mode
9838 for modes provided by autovectorize_vector_modes.
9840 2021-07-13 Jakub Jelinek <jakub@redhat.com>
9841 Richard Biener <rguenther@suse.de>
9843 PR tree-optimization/101419
9844 * tree-pass.h (PROP_objsz): Define.
9845 (make_pass_early_object_sizes): Declare.
9846 * passes.def (pass_all_early_optimizations): Rename pass_object_sizes
9847 there to pass_early_object_sizes, drop parameter.
9848 (pass_all_optimizations): Move pass_object_sizes right after pass_ccp,
9849 drop parameter, move pass_post_ipa_warn right after that.
9850 * tree-object-size.c (pass_object_sizes::execute): Rename to...
9851 (object_sizes_execute): ... this. Add insert_min_max_p argument.
9852 (pass_data_object_sizes): Move after object_sizes_execute.
9853 (pass_object_sizes): Likewise. In execute method call
9854 object_sizes_execute, drop set_pass_param method and insert_min_max_p
9855 non-static data member and its initializer in the ctor.
9856 (pass_data_early_object_sizes, pass_early_object_sizes,
9857 make_pass_early_object_sizes): New.
9858 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Use
9859 (cfun->curr_properties & PROP_objsz) instead of cfun->after_inlining.
9861 2021-07-13 Kito Cheng <kito.cheng@sifive.com>
9864 * config/riscv/constraints.md ("S"): Update description and remove
9866 * doc/md.texi (Machine Constraints): Document the 'S' constraints
9869 2021-07-13 Richard Biener <rguenther@suse.de>
9872 2021-07-12 Richard Biener <rguenther@suse.de>
9874 * tree-vect-slp.c (vect_slp_region): Show the number of
9875 SLP graph entries in the optimization message.
9877 2021-07-13 Michael Meissner <meissner@linux.ibm.com>
9879 * config/rs6000/altivec.md (xxspltiw_v4sf): Change local variable
9881 * config/rs6000/rs6000-protos.h (rs6000_const_f32_to_i32): Change
9882 return type to long.
9883 * config/rs6000/rs6000.c (rs6000_const_f32_to_i32): Change return
9886 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
9888 * gimple-range-fold.cc (fold_using_range::range_of_builtin_ubsan_call):
9889 Query relation between the 2 operands and use it.
9891 2021-07-12 Sergei Trofimovich <siarheit@google.com>
9893 * doc/cfg.texi: Fix s/ei_safe_safe/ei_safe_edge/ typo.
9895 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
9898 * config/i386/predicates.md (vec_setm_sse41_operand):
9899 Rename from vec_setm_operand.
9900 (vec_setm_avx2_operand): New predicate.
9901 * config/i386/sse.md (vec_set<V_128:mode>): Use V_128 mode iterator.
9902 Use vec_setm_sse41_operand as operand 2 predicate.
9903 (vec_set<V_256_512:mode): New expander.
9904 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_sse41_operand
9905 as operand 2 predicate.
9907 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
9909 PR tree-optimization/101335
9910 * range-op.cc (operator_cast::lhs_op1_relation): Delete.
9912 2021-07-12 Andrew Pinski <apinski@marvell.com>
9914 * tree-ssa-phiopt.c (match_simplify_replacement): Move
9915 insert of the sequence before the movement of the
9916 statement. Check if to see if the statement is used
9917 outside of the original phi to see if we should move it.
9919 2021-07-12 Richard Biener <rguenther@suse.de>
9921 * dump-context.h (debug_dump_context::debug_dump_context):
9922 Add FILE * parameter defaulted to stderr.
9923 * dumpfile.c (debug_dump_context::debug_dump_context): Adjust.
9924 * tree-vect-slp.c (dot_slp_tree): New functions.
9926 2021-07-12 Richard Biener <rguenther@suse.de>
9928 PR tree-optimization/101373
9929 * tree-ssa-pre.c (prune_clobbered_mems): Also prune trapping
9930 references when the BB may not return.
9931 (compute_avail): Pass in the function we're working on and
9932 replace cfun references with it. Externally throwing
9933 const calls also possibly terminate the function.
9934 (pass_pre::execute): Pass down the function we're working on.
9935 * gcse.c (compute_hash_table_work): Externally throwing
9936 const/pure calls also need record_last_mem_set_info.
9937 * postreload-gcse.c (record_opr_changes): Looping or externally
9938 throwing const/pure calls also need record_last_mem_set_info.
9940 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
9942 * recog.c (memory_address_addr_space_p): Change the type to bool.
9943 Return true/false instead of 1/0.
9944 (offsettable_memref_p): Ditto.
9945 (offsettable_nonstrict_memref_p): Ditto.
9946 (offsettable_address_addr_space_p): Ditto.
9947 Change the type of addressp indirect function to bool.
9948 * recog.h (memory_address_addr_space_p): Change the type to bool.
9949 (strict_memory_address_addr_space_p): Ditto.
9950 (offsettable_memref_p): Ditto.
9951 (offsettable_nonstrict_memref_p): Ditto.
9952 (offsettable_address_addr_space_p): Ditto.
9953 * reload.c (maybe_memory_address_addr_space_p): Ditto.
9954 (strict_memory_address_addr_space_p): Change the type to bool.
9955 Return true/false instead of 1/0.
9956 (maybe_memory_address_addr_space_p): Change the type to bool.
9958 2021-07-12 Richard Biener <rguenther@suse.de>
9960 * tree-vect-slp.c (vect_slp_region): Show the number of
9961 SLP graph entries in the optimization message.
9963 2021-07-12 Richard Biener <rguenther@suse.de>
9965 PR tree-optimization/101394
9966 * tree-ssa-pre.c (do_pre_regular_insertion): Avoid inserting
9967 copies from abnormals for a full redundancy.
9969 2021-07-12 Richard Biener <rguenther@suse.de>
9971 PR middle-end/101423
9972 * gimple.c (gimple_could_trap_p_1): Internal function calls
9974 * tree-eh.c (tree_could_trap_p): Likewise.
9976 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
9979 * config/arm/arm_neon.h (vmul_n_u32): Replace call to builtin with
9981 (vmulq_n_u32): Likewise.
9982 (vmul_n_f32): Gate __a * __b on __FAST_MATH__.
9983 (vmulq_n_f32): Likewise.
9984 (vmul_n_f16): Likewise.
9985 (vmulq_n_f16): Likewise.
9987 2021-07-12 Martin Liska <mliska@suse.cz>
9990 * gcc.c (check_offload_target_name): Call
9991 candidates_list_and_hint only if we have a candidate.
9993 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
9996 * config/arm/neon.md (vec_init): Move to ...
9997 * config/arm/vec-common.md (vec_init): ... here.
9998 Change the pattern's mode to VDQX and gate it on VALID_MVE_MODE.
10000 2021-07-12 Roger Sayle <roger@nextmovesoftware.com>
10002 PR tree-optimization/101403
10003 * match.pd ((T)bswap(X)>>C): Correctly handle cases where
10004 signedness of the shift is not the same as the signedness of
10005 the type extension.
10007 2021-07-09 Roger Sayle <roger@nextmovesoftware.com>
10008 Uroš Bizjak <ubizjak@gmail.com>
10010 * config/i386/i386.md (*divmodsi4_const): Optimize SImode
10011 divmod of a constant numerator with new define_insn_and_split.
10013 2021-07-09 Iain Sandoe <iain@sandoe.co.uk>
10016 * config/i386/i386-expand.c (ix86_expand_call): If a call is
10017 to a non-local-binding, or local but to a public symbol, then
10018 assume that it might be indirected via the lazy symbol binder.
10019 Mark R10 and R10 as clobbered in that case.
10021 2021-07-09 Eric Botcazou <ebotcazou@adacore.com>
10024 * gcc.c (ASM_DEBUG_DWARF_OPTION): Set again to --gdwarf2 in
10025 the case where HAVE_AS_WORKING_DWARF_N_FLAG is not defined
10026 and HAVE_LD_BROKEN_PE_DWARF5 is defined.
10028 2021-07-09 Uroš Bizjak <ubizjak@gmail.com>
10030 * config/i386/i386.md (*udivmodsi4_pow2_zext_1): Limit the
10031 log2 range of operands[3] to [1,31].
10032 (*udivmodsi4_pow2_zext_2): Ditto. Correct insn RTX pattern.
10034 2021-07-09 Sergei Trofimovich <siarheit@google.com>
10036 * doc/md.texi: Don't split @smallexample in multiple @groups.
10038 2021-07-09 Sergei Trofimovich <siarheit@google.com>
10040 * doc/md.texi: Add missing 'see' word.
10042 2021-07-09 Andrew Pinski <apinski@marvell.com>
10044 * tree-ssa-phiopt.c (phiopt_early_allow): Change arguments
10045 to take sequence and gimple_match_op. Accept the case where
10046 op is a SSA_NAME and one statement in the sequence.
10047 Also allow constants.
10048 (gimple_simplify_phiopt): Always pass a sequence to resimplify.
10049 Update call to phiopt_early_allow. Discard the sequence if not
10052 2021-07-09 Xi Ruoyao <xry111@mengyan1223.wang>
10057 * config/mips/mips.c (mips_const_insns): Use MSA_SUPPORTED_MODE_P
10058 instead of ISA_HAS_MSA.
10059 (mips_expand_vec_unpack): Likewise.
10060 (mips_expand_vector_init): Likewise.
10062 2021-07-09 Kewen Lin <linkw@linux.ibm.com>
10064 * config/rs6000/vsx.md (mods_<mode>): Rename to...
10065 (mod<mode>3): ... this.
10066 (modu_<mode>): Rename to...
10067 (umod<mode>3): ... this.
10068 * config/rs6000/rs6000-builtin.def (MODS_V2DI, MODS_V4SI, MODU_V2DI,
10069 MODU_V4SI): Adjust.
10071 2021-07-08 Jeff Law <jeffreyalaw@gmail.com>
10073 * config/h8300/shiftrotate.md (variable shifts): Expose condition
10074 code handling for the test before the loop.
10076 2021-07-08 Martin Jambor <mjambor@suse.cz>
10079 * ipa-sra.c (class isra_call_summary): New member
10080 m_before_any_store, initialize it in the constructor.
10081 (isra_call_summary::dump): Dump the new field.
10082 (ipa_sra_call_summaries::duplicate): Copy it.
10083 (process_scan_results): Set it.
10084 (isra_write_edge_summary): Stream it.
10085 (isra_read_edge_summary): Likewise.
10086 (param_splitting_across_edge): Only override
10087 safe_to_import_accesses if m_before_any_store is set.
10089 2021-07-08 Martin Sebor <msebor@redhat.com>
10091 PR bootstrap/101374
10092 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref):
10093 Use Object Size Type 0 instead of 1.
10095 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
10097 * tree-vect-loop.c (vectorizable_reduction): Remove always-true
10100 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
10102 * match.pd: Simplify an extend-operate-truncate sequence involving
10105 2021-07-08 Roger Sayle <roger@nextmovesoftware.com>
10106 Richard Biener <rguenther@suse.de>
10108 PR tree-optimization/40210
10109 * match.pd (bswap optimizations): Simplify (bswap(x)>>C1)&C2 as
10110 (x>>C3)&C2 when possible. Simplify bswap(x)>>C1 as ((T)x)>>C2
10111 when possible. Simplify bswap(x)&C1 as (x>>C2)&C1 when 0<=C1<=255.
10113 2021-07-08 Uroš Bizjak <ubizjak@gmail.com>
10116 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
10118 * config/i386/mmx.md (V_32): New mode iterator.
10119 (mov<V_32:mode>): Use V_32 mode iterator.
10120 (*mov<V_32:mode>_internal): Ditto.
10121 (*push<V_32:mode>2_rex64): Ditto.
10122 (*push<V_32:mode>2): Ditto.
10123 (movmisalign<V_32:mode>): Ditto.
10124 (mmx_<any_shiftrt:insn>v1si3): New insn pattern.
10125 (sse4_1_<any_extend:code>v2qiv2hi2): Ditto.
10126 (vec_unpacks_lo_v4qi): New expander.
10127 (vec_unpacks_hi_v4qi): Ditto.
10128 (vec_unpacku_lo_v4qi): Ditto.
10129 (vec_unpacku_hi_v4qi): Ditto.
10130 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add V1SImode.
10131 (VALID_INT_MODE_P): Ditto.
10133 2021-07-08 Michael Meissner <meissner@linux.ibm.com>
10136 * config/rs6000/rs6000.md (udivti3): New insn.
10137 (divti3): New insn.
10138 (umodti3): New insn.
10139 (modti3): New insn.
10141 2021-07-07 Martin Sebor <msebor@redhat.com>
10143 PR tree-optimization/100137
10144 PR tree-optimization/99121
10145 PR tree-optimization/97027
10146 * builtins.c (access_ref::access_ref): Also set offmax.
10147 (access_ref::offset_in_range): Define new function.
10148 (access_ref::add_offset): Set offmax.
10149 (access_ref::inform_access): Handle access_none.
10150 (handle_mem_ref): Clear ostype.
10151 (compute_objsize_r): Handle ASSERT_EXPR.
10152 * builtins.h (struct access_ref): Add offmax member.
10153 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Use
10154 compute_objsize() and simplify.
10156 2021-07-07 Peter Bergner <bergner@linux.ibm.com>
10158 * config/rs6000/rs6000-call.c (mma_init_builtins): Use VSX_BUILTIN_LXVP
10159 and VSX_BUILTIN_STXVP.
10161 2021-07-07 Martin Sebor <msebor@redhat.com>
10164 * config/aarch64/aarch64.c (aarch64_simd_lane_bounds): Remove
10165 a stray %K from error_at() missed in r12-2088.
10167 2021-07-07 Richard Biener <rguenther@suse.de>
10169 PR tree-optimization/99728
10170 * tree-ssa-loop-im.c (gather_mem_refs_stmt): Record
10172 (mem_refs_may_alias_p): Add assert we handled aggregate
10174 (sm_seq_valid_bb): Give up when running into aggregate copies.
10175 (ref_indep_loop_p): Handle aggregate copies as never
10176 being invariant themselves but allow other refs to be
10177 disambiguated against them.
10178 (can_sm_ref_p): Do not try to apply store-motion to aggregate
10181 2021-07-06 Indu Bhagat <indu.bhagat@oracle.com>
10184 * dwarf2ctf.c (ctf_get_AT_data_member_location): Multiply by 8 to get
10187 2021-07-06 Martin Sebor <msebor@redhat.com>
10189 * gimple-pretty-print.c (percent_G_format): Remove.
10190 * tree-diagnostic.c (default_tree_printer): Remove calls.
10191 * tree-pretty-print.c (percent_K_format): Remove.
10192 * tree-pretty-print.h (percent_K_format): Remove.
10194 2021-07-06 Martin Sebor <msebor@redhat.com>
10196 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_builtin):
10197 Remove %K and use error_at.
10198 (aarch64_expand_fcmla_builtin): Same.
10199 (aarch64_expand_builtin_tme): Same.
10200 (aarch64_expand_builtin_memtag): Same.
10201 * config/arm/arm-builtins.c (arm_expand_acle_builtin): Same.
10202 (arm_expand_builtin): Same.
10203 * config/arm/arm.c (bounds_check): Same.
10205 2021-07-06 Martin Sebor <msebor@redhat.com>
10207 * builtins.c (warn_string_no_nul): Remove %G.
10208 (maybe_warn_for_bound): Same.
10209 (warn_for_access): Same.
10210 (check_access): Same.
10211 (check_strncat_sizes): Same.
10212 (expand_builtin_strncat): Same.
10213 (expand_builtin_strncmp): Same.
10214 (expand_builtin): Same.
10215 (expand_builtin_object_size): Same.
10216 (warn_dealloc_offset): Same.
10217 (maybe_emit_free_warning): Same.
10218 * calls.c (maybe_warn_alloc_args_overflow): Same.
10219 (maybe_warn_nonstring_arg): Same.
10220 (maybe_warn_rdwr_sizes): Same.
10221 * expr.c (expand_expr_real_1): Remove %K.
10222 * gimple-fold.c (gimple_fold_builtin_strncpy): Remove %G.
10223 (gimple_fold_builtin_strncat): Same.
10224 * gimple-ssa-sprintf.c (format_directive): Same.
10225 (handle_printf_call): Same.
10226 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Same.
10227 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
10228 (maybe_diag_access_bounds): Same. Call gimple_location.
10229 (check_bounds_or_overlap): Same.
10230 * trans-mem.c (ipa_tm_scan_irr_block): Remove %K. Simplify.
10231 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Remove %G.
10232 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
10233 (maybe_diag_stxncpy_trunc): Same.
10234 (handle_builtin_stxncpy_strncat): Same.
10235 (maybe_warn_pointless_strcmp): Same.
10236 * tree-ssa-uninit.c (maybe_warn_operand): Same.
10238 2021-07-06 Uroš Bizjak <ubizjak@gmail.com>
10241 * config/i386/predicates.md (vec_setm_operand): Enable
10242 register_operand for TARGET_SSE4_1.
10243 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_operand
10244 as operand 2 predicate. Call ix86_expand_vector_set_var
10245 for non-constant index operand.
10246 (vec_setv4qi): Use vec_setm_mmx_operand as operand 2 predicate.
10247 Call ix86_expand_vector_set_var for non-constant index operand.
10249 2021-07-06 Jeff Law <jeffreyalaw@gmail.com>
10251 * config/h8300/jumpcall.md (*branch): When possible, generate
10252 the comparison in CCZN mode.
10253 * config/h8300/predicates.md (simple_memory_operand): Reject all
10254 auto-increment addressing modes.
10256 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
10258 PR bootstrap/100246
10259 * config/i386/i386.h (struct stringop_algs): Define a CTOR for
10262 2021-07-06 Richard Biener <rguenther@suse.de>
10264 * doc/md.texi (vec_fmaddsub<mode>4): Document.
10265 (vec_fmsubadd<mode>4): Likewise.
10266 * optabs.def (vec_fmaddsub$a4): Add.
10267 (vec_fmsubadd$a4): Likewise.
10268 * internal-fn.def (IFN_VEC_FMADDSUB): Add.
10269 (IFN_VEC_FMSUBADD): Likewise.
10270 * tree-vect-slp-patterns.c (addsub_pattern::recognize):
10271 Refactor to handle IFN_VEC_FMADDSUB and IFN_VEC_FMSUBADD.
10272 (addsub_pattern::build): Likewise.
10273 * tree-vect-slp.c (vect_optimize_slp): CFN_VEC_FMADDSUB
10274 and CFN_VEC_FMSUBADD are not transparent for permutes.
10275 * config/i386/sse.md (vec_fmaddsub<mode>4): New expander.
10276 (vec_fmsubadd<mode>4): Likewise.
10278 2021-07-06 Richard Biener <rguenther@suse.de>
10280 * doc/invoke.texi (fmove-loop-stores): Document.
10281 * common.opt (fmove-loop-stores): New option.
10282 * opts.c (default_options_table): Enable -fmove-loop-stores
10283 at -O1 but not -Og.
10284 * tree-ssa-loop-im.c (pass_lim::execute): Pass
10285 flag_move_loop_stores instead of true to
10286 loop_invariant_motion_in_fun.
10288 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
10290 * doc/install.texi: Document --with-dsymutil.
10292 2021-07-06 Andrew Pinski <apinski@marvell.com>
10294 PR tree-optimization/101256
10295 * dbgcnt.def (phiopt_edge_range): New counter.
10296 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
10297 Check to make sure the new name is defined in the same
10298 bb as the conditional before duplicating range info.
10299 Also add debug counter.
10301 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
10303 PR rtl-optimization/100328
10304 * config/i386/i386-options.c (ix86_option_override_internal):
10305 Set param_ira_consider_dup_in_all_alts to 0.
10307 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
10309 PR rtl-optimization/100328
10310 * doc/invoke.texi (ira-consider-dup-in-all-alts): Document new
10312 * ira.c (ira_get_dup_out_num): Adjust as parameter
10313 param_ira_consider_dup_in_all_alts.
10314 * params.opt (ira-consider-dup-in-all-alts): New.
10315 * ira-conflicts.c (process_regs_for_copy): Add one parameter
10316 single_input_op_has_cstr_p.
10317 (get_freq_for_shuffle_copy): New function.
10318 (add_insn_allocno_copies): Adjust as single_input_op_has_cstr_p.
10319 * ira-int.h (ira_get_dup_out_num): Add one bool parameter.
10321 2021-07-05 Jeff Law <jeffreyalaw@gmail.com>
10323 * config/h8300/shiftrotate.md (shift-by-variable patterns): Update to
10324 generate condition code aware RTL directly.
10326 2021-07-05 Andrew Pinski <apinski@marvell.com>
10328 PR tree-optimization/101039
10329 * match.pd (A CMP 0 ? A : -A): New patterns.
10330 * tree-ssa-phiopt.c (abs_replacement): Delete function.
10331 (tree_ssa_phiopt_worker): Don't call abs_replacement.
10332 Update comment about abs_replacement.
10334 2021-07-05 Andrew Pinski <apinski@marvell.com>
10336 * tree-ssa-phiopt.c (gimple_simplify_phiopt):
10337 If "A ? B : C" fails to simplify, try "(!A) ? C : B".
10339 2021-07-05 Andrew Pinski <apinski@marvell.com>
10341 * tree-ssa-phiopt.c (match_simplify_replacement):
10342 Add early_p argument. Call gimple_simplify_phiopt
10343 instead of gimple_simplify.
10344 (tree_ssa_phiopt_worker): Update call to
10345 match_simplify_replacement and allow unconditionally.
10346 (phiopt_early_allow): New function.
10347 (gimple_simplify_phiopt): New function.
10349 2021-07-05 Andrew Pinski <apinski@marvell.com>
10351 PR middle-end/101237
10352 * fold-const.c (negate_expr_p): Remove call to element_mode
10353 and TREE_MODE/TREE_TYPE when calling HONOR_SIGNED_ZEROS,
10354 HONOR_SIGN_DEPENDENT_ROUNDING, and HONOR_SNANS.
10355 (fold_negate_expr_1): Likewise.
10356 (const_unop): Likewise.
10357 (fold_cond_expr_with_comparison): Likewise.
10358 (fold_binary_loc): Likewise.
10359 (fold_ternary_loc): Likewise.
10360 (tree_call_nonnegative_warnv_p): Likewise.
10361 * match.pd (-(A + B) -> (-B) - A): Likewise.
10363 2021-07-05 Iain Sandoe <iain@sandoe.co.uk>
10365 * configure.ac: Handle --with-dsymutil in the same way as we
10366 do for the assembler and linker. (DEFAULT_DSYMUTIL): New.
10367 Extract the type and version for the dsymutil configured or
10368 found by the default searches.
10369 * config.in: Regenerated.
10370 * configure: Regenerated.
10371 * collect2.c (do_dsymutil): Handle locating dsymutil in the
10372 same way as for the assembler and linker.
10373 * config/darwin.h (DSYMUTIL): Delete.
10374 * gcc.c: Report a configured dsymutil correctly.
10375 * exec-tool.in: Allow for dsymutil.
10377 2021-07-05 Uroš Bizjak <ubizjak@gmail.com>
10379 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
10380 Handle V4QI and V2HI modes.
10381 (expand_vec_perm_blend): Allow 4-byte vector modes with TARGET_SSE4_1.
10382 Handle V4QI mode. Emit mmx_pblendvb32 for 4-byte modes.
10383 (expand_vec_perm_pshufb): Rewrite to use switch statemets.
10384 Handle 4-byte dual operands with TARGET_XOP and single operands
10385 with TARGET_SSSE3. Emit mmx_ppermv32 for TARGET_XOP and
10386 mmx_pshufbv4qi3 for TARGET_SSSE3.
10387 (expand_vec_perm_pblendv): Allow 4-byte vector modes with TARGET_SSE4_1.
10388 (expand_vec_perm_interleave2): Allow 4-byte vector modes.
10389 (expand_vec_perm_pshufb2): Allow 4-byte vector modes with TARGET_SSSE3.
10390 (expand_vec_perm_even_odd_1): Handle V4QI mode.
10391 (expand_vec_perm_broadcast_1): Handle V4QI mode.
10392 (ix86_vectorize_vec_perm_const): Handle V4QI mode.
10393 * config/i386/mmx.md (mmx_ppermv32): New insn pattern.
10394 (mmx_pshufbv4qi3): Ditto.
10395 (*mmx_pblendw32): Ditto.
10396 (*mmx_pblendw64): Rename from *mmx_pblendw.
10397 (mmx_punpckhbw_low): New insn_and_split pattern.
10398 (mmx_punpcklbw_low): Ditto.
10400 2021-07-05 Richard Biener <rguenther@suse.de>
10402 * tree-vect-loop-manip.c (vect_loop_versioning): Do not
10403 set LOOP_C_INFINITE on the vectorized loop.
10405 2021-07-05 Richard Biener <rguenther@suse.de>
10407 PR middle-end/101291
10408 * cfgloopmanip.c (loop_version): Set the loop copy of the
10409 versioned loop to the new loop.
10411 2021-07-04 Iain Sandoe <iain@sandoe.co.uk>
10414 * config.gcc: Ensure that Darwin biarch definitions are
10415 added before i386.h.
10416 * config/i386/darwin.h (TARGET_64BIT): Remove.
10417 (PR80556_WORKAROUND): New.
10418 (REAL_LIBGCC_SPEC): Amend to use PR80556_WORKAROUND.
10419 (DARWIN_SUBARCH_SPEC): New.
10420 * config/i386/darwin32-biarch.h (TARGET_64BIT_DEFAULT,
10421 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
10422 (REAL_LIBGCC_SPEC): Remove.
10423 * config/i386/darwin64-biarch.h (TARGET_64BIT_DEFAULT,
10424 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
10425 (REAL_LIBGCC_SPEC): Remove.
10427 2021-07-03 H.J. Lu <hjl.tools@gmail.com>
10429 PR middle-end/101294
10430 * expr.c (store_constructor): Don't use vec_duplicate on vector.
10432 2021-07-02 Martin Sebor <msebor@redhat.com>
10434 PR middle-end/98871
10435 PR middle-end/98512
10436 * diagnostic.c (get_any_inlining_info): New.
10437 (update_effective_level_from_pragmas): Handle inlining context.
10438 (diagnostic_enabled): Same.
10439 (diagnostic_report_diagnostic): Same.
10440 * diagnostic.h (struct diagnostic_info): Add ctor.
10441 (struct diagnostic_context): Add new member.
10442 * tree-diagnostic.c (set_inlining_locations): New.
10443 (tree_diagnostics_defaults): Set new callback pointer.
10445 2021-07-02 Peter Bergner <bergner@linux.ibm.com>
10447 * config/rs6000/rs6000-builtin.def (BU_MMA_PAIR_LD, BU_MMA_PAIR_ST):
10449 (__builtin_vsx_lxvp, __builtin_vsx_stxvp): New built-ins.
10450 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Expand
10451 lxvp and stxvp built-ins.
10452 (mma_init_builtins): Handle lxvp and stxvp built-ins.
10453 (builtin_function_type): Likewise.
10454 * doc/extend.texi (__builtin_vsx_lxvp, __builtin_mma_stxvp): Document.
10456 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
10458 * config/h8300/h8300-protos.h (compute_a_shift_cc): Accept
10459 additional argument for the code.
10460 * config/h8300/h8300.c (compute_a_shift_cc): Accept additional
10461 argument for the code. Just return if the ZN bits are useful or
10462 not rather than the old style CC_* enums.
10463 * config/h8300/shiftrotate.md (shiftqi_noscratch): Move before
10464 more generic shiftqi patterns.
10465 (shifthi_noscratch, shiftsi_noscratch): Similarly.
10466 (shiftqi_noscratch_set_flags): New pattern.
10467 (shifthi_noscratch_set_flags, shiftsi_noscratch_set_flags): Likewise.
10469 2021-07-02 Andrew MacLeod <amacleod@redhat.com>
10471 PR tree-optimization/101223
10472 * range-op.cc (build_lt): Add -1 for signed values.
10473 (built_gt): Subtract -1 for signed values.
10475 2021-07-02 David Faust <david.faust@oracle.com>
10477 * btfout.c (get_btf_kind): Support BTF_KIND_FLOAT.
10478 (btf_asm_type): Likewise.
10480 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
10482 * config/h8300/h8300-protos.h (output_a_shift): Make first argument
10483 an array of rtx rather than a pointer to rtx. Add code argument.
10484 (compute_a_shift_length): Similarly.
10485 * config/h8300/h8300.c (h8300_shift_costs): Adjust now that the
10486 shift itself isn't an operand. Create dummy operand[0] to carry
10487 a mode and pass a suitable rtx code to compute_a_shift_length.
10488 (get_shift_alg): Adjust operand number of clobber in output templates.
10489 (output_a_shift): Make first argument an array of rtx rather than
10490 a pointer to rtx. Add code argument for the type of shift.
10491 Adjust now that the shift itself is no longer an operand.
10492 (compute_a_shift_length): Similarly.
10493 * config/h8300/shiftrotate.md (shiftqi, shifthi, shiftsi): Use an
10494 iterator rather than nshift_operator.
10495 (shiftqi_noscratch, shifthi_noscratch, shiftsi_noscratch): Likewise.
10496 (shiftqi_clobber_flags): Adjust to API changes in output_a_shift
10497 and compute_a_shift_length.
10498 (shiftqi_noscratch_clobber_flags): Likewise.
10499 (shifthi_noscratch_clobber_flags): Likewise.
10500 (shiftsi_noscratch_clobber_flags): Likewise.
10502 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
10505 * config/darwin.h (DSYMUTIL_SPEC): Do not try to run
10506 dsymutil for BTF/CTF.
10508 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
10511 * config/darwin.h (CTF_INFO_SECTION_NAME): Update the
10512 segment to include BTF.
10513 (BTF_INFO_SECTION_NAME): New.
10515 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
10517 * config/m32r/m32r-protos.h (call_operand): Adjust return type.
10518 (small_data_operand, memreg_operand, small_insn_p): Likewise.
10519 * config/m32r/m32r.c (call_operand): Adjust return type.
10520 (small_data_operand, memreg_operand): Likewise.
10522 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
10524 * config/frv/frv-protos.h (integer_register_operand): Adjust return
10526 (frv_load_operand, gpr_or_fpr_operand, gpr_no_subreg_operand): Likewise.
10527 (fpr_or_int6_operand, gpr_or_int_operand); Likewise.
10528 (gpr_or_int12_operand, gpr_or_int10_operand); Likewise.
10529 (move_source_operand, move_destination_operand): Likewise.
10530 (condexec_source_operand, condexec_dest_operand): Likewise.
10531 (lr_operand, gpr_or_memory_operand, fpr_or_memory_operand): Likewise.
10532 (reg_or_0_operand, fcc_operand, icc_operand, cc_operand): Likewise.
10533 (fcr_operand, icr_operand, cr_operand, call_operand): Likewise.
10534 (fpr_operand, even_reg_operand, odd_reg_operand): Likewise.
10535 (even_gpr_operand, odd_gpr_operand, quad_fpr_operand): Likewise.
10536 (even_fpr_operand, odd_fpr_operand): Likewise.
10537 (dbl_memory_one_insn_operand, dbl_memory_two_insn_operand): Likewise.
10538 (int12_operand, int6_operand, int5_operand, uint5_operand): Likewise.
10539 (uint4_operand, uint1_operand, int_2word_operand): Likewise
10540 (upper_int16_operand, uint16_operand, symbolic_operand): Likewise.
10541 (relational_operator, float_relational_operator): Likewise.
10542 (ccr_eqne_operator, minmax_operator): Likewise.
10543 (condexec_si_binary_operator, condexec_si_media_operator): Likewise.
10544 (condexec_si_divide_operator, condexec_si_unary_operator): Likewise.
10545 (condexec_sf_conv_operator, condexec_sf_add_operator): Likewise.
10546 (intop_compare_operator, acc_operand, even_acc_operand): Likewise.
10547 (quad_acc_operand, accg_operand): Likewise.
10549 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
10551 * config/stormy16/stormy16-protos.h (xstormy16_below_100_symbol): Change
10552 return type to a bool.
10553 (nonimmediate_nonstack_operand): Likewise.
10554 (xstormy16_splittable_below100_operand): Likewise.
10555 * config/stormy16/stormy16.c (xstormy16_below_100_symbol): Fix
10557 (xstormy16_splittable_below100_operand): Likewise.
10559 2021-07-02 Richard Biener <rguenther@suse.de>
10561 PR tree-optimization/101293
10562 * tree-ssa-loop-im.c (mem_ref_hasher::equal): Compare MEM_REF bases
10563 with combined offsets.
10564 (gather_mem_refs_stmt): Hash MEM_REFs as if their offset were
10565 combined with the rest of the offset.
10567 2021-07-02 Eric Botcazou <ebotcazou@adacore.com>
10569 * config/i386/i386.c (asm_preferred_eh_data_format): Always use the
10570 PIC encodings for PE-COFF targets.
10572 2021-07-02 Jakub Jelinek <jakub@redhat.com>
10575 * config/i386/i386-expand.c (ix86_broadcast_from_integer_constant):
10576 Return nullptr for TImode inner mode.
10578 2021-07-02 Richard Biener <rguenther@suse.de>
10580 PR tree-optimization/101280
10581 PR tree-optimization/101173
10582 * gimple-loop-interchange.cc
10583 (tree_loop_interchange::valid_data_dependences): Properly
10584 guard all dependence checks with DDR_REVERSED_P or its
10587 2021-07-02 Hongyu Wang <hongyu.wang@intel.com>
10589 * config/i386/i386-expand.c (ix86_expand_builtin):
10590 Add branch to clear odata when ZF is set for asedecenc_expand
10591 and wideaesdecenc_expand.
10593 2021-07-02 Eugene Rozenfeld <erozen@microsoft.com>
10595 * config/i386/gcc-auto-profile: regenerate
10597 2021-07-02 liuhongt <hongtao.liu@intel.com>
10599 * config/i386/sse.md (trunc<mode><pmov_dst_4>2): Refined to ..
10600 (trunc<mode><pmov_dst_4_lower>2): this.
10602 2021-07-01 David Malcolm <dmalcolm@redhat.com>
10604 * diagnostic.h (diagnostic_context::m_file_cache): New field.
10605 * input.c (class fcache): Rename to...
10606 (class file_cache_slot): ...this, making most members private and
10607 prefixing fields with "m_".
10608 (file_cache_slot::get_file_path): New accessor.
10609 (file_cache_slot::get_use_count): New accessor.
10610 (file_cache_slot::missing_trailing_newline_p): New accessor.
10611 (file_cache_slot::inc_use_count): New.
10612 (fcache_buffer_size): Move to...
10613 (file_cache_slot::buffer_size): ...here.
10614 (fcache_line_record_size): Move to...
10615 (file_cache_slot::line_record_size): ...here.
10616 (fcache_tab): Delete, in favor of global_dc->m_file_cache.
10617 (fcache_tab_size): Move to file_cache::num_file_slots.
10618 (diagnostic_file_cache_init): Update for move of fcache_tab
10619 to global_dc->m_file_cache.
10620 (diagnostic_file_cache_fini): Likewise.
10621 (lookup_file_in_cache_tab): Convert to...
10622 (file_cache::lookup_file): ...this.
10623 (diagnostics_file_cache_forcibly_evict_file): Update for move of
10624 fcache_tab to global_dc->m_file_cache, moving most of
10625 implementation to...
10626 (file_cache::forcibly_evict_file): ...this new function and...
10627 (file_cache_slot::evict): ...this new function.
10628 (evicted_cache_tab_entry): Convert to...
10629 (file_cache::evicted_cache_tab_entry): ...this.
10630 (add_file_to_cache_tab): Convert to...
10631 (file_cache::add_file): ...this, moving bulk of implementation
10633 (file_cache_slot::create): ..this new function.
10634 (file_cache::file_cache): New.
10635 (file_cache::~file_cache): New.
10636 (lookup_or_add_file_to_cache_tab): Convert to...
10637 (file_cache::lookup_or_add_file): ..this new function.
10638 (fcache::fcache): Rename to...
10639 (file_cache_slot::file_cache_slot): ...this, adding "m_" prefixes
10641 (fcache::~fcache): Rename to...
10642 (file_cache_slot::~file_cache_slot): ...this, adding "m_" prefixes
10644 (needs_read): Convert to...
10645 (file_cache_slot::needs_read_p): ...this.
10646 (needs_grow): Convert to...
10647 (file_cache_slot::needs_grow_p): ...this.
10648 (maybe_grow): Convert to...
10649 (file_cache_slot::maybe_grow): ...this.
10650 (read_data): Convert to...
10651 (file_cache_slot::read_data): ...this.
10652 (maybe_read_data): Convert to...
10653 (file_cache_slot::maybe_read_data): ...this.
10654 (get_next_line): Convert to...
10655 (file_cache_slot::get_next_line): ...this.
10656 (goto_next_line): Convert to...
10657 (file_cache_slot::goto_next_line): ...this.
10658 (read_line_num): Convert to...
10659 (file_cache_slot::read_line_num): ...this.
10660 (location_get_source_line): Update for moving of globals to
10661 global_dc->m_file_cache.
10662 (location_missing_trailing_newline): Likewise.
10663 * input.h (class file_cache_slot): New forward decl.
10664 (class file_cache): New.
10666 2021-07-01 Michael Meissner <meissner@linux.ibm.com>
10668 * config/rs6000/rs6000.c (rs6000_maybe_emit_fp_cmove): Add IEEE
10669 128-bit floating point conditional move support.
10670 (have_compare_and_set_mask): Add IEEE 128-bit floating point
10672 * config/rs6000/rs6000.md (mov<mode>cc, IEEE128 iterator): New insn.
10673 (mov<mode>cc_p10, IEEE128 iterator): New insn.
10674 (mov<mode>cc_invert_p10, IEEE128 iterator): New insn.
10675 (fpmask<mode>, IEEE128 iterator): New insn.
10676 (xxsel<mode>, IEEE128 iterator): New insn.
10678 2021-07-01 Iain Sandoe <iain@sandoe.co.uk>
10681 * config/darwin.h (CTF_INFO_SECTION_NAME): New.
10683 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
10685 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
10687 * config/i386/i386-protos.h (ix86_expand_vector_init_duplicate):
10689 * config/i386/sse.md (INT_BROADCAST_MODE): New mode iterator.
10690 (vec_duplicate<mode>): New expander.
10692 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
10695 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
10697 (ix86_byte_broadcast): New function.
10698 (ix86_convert_const_wide_int_to_broadcast): Likewise.
10699 (ix86_expand_move): Convert CONST_WIDE_INT to broadcast if mode
10700 size is 16 bytes or bigger.
10701 (ix86_broadcast_from_integer_constant): New function.
10702 (ix86_expand_vector_move): Convert CONST_WIDE_INT and CONST_VECTOR
10703 to broadcast if mode size is 16 bytes or bigger.
10704 * config/i386/i386-protos.h (ix86_gen_scratch_sse_rtx): New
10706 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): New function.
10708 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
10710 * config/i386/predicates.md (ix86_endbr_immediate_operand):
10711 Return true/false instead of 1/0.
10712 (movq_parallel): Ditto.
10714 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
10716 * recog.c (general_operand): Return true/false instead of 1/0.
10717 (register_operand): Ditto.
10718 (immediate_operand): Ditto.
10719 (const_int_operand): Ditto.
10720 (const_scalar_int_operand): Ditto.
10721 (const_double_operand): Ditto.
10722 (push_operand): Ditto.
10723 (pop_operand): Ditto.
10724 (memory_operand): Ditto.
10725 (indirect_operand): Ditto.
10727 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
10729 * genpreds.c (write_predicate_subfunction):
10730 Change the type of written subfunction to bool.
10731 (write_one_predicate_function):
10732 Change the type of written function to bool.
10733 (write_tm_preds_h): Ditto.
10734 * recog.h (*insn_operand_predicate_fn): Change the type to bool.
10735 * recog.c (general_operand): Change the type to bool.
10736 (address_operand): Ditto.
10737 (register_operand): Ditto.
10738 (pmode_register_operand): Ditto.
10739 (scratch_operand): Ditto.
10740 (immediate_operand): Ditto.
10741 (const_int_operand): Ditto.
10742 (const_scalar_int_operand): Ditto.
10743 (const_double_operand): Ditto.
10744 (nonimmediate_operand): Ditto.
10745 (nonmemory_operand): Ditto.
10746 (push_operand): Ditto.
10747 (pop_operand): Ditto.
10748 (memory_operand): Ditto.
10749 (indirect_operand): Ditto.
10750 (ordered_comparison_operator): Ditto.
10751 (comparison_operator): Ditto.
10752 * config/i386/i386-expand.c (ix86_expand_sse_cmp):
10753 Change the type of indirect predicate function to bool.
10754 * config/rs6000/rs6000.c (easy_vector_constant):
10755 Change the type to bool.
10756 * config/mips/mips-protos.h (m16_based_address_p):
10757 Change the type of operand 3 to bool.
10759 2021-07-01 Richard Biener <rguenther@suse.de>
10761 PR tree-optimization/101280
10762 PR tree-optimization/101173
10763 * gimple-loop-interchange.cc
10764 (tree_loop_interchange::valid_data_dependences): Revert
10765 previous change and instead correctly handle DDR_REVERSED_P
10768 2021-07-01 Richard Biener <rguenther@suse.de>
10770 PR tree-optimization/101278
10771 * tree-ssa-dse.c (dse_classify_store): First check for
10772 uses, then ignore stmt for chaining purposes.
10774 2021-07-01 Richard Biener <rguenther@suse.de>
10776 PR tree-optimization/100778
10777 * tree-vect-slp.c (vect_schedule_slp_node): Do not place trapping
10778 vectorized ops ahead of their scalar BB.
10780 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
10783 * config/i386/i386.md (*nabs<dwi>2_doubleword):
10784 New insn_and_split pattern.
10785 (*nabs<dwi>2_1): Ditto.
10786 * config/i386/i386-features.c
10787 (general_scalar_chain::compute_convert_gain):
10788 Handle (NEG (ABS (...))) RTX. Rewrite src code
10789 scanner as switch statement.
10790 (general_scalar_chain::convert_insn):
10791 Handle (NEG (ABS (...))) RTX.
10792 (general_scalar_to_vector_candidate_p):
10793 Detect (NEG (ABS (...))) RTX. Reorder case statements
10794 for (AND (NOT (...) ...)) fallthrough.
10796 2021-07-01 Richard Biener <rguenther@suse.de>
10798 PR tree-optimization/101178
10799 * tree-vect-slp.c (slpg_vertex::materialize): Remove.
10800 (slpg::perm_in): Add.
10801 (slpg::get_perm_in): Remove.
10802 (slpg::get_perm_materialized): Add.
10803 (vect_optimize_slp): Handle VEC_PERM nodes more optimally
10804 during permute propagation and materialization.
10806 2021-07-01 Jakub Jelinek <jakub@redhat.com>
10809 * dwarf2out.c (loc_list_from_tree_1): Handle COMPOUND_LITERAL_EXPR.
10811 2021-07-01 Jakub Jelinek <jakub@redhat.com>
10813 PR middle-end/94366
10814 * omp-low.c (lower_rec_input_clauses): Rename is_fp_and_or to
10815 is_truth_op, set it for TRUTH_*IF_EXPR regardless of new_var's type,
10816 use boolean_type_node instead of integer_type_node as NE_EXPR type.
10817 (lower_reduction_clauses): Likewise.
10819 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
10821 * config/gcn/gcn.c: Include dwarf2.h.
10822 (gcn_addr_space_debug): New function.
10823 (TARGET_ADDR_SPACE_DEBUG): New hook.
10825 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
10827 * common/config/gcn/gcn-common.c
10828 (gcn_option_optimization_table): Change OPT_fomit_frame_pointer to -O3.
10829 * config/gcn/gcn.c (gcn_expand_prologue): Prefer the frame pointer
10831 (gcn_expand_prologue): Prefer the frame pointer when emitting CFI.
10832 (gcn_frame_pointer_rqd): New function.
10833 (TARGET_FRAME_POINTER_REQUIRED): New hook.
10835 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
10837 * config/gcn/gcn.c (move_callee_saved_registers): Emit CFI notes for
10838 prologue register saves.
10839 (gcn_debug_unwind_info): Use UI_DWARF2.
10840 (gcn_dwarf_register_number): Map DWARF_LINK_REGISTER to DWARF PC.
10841 (gcn_dwarf_register_span): DWARF_LINK_REGISTER doesn't span.
10842 * config/gcn/gcn.h: (DWARF_FRAME_RETURN_COLUMN): New define.
10843 (DWARF_LINK_REGISTER): New define.
10844 (FIRST_PSEUDO_REGISTER): Increment.
10845 (FIXED_REGISTERS): Add entry for DWARF_LINK_REGISTER.
10846 (CALL_USED_REGISTERS): Likewise.
10847 (REGISTER_NAMES): Likewise.
10849 2021-06-30 Richard Biener <rguenther@suse.de>
10851 PR tree-optimization/101267
10852 * tree-vect-stmts.c (vect_check_scalar_mask): Adjust
10853 API and use SLP compatible interface of vect_is_simple_use.
10854 Reject not vectorized SLP defs for callers that do not support
10856 (vect_check_store_rhs): Handle masked stores and pass down
10857 the appropriate operator index.
10858 (vectorizable_call): Adjust.
10859 (vectorizable_store): Likewise.
10860 (vectorizable_load): Likewise. Handle SLP pecularity of
10862 (vect_is_simple_use): Remove special-casing of masked stores.
10864 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
10866 * common.opt (foffload): Remove help as Driver only.
10867 * gcc.c (display_help): Add -foffload.
10869 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
10871 * gcc.c (close_at_file, execute): Replace alloca by XALLOCAVEC.
10872 (check_offload_target_name): Fix splitting OFFLOAD_TARGETS into
10873 a candidate list; better inform no offload target is configured
10874 and fix hint extraction when passed target is not '\0' at [len].
10875 * common.opt (foffload): Add tailing '.'.
10876 (foffload-options): Likewise; fix flag name in the help string.
10878 2021-06-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
10881 * config/arm/arm_neon.h: Move vabs intrinsics before vcage_f32.
10882 (vcage_f32): Gate comparison on __FAST_MATH__.
10883 (vcageq_f32): Likewise.
10884 (vcale_f32): Likewise.
10885 (vcaleq_f32): Likewise.
10886 (vcagt_f32): Likewise.
10887 (vcagtq_f32): Likewise.
10888 (vcalt_f32): Likewise.
10889 (vcaltq_f32): Likewise.
10890 (vcage_f16): Likewise.
10891 (vcageq_f16): Likewise.
10892 (vcale_f16): Likewise.
10893 (vcaleq_f16): Likewise.
10894 (vcagt_f16): Likewise.
10895 (vcagtq_f16): Likewise.
10896 (vcalt_f16): Likewise.
10897 (vcaltq_f16): Likewise.
10899 2021-06-30 Richard Biener <rguenther@suse.de>
10901 PR tree-optimization/101264
10902 * tree-vect-slp.c (vect_optimize_slp): Propagate the
10903 computed perm_in to all "any" permute successors
10904 we cannot de-duplicate immediately.
10906 2021-06-30 liuhongt <hongtao.liu@intel.com>
10909 * config/i386/sse.md
10910 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>):
10912 (avx512f_sfixupimm<mode><maskz_scalar_name><round_saeonly_name>):
10914 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>"): Refined.
10915 * config/i386/subst.md (maskz_scalar): New define_subst.
10916 (maskz_scalar_name): New subst_attr.
10917 (maskz_scalar_op5): Ditto.
10918 (round_saeonly_maskz_scalar_op5): Ditto.
10919 (round_saeonly_maskz_scalar_operand5): Ditto.
10921 2021-06-30 David Edelsohn <dje.gcc@gmail.com>
10923 * config/rs6000/rs6000.c (rs6000_xcoff_section_type_flags):
10924 Increase code CSECT alignment to at least 32 bytes.
10925 * config/rs6000/xcoff.h (TEXT_SECTION_ASM_OP): Add 32 byte
10926 alignment designation.
10928 2021-06-29 Sergei Trofimovich <siarheit@google.com>
10930 * doc/generic.texi: Fix s/net yet/not yet/ typo.
10932 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
10934 PR tree-optimization/101254
10935 * range-op.cc (operator_minus::op1_op2_relation_effect): Check for
10936 wrapping/non-wrapping when setting the result range.
10938 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
10940 * value-query.cc (gimple_range_global): Allow phis.
10942 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
10944 * vr-values.c (vr_values::vrp_stmt_computes_nonzero): Use stmt.
10945 (simplify_using_ranges::op_with_boolean_value_range_p): Add a
10946 statement for location context.
10947 (check_for_binary_op_overflow): Ditto.
10948 (simplify_using_ranges::get_vr_for_comparison): Ditto.
10949 (simplify_using_ranges::compare_name_with_value): Ditto.
10950 (simplify_using_ranges::compare_names): Ditto.
10951 (vrp_evaluate_conditional_warnv_with_ops_using_ranges): Ditto.
10952 (simplify_using_ranges::simplify_truth_ops_using_ranges): Ditto.
10953 (simplify_using_ranges::simplify_min_or_max_using_ranges): Ditto.
10954 (simplify_using_ranges::simplify_internal_call_using_ranges): Ditto.
10955 (simplify_using_ranges::two_valued_val_range_p): Ditto.
10956 (simplify_using_ranges::simplify): Ditto.
10957 * vr-values.h: Adjust prototypes.
10959 2021-06-29 Uroš Bizjak <ubizjak@gmail.com>
10962 * config/i386/mmx.md (vec_addsubv2sf3): New insn pattern.
10964 2021-06-29 Julian Brown <julian@codesourcery.com>
10966 * config/gcn/gcn.c (gcn_init_libfuncs): New function.
10967 (TARGET_INIT_LIBFUNCS): Define target hook using above function.
10968 * config/gcn/gcn.h (UNITS_PER_WORD): Define to 8 for IN_LIBGCC2, 4
10970 (LIBGCC2_UNITS_PER_WORD, BITS_PER_WORD): Remove definitions.
10971 (MAX_FIXED_MODE_SIZE): Change to 128.
10973 2021-06-29 Julian Brown <julian@codesourcery.com>
10975 * config/gcn/gcn.md (UNSPEC_FLBIT_INT): New unspec constant.
10976 (s_mnemonic): Add clrsb.
10977 (gcn_flbit<mode>_int): Add insn pattern for SImode/DImode.
10978 (clrsb<mode>2): Add expander for SImode/DImode.
10980 2021-06-29 Julian Brown <julian@codesourcery.com>
10982 * config/gcn/gcn.md (<su>mulsidi3, <su>mulsidi3_reg, <su>mulsidi3_imm,
10983 muldi3): Add patterns.
10985 2021-06-29 Julian Brown <julian@codesourcery.com>
10987 * config/gcn/gcn.md (<su>mulsi3_highpart): Change to expander.
10988 (<su>mulsi3_highpart_reg, <su>mulsi3_highpart_imm): New patterns.
10990 2021-06-29 Julian Brown <julian@codesourcery.com>
10992 * config/gcn/gcn.md (mulsi3): Make s_mulk_i32 variant clobber SCC.
10994 2021-06-29 Joseph Myers <joseph@codesourcery.com>
10996 * btfout.c, ctfout.c: Include "memmodel.h".
10998 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
11000 * gcc.c (check_offload_target_name): Cast len argument to
11001 %q.*s to 'int'; avoid -Wstringop-truncation warning.
11003 2021-06-29 Richard Biener <rguenther@suse.de>
11005 * tree-vect-slp.c (vect_optimize_slp): Forward propagate
11006 to "any" permute nodes and relax "any" permute proapgation
11007 during iterative backward propagation.
11009 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
11012 * common.opt (-foffload=): Update description.
11013 (-foffload-options=): New.
11014 * doc/invoke.texi (C Language Options): Document
11015 -foffload and -foffload-options.
11016 * gcc.c (check_offload_target_name): New, split off from
11017 handle_foffload_option.
11018 (check_foffload_target_names): New.
11019 (handle_foffload_option): Handle -foffload=default.
11020 (driver_handle_option): Update for -foffload-options.
11021 * lto-opts.c (lto_write_options): Use -foffload-options
11022 instead of -foffload.
11023 * lto-wrapper.c (merge_and_complain, append_offload_options):
11025 * opts.c (common_handle_option): Likewise.
11027 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
11029 * doc/invoke.texi (C Language Options): Sort options
11030 alphabetically in optlist and also the description itself.
11031 Remove leftover -fallow-single-precision from and add missing
11032 -fgnu-tm to the optlist.
11034 2021-06-29 Richard Biener <rguenther@suse.de>
11036 * tree-vect-slp.c (slpg_vertex::visited): Remove.
11037 (vect_slp_perms_eq): Handle -1 permutes.
11038 (vect_optimize_slp): Rewrite permute propagation.
11040 2021-06-29 Jakub Jelinek <jakub@redhat.com>
11043 * match.pd ((intptr_t)x eq/ne CST to x eq/ne (typeof x) CST): Don't
11044 perform the optimization in GENERIC when sanitizing and x has a
11047 2021-06-29 Richard Biener <rguenther@suse.de>
11049 PR tree-optimization/101242
11050 * tree-vect-slp.c (vect_slp_build_vertices): Force-add
11051 PHIs with not represented initial values as leafs.
11053 2021-06-29 Jan-Benedict Glaw <jbglaw@getslash.de>
11055 * config/pdp11/pdp11.h (ASM_OUTPUT_SKIP): Fix signedness warning.
11056 * config/pdp11/pdp11.c (pdp11_asm_print_operand_punct_valid_p): Remove
11057 "register" keyword.
11058 (pdp11_initial_elimination_offset) Remove unused variable.
11059 (pdp11_cmp_length) Ditto.
11060 (pdp11_insn_cost): Ditto, and fix signedness warning.
11062 2021-06-29 David Edelsohn <dje.gcc@gmail.com>
11064 * btfout.c: Include tm_p.h.
11067 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
11069 * config/bpf/bpf.c (bpf_expand_prologue): Do not mark insns as
11071 (bpf_expand_epilogue): Likewise.
11072 * config/bpf/bpf.h (DWARF2_FRAME_INFO): Define to 0.
11073 Do not define DBX_DEBUGGING_INFO.
11075 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
11077 * doc/invoke.texi: Document the CTF and BTF debug info options.
11079 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
11080 David Faust <david.faust@oracle.com>
11081 Jose E. Marchesi <jose.marchesi@oracle.com>
11082 Weimin Pan <weimin.pan@oracle.com>
11084 * Makefile.in: Add ctfc.*, ctfout.c and btfout.c files to
11085 GTFILES. Add new object files.
11086 * common.opt: Add CTF and BTF debug info options.
11087 * btfout.c: New file.
11088 * ctfc.c: Likewise.
11089 * ctfc.h: Likewise.
11090 * ctfout.c: Likewise.
11091 * dwarf2ctf.c: Likewise.
11092 * dwarf2ctf.h: Likewise.
11093 * dwarf2cfi.c (dwarf2out_do_frame): Acknowledge CTF_DEBUG and
11095 * dwarf2out.c (dwarf2out_source_line): Likewise.
11096 (dwarf2out_finish): Skip emitting DWARF if CTF or BTF are to
11098 (debug_format_do_cu): New function.
11099 (dwarf2out_early_finish): Traverse DIEs and emit CTF/BTF for
11101 Include dwarf2ctf.c.
11102 * final.c (dwarf2_debug_info_emitted_p): Acknowledge DWARF-based debug
11104 * flag-types.h (enum debug_info_type): Add CTF_DEBUG and BTF_DEBUG.
11105 (CTF_DEBUG): New bitmask.
11106 (BTF_DEBUG): Likewise.
11107 (enum ctf_debug_info_levels): New enum.
11108 * gengtype.c (open_base_files): Handle ctfc.h.
11109 (main): Handle uint32_t type.
11110 * flags.h (btf_debuginfo_p): New definition.
11111 (dwarf_based_debuginfo_p): Likewise.
11112 * opts.c (debug_type_names): Add entries for CTF and BTF.
11113 (btf_debuginfo_p): New function.
11114 (dwarf_based_debuginfo_p): Likewise.
11115 (common_handle_option): Handle -gctfN and -gbtf options.
11116 (set_debug_level): Set CTF_DEBUG, BTF_DEBUG whenever appropriate.
11117 * toplev.c (process_options): Inform the user and ignore -gctfLEVEL if
11120 2021-06-28 Jose E. Marchesi <jose.marchesi@oracle.com>
11122 * dwarf2out.c (AT_class): Function is no longer static.
11123 (AT_int): Likewise.
11124 (AT_unsigned): Likewise.
11125 (AT_loc): Likewise.
11126 (get_AT): Likewise.
11127 (get_AT_string): Likewise.
11128 (get_AT_flag): Likewise.
11129 (get_AT_unsigned): Likewise.
11130 (get_AT_ref): Likewise.
11131 (new_die_raw): Likewise.
11132 (lookup_decl_die): Likewise.
11133 (base_type_die): Likewise.
11134 (add_name_attribute): Likewise.
11135 (add_AT_int): Likewise.
11136 (add_AT_unsigned): Likewise.
11137 (add_AT_loc): Likewise.
11138 (dw_get_die_tag): New function.
11139 (dw_get_die_child): Likewise.
11140 (dw_get_die_sib): Likewise.
11141 (struct dwarf_file_data): Move from here to dwarf2out.h
11142 (struct dw_attr_struct): Likewise.
11143 * dwarf2out.h: Analogous changes.
11145 2021-06-28 Martin Jambor <mjambor@suse.cz>
11148 * ipa-param-manipulation.h (class ipa_param_body_adjustments): New
11149 members m_dead_stmts and m_dead_ssas.
11150 * ipa-param-manipulation.c
11151 (ipa_param_body_adjustments::mark_dead_statements): New function.
11152 (ipa_param_body_adjustments::common_initialization): Call it on
11153 all removed but not split parameters.
11154 (ipa_param_body_adjustments::ipa_param_body_adjustments): Initialize
11156 (ipa_param_body_adjustments::modify_call_stmt): Remove arguments that
11158 * tree-inline.c (remap_gimple_stmt): Do not copy dead statements, reset
11159 dead debug statements.
11160 (copy_phis_for_bb): Do not copy dead PHI nodes.
11162 2021-06-28 Martin Jambor <mjambor@suse.cz>
11165 * symtab-clones.h (clone_info): Removed member param_adjustments.
11166 * ipa-param-manipulation.h: Adjust initial comment to reflect how we
11167 deal with pass-through splits now.
11168 (ipa_param_performed_split): Removed.
11169 (ipa_param_adjustments::modify_call): Adjusted parameters.
11170 (class ipa_param_body_adjustments): Adjusted parameters of
11171 register_replacement, modify_gimple_stmt and modify_call_stmt.
11172 (ipa_verify_edge_has_no_modifications): Declare.
11173 (ipa_edge_modifications_finalize): Declare.
11174 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Remove
11175 performed_splits processing, pas only edge to padjs->modify_call,
11176 check that call arguments were not modified if they should not have
11178 * cgraphclones.c (cgraph_node::create_clone): Do not copy performed
11180 * ipa-param-manipulation.c (struct pass_through_split_map): New type.
11181 (ipa_edge_modification_info): Likewise.
11182 (ipa_edge_modification_sum): Likewise.
11183 (ipa_edge_modifications): New edge summary.
11184 (ipa_verify_edge_has_no_modifications): New function.
11185 (transitive_split_p): Removed.
11186 (transitive_split_map): Likewise.
11187 (init_transitive_splits): Likewise.
11188 (ipa_param_adjustments::modify_call): Adjusted to use the new edge
11189 summary instead of performed_splits.
11190 (ipa_param_body_adjustments::register_replacement): Drop dummy
11191 parameter, set base_index of the created ipa_param_body_replacement.
11192 (phi_arg_will_live_p): New function.
11193 (ipa_param_body_adjustments::common_initialization): Do not create
11194 IPA_SRA dummy decls.
11195 (simple_tree_swap_info): Removed.
11196 (remap_split_decl_to_dummy): Likewise.
11197 (record_argument_state_1): New function.
11198 (record_argument_state): Likewise.
11199 (ipa_param_body_adjustments::modify_call_stmt): New parameter
11200 orig_stmt. Do not work with dummy decls, save necessary info about
11201 changes to ipa_edge_modifications.
11202 (ipa_param_body_adjustments::modify_gimple_stmt): New parameter
11203 orig_stmt, pass it to modify_call_stmt.
11204 (ipa_param_body_adjustments::modify_cfun_body): Adjust call to
11205 modify_gimple_stmt.
11206 (ipa_edge_modifications_finalize): New function.
11207 * tree-inline.c (remap_gimple_stmt): Pass original statement to
11208 modify_gimple_stmt.
11209 (copy_phis_for_bb): Do not copy dead PHI nodes.
11210 (expand_call_inline): Do not remap performed_splits.
11211 (update_clone_info): Likewise.
11212 * toplev.c: Include ipa-param-manipulation.h.
11213 (toplev::finalize): Call ipa_edge_modifications_finalize.
11215 2021-06-28 Andrew Pinski <apinski@marvell.com>
11217 * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Duplicate range
11218 info if we're the only things setting the target PHI.
11219 (value_replacement): Don't duplicate range here.
11220 (minmax_replacement): Likewise.
11222 2021-06-28 Richard Biener <rguenther@suse.de>
11224 PR tree-optimization/101229
11225 * gimple-walk.c (gimple_walk_op): Handle PHIs.
11227 2021-06-28 Martin Liska <mliska@suse.cz>
11229 * config/v850/v850.c (construct_dispose_instruction): Allocate
11231 (construct_prepare_instruction): Likewise.
11233 2021-06-28 Martin Liska <mliska@suse.cz>
11235 * config/v850/v850.c (v850_option_override): Build default
11237 (v850_can_inline_p): New. Allow MASK_PROLOG_FUNCTION to be
11238 ignored for inlining.
11239 (TARGET_CAN_INLINE_P): New.
11241 2021-06-28 Richard Biener <rguenther@suse.de>
11243 PR tree-optimization/101207
11244 * tree-vect-slp.c (vect_optimize_slp): Do BB reduction
11245 permute eliding for load permutations properly.
11247 2021-06-28 Richard Biener <rguenther@suse.de>
11249 PR tree-optimization/101173
11250 * gimple-loop-interchange.cc
11251 (tree_loop_interchange::valid_data_dependences): Disallow outer
11252 loop dependence distance of zero.
11254 2021-06-28 liuhongt <hongtao.liu@intel.com>
11257 * config/i386/sse.md (*avx_cmp<mode>3_lt): New
11258 define_insn_and_split.
11259 (*avx_cmp<mode>3_ltint): Ditto.
11260 (*avx2_pcmp<mode>3_3): Ditto.
11261 (*avx2_pcmp<mode>3_4): Ditto.
11262 (*avx2_pcmp<mode>3_5): Ditto.
11264 2021-06-28 liuhongt <hongtao.liu@intel.com>
11266 * config/i386/i386-builtin.def (IX86_BUILTIN_BLENDVPD256,
11267 IX86_BUILTIN_BLENDVPS256, IX86_BUILTIN_PBLENDVB256,
11268 IX86_BUILTIN_BLENDVPD, IX86_BUILTIN_BLENDVPS,
11269 IX86_BUILTIN_PBLENDVB128): Replace icode with
11271 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold blendv
11273 * config/i386/sse.md (*<sse4_1_avx2>_pblendvb_lt_subreg_not):
11274 New pre_reload splitter.
11276 2021-06-27 Andrew Pinski <apinski@marvell.com>
11278 PR middle-end/101230
11279 * fold-const.c (fold_ternary_loc): Check
11280 the return value of invert_tree_comparison.
11282 2021-06-27 David Edelsohn <dje.gcc@gmail.com>
11284 * config.gcc: Add SPDX License Identifier.
11285 (powerpc-ibm-aix789): Default to aix73.h.
11286 (powerpc-ibm-aix7.2.*.*): New stanza.
11287 * config/rs6000/aix72.h: Add SPDX License Identifier.
11288 * config/rs6000/aix73.h: New file.
11290 2021-06-26 Jason Merrill <jason@redhat.com>
11292 * except.c: #include "dwarf2.h" instead of "dwarf2out.h".
11294 2021-06-26 Andrew Pinski <apinski@marvell.com>
11296 * genmatch.c (lower_cond): Copy for_subst_vec
11297 for the simplify also.
11298 (lower): Swap the order for lower_for and lower_cond.
11300 2021-06-26 Andrew Pinski <apinski@marvell.com>
11302 * tree-ssa-phiopt.c (match_simplify_replacement): Reset
11303 flow senatitive info on the moved ssa set.
11305 2021-06-26 Andrew Pinski <apinski@marvell.com>
11307 * fold-const.c (fold_cond_expr_with_comparison):
11308 Exand arg0 into comp_code, arg00, and arg01.
11309 (fold_ternary_loc): Use invert_tree_comparison
11310 instead of fold_invert_truthvalue for the case
11311 where we have A CMP B ? C : A.
11313 2021-06-25 Martin Sebor <msebor@redhat.com>
11315 PR middle-end/101216
11316 * calls.c (maybe_warn_rdwr_sizes): Use the no_warning constant.
11318 2021-06-25 Jeff Law <jeffreyalaw@gmail.com>
11320 * config/h8300/h8300.c (select_cc_mode): Handle ASHIFTRT and LSHIFTRT.
11322 2021-06-25 Richard Biener <rguenther@suse.de>
11324 PR tree-optimization/101202
11325 * tree-vect-slp.c (vect_optimize_slp): Explicitely handle
11328 2021-06-25 Richard Biener <rguenther@suse.de>
11330 * tree-vect-slp-patterns.c (addsub_pattern::build): Copy
11331 STMT_VINFO_REDUC_DEF from the original representative.
11333 2021-06-25 Martin Sebor <msebor@redhat.com>
11335 * builtins.c (warn_string_no_nul): Replace uses of TREE_NO_WARNING,
11336 gimple_no_warning_p and gimple_set_no_warning with
11337 warning_suppressed_p, and suppress_warning.
11339 (maybe_warn_for_bound): Same.
11340 (warn_for_access): Same.
11341 (check_access): Same.
11342 (expand_builtin_strncmp): Same.
11343 (fold_builtin_varargs): Same.
11344 * calls.c (maybe_warn_nonstring_arg): Same.
11345 (maybe_warn_rdwr_sizes): Same.
11346 * cfgexpand.c (expand_call_stmt): Same.
11347 * cgraphunit.c (check_global_declaration): Same.
11348 * fold-const.c (fold_undefer_overflow_warnings): Same.
11349 (fold_truth_not_expr): Same.
11350 (fold_unary_loc): Same.
11351 (fold_checksum_tree): Same.
11352 * gimple-array-bounds.cc (array_bounds_checker::check_array_ref): Same.
11353 (array_bounds_checker::check_mem_ref): Same.
11354 (array_bounds_checker::check_addr_expr): Same.
11355 (array_bounds_checker::check_array_bounds): Same.
11356 * gimple-expr.c (copy_var_decl): Same.
11357 * gimple-fold.c (gimple_fold_builtin_strcpy): Same.
11358 (gimple_fold_builtin_strncat): Same.
11359 (gimple_fold_builtin_stxcpy_chk): Same.
11360 (gimple_fold_builtin_stpcpy): Same.
11361 (gimple_fold_builtin_sprintf): Same.
11362 (fold_stmt_1): Same.
11363 * gimple-ssa-isolate-paths.c (diag_returned_locals): Same.
11364 * gimple-ssa-nonnull-compare.c (do_warn_nonnull_compare): Same.
11365 * gimple-ssa-sprintf.c (handle_printf_call): Same.
11366 * gimple-ssa-store-merging.c (imm_store_chain_info::output_merged_store): Same.
11367 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
11368 * gimple-ssa-warn-restrict.h: Adjust declarations.
11369 (maybe_diag_access_bounds): Replace uses of TREE_NO_WARNING,
11370 gimple_no_warning_p and gimple_set_no_warning with
11371 warning_suppressed_p, and suppress_warning.
11372 (check_call): Same.
11373 (check_bounds_or_overlap): Same.
11374 * gimple.c (gimple_build_call_from_tree): Same.
11375 * gimplify.c (gimplify_return_expr): Same.
11376 (gimplify_cond_expr): Same.
11377 (gimplify_modify_expr_complex_part): Same.
11378 (gimplify_modify_expr): Same.
11379 (gimple_push_cleanup): Same.
11380 (gimplify_expr): Same.
11381 * omp-expand.c (expand_omp_for_generic): Same.
11382 (expand_omp_taskloop_for_outer): Same.
11383 * omp-low.c (lower_rec_input_clauses): Same.
11384 (lower_lastprivate_clauses): Same.
11385 (lower_send_clauses): Same.
11386 (lower_omp_target): Same.
11387 * tree-cfg.c (pass_warn_function_return::execute): Same.
11388 * tree-complex.c (create_one_component_var): Same.
11389 * tree-inline.c (remap_gimple_op_r): Same.
11390 (copy_tree_body_r): Same.
11391 (declare_return_variable): Same.
11392 (expand_call_inline): Same.
11393 * tree-nested.c (lookup_field_for_decl): Same.
11394 * tree-sra.c (create_access_replacement): Same.
11395 (generate_subtree_copies): Same.
11396 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Same.
11397 * tree-ssa-forwprop.c (combine_cond_expr_cond): Same.
11398 * tree-ssa-loop-ch.c (ch_base::copy_headers): Same.
11399 * tree-ssa-loop-im.c (execute_sm): Same.
11400 * tree-ssa-phiopt.c (cond_store_replacement): Same.
11401 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
11402 (handle_builtin_strcpy): Same.
11403 (maybe_diag_stxncpy_trunc): Same.
11404 (handle_builtin_stxncpy_strncat): Same.
11405 (handle_builtin_strcat): Same.
11406 * tree-ssa-uninit.c (get_no_uninit_warning): Same.
11407 (set_no_uninit_warning): Same.
11408 (uninit_undefined_value_p): Same.
11409 (warn_uninit): Same.
11410 (maybe_warn_operand): Same.
11411 * tree-vrp.c (compare_values_warnv): Same.
11412 * vr-values.c (vr_values::extract_range_for_var_from_comparison_expr): Same.
11413 (test_for_singularity): Same.
11414 * gimple.h (warning_suppressed_p): New function.
11415 (suppress_warning): Same.
11416 (copy_no_warning): Same.
11417 (gimple_set_block): Call gimple_set_location.
11418 (gimple_set_location): Call copy_warning.
11420 2021-06-25 Martin Sebor <msebor@redhat.com>
11422 * tree.h (warning_suppressed_at, copy_warning,
11423 warning_suppressed_p, suppress_warning): New functions.
11425 2021-06-25 Martin Sebor <msebor@redhat.com>
11427 * Makefile.in (OBJS-libcommon): Add diagnostic-spec.o.
11428 * gengtype.c (open_base_files): Add diagnostic-spec.h.
11429 * diagnostic-spec.c: New file.
11430 * diagnostic-spec.h: New file.
11431 * tree.h (no_warning, all_warnings, suppress_warning_at): New
11433 * warning-control.cc: New file.
11435 2021-06-25 liuhongt <hongtao.liu@intel.com>
11438 * config/i386/i386.c (x86_order_regs_for_local_alloc):
11441 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
11443 PR tree-optimization/101189
11444 * gimple-range-fold.cc (fold_using_range::range_of_range_op): Pass
11445 LHS range of condition to postfold routine.
11446 (fold_using_range::postfold_gcond_edges): Only process the TRUE or
11447 FALSE edge if the LHS range supports it being taken.
11448 * gimple-range-fold.h (postfold_gcond_edges): Add range parameter.
11450 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
11452 * value-relation.cc (equiv_oracle::dump): Do not dump NULL blocks.
11453 (relation_oracle::find_relation_block): Check correct bitmap.
11454 (relation_oracle::dump): Do not dump NULL blocks.
11456 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
11458 * gimple-range-cache.cc (ranger_cache::propagate_cache): Call
11459 range_on_edge instead of manually calculating.
11461 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
11463 * range-op.cc: Fix comment.
11465 2021-06-24 Uroš Bizjak <ubizjak@gmail.com>
11468 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
11469 Handle V8QI and V4HI modes.
11470 * config/i386/mmx.md (sse4_1_<any_extend:code>v4qiv4hi2):
11472 (sse4_1_<any_extend:code>v4qiv4hi2): Ditto.
11473 (mmxpackmode): New mode attribute.
11474 (vec_pack_trunc_<mmxpackmode:mode>): New expander.
11475 (mmxunpackmode): New mode attribute.
11476 (vec_unpacks_lo_<mmxunpackmode:mode>): New expander.
11477 (vec_unpacks_hi_<mmxunpackmode:mode>): Ditto.
11478 (vec_unpacku_lo_<mmxunpackmode:mode>): Ditto.
11479 (vec_unpacku_hi_<mmxunpackmode:mode>): Ditto.
11480 * config/i386/i386.md (extsuffix): Move from ...
11481 * config/i386/sse.md: ... here.
11483 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
11485 * dwarf2out.c (dwarf2out_assembly_start): Emit .file 0 marker here..
11486 (dwarf2out_finish): ...instead of here.
11488 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
11490 * configure.ac (--gdwarf-5 option): Use objdump instead of readelf.
11491 (working --gdwarf-4/--gdwarf-5 for all sources): Likewise.
11492 (--gdwarf-4 not refusing generated .debug_line): Adjust for Windows.
11493 * configure: Regenerate.
11495 2021-06-24 Richard Biener <rguenther@suse.de>
11497 * config/i386/sse.md (vec_addsubv4df3, vec_addsubv2df3,
11498 vec_addsubv8sf3, vec_addsubv4sf3): Merge into ...
11499 (vec_addsub<mode>3): ... using a new addsub_cst mode attribute.
11501 2021-06-24 Richard Biener <rguenther@suse.de>
11503 * config/i386/sse.md (avx_addsubv4df3): Rename to
11505 (avx_addsubv8sf3): Rename to vec_addsubv8sf3.
11506 (sse3_addsubv2df3): Rename to vec_addsubv2df3.
11507 (sse3_addsubv4sf3): Rename to vec_addsubv4sf3.
11508 * config/i386/i386-builtin.def: Adjust.
11509 * internal-fn.def (VEC_ADDSUB): New internal optab fn.
11510 * optabs.def (vec_addsub_optab): New optab.
11511 * tree-vect-slp-patterns.c (class addsub_pattern): New.
11512 (slp_patterns): Add addsub_pattern.
11513 * tree-vect-slp.c (vect_optimize_slp): Disable propagation
11514 across CFN_VEC_ADDSUB.
11515 * tree-vectorizer.h (vect_pattern::vect_pattern): Make
11517 * doc/md.texi (vec_addsub<mode>3): Document.
11519 2021-06-24 Jakub Jelinek <jakub@redhat.com>
11521 PR middle-end/101170
11522 * df-scan.c (df_ref_record): For paradoxical big-endian SUBREGs
11523 where regno + subreg_regno_offset wraps around use 0 as starting
11526 2021-06-24 Jakub Jelinek <jakub@redhat.com>
11528 PR middle-end/101172
11529 * stor-layout.c (finish_bitfield_representative): If nextf has
11530 error_mark_node type, set repr type to error_mark_node too.
11532 2021-06-24 Ilya Leoshkevich <iii@linux.ibm.com>
11534 * config/s390/s390.c (s390_function_profiler): Ignore labelno
11536 * config/s390/s390.h (NO_PROFILE_COUNTERS): Define.
11538 2021-06-24 Richard Biener <rguenther@suse.de>
11540 * tree-vect-slp.c (vect_optimize_slp): Do not propagate
11541 across operations that have different semantics on different
11544 2021-06-24 Jakub Jelinek <jakub@redhat.com>
11546 * tree.h (OMP_CLAUSE_MAP_IN_REDUCTION): Document meaning for OpenMP.
11547 * gimplify.c (gimplify_scan_omp_clauses): For OpenMP map clauses
11548 with OMP_CLAUSE_MAP_IN_REDUCTION flag partially defer gimplification
11549 of non-decl OMP_CLAUSE_DECL. For OMP_CLAUSE_IN_REDUCTION on
11550 OMP_TARGET user outer_ctx instead of ctx for placeholders and
11551 initializer/combiner gimplification.
11552 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_MAP_IN_REDUCTION
11553 on target constructs.
11554 (lower_rec_input_clauses): Likewise.
11555 (lower_omp_target): Likewise.
11556 * omp-expand.c (expand_omp_target): Temporarily ignore nowait clause
11557 on target if in_reduction is present.
11559 2021-06-24 Kewen Lin <linkw@linux.ibm.com>
11561 * tree-predcom.c (class pcom_worker): New class.
11562 (release_chain): Renamed to...
11563 (pcom_worker::release_chain): ...this.
11564 (release_chains): Renamed to...
11565 (pcom_worker::release_chains): ...this.
11566 (aff_combination_dr_offset): Renamed to...
11567 (pcom_worker::aff_combination_dr_offset): ...this.
11568 (determine_offset): Renamed to...
11569 (pcom_worker::determine_offset): ...this.
11570 (class comp_ptrs): New class.
11571 (split_data_refs_to_components): Renamed to...
11572 (pcom_worker::split_data_refs_to_components): ...this,
11573 and update with class comp_ptrs.
11574 (suitable_component_p): Renamed to...
11575 (pcom_worker::suitable_component_p): ...this.
11576 (filter_suitable_components): Renamed to...
11577 (pcom_worker::filter_suitable_components): ...this.
11578 (valid_initializer_p): Renamed to...
11579 (pcom_worker::valid_initializer_p): ...this.
11580 (find_looparound_phi): Renamed to...
11581 (pcom_worker::find_looparound_phi): ...this.
11582 (add_looparound_copies): Renamed to...
11583 (pcom_worker::add_looparound_copies): ...this.
11584 (determine_roots_comp): Renamed to...
11585 (pcom_worker::determine_roots_comp): ...this.
11586 (determine_roots): Renamed to...
11587 (pcom_worker::determine_roots): ...this.
11588 (single_nonlooparound_use): Renamed to...
11589 (pcom_worker::single_nonlooparound_use): ...this.
11590 (remove_stmt): Renamed to...
11591 (pcom_worker::remove_stmt): ...this.
11592 (execute_pred_commoning_chain): Renamed to...
11593 (pcom_worker::execute_pred_commoning_chain): ...this.
11594 (execute_pred_commoning): Renamed to...
11595 (pcom_worker::execute_pred_commoning): ...this.
11596 (struct epcc_data): New member worker.
11597 (execute_pred_commoning_cbck): Call execute_pred_commoning
11598 with pcom_worker pointer.
11599 (find_use_stmt): Renamed to...
11600 (pcom_worker::find_use_stmt): ...this.
11601 (find_associative_operation_root): Renamed to...
11602 (pcom_worker::find_associative_operation_root): ...this.
11603 (find_common_use_stmt): Renamed to...
11604 (pcom_worker::find_common_use_stmt): ...this.
11605 (combinable_refs_p): Renamed to...
11606 (pcom_worker::combinable_refs_p): ...this.
11607 (reassociate_to_the_same_stmt): Renamed to...
11608 (pcom_worker::reassociate_to_the_same_stmt): ...this.
11609 (stmt_combining_refs): Renamed to...
11610 (pcom_worker::stmt_combining_refs): ...this.
11611 (combine_chains): Renamed to...
11612 (pcom_worker::combine_chains): ...this.
11613 (try_combine_chains): Renamed to...
11614 (pcom_worker::try_combine_chains): ...this.
11615 (prepare_initializers_chain): Renamed to...
11616 (pcom_worker::prepare_initializers_chain): ...this.
11617 (prepare_initializers): Renamed to...
11618 (pcom_worker::prepare_initializers): ...this.
11619 (prepare_finalizers_chain): Renamed to...
11620 (pcom_worker::prepare_finalizers_chain): ...this.
11621 (prepare_finalizers): Renamed to...
11622 (pcom_worker::prepare_finalizers): ...this.
11623 (tree_predictive_commoning_loop): Renamed to...
11624 (pcom_worker::tree_predictive_commoning_loop): ...this, adjust
11625 some calls and remove some cleanup code.
11626 (tree_predictive_commoning): Adjusted to use pcom_worker instance.
11627 (static variable looparound_phis): Remove.
11628 (static variable name_expansions): Remove.
11630 2021-06-24 Richard Biener <rguenther@suse.de>
11632 * tree-vect-slp.c (slpg_vertex): New struct.
11633 (vect_slp_build_vertices): Adjust.
11634 (vect_optimize_slp): Likewise. Maintain an outgoing permute
11635 and a materialized one.
11637 2021-06-24 Richard Biener <rguenther@suse.de>
11639 PR tree-optimization/101105
11640 * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
11641 Only ignore steps when they are equal or scalar order is preserved.
11643 2021-06-24 liuhongt <hongtao.liu@intel.com>
11646 * config/i386/i386-expand.c (ix86_expand_vec_interleave):
11647 Adjust comments for ix86_expand_vecop_qihi2.
11648 (ix86_expand_vecmul_qihi): Renamed to ..
11649 (ix86_expand_vecop_qihi2): Adjust function prototype to
11650 support shift operation, add static to definition.
11651 (ix86_expand_vec_shift_qihi_constant): Add static to definition.
11652 (ix86_expand_vecop_qihi): Call ix86_expand_vecop_qihi2 and
11653 ix86_expand_vec_shift_qihi_constant.
11654 * config/i386/i386-protos.h (ix86_expand_vecmul_qihi): Deleted.
11655 (ix86_expand_vec_shift_qihi_constant): Deleted.
11656 * config/i386/sse.md (VI12_256_512_AVX512VL): New mode
11658 (mulv8qi3): Call ix86_expand_vecop_qihi directly, add
11659 condition TARGET_64BIT.
11660 (mul<mode>3): Ditto.
11661 (<insn><mode>3): Ditto.
11662 (vlshr<mode>3): Extend to support avx512 vlshr.
11663 (v<insn><mode>3): New expander for
11665 (v<insn>v8qi3): Ditto.
11666 (vashrv8hi3<mask_name>): Renamed to ..
11667 (vashr<mode>3): And extend to support V16QImode for avx512.
11668 (vashrv16qi3): Deleted.
11669 (vashrv2di3<mask_name>): Extend expander to support avx512
11672 2021-06-23 Dimitar Dimitrov <dimitar@dinux.eu>
11674 * doc/lto.texi (Design Overview): Update that slim objects are
11677 2021-06-23 Aaron Sawdey <acsawdey@linux.ibm.com>
11679 * config/rs6000/rs6000-cpus.def: Take OPTION_MASK_PCREL_OPT out
11680 of OTHER_POWER10_MASKS so it will not be enabled by default.
11682 2021-06-23 Richard Biener <rguenther@suse.de>
11683 Martin Jambor <mjambor@suse.cz>
11685 * tree-inline.c (setup_one_parameter): Set TREE_READONLY of the
11686 param replacement unconditionally. Adjust comment.
11688 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
11690 * Makefile.in (OBJS): Add gimple-range-fold.o
11691 * gimple-range-fold.cc: New.
11692 * gimple-range-fold.h: New.
11693 * gimple-range-gori.cc (gimple_range_calc_op1): Move to here.
11694 (gimple_range_calc_op2): Ditto.
11695 * gimple-range-gori.h: Move prototypes to here.
11696 * gimple-range.cc: Adjust include files.
11697 (fur_source:fur_source): Relocate to gimple-range-fold.cc.
11698 (fur_source::get_operand): Ditto.
11699 (fur_source::get_phi_operand): Ditto.
11700 (fur_source::query_relation): Ditto.
11701 (fur_source::register_relation): Ditto.
11702 (class fur_edge): Ditto.
11703 (fur_edge::fur_edge): Ditto.
11704 (fur_edge::get_operand): Ditto.
11705 (fur_edge::get_phi_operand): Ditto.
11706 (fur_stmt::fur_stmt): Ditto.
11707 (fur_stmt::get_operand): Ditto.
11708 (fur_stmt::get_phi_operand): Ditto.
11709 (fur_stmt::query_relation): Ditto.
11710 (class fur_depend): Relocate to gimple-range-fold.h.
11711 (fur_depend::fur_depend): Relocate to gimple-range-fold.cc.
11712 (fur_depend::register_relation): Ditto.
11713 (fur_depend::register_relation): Ditto.
11714 (class fur_list): Ditto.
11715 (fur_list::fur_list): Ditto.
11716 (fur_list::get_operand): Ditto.
11717 (fur_list::get_phi_operand): Ditto.
11718 (fold_range): Ditto.
11719 (adjust_pointer_diff_expr): Ditto.
11720 (gimple_range_adjustment): Ditto.
11721 (gimple_range_base_of_assignment): Ditto.
11722 (gimple_range_operand1): Ditto.
11723 (gimple_range_operand2): Ditto.
11724 (gimple_range_calc_op1): Relocate to gimple-range-gori.cc.
11725 (gimple_range_calc_op2): Ditto.
11726 (fold_using_range::fold_stmt): Relocate to gimple-range-fold.cc.
11727 (fold_using_range::range_of_range_op): Ditto.
11728 (fold_using_range::range_of_address): Ditto.
11729 (fold_using_range::range_of_phi): Ditto.
11730 (fold_using_range::range_of_call): Ditto.
11731 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
11732 (fold_using_range::range_of_builtin_call): Ditto.
11733 (fold_using_range::range_of_cond_expr): Ditto.
11734 (fold_using_range::range_of_ssa_name_with_loop_info): Ditto.
11735 (fold_using_range::relation_fold_and_or): Ditto.
11736 (fold_using_range::postfold_gcond_edges): Ditto.
11737 * gimple-range.h: Add gimple-range-fold.h to include files. Change
11738 GIMPLE_RANGE_STMT_H to GIMPLE_RANGE_H.
11739 (gimple_range_handler): Relocate to gimple-range-fold.h.
11740 (gimple_range_ssa_p): Ditto.
11741 (range_compatible_p): Ditto.
11742 (class fur_source): Ditto.
11743 (class fur_stmt): Ditto.
11744 (class fold_using_range): Ditto.
11745 (gimple_range_calc_op1): Relocate to gimple-range-gori.h
11746 (gimple_range_calc_op2): Ditto.
11748 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
11750 PR tree-optimization/101148
11751 PR tree-optimization/101014
11752 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
11753 (ranger_cache::~ranger_cache): Adjust.
11754 (ranger_cache::block_range): Check if propagation disallowed.
11755 (ranger_cache::propagate_cache): Disallow propagation if new value
11756 can't be stored properly.
11757 * gimple-range-cache.h (ranger_cache::m_propfail): New member.
11759 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
11761 * gimple-range-cache.cc (class ssa_block_ranges): Adjust prototype.
11762 (sbr_vector::set_bb_range): Return true.
11763 (class sbr_sparse_bitmap): Adjust.
11764 (sbr_sparse_bitmap::set_bb_range): Return value.
11765 (block_range_cache::set_bb_range): Return value.
11766 (ranger_cache::propagate_cache): Use return value to print msg.
11767 * gimple-range-cache.h (class block_range_cache): Adjust.
11769 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
11771 * gimple-range.cc (dump_bb): Use range_on_edge from the cache.
11773 2021-06-23 Jeff Law <jeffreyalaw@gmail.com>
11775 * config/h8300/logical.md (<code><mode>3<ccnz>): Use <cczn>
11776 so this pattern can be used for test/compare removal. Pass
11777 current insn to compute_logical_op_length and output_logical_op.
11778 * config/h8300/h8300.c (compute_logical_op_cc): Remove.
11779 (h8300_and_costs): Add argument to compute_logical_op_length.
11780 (output_logical_op): Add new argument. Use it to determine if the
11781 condition codes are used and adjust the output accordingly.
11782 (compute_logical_op_length): Add new argument and update length
11783 computations when condition codes are used.
11784 * config/h8300/h8300-protos.h (compute_logical_op_length): Update
11786 (output_logical_op): Likewise.
11788 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
11791 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
11792 Handle 64bit modes for TARGET_XOP. Use indirect gen_* functions.
11793 * config/i386/mmx.md (mmx_ppermv64): New insn pattern.
11794 * config/i386/i386.md (unspec): Move UNSPEC_XOP_PERMUTE from ...
11795 * config/i386/sse.md (unspec): ... here.
11797 2021-06-23 Martin Liska <mliska@suse.cz>
11800 * optc-save-gen.awk: Put back arm_fp16_format to
11803 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
11806 * config/i386/i386.md (bsr_rex64): Add zero-flag setting RTX.
11809 (clz<mode>2): Update RTX pattern for additions.
11811 2021-06-23 Jakub Jelinek <jakub@redhat.com>
11813 PR middle-end/101167
11814 * omp-low.c (lower_omp_regimplify_p): Regimplify also PARM_DECLs
11815 and RESULT_DECLs that have DECL_HAS_VALUE_EXPR_P set.
11817 2021-06-22 Sergei Trofimovich <siarheit@google.com>
11819 * doc/rtl.texi: drop unbalanced parenthesis.
11821 2021-06-22 Richard Biener <rguenther@suse.de>
11823 PR middle-end/101156
11824 * gimplify.c (gimplify_expr): Remove premature incorrect
11827 2021-06-22 Jakub Jelinek <jakub@redhat.com>
11829 PR tree-optimization/101159
11830 * tree-vect-patterns.c (vect_recog_popcount_pattern): Fix some
11833 2021-06-22 Jakub Jelinek <jakub@redhat.com>
11835 PR middle-end/101160
11836 * function.c (assign_parms): For decl_result with TYPE_EMPTY_P type
11837 clear crtl->return_rtx instead of keeping it referencing a pseudo.
11839 2021-06-22 Jakub Jelinek <jakub@redhat.com>
11840 Andrew Pinski <apinski@marvell.com>
11842 PR tree-optimization/101162
11843 * fold-const.c (range_check_type): Handle OFFSET_TYPE like pointer
11846 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11848 * range-op.cc (range_relational_tests): New.
11849 (range_op_tests): Call range_relational_tests.
11851 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11853 * range-op.cc (operator_cast::lhs_op1_relation): New.
11854 (operator_identity::lhs_op1_relation): Mew.
11856 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11858 * range-op.cc (operator_minus::op1_op2_relation_effect): New.
11860 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11862 * range-op.cc (operator_plus::lhs_op1_relation): New.
11863 (operator_plus::lhs_op2_relation): New.
11865 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11867 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a
11868 relation_oracle if dominators exist.
11869 (ranger_cache::~ranger_cache): Dispose of oracle.
11870 (ranger_cache::dump_bb): Dump oracle.
11871 * gimple-range.cc (fur_source::fur_source): New.
11872 (fur_source::get_operand): Use mmeber query.
11873 (fur_source::get_phi_operand): Use member_query.
11874 (fur_source::query_relation): New.
11875 (fur_source::register_dependency): Delete.
11876 (fur_source::register_relation): New.
11877 (fur_edge::fur_edge): Adjust.
11878 (fur_edge::get_phi_operand): Fix comment.
11879 (fur_edge::query): Delete.
11880 (fur_stmt::fur_stmt): Adjust.
11881 (fur_stmt::query): Delete.
11882 (fur_depend::fur_depend): Adjust.
11883 (fur_depend::register_relation): New.
11884 (fur_depend::register_relation): New.
11885 (fur_list::fur_list): Adjust.
11886 (fur_list::get_operand): Use member query.
11887 (fold_using_range::range_of_range_op): Process and query relations.
11888 (fold_using_range::range_of_address): Adjust dependency call.
11889 (fold_using_range::range_of_phi): Ditto.
11890 (gimple_ranger::gimple_ranger): New. Use ranger_ache oracle.
11891 (fold_using_range::relation_fold_and_or): New.
11892 (fold_using_range::postfold_gcond_edges): New.
11893 * gimple-range.h (class gimple_ranger): Adjust.
11894 (class fur_source): Adjust members.
11895 (class fur_stmt): Ditto.
11896 (class fold_using_range): Ditto.
11898 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11900 * range-op.cc (range_operator::wi_fold): Apply relation effect.
11901 (range_operator::fold_range): Adjust and apply relation effect.
11902 (*::fold_range): Add relation parameters.
11903 (*::op1_range): Ditto.
11904 (*::op2_range): Ditto.
11905 (range_operator::lhs_op1_relation): New.
11906 (range_operator::lhs_op2_relation): New.
11907 (range_operator::op1_op2_relation): New.
11908 (range_operator::op1_op2_relation_effect): New.
11909 (relop_early_resolve): New.
11910 (operator_equal::op1_op2_relation): New.
11911 (operator_equal::fold_range): Call relop_early_resolve.
11912 (operator_not_equal::op1_op2_relation): New.
11913 (operator_not_equal::fold_range): Call relop_early_resolve.
11914 (operator_lt::op1_op2_relation): New.
11915 (operator_lt::fold_range): Call relop_early_resolve.
11916 (operator_le::op1_op2_relation): New.
11917 (operator_le::fold_range): Call relop_early_resolve.
11918 (operator_gt::op1_op2_relation): New.
11919 (operator_gt::fold_range): Call relop_early_resolve.
11920 (operator_ge::op1_op2_relation): New.
11921 (operator_ge::fold_range): Call relop_early_resolve.
11922 * range-op.h (class range_operator): Adjust parameters and methods.
11924 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11926 * Makefile.in (OBJS): Add value-relation.o.
11927 * gimple-range.h: Adjust include files.
11928 * tree-data-ref.c: Adjust include file order.
11929 * value-query.cc (range_query::get_value_range): Default to no oracle.
11930 (range_query::query_relation): New.
11931 (range_query::query_relation): New.
11932 * value-query.h (class range_query): Adjust.
11933 * value-relation.cc: New.
11934 * value-relation.h: New.
11936 2021-06-22 Richard Biener <rguenther@suse.de>
11938 PR tree-optimization/101151
11939 * tree-ssa-sink.c (statement_sink_location): Expand irreducible
11942 2021-06-22 Jojo R <rjiejie@linux.alibaba.com>
11944 * config/riscv/riscv.c (thead_c906_tune_info): New.
11945 (riscv_tune_info_table): Use new tune.
11947 2021-06-22 Richard Biener <rguenther@suse.de>
11949 PR tree-optimization/101158
11950 * tree-vect-slp.c (vect_build_slp_tree_1): Move same operand
11951 checking after checking for matching operation.
11953 2021-06-22 Richard Biener <rguenther@suse.de>
11955 PR tree-optimization/101159
11956 * tree-vect-patterns.c (vect_recog_popcount_pattern): Add
11957 missing NULL vectype check.
11959 2021-06-22 Richard Biener <rguenther@suse.de>
11961 PR tree-optimization/101154
11962 * tree-vect-slp.c (vect_build_slp_tree_2): Fix out-of-bound access.
11964 2021-06-22 Jakub Jelinek <jakub@redhat.com>
11967 * config/i386/i386-protos.h (ix86_last_zero_store_uid): Declare.
11968 * config/i386/i386-expand.c (ix86_last_zero_store_uid): New variable.
11969 * config/i386/i386.c (ix86_expand_prologue): Clear it.
11970 * config/i386/i386.md (peephole2s for 1/2/4 stores of const0_rtx):
11971 Remove "" from match_operand. Emit new insns using emit_move_insn and
11972 set ix86_last_zero_store_uid to INSN_UID of the last store.
11973 Add peephole2s for 1/2/4 stores of const0_rtx following previous
11976 2021-06-22 Martin Liska <mliska@suse.cz>
11978 * auto-profile.c (AUTO_PROFILE_VERSION): Bump as string format
11981 2021-06-22 Martin Liska <mliska@suse.cz>
11983 * gcov-io.h: Remove padding entries.
11985 2021-06-22 liuhongt <hongtao.liu@intel.com>
11987 PR tree-optimization/97770
11988 * tree-vect-patterns.c (vect_recog_popcount_pattern):
11990 (vect_recog_func vect_vect_recog_func_ptrs): Add new pattern.
11992 2021-06-22 liuhongt <hongtao.liu@intel.com>
11995 * config/i386/i386-builtin.def (BDESC): Adjust builtin name.
11996 * config/i386/sse.md (<avx512>_expand<mode>_mask): Rename to ..
11997 (expand<mode>_mask): this ..
11998 (*expand<mode>_mask): New pre_reload splitter to transform
11999 v{,p}expand* to vmov* when mask is zero, all ones, or has all
12000 ones in it's lower part, otherwise still generate
12003 2021-06-22 liuhongt <hongtao.liu@intel.com>
12006 * config/i386/i386-expand.c
12007 (ix86_expand_special_args_builtin): Keep constm1_operand only
12008 if it satisfies insn's operand predicate.
12010 2021-06-21 Jason Merrill <jason@redhat.com>
12013 * df-scan.c (df_ref_record): Check that regno < endregno.
12014 * function.c (assign_parms, expand_function_end): Do nothing with a
12015 TYPE_EMPTY_P result.
12017 2021-06-21 Richard Biener <rguenther@suse.de>
12019 PR tree-optimization/101120
12020 * tree-vect-data-refs.c (bump_vector_ptr): Fold the
12022 * tree-vect-slp.c (vect_transform_slp_perm_load): Add
12023 DR chain DCE capability.
12024 * tree-vectorizer.h (vect_transform_slp_perm_load): Adjust.
12025 * tree-vect-stmts.c (vectorizable_load): Remove unused
12026 loads in the DR chain for SLP.
12028 2021-06-21 Jakub Jelinek <jakub@redhat.com>
12030 PR inline-asm/100785
12031 * gimplify.c (gimplify_asm_expr): Don't diagnose errors if
12032 output or input operands were already error_mark_node.
12033 * cfgexpand.c (expand_asm_stmt): If errors are emitted,
12034 remove all inputs, outputs and clobbers from the asm and
12035 set template to "".
12037 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
12039 * config/arm/arm_neon.h (vceq_s8): Replace builtin with __a == __b.
12040 (vceq_s16): Likewise.
12041 (vceq_s32): Likewise.
12042 (vceq_u8): Likewise.
12043 (vceq_u16): Likewise.
12044 (vceq_u32): Likewise.
12045 (vceq_p8): Likewise.
12046 (vceqq_s8): Likewise.
12047 (vceqq_s16): Likewise.
12048 (vceqq_s32): Likewise.
12049 (vceqq_u8): Likewise.
12050 (vceqq_u16): Likewise.
12051 (vceqq_u32): Likewise.
12052 (vceqq_p8): Likewise.
12053 (vceq_f32): Gate __a == __b on __FAST_MATH__.
12054 (vceqq_f32): Likewise.
12055 (vceq_f16): Likewise.
12056 (vceqq_f16): Likewise.
12058 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
12061 * config/arm/iterators.md (NEON_VACMP): Remove.
12062 * config/arm/neon.md (neon_vca<cmp_op><mode>): Use GLTE instead of GTGE
12064 (neon_vca<cmp_op><mode>_insn): Likewise.
12065 (neon_vca<cmp_op_unsp><mode>_insn_unspec): Use NEON_VAGLTE instead of
12068 2021-06-21 Richard Biener <rguenther@suse.de>
12070 PR tree-optimization/101121
12071 * tree-vect-slp.c (vect_build_slp_tree_2): To not fail fatally
12072 when we just lack a stmt with the desired op when doing permutation.
12073 (vect_build_slp_tree): When caching a failed SLP build attempt
12074 assert that at least one lane is marked as not matching.
12076 2021-06-21 liuhongt <hongtao.liu@intel.com>
12079 * config/i386/i386.md: (*anddi_1): Disparage slightly the mask
12080 register alternative.
12081 (*and<mode>_1): Ditto.
12083 (*andn<mode>_1): Ditto.
12084 (*<code><mode>_1): Ditto.
12085 (*<code>qi_1): Ditto.
12086 (*one_cmpl<mode>2_1): Ditto.
12087 (*one_cmplsi2_1_zext): Ditto.
12088 (*one_cmplqi2_1): Ditto.
12089 * config/i386/i386.c (x86_order_regs_for_local_alloc): Change
12090 the order of mask registers to be before general registers.
12092 2021-06-21 Roger Sayle <roger@nextmovesoftware.com>
12095 * config/i386/i386.md: New define_peephole2s to shrink writing
12096 1, 2 or 4 consecutive zeros to memory when optimizing for size.
12098 2021-06-18 Jeff Law <jeffreyalaw@gmail.com>
12100 * config/h8300/h8300.c (h8300_select_cc_mode): Handle SYMBOL_REF.
12101 * config/h8300/logical.md (<code><mode>3 logcial expander): Generate
12102 more efficient code when the source can be trivially simplified.
12104 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
12106 * gimple-range-cache.cc (ranger_cache::range_of_def): Calculate
12107 a range if global is not available.
12108 (ranger_cache::entry_range): Fallback to range_of_def.
12109 * gimple-range-cache.h (range_of_def): Adjust prototype.
12111 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
12113 PR tree-optimization/101014
12114 * gimple-range-cache.cc (ranger_cache::ranger_cache): Remove poor
12116 (ranger_cache::~ranger_cache): Ditto.
12117 (ranger_cache::enable_new_values): Delete.
12118 (ranger_cache::push_poor_value): Delete.
12119 (ranger_cache::range_of_def): Remove poor value processing.
12120 (ranger_cache::entry_range): Ditto.
12121 (ranger_cache::fill_block_cache): Ditto.
12122 * gimple-range-cache.h (class ranger_cache): Remove poor value members.
12123 * gimple-range.cc (gimple_ranger::range_of_expr): Remove call.
12124 * gimple-range.h (class gimple_ranger): Adjust.
12126 2021-06-18 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
12129 * common/config/arm/arm-common.c (arm_canon_arch_option_1): New function
12130 derived from arm_canon_arch.
12131 (arm_canon_arch_option): Call it.
12132 (arm_canon_arch_multilib_option): New function.
12133 * config/arm/arm-cpus.in (IGNORE_FOR_MULTILIB): New fgroup.
12134 * config/arm/arm.h (arm_canon_arch_multilib_option): New prototype.
12135 (CANON_ARCH_MULTILIB_SPEC_FUNCTION): New macro.
12136 (MULTILIB_ARCH_CANONICAL_SPECS): New macro.
12137 (DRIVER_SELF_SPECS): Add MULTILIB_ARCH_CANONICAL_SPECS.
12138 * config/arm/arm.opt (mlibarch): New option.
12139 * config/arm/t-rmprofile (MULTILIB_MATCHES): For armv8*-m, replace use
12140 of march on RHS with mlibarch.
12142 2021-06-18 Marcel Vollweiler <marcel@codesourcery.com>
12144 * config.in: Regenerate.
12145 * config/gcn/gcn.c (print_operand_address): Fix for global_load assembler
12147 * configure: Regenerate.
12148 * configure.ac: Fix for global_load assembler functions.
12150 2021-06-18 Richard Biener <rguenther@suse.de>
12152 PR tree-optimization/101112
12153 * tree-vect-slp.c (vect_slp_linearize_chain): Fix condition
12154 to lookup a pattern stmt def.
12156 2021-06-18 Jakub Jelinek <jakub@redhat.com>
12158 PR middle-end/101062
12159 * stor-layout.c (finish_bitfield_layout): Don't add bitfield
12160 representatives in QUAL_UNION_TYPE.
12162 2021-06-18 Andrew Pinski <apinski@marvell.com>
12164 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
12165 Add counting of how many times it is done.
12166 (factor_out_conditional_conversion): Likewise.
12167 (match_simplify_replacement): Likewise.
12168 (value_replacement): Likewise.
12169 (spaceship_replacement): Likewise.
12170 (cond_store_replacement): Likewise.
12171 (cond_if_else_store_replacement_1): Likewise.
12172 (hoist_adjacent_loads): Likewise.
12174 2021-06-18 Andrew Pinski <apinski@marvell.com>
12176 * tree-cfg.c (verify_gimple_assign_unary): Reject point and offset
12177 types on NEGATE_EXPR, ABS_EXPR, BIT_NOT_EXPR, PAREN_EXPR and CNONJ_EXPR.
12178 (verify_gimple_assign_binary): Reject point and offset types on
12179 MULT_EXPR, MULT_HIGHPART_EXPR, TRUNC_DIV_EXPR, CEIL_DIV_EXPR,
12180 FLOOR_DIV_EXPR, ROUND_DIV_EXPR, TRUNC_MOD_EXPR, CEIL_MOD_EXPR,
12181 FLOOR_MOD_EXPR, ROUND_MOD_EXPR, RDIV_EXPR, and EXACT_DIV_EXPR.
12183 2021-06-18 Michael Meissner <meissner@linux.ibm.com>
12185 * config/rs6000/rs6000.c (rs6000_emit_minmax): Add support for ISA
12186 3.1 IEEE 128-bit floating point xsmaxcqp/xsmincqp instructions.
12187 * config/rs6000/rs6000.md (s<minmax><mode>3, IEEE128 iterator):
12190 2021-06-17 Aaron Sawdey <acsawdey@linux.ibm.com>
12192 * config/rs6000/genfusion.pl (gen_logical_addsubf): Add
12193 earlyclobber to alts 0/1.
12194 (gen_addadd): Add earlyclobber to alts 0/1.
12195 * config/rs6000/fusion.md: Regenerate file.
12197 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12199 * cfgloopanal.c (get_loop_hot_path): Make path an auto_vec.
12201 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
12203 * gimple-range-cache.cc: Comment cleanups.
12204 * gimple-range-gori.cc: Comment cleanups.
12205 * gimple-range.cc: Comment/spacing cleanups
12206 * value-range.h: Comment cleanups.
12208 2021-06-17 H.J. Lu <hjl.tools@gmail.com>
12211 * calls.c (expand_call): Replace PUSH_ARGS with
12212 targetm.calls.push_argument (0).
12213 (emit_library_call_value_1): Likewise.
12214 * defaults.h (PUSH_ARGS): Removed.
12215 (PUSH_ARGS_REVERSED): Replace PUSH_ARGS with
12216 targetm.calls.push_argument (0).
12217 * expr.c (block_move_libcall_safe_for_call_parm): Likewise.
12218 (emit_push_insn): Pass the number bytes to push to
12219 targetm.calls.push_argument and pass 0 if ARGS_ADDR is 0.
12220 * hooks.c (hook_bool_uint_true): New.
12221 * hooks.h (hook_bool_uint_true): Likewise.
12222 * rtlanal.c (nonzero_bits1): Replace PUSH_ARGS with
12223 targetm.calls.push_argument (0).
12224 * target.def (push_argument): Add a targetm.calls hook.
12225 * targhooks.c (default_push_argument): New.
12226 * targhooks.h (default_push_argument): Likewise.
12227 * config/bpf/bpf.h (PUSH_ARGS): Removed.
12228 * config/cr16/cr16.c (TARGET_PUSH_ARGUMENT): New.
12229 * config/cr16/cr16.h (PUSH_ARGS): Removed.
12230 * config/i386/i386.c (ix86_push_argument): New.
12231 (TARGET_PUSH_ARGUMENT): Likewise.
12232 * config/i386/i386.h (PUSH_ARGS): Removed.
12233 * config/m32c/m32c.c (TARGET_PUSH_ARGUMENT): New.
12234 * config/m32c/m32c.h (PUSH_ARGS): Removed.
12235 * config/nios2/nios2.h (PUSH_ARGS): Likewise.
12236 * config/pru/pru.h (PUSH_ARGS): Likewise.
12237 * doc/tm.texi.in: Remove PUSH_ARGS documentation. Add
12238 TARGET_PUSH_ARGUMENT hook.
12239 * doc/tm.texi: Regenerated.
12241 2021-06-17 Uroš Bizjak <ubizjak@gmail.com>
12244 * config/i386/i386-expand.c (expand_vector_set_var):
12245 Handle V2FS mode remapping. Pass TARGET_MMX_WITH_SSE to
12246 ix86_expand_vector_init_duplicate.
12247 (ix86_expand_vector_init_duplicate): Emit insv_1 for
12248 QImode for !TARGET_PARTIAL_REG_STALL.
12249 * config/i386/predicates.md (vec_setm_mmx_operand): New predicate.
12250 * config/i386/mmx.md (vec_setv2sf): Use vec_setm_mmx_operand
12251 as operand 2 predicate. Call ix86_expand_vector_set_var
12252 for non-constant index operand.
12253 (vec_setv2si): Ditto.
12254 (vec_setv4hi): Ditto.
12255 (vec_setv8qi): ditto.
12257 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
12259 PR tree-optimization/100790
12260 * gimple-range.cc (range_of_builtin_call): Cleanup clz and ctz
12263 2021-06-17 Martin Liska <mliska@suse.cz>
12265 * doc/invoke.texi: Use consistently -O1 instead of -O.
12267 2021-06-17 Martin Liska <mliska@suse.cz>
12269 * gcov-io.h: Update documentation entry about string format.
12271 2021-06-17 Marius Hillenbrand <mhillen@linux.ibm.com>
12274 * config/s390/vecintrin.h (vec_doublee): Fix to use
12275 __builtin_s390_vflls.
12276 (vec_floate): Fix to use __builtin_s390_vflrd.
12278 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12280 * dominance.c (get_dominated_to_depth): Return auto_vec<basic_block>.
12281 * dominance.h (get_dominated_to_depth): Likewise.
12282 (get_all_dominated_blocks): Likewise.
12283 * cfgcleanup.c (delete_unreachable_blocks): Adjust.
12284 * gcse.c (hoist_code): Likewise.
12285 * tree-cfg.c (remove_edge_and_dominated_blocks): Likewise.
12286 * tree-parloops.c (oacc_entry_exit_ok): Likewise.
12287 * tree-ssa-dce.c (eliminate_unnecessary_stmts): Likewise.
12288 * tree-ssa-phiprop.c (pass_phiprop::execute): Likewise.
12290 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12292 * dominance.c (get_dominated_by_region): Return auto_vec<basic_block>.
12293 * dominance.h (get_dominated_by_region): Likewise.
12294 * tree-cfg.c (gimple_duplicate_sese_region): Adjust.
12295 (gimple_duplicate_sese_tail): Likewise.
12296 (move_sese_region_to_fn): Likewise.
12298 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12300 * dominance.c (get_dominated_by): Return auto_vec<basic_block>.
12301 * dominance.h (get_dominated_by): Likewise.
12302 * auto-profile.c (afdo_find_equiv_class): Adjust.
12303 * cfgloopmanip.c (duplicate_loop_to_header_edge): Likewise.
12304 * loop-unroll.c (unroll_loop_runtime_iterations): Likewise.
12305 * tree-cfg.c (test_linear_chain): Likewise.
12306 (test_diamond): Likewise.
12308 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12310 * cfgloop.h (get_loop_hot_path): Return auto_vec<basic_block>.
12311 * cfgloopanal.c (get_loop_hot_path): Likewise.
12312 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Likewise.
12314 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12316 * cgraph.c (cgraph_node::collect_callers): Return
12317 auto_vec<cgraph_edge *>.
12318 * cgraph.h (cgraph_node::collect_callers): Likewise.
12319 * ipa-cp.c (create_specialized_node): Adjust.
12320 (decide_about_value): Likewise.
12321 (decide_whether_version_node): Likewise.
12322 * ipa-sra.c (process_isra_node_results): Likewise.
12324 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12326 * vec.h (vl_ptr>::using_auto_storage): Handle null m_vec.
12327 (auto_vec<T, 0>::auto_vec): Define move constructor, and delete copy
12329 (auto_vec<T, 0>::operator=): Define move assignment and delete copy
12332 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
12334 * gimple-range.cc (debug_seed_ranger): New.
12335 (dump_ranger): New.
12336 (debug_ranger): New.
12338 2021-06-17 Richard Biener <rguenther@suse.de>
12340 PR tree-optimization/54400
12341 * tree-vectorizer.h (enum slp_instance_kind): Add
12342 slp_inst_kind_bb_reduc.
12343 (reduction_fn_for_scalar_code): Declare.
12344 * tree-vect-data-refs.c (vect_slp_analyze_instance_dependence):
12345 Check SLP_INSTANCE_KIND instead of looking at the
12347 (vect_slp_analyze_instance_alignment): Likewise.
12348 * tree-vect-loop.c (reduction_fn_for_scalar_code): Export.
12349 * tree-vect-slp.c (vect_slp_linearize_chain): Split out
12350 chain linearization from vect_build_slp_tree_2 and generalize
12351 for the use of BB reduction vectorization.
12352 (vect_build_slp_tree_2): Adjust accordingly.
12353 (vect_optimize_slp): Elide permutes at the root of BB reduction
12355 (vectorizable_bb_reduc_epilogue): New function.
12356 (vect_slp_prune_covered_roots): Likewise.
12357 (vect_slp_analyze_operations): Use them.
12358 (vect_slp_check_for_constructors): Recognize associatable
12359 chains for BB reduction vectorization.
12360 (vectorize_slp_instance_root_stmt): Generate code for the
12361 BB reduction epilogue.
12363 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
12365 * gimple-range-gori.cc (gori_compute::has_edge_range_p): Check with
12367 (gori_compute::may_recompute_p): New.
12368 (gori_compute::outgoing_edge_range_p): Perform recomputations.
12369 * gimple-range-gori.h (class gori_compute): Add prototype.
12371 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
12373 * gimple-range-cache.cc (ranger_cache::range_on_edge): Always return
12374 true when a range can be calculated.
12375 * gimple-range.cc (gimple_ranger::dump_bb): Check has_edge_range_p.
12377 2021-06-16 Martin Sebor <msebor@redhat.com>
12379 * doc/invoke.texi (-Wmismatched-dealloc, -Wmismatched-new-delete):
12380 Correct documented defaults.
12382 2021-06-16 Andrew MacLeod <amacleod@redhat.com>
12384 * gimple-range-cache.cc (ranger_cache::ranger_cache): Initialize
12385 m_new_value_p directly.
12387 2021-06-16 Uroš Bizjak <ubizjak@gmail.com>
12390 * config/i386/i386-expand.c (expand_vec_perm_2perm_pblendv):
12391 Handle 64bit modes for TARGET_SSE4_1.
12392 (expand_vec_perm_pshufb2): Handle 64bit modes for TARGET_SSSE3.
12393 (expand_vec_perm_even_odd_pack): Handle V4HI mode.
12394 (expand_vec_perm_even_odd_1) <case E_V4HImode>: Expand via
12395 expand_vec_perm_pshufb2 for TARGET_SSSE3 and via
12396 expand_vec_perm_even_odd_pack for TARGET_SSE4_1.
12397 * config/i386/mmx.md (mmx_packusdw): New insn pattern.
12399 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
12401 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn<mode>):
12402 Change to an expander that emits the correct instruction
12403 depending on endianness.
12404 (aarch64_<sur><addsub>hn<mode>_insn_le): Define.
12405 (aarch64_<sur><addsub>hn<mode>_insn_be): Define.
12407 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
12409 * config/aarch64/aarch64-simd-builtins.def: Split generator
12410 for aarch64_<su>qmovn builtins into scalar and vector
12412 * config/aarch64/aarch64-simd.md (aarch64_<su>qmovn<mode>_insn_le):
12414 (aarch64_<su>qmovn<mode>_insn_be): Define.
12415 (aarch64_<su>qmovn<mode>): Split into scalar and vector
12416 variants. Change vector variant to an expander that emits the
12417 correct instruction depending on endianness.
12419 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
12421 * config/aarch64/aarch64-simd-builtins.def: Split generator
12422 for aarch64_sqmovun builtins into scalar and vector variants.
12423 * config/aarch64/aarch64-simd.md (aarch64_sqmovun<mode>):
12424 Split into scalar and vector variants. Change vector variant
12425 to an expander that emits the correct instruction depending
12427 (aarch64_sqmovun<mode>_insn_le): Define.
12428 (aarch64_sqmovun<mode>_insn_be): Define.
12430 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
12432 * config/aarch64/aarch64-simd.md (aarch64_xtn<mode>_insn_le):
12433 Define - modeling zero-high-half semantics.
12434 (aarch64_xtn<mode>): Change to an expander that emits the
12435 appropriate instruction depending on endianness.
12436 (aarch64_xtn<mode>_insn_be): Define - modeling zero-high-half
12438 (aarch64_xtn2<mode>_le): Rename to...
12439 (aarch64_xtn2<mode>_insn_le): This.
12440 (aarch64_xtn2<mode>_be): Rename to...
12441 (aarch64_xtn2<mode>_insn_be): This.
12442 (vec_pack_trunc_<mode>): Emit truncation instruction instead
12444 * config/aarch64/iterators.md (Vnarrowd): Add Vnarrowd mode
12445 attribute iterator.
12447 2021-06-16 Martin Jambor <mjambor@suse.cz>
12449 PR tree-optimization/100453
12450 * tree-sra.c (create_access): Disqualify any const candidates
12451 which are written to.
12452 (sra_modify_expr): Do not store sub-replacements back to a const base.
12453 (handle_unscalarized_data_in_subtree): Likewise.
12454 (sra_modify_assign): Likewise. Earlier, use TREE_READONLy test
12455 instead of constant_decl_p.
12457 2021-06-16 Jakub Jelinek <jakub@redhat.com>
12459 PR middle-end/101062
12460 * stor-layout.c (finish_bitfield_representative): For fields in unions
12461 assume nextf is always NULL.
12462 (finish_bitfield_layout): Compute bit field representatives also in
12463 unions, but handle it as if each bitfield was the only field in the
12466 2021-06-16 Richard Biener <rguenther@suse.de>
12468 PR tree-optimization/101088
12469 * tree-ssa-loop-im.c (sm_seq_valid_bb): Only look for
12470 supported refs on edges. Do not assert same ref but
12471 different kind stores are unsuported but mark them so.
12472 (hoist_memory_references): Only look for supported refs
12475 2021-06-16 Roger Sayle <roger@nextmovesoftware.com>
12477 PR rtl-optimization/46235
12478 * config/i386/i386.md: New define_split for bt followed by cmov.
12479 (*bt<mode>_setcqi): New define_insn_and_split for bt followed by setc.
12480 (*bt<mode>_setncqi): New define_insn_and_split for bt then setnc.
12481 (*bt<mode>_setnc<mode>): New define_insn_and_split for bt followed
12482 by setnc with zero extension.
12484 2021-06-16 Richard Biener <rguenther@suse.de>
12486 PR tree-optimization/101083
12487 * tree-vect-slp.c (vect_slp_build_two_operator_nodes): Get
12488 vectype as argument.
12489 (vect_build_slp_tree_2): Adjust.
12491 2021-06-15 Martin Sebor <msebor@redhat.com>
12493 PR middle-end/100876
12494 * builtins.c: (gimple_call_return_array): Account for size_t
12495 mangling as either unsigned int or unsigned long
12497 2021-06-15 Jeff Law <jeffreyalaw@gmail.com>
12499 * compare-elim.c (try_eliminate_compare): Run DCE to clean things
12500 up before eliminating comparisons.
12502 2021-06-15 Aldy Hernandez <aldyh@redhat.com>
12504 * range-op.cc (operator_bitwise_or::wi_fold): Make sure
12505 nonzero|X is nonzero.
12506 (range_op_bitwise_and_tests): Add tests for above.
12508 2021-06-15 Carl Love <cel@us.ibm.com>
12511 * config/rs6000/rs6000-builtin.def (VCMPEQUT): Fix the ICODE for the
12513 (VRLQ, VSLQ, VSRQ, VSRAQ): Remove unused BU_P10_OVERLOAD_2
12516 2021-06-15 Tobias Burnus <tobias@codesourcery.com>
12519 * gimplify.c (enum gimplify_defaultmap_kind): Add GDMK_SCALAR_TARGET.
12520 (struct gimplify_omp_ctx): Extend defaultmap array by one.
12521 (new_omp_context): Init defaultmap[GDMK_SCALAR_TARGET].
12522 (omp_notice_variable): Update type classification for Fortran.
12523 (gimplify_scan_omp_clauses): Update calls for new argument; handle
12524 GDMK_SCALAR_TARGET; for Fortran, GDMK_POINTER avoid GOVD_MAP_0LEN_ARRAY.
12525 * langhooks-def.h (lhd_omp_scalar_p): Add 'ptr_ok' argument.
12526 * langhooks.c (lhd_omp_scalar_p): Likewise.
12527 (LANG_HOOKS_OMP_ALLOCATABLE_P, LANG_HOOKS_OMP_SCALAR_TARGET_P): New.
12528 (LANG_HOOKS_DECLS): Add them.
12529 * langhooks.h (struct lang_hooks_for_decls): Add new hooks, update
12530 omp_scalar_p pointer type to include the new bool argument.
12532 2021-06-15 David Malcolm <dmalcolm@redhat.com>
12534 * doc/analyzer.texi
12535 (Special Functions for Debugging the Analyzer): Add
12536 __analyzer_dump_capacity.
12538 2021-06-15 Jakub Jelinek <jakub@redhat.com>
12541 * expr.c (expand_expr_real_2) <case VEC_PACK_FIX_TRUNC_EXPR,
12542 case VEC_PACK_TRUNC_EXPR>: Clear subtarget when changing mode.
12544 2021-06-15 Richard Biener <rguenther@suse.de>
12546 * cfgloopanal.c (mark_irreducible_loops): Use a dominance
12547 check to identify loop latches.
12548 * cfgloop.c (verify_loop_structure): Likewise.
12549 * loop-init.c (apply_loop_flags): Allow marked irreducible
12550 regions even with multiple latches.
12551 * predict.c (rebuild_frequencies): Simplify.
12553 2021-06-15 Richard Biener <rguenther@suse.de>
12555 * tree-ssa-threadupdate.c
12556 (jump_thread_path_registry::mark_threaded_blocks): Assert we
12557 have marked irreducible regions.
12559 2021-06-14 Martin Sebor <msebor@redhat.com>
12562 * builtins.c (gimple_call_return_array): Check for attribute fn spec.
12563 Handle calls to placement new.
12564 (ndecl_dealloc_argno): Avoid placement delete.
12566 2021-06-14 Peter Bergner <bergner@linux.ibm.com>
12569 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Use
12570 create_tmp_reg_or_ssa_name().
12572 2021-06-14 Andrew MacLeod <amacleod@redhat.com>
12574 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
12575 (ranger_cache::enable_new_values): Set to specified value and
12576 return the old value.
12577 (ranger_cache::disable_new_values): Delete.
12578 (ranger_cache::fill_block_cache): Disable non 1st order derived
12580 * gimple-range-cache.h (ranger_cache): Adjust prototypes.
12581 * gimple-range.cc (gimple_ranger::range_of_expr): Adjust.
12583 2021-06-14 Uroš Bizjak <ubizjak@gmail.com>
12586 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const):
12587 Return true early when testing with V2HImode.
12588 * config/i386/mmx.md (*punpckwd): Split to sse2_pshuflw_1.
12590 2021-06-14 Christophe Lyon <christophe.lyon@linaro.org>
12592 * config/arm/mve.md (mve_vec_unpack<US>_lo_<mode>): New pattern.
12593 (mve_vec_unpack<US>_hi_<mode>): New pattern.
12594 (@mve_vec_pack_trunc_lo_<mode>): New pattern.
12595 (mve_vmovntq_<supf><mode>): Prefix with '@'.
12596 * config/arm/neon.md (vec_unpack<US>_hi_<mode>): Move to
12598 (vec_unpack<US>_lo_<mode>): Likewise.
12599 (vec_pack_trunc_<mode>): Rename to
12600 neon_quad_vec_pack_trunc_<mode>.
12601 * config/arm/vec-common.md (vec_unpack<US>_hi_<mode>): New
12603 (vec_unpack<US>_lo_<mode>): New.
12604 (vec_pack_trunc_<mode>): New.
12606 2021-06-14 Richard Biener <rguenther@suse.de>
12608 PR tree-optimization/100934
12609 * tree-ssa-dom.c (pass_dominator::execute): Properly
12610 mark irreducible regions.
12612 2021-06-14 Martin Liska <mliska@suse.cz>
12614 * doc/invoke.texi: Put r{...} on the same line as @item.
12616 2021-06-14 Martin Liska <mliska@suse.cz>
12618 * doc/invoke.texi: Add missing newline.
12620 2021-06-14 Martin Liska <mliska@suse.cz>
12622 * doc/invoke.texi: Remove '+' charasters.
12624 2021-06-14 Claudiu Zissulescu <claziss@synopsys.com>
12626 * config.gcc (arc): Add support for with_cpu option.
12627 * config/arc/arc.h (OPTION_DEFAULT_SPECS): Add fpu.
12629 2021-06-14 Richard Biener <rguenther@suse.de>
12631 PR tree-optimization/101031
12632 * tree-ssa-strlen.c (maybe_invalidate): Increment max_size
12633 instead of size when accounting for a possibly string
12636 2021-06-14 Martin Liska <mliska@suse.cz>
12638 * gimple-ssa-evrp.c (pointer_equiv_analyzer::~pointer_equiv_analyzer): Use delete[].
12640 2021-06-14 Aldy Hernandez <aldyh@redhat.com>
12642 * value-query.cc (gimple_range_global): Call get_range_global
12643 if called after inlining.
12645 2021-06-13 Uroš Bizjak <ubizjak@gmail.com>
12648 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
12649 Emit constant permutation insn directly from here.
12651 2021-06-13 Trevor Saunders <tbsaunde@tbsaunde.org>
12653 * attribs.c (find_attribute_namespace): Iterate over vec<> with
12655 * auto-profile.c (afdo_find_equiv_class): Likewise.
12656 * gcc.c (do_specs_vec): Likewise.
12657 (do_spec_1): Likewise.
12658 (driver::set_up_specs): Likewise.
12659 * gimple-loop-jam.c (any_access_function_variant_p): Likewise.
12660 * gimple-ssa-store-merging.c (compatible_load_p): Likewise.
12661 (imm_store_chain_info::try_coalesce_bswap): Likewise.
12662 (imm_store_chain_info::coalesce_immediate_stores): Likewise.
12663 (get_location_for_stmts): Likewise.
12664 * graphite-poly.c (print_iteration_domains): Likewise.
12665 (free_poly_bb): Likewise.
12666 (remove_gbbs_in_scop): Likewise.
12667 (free_scop): Likewise.
12668 (dump_gbb_cases): Likewise.
12669 (dump_gbb_conditions): Likewise.
12670 (print_pdrs): Likewise.
12671 (print_scop): Likewise.
12672 * ifcvt.c (cond_move_process_if_block): Likewise.
12673 * lower-subreg.c (decompose_multiword_subregs): Likewise.
12674 * regcprop.c (pass_cprop_hardreg::execute): Likewise.
12675 * sanopt.c (sanitize_rewrite_addressable_params): Likewise.
12676 * sel-sched-dump.c (dump_insn_vector): Likewise.
12677 * store-motion.c (store_ops_ok): Likewise.
12678 (store_killed_in_insn): Likewise.
12679 * timevar.c (timer::named_items::print): Likewise.
12680 * tree-cfgcleanup.c (cleanup_control_flow_pre): Likewise.
12681 (cleanup_tree_cfg_noloop): Likewise.
12682 * tree-data-ref.c (dump_data_references): Likewise.
12683 (print_dir_vectors): Likewise.
12684 (print_dist_vectors): Likewise.
12685 (dump_data_dependence_relations): Likewise.
12686 (dump_dist_dir_vectors): Likewise.
12687 (dump_ddrs): Likewise.
12688 (create_runtime_alias_checks): Likewise.
12689 (free_subscripts): Likewise.
12690 (save_dist_v): Likewise.
12691 (save_dir_v): Likewise.
12692 (invariant_access_functions): Likewise.
12693 (same_access_functions): Likewise.
12694 (access_functions_are_affine_or_constant_p): Likewise.
12695 (find_data_references_in_stmt): Likewise.
12696 (graphite_find_data_references_in_stmt): Likewise.
12697 (free_dependence_relations): Likewise.
12698 (free_data_refs): Likewise.
12699 * tree-inline.c (copy_debug_stmts): Likewise.
12700 * tree-into-ssa.c (dump_currdefs): Likewise.
12701 (rewrite_update_phi_arguments): Likewise.
12702 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
12703 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
12705 (vect_slp_analyze_node_dependences): Likewise.
12706 (vect_slp_analyze_instance_dependence): Likewise.
12707 (vect_record_base_alignments): Likewise.
12708 (vect_get_peeling_costs_all_drs): Likewise.
12709 (vect_peeling_supportable): Likewise.
12710 * tree-vectorizer.c (vec_info::~vec_info): Likewise.
12711 (vec_info::free_stmt_vec_infos): Likewise.
12713 2021-06-13 Jeff Law <jeffreyalaw@gmail.com>
12715 * config/h8300/logical.md (<code>qi3_1<cczn>): New pattern.
12716 (andqi3_1<cczn>): Removed.
12717 (<ors>qi3_1): Do not split for IOR/XOR a single bit.
12718 (H8/SX bit logicals): Split out from other patterns.
12719 * config/h8300/multiply.md (mulqihi3_const<cczn>): Renamed from
12720 mulqihi3_const_clobber_flags.
12721 (mulqihi3<cczn>, mulhisi3_const<cczn>, mulhisi3<cczn>): Similarly
12723 2021-06-13 H.J. Lu <hjl.tools@gmail.com>
12726 * config/i386/i386.c (ix86_expand_prologue): Set red_zone_used
12727 to true if red zone is used.
12728 (ix86_output_indirect_jmp): Replace ix86_red_zone_size with
12729 ix86_red_zone_used.
12730 * config/i386/i386.h (machine_function): Add red_zone_used.
12731 (ix86_red_zone_size): Removed.
12732 (ix86_red_zone_used): New.
12733 * config/i386/i386.md (peephole2 patterns): Replace
12734 ix86_red_zone_size with ix86_red_zone_used.
12736 2021-06-12 Jason Merrill <jason@redhat.com>
12738 * doc/extend.texi (unused variable attribute): Applies to
12739 structure fields as well.
12741 2021-06-12 Eugene Rozenfeld <erozen@microsoft.com>
12743 * auto-profile.c (read_profile): fix a typo in an error string
12745 2021-06-11 Thomas Schwinge <thomas@codesourcery.com>
12747 * tree-pretty-print.h (dump_omp_clauses): Add 'bool = true'
12749 * tree-pretty-print.c (dump_omp_clauses): Update.
12750 (dump_generic_node) <OMP_CLAUSE>: Use it.
12752 2021-06-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
12755 * config/arm/arm_mve.h (__arm_vld1q): Change __ARM_mve_coerce(p0,
12756 int8_t const *) to __ARM_mve_coerce1(p0, int8_t *) in the argument for
12757 the polymorphic variants matching code.
12758 (__arm_vld1q_z): Likewise.
12759 (__arm_vld2q): Likewise.
12760 (__arm_vld4q): Likewise.
12761 (__arm_vldrbq_gather_offset): Likewise.
12762 (__arm_vldrbq_gather_offset_z): Likewise.
12764 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
12766 PR tree-optimization/96392
12767 * fold-const.h (tree_expr_maybe_real_minus_zero_p): Fix prototype.
12769 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
12771 PR tree-optimization/96392
12772 * fold-const.c (fold_real_zero_addition_p): Take both arguments
12773 of the addition or subtraction, not just the zero. Use this
12774 other argument in tests for signaling NaNs and signed zeros.
12775 (tree_expr_maybe_real_minus_zero_p): New predicate.
12776 * fold-const.h (fold_real_zero_addition_p): Update prototype.
12777 (tree_expr_maybe_real_minus_zero_p): New function prototype.
12778 * match.pd: Update calls to fold_real_zero_addition_p.
12779 Replace HONOR_NANS with tree_expr_maybe_nan_p.
12780 Replace HONOR_SIGNED_ZEROS with tree_expr_maybe_real_minus_zero_p.
12781 Replace HONOR_SNANS with tree_expr_maybe_signaling_nan_p.
12782 * tree-ssa-reassoc.c (eliminate_using_constants): Update
12783 call to fold_real_zero_addition_p.
12785 2021-06-11 Richard Biener <rguenther@suse.de>
12787 PR tree-optimization/101025
12788 * tree-ssa-loop-im.c (sm_seq_valid_bb): Make sure to process
12789 all refs that require dependence checking.
12791 2021-06-11 Richard Biener <rguenther@suse.de>
12793 PR tree-optimization/101028
12794 * tree-vect-slp.c (vect_build_slp_tree_2): When SLP
12795 reassoc discovery fails fatally, mark appropriate lanes
12798 2021-06-11 Richard Biener <rguenther@suse.de>
12800 PR tree-optimization/101026
12801 * tree-vect-slp.c (vect_build_slp_tree_2): Make sure we
12802 have a representative for the associated chain nodes.
12804 2021-06-11 Jakub Jelinek <jakub@redhat.com>
12806 PR rtl-optimization/101008
12807 * simplify-rtx.c (relational_result): New function.
12808 (simplify_logical_relational_operation,
12809 simplify_relational_operation): Use it.
12811 2021-06-11 Jakub Jelinek <jakub@redhat.com>
12814 * config/i386/sse.md (*vec_concat<mode>_0_1): Require TARGET_SSE2.
12816 2021-06-11 Uroš Bizjak <ubizjak@gmail.com>
12819 * config/i386/i386-expand.c (expand_vec_perm_pshufb): Return
12820 false if the permutation can be implemented with constant
12821 permutation instruction in wider mode.
12822 (canonicalize_vector_int_perm): Move above expand_vec_perm_pshufb.
12823 Handle V8QImode and V4HImode.
12825 2021-06-11 Martin Liska <mliska@suse.cz>
12827 PR gcov-profile/100788
12828 * common.opt: Add new option.
12829 * coverage.c (coverage_begin_function): Emit warning instead on
12830 the internal compiler error.
12831 * doc/invoke.texi: Document the option.
12832 * toplev.c (process_options): Enable it by default.
12834 2021-06-11 Richard Biener <rguenther@suse.de>
12836 PR middle-end/101009
12837 * tree-data-ref.c (build_classic_dist_vector_1): Make sure
12838 to set *init_b to true when we encounter a constant equal
12840 (compute_affine_dependence): Also dump the actual DR_REF.
12842 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
12844 PR tree-optimization/100984
12845 * gimple-ssa-evrp.c (ssa_equiv_stack): Use auto_vec for
12846 replacements table.
12847 (ssa_equiv_stack::~ssa_equiv_stack): Remove.
12849 2021-06-11 Kewen Lin <linkw@linux.ibm.com>
12851 * config/rs6000/rs6000.md
12852 (floatsi<SFDF:mode>2_lfiwax_<QHI:mode>_mem_zext): New
12853 define_insn_and_split.
12855 2021-06-11 Richard Biener <rguenther@suse.de>
12857 * tree-vect-slp.c (vect_build_slp_tree_2): Use stablesort
12858 to sort operands of the associative chain.
12860 2021-06-11 Richard Biener <rguenther@suse.de>
12862 * system.h (gcc_stablesort_r): Declare.
12863 * sort.cc (gcc_sort_r): Support stable sort.
12864 (gcc_stablesort_r): Define.
12865 * vec.h (vec<>::stablesort): Add.
12867 2021-06-10 Uroš Bizjak <ubizjak@gmail.com>
12870 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
12871 Handle V2SF mode. Emit SHUFPS to fixup unpack-high for V2SF mode.
12872 (expand_vec_perm_blend): Handle 64bit modes for TARGET_SSE4_1.
12873 (expand_vec_perm_pshufb): Handle 64bit modes for TARGET_SSSE3.
12874 (expand_vec_perm_pblendv): Handle 64bit modes for TARGET_SSE4_1.
12875 (expand_vec_perm_interleave2): Handle 64bit modes.
12876 (expand_vec_perm_even_odd_pack): Handle V8QI mode.
12877 (expand_vec_perm_even_odd_1): Ditto.
12878 (ix86_vectorize_vec_perm_const): Ditto.
12879 * config/i386/i386.md (UNSPEC_PSHUFB): Move from ...
12880 * config/i386/sse.md: ... here.
12881 * config/i386/mmx.md (*vec_interleave_lowv2sf):
12882 New insn_and_split pattern.
12883 (*vec_interleave_highv2sf): Ditto.
12884 (mmx_pshufbv8qi3): New insn pattern.
12885 (*mmx_pblendw): Ditto.
12887 2021-06-10 Peter Bergner <bergner@linux.ibm.com>
12889 * config/rs6000/rs6000-builtin.def (build_pair): New built-in.
12890 (build_acc): Likewise.
12891 * config/rs6000/rs6000-call.c (mma_expand_builtin): Swap assemble
12892 source operands in little-endian mode.
12893 (rs6000_gimple_fold_mma_builtin): Handle VSX_BUILTIN_BUILD_PAIR.
12894 (mma_init_builtins): Likewise.
12895 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Handle endianness
12896 ordering for the MMA assemble and build source operands.
12897 * doc/extend.texi (__builtin_vsx_build_acc, __builtin_mma_build_pair):
12899 (__builtin_mma_assemble_acc, __builtin_mma_assemble_pair): Remove
12902 2021-06-10 Jeff Law <jeffreyalaw@gmail.com>
12904 * config/h8300/h8300.c (select_cc_mode): Handle MEM. Use
12906 * config/h8300/extensions.md: Replace _clobber_flags patterns
12909 2021-06-10 Robin Dapp <rdapp@linux.ibm.com>
12911 * config/s390/vector.md (vcond_mask_<mode><mode>): Change to
12912 (vcond_mask_<mode><tointvec>): this.
12914 2021-06-10 Andrew Stubbs <ams@codesourcery.com>
12915 Thomas Schwinge <thomas@codesourcery.com>
12917 * omp-builtins.def (BUILT_IN_GOACC_ENTER_EXIT_DATA): Split into...
12918 (BUILT_IN_GOACC_ENTER_DATA, BUILT_IN_GOACC_EXIT_DATA): ... these.
12919 * gimple.h (enum gf_mask): Split
12920 'GF_OMP_TARGET_KIND_OACC_ENTER_EXIT_DATA' into
12921 'GF_OMP_TARGET_KIND_OACC_ENTER_DATA' and
12922 'GF_OMP_TARGET_KIND_OACC_EXIT_DATA'.
12923 (is_gimple_omp_oacc): Update.
12924 * gimple-pretty-print.c (dump_gimple_omp_target): Likewise.
12925 * gimplify.c (gimplify_omp_target_update): Likewise.
12926 * omp-expand.c (expand_omp_target, build_omp_regions_1)
12927 (omp_make_gimple_edges): Likewise.
12928 * omp-low.c (check_omp_nesting_restrictions, lower_omp_target):
12931 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
12933 * value-query.cc (value_query::value_on_edge): Rename name to
12935 (range_query::range_on_edge): Same.
12936 (range_query::value_of_expr): Same.
12937 (range_query::value_on_edge): Same.
12938 * value-query.h (class value_query): Same.
12939 (class range_query): Same.
12941 2021-06-10 Richard Biener <rguenther@suse.de>
12943 PR tree-optimization/101003
12944 * tree-vect-slp.c (vect_build_slp_tree_2): Appropriately
12945 use the pattern stmt defs when linearizing a chain.
12947 2021-06-10 Jakub Jelinek <jakub@redhat.com>
12950 * ifcvt.c (noce_get_alt_condition, noce_try_abs): Use
12951 prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
12953 2021-06-10 Clement Chigot <clement.chigot@atos.net>
12955 * config/rs6000/aix71.h (ASM_CPU_SPEC): Add Power10 directive.
12956 * config/rs6000/aix72.h (ASM_CPU_SPEC): Likewise.
12958 2021-06-09 Andrew Pinski <apinski@marvell.com>
12960 PR tree-optimization/100925
12961 * match.pd (a ? CST1 : CST2): Limit transformations
12962 that would produce a negative to integeral types only.
12963 Change !POINTER_TYPE_P to INTEGRAL_TYPE_P also.
12965 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
12968 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
12970 * doc/tm.texi: Correctly update.
12972 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
12974 * doc/tm.texi: Correctly update.
12976 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
12979 * doc/tm.texi.in (Trampolines): Add a missing blank line.
12981 2021-06-09 Paul Eggert <eggert@cs.ucla.edu>
12984 * doc/invoke.texi (Code Gen Options); Document that -fno-trampolines
12985 and -ftrampolines work only with Ada.
12986 * doc/tm.texi.in (Trampolines): Likewise.
12987 * doc/tm.texi: Regenerated.
12989 2021-06-09 Carl Love <cel@us.ibm.com>
12991 * config/rs6000/altivec.h (vec_signextll, vec_signexti, vec_signextq):
12992 Add define for new builtins.
12993 * config/rs6000/altivec.md(altivec_vreveti2): Add define_expand.
12994 * config/rs6000/rs6000-builtin.def (VSIGNEXTI, VSIGNEXTLL): Add
12995 overloaded builtin definitions.
12996 (VSIGNEXTSB2W, VSIGNEXTSH2W, VSIGNEXTSB2D, VSIGNEXTSH2D,VSIGNEXTSW2D,
12997 VSIGNEXTSD2Q): Add builtin expansions.
12998 (SIGNEXT): Add P10 overload definition.
12999 * config/rs6000/rs6000-call.c (P9V_BUILTIN_VEC_VSIGNEXTI, P9V_BUILTIN_VEC_VSIGNEXTLL,
13000 P10_BUILTIN_VEC_SIGNEXT): Add overloaded argument definitions.
13001 * config/rs6000/vsx.md (vsx_sign_extend_v2di_v1ti): Add define_insn.
13002 (vsignextend_v2di_v1ti, vsignextend_qi_<mode>, vsignextend_hi_<mode>,
13003 vsignextend_si_v2di)[VIlong]: Add define_expand.
13004 Make define_insn vsx_sign_extend_si_v2di visible.
13005 * doc/extend.texi: Add documentation for the vec_signexti,
13006 vec_signextll builtins and vec_signextq.
13008 2021-06-09 Carl Love <cel@us.ibm.com>
13010 * config/rs6000/rs6000.c (__fixkfti, __fixunskfti, __floattikf,
13011 __floatuntikf): Names changed to __fixkfti_sw, __fixunskfti_sw,
13012 __floattikf_sw, __floatuntikf_sw respectively.
13013 * config/rs6000/rs6000.md (floatti<mode>2, floatunsti<mode>2,
13014 fix_trunc<mode>ti2, fixuns_trunc<mode>ti2): Add
13015 define_insn for mode IEEE 128.
13017 2021-06-09 Carl Love <cel@us.ibm.com>
13019 * config/rs6000/altivec.md (altivec_vslq, altivec_vsrq):
13020 Rename to altivec_vslq_<mode>, altivec_vsrq_<mode>, mode VEC_TI.
13021 * config/rs6000/vector.md (VEC_TI): Was named VSX_TI in vsx.md.
13022 (vashlv1ti3): Change to vashl<mode>3, mode VEC_TI.
13023 (vlshrv1ti3): Change to vlshr<mode>3, mode VEC_TI.
13024 * config/rs6000/vsx.md (VSX_TI): Remove define_mode_iterator. Update
13025 uses of VSX_TI to VEC_TI.
13027 2021-06-09 Carl Love <cel@us.ibm.com>
13029 * config/rs6000/dfp.md (floattitd2, fixtdti2): New define_insns.
13031 2021-06-09 Carl Love <cel@us.ibm.com>
13033 * config/rs6000/altivec.h (vec_dive, vec_mod): Add define for new
13035 * config/rs6000/altivec.md (UNSPEC_VMULEUD, UNSPEC_VMULESD,
13036 UNSPEC_VMULOUD, UNSPEC_VMULOSD): New unspecs.
13037 (altivec_eqv1ti, altivec_gtv1ti, altivec_gtuv1ti, altivec_vmuleud,
13038 altivec_vmuloud, altivec_vmulesd, altivec_vmulosd, altivec_vrlq,
13039 altivec_vrlqmi, altivec_vrlqmi_inst, altivec_vrlqnm,
13040 altivec_vrlqnm_inst, altivec_vslq, altivec_vsrq, altivec_vsraq,
13041 altivec_vcmpequt_p, altivec_vcmpgtst_p, altivec_vcmpgtut_p): New
13043 (vec_widen_umult_even_v2di, vec_widen_smult_even_v2di,
13044 vec_widen_umult_odd_v2di, vec_widen_smult_odd_v2di, altivec_vrlqmi,
13045 altivec_vrlqnm): New define_expands.
13046 * config/rs6000/rs6000-builtin.def (VCMPEQUT_P, VCMPGTST_P,
13047 VCMPGTUT_P): Add macro expansions.
13048 (BU_P10V_AV_P): Add builtin predicate definition.
13049 (VCMPGTUT, VCMPGTST, VCMPEQUT, CMPNET, CMPGE_1TI,
13050 CMPGE_U1TI, CMPLE_1TI, CMPLE_U1TI, VNOR_V1TI_UNS, VNOR_V1TI, VCMPNET_P,
13051 VCMPAET_P, VMULEUD, VMULESD, VMULOUD, VMULOSD, VRLQ,
13052 VSLQ, VSRQ, VSRAQ, VRLQNM, DIV_V1TI, UDIV_V1TI, DIVES_V1TI, DIVEU_V1TI,
13053 MODS_V1TI, MODU_V1TI, VRLQMI): New macro expansions.
13054 (VRLQ, VSLQ, VSRQ, VSRAQ, DIVE, MOD): New overload expansions.
13055 * config/rs6000/rs6000-call.c (P10_BUILTIN_VCMPEQUT,
13056 P10V_BUILTIN_CMPGE_1TI, P10V_BUILTIN_CMPGE_U1TI,
13057 P10V_BUILTIN_VCMPGTUT, P10V_BUILTIN_VCMPGTST,
13058 P10V_BUILTIN_CMPLE_1TI, P10V_BUILTIN_VCMPLE_U1TI,
13059 P10V_BUILTIN_DIV_V1TI, P10V_BUILTIN_UDIV_V1TI,
13060 P10V_BUILTIN_VMULESD, P10V_BUILTIN_VMULEUD,
13061 P10V_BUILTIN_VMULOSD, P10V_BUILTIN_VMULOUD,
13062 P10V_BUILTIN_VNOR_V1TI, P10V_BUILTIN_VNOR_V1TI_UNS,
13063 P10V_BUILTIN_VRLQ, P10V_BUILTIN_VRLQMI,
13064 P10V_BUILTIN_VRLQNM, P10V_BUILTIN_VSLQ,
13065 P10V_BUILTIN_VSRQ, P10V_BUILTIN_VSRAQ,
13066 P10V_BUILTIN_VCMPGTUT_P, P10V_BUILTIN_VCMPGTST_P,
13067 P10V_BUILTIN_VCMPEQUT_P, P10V_BUILTIN_VCMPGTUT_P,
13068 P10V_BUILTIN_VCMPGTST_P, P10V_BUILTIN_CMPNET,
13069 P10V_BUILTIN_VCMPNET_P, P10V_BUILTIN_VCMPAET_P,
13070 P10V_BUILTIN_DIVES_V1TI, P10V_BUILTIN_MODS_V1TI,
13071 P10V_BUILTIN_MODU_V1TI):
13072 New overloaded definitions.
13073 (rs6000_gimple_fold_builtin) [P10V_BUILTIN_VCMPEQUT,
13074 P10V_BUILTIN_CMPNET, P10V_BUILTIN_CMPGE_1TI,
13075 P10V_BUILTIN_CMPGE_U1TI, P10V_BUILTIN_VCMPGTUT,
13076 P10V_BUILTIN_VCMPGTST, P10V_BUILTIN_CMPLE_1TI,
13077 P10V_BUILTIN_CMPLE_U1TI]: New case statements.
13078 (rs6000_init_builtins) [bool_V1TI_type_node, int_ftype_int_v1ti_v1ti]:
13080 (altivec_init_builtins): New E_V1TImode case statement.
13081 (builtin_function_type)[P10_BUILTIN_128BIT_VMULEUD,
13082 P10_BUILTIN_128BIT_VMULOUD, P10_BUILTIN_128BIT_DIVEU_V1TI,
13083 P10_BUILTIN_128BIT_MODU_V1TI, P10_BUILTIN_CMPGE_U1TI,
13084 P10_BUILTIN_VCMPGTUT, P10_BUILTIN_VCMPEQUT]: New case statements.
13085 * config/rs6000/rs6000.c (rs6000_handle_altivec_attribute) [E_TImode,
13086 E_V1TImode]: New case statements.
13087 * config/rs6000/rs6000.h (rs6000_builtin_type_index): New enum
13088 value RS6000_BTI_bool_V1TI.
13089 * config/rs6000/vector.md (vector_gtv1ti,vector_nltv1ti,
13090 vector_gtuv1ti, vector_nltuv1ti, vector_ngtv1ti, vector_ngtuv1ti,
13091 vector_eq_v1ti_p, vector_ne_v1ti_p, vector_ae_v1ti_p,
13092 vector_gt_v1ti_p, vector_gtu_v1ti_p, vrotlv1ti3, vashlv1ti3,
13093 vlshrv1ti3, vashrv1ti3): New define_expands.
13094 * config/rs6000/vsx.md (UNSPEC_VSX_DIVSQ, UNSPEC_VSX_DIVUQ,
13095 UNSPEC_VSX_DIVESQ, UNSPEC_VSX_DIVEUQ, UNSPEC_VSX_MODSQ,
13096 UNSPEC_VSX_MODUQ): New unspecs.
13097 (mulv2di3, vsx_div_v1ti, vsx_udiv_v1ti, vsx_dives_v1ti,
13098 vsx_diveu_v1ti, vsx_mods_v1ti, vsx_modu_v1ti, xxswapd_v1ti): New
13100 (vcmpnet): New define_expand.
13101 * doc/extend.texi: Add documentation for the new builtins vec_rl,
13102 vec_rlmi, vec_rlnm, vec_sl, vec_sr, vec_sra, vec_mule, vec_mulo,
13103 vec_div, vec_dive, vec_mod, vec_cmpeq, vec_cmpne, vec_cmpgt, vec_cmplt,
13104 vec_cmpge, vec_cmple, vec_all_eq, vec_all_ne, vec_all_gt, vec_all_lt,
13105 vec_all_ge, vec_all_le, vec_any_eq, vec_any_ne, vec_any_gt, vec_any_lt,
13106 vec_any_ge, vec_any_le.
13108 2021-06-09 Carl Love <cel@us.ibm.com>
13110 * config/rs6000/altivec.md (altivec_vrl<VI_char>mi): Fix
13111 bug in argument generation.
13113 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
13115 * config/arm/iterators.md (<supf>): Remove VCLZQ_U, VCLZQ_S.
13117 * config/arm/mve.md (mve_vclzq_<supf><mode>): Add '@' prefix,
13118 remove <supf> iterator.
13119 (mve_vclzq_u<mode>): New.
13120 * config/arm/neon.md (clz<mode>2): Rename to neon_vclz<mode>.
13121 (neon_vclz<mode): Move to ...
13122 * config/arm/unspecs.md (VCLZQ_U, VCLZQ_S): Remove.
13123 * config/arm/vec-common.md: ... here. Add support for MVE.
13125 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
13127 * config/arm/mve.md (mve_vhaddq_<supf><mode>): Prefix with '@'.
13128 (@mve_vrhaddq_<supf><mode): Likewise.
13129 * config/arm/neon.md (neon_v<r>hadd<sup><mode>): Likewise.
13130 * config/arm/vec-common.md (avg<mode>3_floor, uavg<mode>3_floor)
13131 (avg<mode>3_ceil", uavg<mode>3_ceil): New patterns.
13133 2021-06-09 imba-tjd <109224573@qq.com>
13135 * doc/invoke.texi: Fix typo.
13137 2021-06-09 Roger Sayle <roger@nextmovesoftware.com>
13139 PR middle-end/53267
13140 * fold-const-call.c (fold_const_call_sss) [CASE_CFN_FMOD]:
13141 Support evaluation of fmod/fmodf/fmodl at compile-time.
13143 2021-06-09 Richard Biener <rguenther@suse.de>
13145 PR tree-optimization/100981
13146 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
13147 gimple_get_lhs to also handle calls.
13148 * tree-vect-slp-patterns.c (complex_pattern::build): Transfer
13151 2021-06-09 Richard Biener <rguenther@suse.de>
13153 PR tree-optimization/97832
13154 * tree-vectorizer.h (_slp_tree::failed): New.
13155 * tree-vect-slp.c (_slp_tree::_slp_tree): Initialize
13157 (_slp_tree::~_slp_tree): Free failed.
13158 (vect_build_slp_tree): Retain failed nodes and record
13159 matches in them, copying that back out when running
13160 into a cached fail. Dump start and end of discovery.
13161 (dt_sort_cmp): New.
13162 (vect_build_slp_tree_2): Handle associatable chains
13163 together doing more aggressive operand swapping.
13165 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
13168 * config.gcc (gcc_cv_initfini_array): Set to yes for Linux and
13170 * doc/install.texi: Require glibc 2.1 and binutils 2.12 for
13171 Linux and GNU targets.
13173 2021-06-09 Richard Biener <rguenther@suse.de>
13175 * tree-vect-stmts.c (vect_is_simple_use): Always get dt
13178 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
13180 * config/arc/arc.md (loop_end): Change it to
13181 define_insn_and_split.
13183 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
13185 * config/arc/arc.md (maddhisi4): Use VMAC2H instruction.
13186 (machi): New pattern.
13187 (umaddhisi4): Use VMAC2HU instruction.
13188 (umachi): New pattern.
13190 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
13192 * config/arc/arc-protos.h (arc_split_move_p): New prototype.
13193 * config/arc/arc.c (arc_split_move_p): New function.
13194 (arc_split_move): Clean up.
13195 * config/arc/arc.md (movdi_insn): Clean up, use arc_split_move_p.
13196 (movdf_insn): Likewise.
13197 * config/arc/simdext.md (mov<VWH>_insn): Likewise.
13199 2021-06-09 Uroš Bizjak <ubizjak@gmail.com>
13202 * config/i386/i386.c (print_operand_address_as): Rename "no_rip"
13203 argument to "raw". Do not emit segment overrides when "raw" is true.
13205 2021-06-09 Martin Liska <mliska@suse.cz>
13207 * doc/gcov.texi: Create a proper JSON files.
13208 * doc/invoke.texi: Remove dots in order to make it a valid
13211 2021-06-09 Xionghu Luo <luoxhu@linux.ibm.com>
13213 * config/rs6000/rs6000-p8swap.c (pattern_is_rotate64): New.
13214 (insn_is_load_p): Use pattern_is_rotate64.
13215 (insn_is_swap_p): Likewise.
13216 (quad_aligned_load_p): Likewise.
13217 (const_load_sequence_p): Likewise.
13218 (replace_swapped_aligned_load): Likewise.
13219 (recombine_lvx_pattern): Likewise.
13220 (recombine_stvx_pattern): Likewise.
13222 2021-06-09 Andrew MacLeod <amacleod@redhat.com>
13224 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Use a
13225 fur_stmt source record.
13226 * gimple-range.cc (fur_source::get_operand): Generic range query.
13227 (fur_source::get_phi_operand): New.
13228 (fur_source::register_dependency): New.
13229 (fur_source::query): New.
13230 (class fur_edge): New. Edge source for operands.
13231 (fur_edge::fur_edge): New.
13232 (fur_edge::get_operand): New.
13233 (fur_edge::get_phi_operand): New.
13234 (fur_edge::query): New.
13235 (fur_stmt::fur_stmt): New.
13236 (fur_stmt::get_operand): New.
13237 (fur_stmt::get_phi_operand): New.
13238 (fur_stmt::query): New.
13239 (class fur_depend): New. Statement source and process dependencies.
13240 (fur_depend::fur_depend): New.
13241 (fur_depend::register_dependency): New.
13242 (class fur_list): New. List source for operands.
13243 (fur_list::fur_list): New.
13244 (fur_list::get_operand): New.
13245 (fur_list::get_phi_operand): New.
13246 (fold_range): New. Instantiate appropriate fur_source class and fold.
13247 (fold_using_range::range_of_range_op): Use new API.
13248 (fold_using_range::range_of_address): Ditto.
13249 (fold_using_range::range_of_phi): Ditto.
13250 (imple_ranger::fold_range_internal): Use fur_depend class.
13251 (fold_using_range::range_of_ssa_name_with_loop_info): Use new API.
13252 * gimple-range.h (class fur_source): Now a base class.
13253 (class fur_stmt): New.
13254 (fold_range): New prototypes.
13255 (fur_source::fur_source): Delete.
13257 2021-06-08 Andrew Pinski <apinski@marvell.com>
13259 PR tree-optimization/25290
13260 * tree-ssa-phiopt.c (xor_replacement): Delete.
13261 (tree_ssa_phiopt_worker): Delete use of xor_replacement.
13262 (match_simplify_replacement): Allow one cheap preparation
13263 statement that can be moved to before the if.
13265 2021-06-08 Pat Haugen <pthaugen@linux.ibm.com>
13267 * config/rs6000/power10.md (power10-fused-load, power10-fused-store,
13268 power10-fused_alu, power10-fused-vec, power10-fused-branch): New.
13270 2021-06-08 Jeff Law <jeffreyalaw@gmail.com>
13272 * config/h8300/logical.md (andqi3_1): Move BCLR case into define_insn_and_split.
13273 Create length attribute on define_insn_and_split. Only split for cases which we
13275 (andqi3_1<cczn>): Renamed from andqi3_1_clobber_flags. Only handle AND here and
13276 fix length computation.
13277 (b<code><mode>msx): Combine QImode and HImode H8/SX patterns using iterator.
13279 2021-06-08 Richard Biener <rguenther@suse.de>
13281 PR tree-optimization/100923
13282 * tree-ssa-sccvn.c (valueize_refs_1): Take a pointer to
13283 the operand vector to be valueized.
13284 (valueize_refs): Likewise.
13285 (valueize_shared_reference_ops_from_ref): Adjust.
13286 (valueize_shared_reference_ops_from_call): Likewise.
13287 (vn_reference_lookup_3): Likewise.
13288 (vn_reference_lookup_pieces): Likewise. Re-valueize
13289 with honoring availability when we are about to create
13290 the ao_ref and valueized before.
13291 (vn_reference_lookup): Likewise.
13292 (vn_reference_insert_pieces): Adjust.
13294 2021-06-08 Richard Biener <rguenther@suse.de>
13296 * tree-vectorizer.h (_slp_instance::root_stmt): Change to...
13297 (_slp_instance::root_stmts): ... a vector.
13298 (SLP_INSTANCE_ROOT_STMT): Rename to ...
13299 (SLP_INSTANCE_ROOT_STMTS): ... this.
13300 (slp_root::root): Change to...
13301 (slp_root::roots): ... a vector.
13302 (slp_root::slp_root): Adjust.
13303 * tree-vect-slp.c (_slp_instance::location): Adjust.
13304 (vect_free_slp_instance): Release the root stmt vector.
13305 (vect_build_slp_instance): Adjust.
13306 (vect_analyze_slp): Likewise.
13307 (_bb_vec_info::~_bb_vec_info): Likewise.
13308 (vect_slp_analyze_operations): Likewise.
13309 (vect_bb_vectorization_profitable_p): Likewise. Adjust
13310 costs for the root stmt.
13311 (vect_slp_check_for_constructors): Gather all BIT_INSERT_EXPRs
13313 (vect_slp_analyze_bb_1): Simplify by marking all root stmts
13315 (vectorize_slp_instance_root_stmt): Adjust.
13316 (vect_schedule_slp): Likewise.
13318 2021-06-08 Aldy Hernandez <aldyh@redhat.com>
13320 * gimple-ssa-evrp.c (class ssa_equiv_stack): New.
13321 (ssa_equiv_stack::ssa_equiv_stack): New.
13322 (ssa_equiv_stack::~ssa_equiv_stack): New.
13323 (ssa_equiv_stack::enter): New.
13324 (ssa_equiv_stack::leave): New.
13325 (ssa_equiv_stack::push_replacement): New.
13326 (ssa_equiv_stack::get_replacement): New.
13327 (is_pointer_ssa): New.
13328 (class pointer_equiv_analyzer): New.
13329 (pointer_equiv_analyzer::pointer_equiv_analyzer): New.
13330 (pointer_equiv_analyzer::~pointer_equiv_analyzer): New.
13331 (pointer_equiv_analyzer::set_global_equiv): New.
13332 (pointer_equiv_analyzer::set_cond_equiv): New.
13333 (pointer_equiv_analyzer::get_equiv): New.
13334 (pointer_equiv_analyzer::enter): New.
13335 (pointer_equiv_analyzer::leave): New.
13336 (pointer_equiv_analyzer::get_equiv_expr): New.
13337 (pta_valueize): New.
13338 (pointer_equiv_analyzer::visit_stmt): New.
13339 (pointer_equiv_analyzer::visit_edge): New.
13340 (hybrid_folder::value_of_expr): Call PTA.
13341 (hybrid_folder::value_on_edge): Same.
13342 (hybrid_folder::pre_fold_bb): New.
13343 (hybrid_folder::post_fold_bb): New.
13344 (hybrid_folder::pre_fold_stmt): New.
13345 (rvrp_folder::pre_fold_bb): New.
13346 (rvrp_folder::post_fold_bb): New.
13347 (rvrp_folder::pre_fold_stmt): New.
13348 (rvrp_folder::value_of_expr): Call PTA.
13349 (rvrp_folder::value_on_edge): Same.
13351 2021-06-08 Jakub Jelinek <jakub@redhat.com>
13354 * tree-inline.c (copy_tree_body_r): For OMP_CLAUSE_DEPEND don't
13355 check TREE_CODE if OMP_CLAUSE_DECL is NULL.
13357 2021-06-08 Richard Biener <rguenther@suse.de>
13359 PR middle-end/100951
13360 * tree-vect-generic.c (expand_vector_piecewise): Build a
13361 VECTOR_CST if all elements are constant.
13362 (expand_vector_condition): Likewise.
13363 (lower_vec_perm): Likewise.
13364 (expand_vector_conversion): Likewise.
13366 2021-06-08 Martin Liska <mliska@suse.cz>
13368 * doc/invoke.texi: Document new param evrp-sparse-threshold.
13370 2021-06-08 Martin Liska <mliska@suse.cz>
13372 * genautomata.c (create_automata): Fix typo.
13374 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
13376 PR tree-optimization/100794
13377 * tree-predcom.c (tree_predictive_commoning_loop): Add parameter
13378 allow_unroll_p and only allow unrolling when it's true.
13379 (tree_predictive_commoning): Add parameter allow_unroll_p and
13381 (run_tree_predictive_commoning): Likewise.
13382 (pass_predcom::gate): Check flag_tree_loop_vectorize and
13383 global_options_set.x_flag_predictive_commoning.
13384 (pass_predcom::execute): Adjust for allow_unroll_p.
13386 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
13388 * tree-predcom.c (execute_pred_commoning): Remove update_ssa call.
13389 (tree_predictive_commoning_loop): Factor some cleanup stuffs into
13390 lambda function cleanup, remove scev_reset call, and adjust return
13392 (tree_predictive_commoning): Adjust for different changed values,
13393 only set flag TODO_update_ssa_only_virtuals if changed.
13394 (pass_data pass_data_predcom): Remove TODO_update_ssa_only_virtuals
13395 from todo_flags_finish.
13397 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
13399 * gimple-range-cache.cc (class sbr_sparse_bitmap): New.
13400 (sbr_sparse_bitmap::sbr_sparse_bitmap): New.
13401 (sbr_sparse_bitmap::bitmap_set_quad): New.
13402 (sbr_sparse_bitmap::bitmap_get_quad): New.
13403 (sbr_sparse_bitmap::set_bb_range): New.
13404 (sbr_sparse_bitmap::get_bb_range): New.
13405 (sbr_sparse_bitmap::bb_range_p): New.
13406 (block_range_cache::block_range_cache): initialize bitmap obstack.
13407 (block_range_cache::~block_range_cache): Destruct obstack.
13408 (block_range_cache::set_bb_range): Decide when to utilze the
13409 sparse on entry cache.
13410 * gimple-range-cache.h (block_range_cache): Add bitmap obstack.
13411 * params.opt (-param=evrp-sparse-threshold): New.
13413 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
13415 * bitmap.c (bitmap_set_aligned_chunk): New.
13416 (bitmap_get_aligned_chunk): New.
13417 (test_aligned_chunk): New.
13418 (bitmap_c_tests): Call test_aligned_chunk.
13419 * bitmap.h (bitmap_set_aligned_chunk, bitmap_get_aligned_chunk): New.
13421 2021-06-07 Uroš Bizjak <ubizjak@gmail.com>
13424 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
13426 (ix86_expand_vector_init_one_nonzero): Ditto.
13427 (ix86_expand_vector_init_one_var): Ditto.
13428 (ix86_expand_vector_init_general): Ditto.
13429 * config/i386/mmx.md (vec_initv4qiqi): New expander.
13431 2021-06-07 Jeff Law <jeffreyalaw@gmail.com>
13433 * config/h8300/movepush.md: Change most _clobber_flags
13434 patterns to instead use <cczn> subst.
13435 (movsi_cczn): New pattern with usable CC cases split out.
13436 (movsi_h8sx_cczn): Likewise.
13438 2021-06-07 Martin Liska <mliska@suse.cz>
13440 * common/common-target.def: Split long lines and replace them
13442 * target.def: Likewise.
13443 * doc/tm.texi: Re-generated.
13445 2021-06-07 Jakub Jelinek <jakub@redhat.com>
13448 * fold-const.c (fold_read_from_vector): Return NULL if trying to
13449 read from a CONSTRUCTOR with vector type elements.
13451 2021-06-07 Jakub Jelinek <jakub@redhat.com>
13453 PR middle-end/100898
13454 * tree-inline.c (copy_bb): Only use gimple_call_arg_ptr if memcpy
13455 should copy any arguments. Don't call gimple_call_num_args
13456 on id->call_stmt or call_stmt more than once.
13458 2021-06-07 liuhongt <hongtao.liu@intel.com>
13461 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): Refine
13463 (<insn>v4siv4di2): Delete constraints for define_expand.
13465 2021-06-07 liuhongt <hongtao.liu@intel.com>
13468 * config/i386/i386-expand.c (ix86_expand_builtin): Remove
13469 assignment of cfun->machine->has_explicit_vzeroupper.
13470 * config/i386/i386-features.c
13471 (ix86_add_reg_usage_to_vzerouppers): Delete.
13472 (ix86_add_reg_usage_to_vzeroupper): Ditto.
13473 (rest_of_handle_insert_vzeroupper): Remove
13474 ix86_add_reg_usage_to_vzerouppers, add df_analyze at the end
13476 (gate): Remove cfun->machine->has_explicit_vzeroupper.
13477 * config/i386/i386-protos.h (ix86_expand_avx_vzeroupper):
13479 * config/i386/i386.c (ix86_insn_callee_abi): New function.
13480 (ix86_initialize_callee_abi): Ditto.
13481 (ix86_expand_avx_vzeroupper): Ditto.
13482 (ix86_hard_regno_call_part_clobbered): Adjust for vzeroupper
13484 (TARGET_INSN_CALLEE_ABI): Define as ix86_insn_callee_abi.
13485 (ix86_emit_mode_set): Call ix86_expand_avx_vzeroupper
13487 * config/i386/i386.h (struct GTY(()) machine_function): Delete
13488 has_explicit_vzeroupper.
13489 * config/i386/i386.md (enum unspec): New member
13491 (ABI_DEFAULT,ABI_VZEROUPPER,ABI_UNKNOWN): New
13492 define_constants for insn callee abi index.
13493 * config/i386/predicates.md (vzeroupper_pattern): Adjust.
13494 * config/i386/sse.md (UNSPECV_VZEROUPPER): Deleted.
13495 (avx_vzeroupper): Call ix86_expand_avx_vzeroupper.
13496 (*avx_vzeroupper): Rename to ..
13497 (avx_vzeroupper_callee_abi): .. this, and adjust pattern as
13498 call_insn which has a special vzeroupper ABI.
13499 (*avx_vzeroupper_1): Deleted.
13501 2021-06-07 liuhongt <hongtao.liu@intel.com>
13504 * df-scan.c (df_get_call_refs): When call_insn is a fake call,
13505 it won't use stack pointer reg.
13506 * final.c (leaf_function_p): When call_insn is a fake call, it
13507 won't affect caller as a leaf function.
13508 * reg-stack.c (callee_clobbers_any_stack_reg): New.
13509 (subst_stack_regs): When call_insn doesn't clobber any stack
13510 reg, don't clear the arguments.
13511 * rtl.c (shallow_copy_rtx): Don't clear flag used when orig is
13513 * shrink-wrap.c (requires_stack_frame_p): No need for stack
13514 frame for a fake call.
13515 * rtl.h (FAKE_CALL_P): New macro.
13517 2021-06-06 Eric Botcazou <ebotcazou@adacore.com>
13519 * config/sparc/sparc-protos.h (order_regs_for_local_alloc): Rename
13521 (sparc_order_regs_for_local_alloc): ...this.
13522 (sparc_leaf_reg_remap): Declare.
13523 * config/sparc/sparc.h (ADJUST_REG_ALLOC_ORDER): Adjust.
13524 (LEAF_REG_REMAP): Reimplement as call to sparc_leaf_reg_remap.
13525 * config/sparc/sparc.c (leaf_reg_remap): Delete.
13526 (order_regs_for_local_alloc): Rename to...
13527 (sparc_order_regs_for_local_alloc): ...this.
13528 (sparc_leaf_reg_remap): New function.
13529 (sparc_conditional_register_usage): Do not modify leaf_reg_remap.
13531 2021-06-06 David Edelsohn <dje.gcc@gmail.com>
13533 * config/rs6000/rs6000.c (rs6000_xcoff_asm_output_aligned_decl_common):
13534 Use assemble_name to output BSS section name.
13536 2021-06-06 Uroš Bizjak <ubizjak@gmail.com>
13538 * config/i386/constraints.md (Bs):
13539 Remove boolean operators from match_test RTX.
13542 (M): Use "mode" variable instead of GET_MODE (op) in match_test RTX.
13545 2021-06-06 Martin Liska <mliska@suse.cz>
13547 * doc/extend.texi: Add missing @headitem.
13548 * doc/invoke.texi: Likewise.
13549 * doc/objc.texi: Likewise.
13551 2021-06-06 Martin Liska <mliska@suse.cz>
13553 * genhooks.c (emit_findices): Remove unused function.
13554 (emit_documentation): Do not call emit_findices
13555 and do not search for @Fcode directives.
13557 2021-06-06 Martin Liska <mliska@suse.cz>
13559 * doc/invoke.texi: Remove extra character.
13561 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
13563 * config/sh/sh.md (doloop_end_split): Fix empty split condition.
13565 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
13567 * config/sparc/sparc.md (*snedi<W:mode>_zero_vis3,
13568 *neg_snedi<W:mode>_zero_subxc, *plus_snedi<W:mode>_zero,
13569 *plus_plus_snedi<W:mode>_zero, *minus_snedi<W:mode>_zero,
13570 *minus_minus_snedi<W:mode>_zero): Fix empty split condition.
13572 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
13574 * config/or1k/or1k.md (*movdi): Fix empty split condition.
13576 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
13578 * config/mips/mips.md (<anonymous>, bswapsi2, bswapdi2): Fix empty
13581 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
13583 * config/m68k/m68k.md (*zero_extend_inc, *zero_extend_dec,
13584 *zero_extendsidi2): Fix empty split condition.
13586 2021-06-05 Jeff Law <jeffreyalaw@gmail.com>
13588 * config/h8300/addsub.md: Fix split condition in define_insn_and_split
13590 * config/h8300/bitfield.md: Likewise.
13591 * config/h8300/combiner.md: Likewise.
13592 * config/h8300/divmod.md: Likewise.
13593 * config/h8300/extensions.md: Likewise.
13594 * config/h8300/jumpcall.md: Likewise.
13595 * config/h8300/movepush.md: Likewise.
13596 * config/h8300/multiply.md: Likewise.
13597 * config/h8300/other.md: Likewise.
13598 * config/h8300/shiftrotate.md: Likewise.
13599 * config/h8300/logical.md: Likewise. Fix split pattern to use
13600 code iterator that somehow slipped through.
13602 2021-06-04 Tobias Burnus <tobias@codesourcery.com>
13604 PR middle-end/100905
13605 * tree-nested.c (convert_nonlocal_omp_clauses,
13606 convert_local_omp_clauses): Handle OMP_CLAUSE_BIND.
13608 2021-06-04 Martin Sebor <msebor@redhat.com>
13610 PR middle-end/100732
13611 * gimple-fold.c (gimple_fold_builtin_sprintf): Avoid folding calls
13612 with either source or destination argument of invalid type.
13613 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Avoid checking
13614 calls with arguments of invalid type.
13616 2021-06-04 Martin Sebor <msebor@redhat.com>
13618 * attribs.c (init_attr_rdwr_indices): Use VLA bounds in the expected
13620 (attr_access::vla_bounds): Also handle VLA bounds.
13622 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
13624 * config/i386/predicates.md (GOT_memory_operand):
13625 Implement using match_code RTXes.
13626 (GOT32_symbol_operand): Ditto.
13628 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
13631 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
13633 (ix86_expand_vector_init_general): Ditto.
13634 Use SImode instead of word_mode for logic operations
13635 when GET_MODE_SIZE (mode) < UNITS_PER_WORD.
13636 (expand_vec_perm_even_odd_1): Assert that V2HI mode should be
13637 implemented by expand_vec_perm_1.
13638 (expand_vec_perm_broadcast_1): Assert that V2HI and V4HI modes
13639 should be implemented using standard shuffle patterns.
13640 (ix86_vectorize_vec_perm_const): Handle V2HImode. Add V4HI and
13641 V2HI modes to modes, implementable with shuffle for one operand.
13642 * config/i386/mmx.md (*punpckwd): New insn_and_split pattern.
13643 (*pshufw_1): New insn pattern.
13644 (*vec_dupv2hi): Ditto.
13645 (vec_initv2hihi): New expander.
13647 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
13649 * config/arm/vfp.md (no_literal_pool_df_immediate,
13650 no_literal_pool_sf_immediate): Fix empty split condition.
13652 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
13654 * config/i386/i386.md (*load_tp_x32_zext, *add_tp_x32_zext,
13655 *tls_dynamic_gnu2_combine_32): Fix empty split condition.
13656 * config/i386/sse.md (*<sse2_avx2>_pmovmskb_lt,
13657 *<sse2_avx2>_pmovmskb_zext_lt, *sse2_pmovmskb_ext_lt,
13658 *<sse4_1_avx2>_pblendvb_lt): Likewise.
13660 2021-06-04 Jakub Jelinek <jakub@redhat.com>
13663 * config/i386/i386-expand.c (ix86_expand_vector_init): Handle
13664 concatenation from half-sized modes with TImode elements.
13666 2021-06-04 Claudiu Zissulescu <claziss@synopsys.com>
13668 * config/arc/arc.c (arc_override_options): Disable millicode
13669 thunks when RF16 is on.
13671 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
13673 * config/rs6000/rs6000.h (PROMOTE_MODE): Remove.
13675 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
13677 * config/rs6000/rs6000-call.c (rs6000_promote_function_mode):
13678 Replace PROMOTE_MODE marco with its content.
13680 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
13682 * config/cris/cris.md (*addi_reload): Fix empty split condition.
13684 2021-06-03 Jim Wilson <jimw@sifive.com>
13686 * config.gcc (riscv*-*-*): If --with-riscv-attribute not used,
13687 turn it on for all riscv targets.
13689 2021-06-03 Uroš Bizjak <ubizjak@gmail.com>
13692 * config/i386/i386-expand.c (ix86_expand_vector_set):
13693 Handle V2HI and V4QI modes.
13694 (ix86_expand_vector_extract): Ditto.
13695 * config/i386/mmx.md (*pinsrw): New insn pattern.
13698 (*pextrw_zext): Ditto.
13700 (*pextrb_zext): Ditto.
13701 (vec_setv2hi): New expander.
13702 (vec_extractv2hihi): Ditto.
13703 (vec_setv4qi): Ditto.
13704 (vec_extractv4qiqi): Ditto.
13705 (vec_setv8qi): Enable only for TARGET_SSE4_1.
13706 (vec_extractv8qiqi): Ditto.
13708 2021-06-03 Aaron Sawdey <acsawdey@linux.ibm.com>
13710 * config/rs6000/genfusion.pl (gen_logical_addsubf): Fix input
13711 order to subf instruction.
13712 * config/rs6000/fusion.md: Regenerate.
13714 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
13716 * calls.c (get_size_range): Use range_of_expr instead of
13717 determine_value_range.
13718 * tree-affine.c (expr_to_aff_combination): Same.
13719 * tree-data-ref.c (split_constant_offset): Same.
13720 * tree-vrp.c (determine_value_range_1): Remove.
13721 (determine_value_range): Remove.
13722 * tree-vrp.h (determine_value_range): Remove.
13724 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
13726 * function-tests.c (test_ranges): Call gimple_range_tests.
13727 * gimple-range-cache.cc (ranger_cache::range_of_expr): Pass stmt
13729 * gimple-range.cc (fur_source::get_operand): Do not call
13730 get_tree_range or gimple_range_global.
13732 (get_tree_range): Move to value-query.cc.
13733 Call get_arith_expr_range.
13734 (gimple_ranger::range_of_expr): Add argument to get_tree_range.
13735 Include gimple-range-tests.cc.
13736 * gimple-range.h (fold_range): Add argument.
13737 (get_tree_range): Remove.
13738 * selftest.h (gimple_range_tests): New.
13739 * value-query.cc (global_range_query::range_of_expr): Add
13741 (range_query::get_tree_range): Move from gimple-range.cc.
13742 * value-query.h (class range_query): Add get_tree_range and
13743 get_arith_expr_range. Make fur_source a friend.
13744 * vr-values.c (vr_values::range_of_expr): Pass stmt to
13746 * gimple-range-tests.cc: New file.
13748 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
13750 * gimple-range.cc (gimple_ranger::export_global_ranges): Call
13751 update_global_range.
13752 * value-query.cc (update_global_range): New.
13753 * value-query.h (update_global_range): New.
13755 2021-06-03 David Malcolm <dmalcolm@redhat.com>
13757 * diagnostic-show-locus.c (diagnostic_show_locus): Don't reject
13758 printing the same location twice if there are fix-it hints,
13759 multiple locations, or a label.
13761 2021-06-03 Andre Vieira <andre.simoesdiasvieira@arm.com>
13763 * tree-vect-loop.c (vect_transform_loop): Use main loop's various'
13764 thresholds to narrow the upper bound on epilogue iterations.
13766 2021-06-03 Christophe Lyon <christophe.lyon@linaro.org>
13768 * config/arm/mve.md (mve_vabsq_f<mode>): Use 'abs' instead of unspec.
13769 (mve_vabsq_s<mode>): Likewise.
13770 * config/arm/neon.md (abs<mode>2): Rename to neon_abs<mode>2.
13771 * config/arm/unspecs.md (VABSQ_F, VABSQ_S): Delete.
13772 * config/arm/vec-common.md (neg<mode>2): Rename to
13773 <absneg_str><mode>2.
13775 2021-06-03 Claudiu Zissulescu <claziss@synopsys.com>
13777 * common/config/arc/arc-common.c (arc_option_optimization_table):
13778 Remove malign-call.
13779 * config/arc/arc.c (arc_unalign_branch_p): Remove unused function.
13780 * config/arc/arc.h (TARGET_MIXED_CODE): Remove macro.
13781 (INDEX_REG_CLASS): Only refer to GENERAL_REGS.
13782 * config/arc/arc.md (abssi2_mixed): Remove pattern.
13783 * config/arc/arc.opt (munalign-prob-threshold): Mark it obsolete.
13784 (malign-call): Likewise.
13785 (mmixed-code): Likewise.
13786 * doc/invoke.texi (ARC): Update doc.
13788 2021-06-03 Martin Liska <mliska@suse.cz>
13790 * common.opt: Use proper Enum values.
13791 * opts.c (COVERAGE_SANITIZER_OPT): Remove.
13792 (parse_sanitizer_options): Handle only sanitizer_opts.
13793 (common_handle_option): Just assign value.
13795 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
13798 * tree-inline.c (inline_forbidden_p): Remove test on return type.
13800 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
13802 * dwarf2out.c (loc_list_from_tree_1) <FUNCTION_DECL>: Also generate
13803 DW_OP_GNU_variable_value referencing an existing DIE at file scope.
13804 (type_byte_size): Inline into...
13805 (add_byte_size_attribute): ...this and call add_scalar_info.
13807 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
13809 * dwarf2out.c (mem_loc_descriptor) <UDIV>: Fix typo.
13810 (typed_binop_from_tree): New function.
13811 (loc_list_from_tree_1) <EXACT_DIV_EXPR>: For an unsigned type,
13812 turn a divide by a power of 2 into a shift.
13813 <CEIL_DIV_EXPR>: For an unsigned type, use a signed divide if the
13814 size of the mode is lower than DWARF2_ADDR_SIZE; otherwise, do a
13815 typed divide by calling typed_binop_from_tree.
13817 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
13819 * dwarf2out.c (scompare_loc_descriptor): Fix head comment.
13820 (is_handled_procedure_type): Likewise.
13821 (struct loc_descr_context): Add strict_signedness field.
13822 (resolve_args_picking_1): Deal with DW_OP_[GNU_]deref_type,
13823 DW_OP_[GNU_]convert and DW_OP_[GNU_]reinterpret.
13824 (resolve_args_picking): Minor tweak.
13825 (function_to_dwarf_procedure): Initialize strict_signedness field.
13826 (type_byte_size): Likewise.
13827 (field_byte_offset): Likewise.
13828 (gen_descr_array_type_die): Likewise.
13829 (gen_variant_part): Likewise.
13830 (loc_list_from_tree_1) <CALL_EXPR>: Tidy up and set strict_signedness
13831 to true when a context is present before evaluating the arguments.
13832 <COND_EXPR>: Do not generate a useless comparison with zero.
13833 When dereferencing an address, if strict_signedness is true and the
13834 type is small and signed, use DW_OP_deref_type to do the dereference
13835 and then DW_OP_convert to convert back to the generic type.
13837 2021-06-03 Jakub Jelinek <jakub@redhat.com>
13840 * tree-inline.c (copy_tree_body_r): Handle iterators on
13841 OMP_CLAUSE_AFFINITY or OMP_CLAUSE_DEPEND.
13843 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
13845 * config/arc/arc.md (*bbit_di): Remove.
13847 2021-06-02 Christoph Muellner <cmuellner@gcc.gnu.org>
13849 PR rtl-optimization/100264
13850 * ree.c (get_sub_rtx): Ignore SET expressions without register
13851 destinations and remove assertion, as it is not valid anymore
13852 with this new behaviour.
13853 (merge_def_and_ext): Eliminate destination check for register
13854 as such SET expressions can't occur anymore.
13855 (combine_reaching_defs): Likewise.
13857 2021-06-02 Jakub Jelinek <jakub@redhat.com>
13860 * config/xtensa/xtensa.h (LEAF_REG_REMAP): Cast REGNO to int to avoid
13861 -Wtype-limits warnings.
13862 (DWARF_FRAME_REGISTER): Rewrite into ternary operator with addition
13863 in operands to avoid -Wsign-compare warnings.
13865 2021-06-02 Pat Haugen <pthaugen@linux.ibm.com>
13867 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): Use
13870 2021-06-02 Vineet Gupta <vgupta@synopsys.com>
13872 * config/arc/arc.h (TARGET_CPU_DEFAULT): Change to hs38_linux.
13874 2021-06-02 Ilya Leoshkevich <iii@linux.ibm.com>
13876 * config/s390/s390.md(*ashrdi3_31<setcc><cconly>): Use a single
13878 * config/s390/subst.md(cconly_subst): Use a single constraint
13879 in (match_scratch).
13881 2021-06-02 Martin Liska <mliska@suse.cz>
13883 * ipa-icf.h: Use auto_vec for memory_access_types.
13885 2021-06-02 Jeff Law <jeffreyalaw@gmail.com>
13887 * config/h8300/h8300-protos.h (compute_a_shift_length): Drop unused
13888 argument from prototype.
13889 (output_logical_op): Add rtx_code argument.
13890 (compute_logical_op_length): Likewise.
13891 * config/h8300/h8300.c (h8300_and_costs): Pass additional argument
13892 to compute_a_shift_length.
13893 (output_logical_op); New argument with the rtx code rather than
13894 extracting it from an operand. Handle QImode too.
13895 (compute_logical_op_length): Similary.
13896 (compute_a_shift_length): Drop unused argument.
13897 * config/h8300/h8300.md (logicals): New code iterator.
13898 * config/h8300/logical.md (<code><mode>3 expander): Combine
13899 the "and" expander with the "ior"/"xor" expander.
13900 (bclr<mode>msx): Combine the QI/HI mode patterns.
13901 (<logical><mode>3 insns): Use code iterator rather than match_operator.
13902 Handle QImode as well. Update call to output_logical_op and
13903 compute_logical_op_length to pass in rtx_code
13904 Fix split condition on all define_insn_and_split patterns.
13905 (one_cmpl<mode>2<cczn>): Use <cczn> to support both clobbering
13906 the flags and setting ZN via existing define_subst.
13907 * config/h8300/shiftrotate.md: Drop unused argument from
13908 calls to compute_a_shift_length.
13909 Signed-off-by: Jeff Law <jeffreyalaw@gmail.com>
13911 2021-06-01 Andrew Pinski <apinski@marvell.com>
13913 PR tree-optimization/25290
13914 * tree-ssa-phiopt.c (match_simplify_replacement):
13916 (tree_ssa_phiopt_worker): Use match_simplify_replacement.
13917 (two_value_replacement): Change the comment about
13918 conditional_replacement.
13919 (conditional_replacement): Delete.
13921 2021-06-01 Andrew Pinski <apinski@marvell.com>
13923 PR tree-optimization/95481
13924 * tree-tailcall.c (find_tail_calls): Handle empty typed
13927 2021-06-01 Andrew Pinski <apinski@marvell.com>
13929 * gimplify.c (zero_sized_field_decl): Delete
13930 (zero_sized_type): Delete
13931 (gimplify_init_ctor_eval): Use is_empty_type instead
13932 of zero_sized_field_decl.
13933 (gimplify_modify_expr): Use is_empty_type instead of
13936 2021-06-01 Jason Merrill <jason@redhat.com>
13939 * tree.h (CALL_FROM_NEW_OR_DELETE_P): Adjust comment.
13941 2021-06-01 Jason Merrill <jason@redhat.com>
13944 * diagnostic.h (warning_enabled_at): Declare.
13945 * diagnostic.c (diagnostic_enabled): Factor out from...
13946 (diagnostic_report_diagnostic): ...here.
13947 (warning_enabled_at): New.
13949 2021-06-01 Aldy Hernandez <aldyh@redhat.com>
13951 * gimple-ssa-evrp.c: Enable exporting of global ranges.
13953 2021-06-01 Martin Liska <mliska@suse.cz>
13956 * doc/invoke.texi: Mention that -fgcse-after-reload
13957 is enabled with -O3.
13959 2021-06-01 liuhongt <hongtao.liu@intel.com>
13961 PR tree-optimization/98365
13962 * tree-if-conv.c (strip_nop_cond_scalar_reduction): New function.
13963 (is_cond_scalar_reduction): Handle nop_expr in cond scalar reduction.
13964 (convert_scalar_cond_reduction): Ditto.
13965 (predicate_scalar_phi): Ditto.
13967 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
13969 PR tree-optimization/100781
13970 * gimple-range-cache.cc (ranger_cache::ranger_cache): Enable new
13971 value calculation by default.
13972 (ranger_cache::enable_new_values): New.
13973 (ranger_cache::disable_new_values): New.
13974 (ranger_cache::push_poor_value): Check if new values are allowed.
13975 * gimple-range-cache.h (class ranger_cache): New member/methods.
13976 * gimple-range.cc (gimple_ranger::range_of_expr): Check for debug
13977 statement, and disable/renable new value calculation.
13979 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
13981 * gimple-range-cache.cc (ranger_cache::ssa_range_in_bb): Delete.
13982 (ranger_cache::range_of_def): New.
13983 (ranger_cache::entry_range): New.
13984 (ranger_cache::exit_range): New.
13985 (ranger_cache::range_of_expr): Adjust.
13986 (ranger_cache::range_on_edge): Adjust.
13987 (ranger_cache::propagate_cache): Call exit_range directly.
13988 * gimple-range-cache.h (class ranger_cache): Adjust.
13990 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
13992 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust for
13993 gori_compute being a member rather than base class.
13994 dervied call to member call.
13995 (ranger_cache::dump): No longer dump gori_map.
13996 (ranger_cache::dump_bb): New.
13997 (ranger_cache::get_non_stale_global_range): Adjust for gori_compute
13998 being a member rather than base class.
13999 (ranger_cache::set_global_range): Ditto.
14000 (ranger_cache::ssa_range_in_bb): Ditto.
14001 (ranger_cache::range_of_expr): New.
14002 (ranger_cache::range_on_edge): New.
14003 (ranger_cache::block_range): Adjust for gori_computes. Debug changes.
14004 (ranger_cache::propagate_cache): Adjust debugging output.
14005 (ranger_cache::fill_block_cache): Adjust for gori_computes. Debug
14007 * gimple-range-cache.h (class ranger_cache): Make gori_compute a
14008 member, and inherit from range_query instead.
14009 (ranger_cache::dump_bb): New. split from dump.
14010 * gimple-range-gori.cc (gori_compute::ssa_range_in_bb): Delete.
14011 (gori_compute::expr_range_at_stmt): Delete.
14012 (gori_compute::compute_name_range_op): Delete.
14013 (gori_compute::compute_operand_range_switch): Add fur_source.
14014 (gori_compute::compute_operand_range): Add fur_source param, inline
14015 old compute_name_range_op and optimize_logical_operands.
14016 (struct tf_range): Delete.
14017 (gori_compute::logical_combine): Adjust
14018 (gori_compute::optimize_logical_operands): Delete.
14019 (gori_compute::compute_logical_operands_in_chain): Delete.
14020 (gori_compute::compute_logical_operands): Adjust.
14021 (gori_compute::compute_operand1_range): Adjust to fur_source.
14022 (gori_compute::compute_operand2_range): Ditto.
14023 (gori_compute::compute_operand1_and_operand2_range): Ditto.
14024 (gori_compute::outgoing_edge_range_p): Add range_query parameter,
14025 and adjust to fur_source.
14026 * gimple-range-gori.h (class gori_compute): Simplify and adjust to
14027 range_query and fur_source.
14028 * gimple-range.cc (gimple_ranger::range_on_edge): Query range_on_edge
14029 from the ranger_cache..
14030 (gimple_ranger::fold_range_internal): Adjust to base class change of
14032 (gimple_ranger::dump_bb): Adjust dump.
14033 * gimple-range.h (gimple_ranger):export gori computes object.
14035 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
14037 PR tree-optimization/100774
14038 * gimple-range-cache.cc (ranger_cache::get_non_stale_global_range):
14039 Constant values are also not stale.
14040 (ranger_cache::set_global_range): Range invariant values should also
14041 have the correct timestamp.
14043 2021-05-31 Martin Liska <mliska@suse.cz>
14045 * tree-streamer-in.c (unpack_ts_function_decl_value_fields):
14046 Unpack FUNCTION_DECL_DECL_TYPE.
14047 * tree-streamer-out.c (pack_ts_function_decl_value_fields):
14048 Stream FUNCTION_DECL_DECL_TYPE instead of
14049 DECL_IS_OPERATOR_NEW_P.
14050 * tree.h (set_function_decl_type): Use FUNCTION_DECL_DECL_TYPE
14052 (DECL_IS_OPERATOR_NEW_P): Likewise.
14053 (DECL_IS_OPERATOR_DELETE_P): Likewise.
14054 (DECL_LAMBDA_FUNCTION_P): Likewise.
14056 2021-05-31 Richard Biener <rguenther@suse.de>
14059 * internal-fn.c (expand_SHUFFLEVECTOR): Define.
14060 * internal-fn.def (SHUFFLEVECTOR): New.
14061 * internal-fn.h (expand_SHUFFLEVECTOR): Declare.
14062 * doc/extend.texi: Document __builtin_shufflevector.
14064 2021-05-31 Peter Bergner <bergner@linux.ibm.com>
14067 * config/rs6000/predicates.md(mma_assemble_input_operand): Allow
14068 indexed form addresses.
14070 2021-05-29 Jeff Law <jlaw@tachyum.com>
14072 * config/h8300/h8300.c (h8300_emit_stack_adjustment): Drop unused
14073 parameter. Call callers fixed.
14075 (output_plussi): Add FALLTHRU markers.
14076 (h8300_shift_needs_scratch_p): Add gcc_unreachable marker.
14078 2021-05-29 Jakub Jelinek <jakub@redhat.com>
14080 PR middle-end/99928
14081 * gimplify.c (gimplify_scan_omp_clauses): For taskloop simd
14082 combined with parallel, make sure to add shared clause to
14083 parallel for explicit linear clause.
14085 2021-05-29 Aldy Hernandez <aldyh@redhat.com>
14087 PR tree-optimization/100787
14088 * gimple-ssa-evrp.c: Disable exporting of global ranges.
14090 2021-05-28 Jason Merrill <jason@redhat.com>
14092 * tree-iterator.h (struct tree_stmt_iterator): Add operator++,
14093 operator--, operator*, operator==, and operator!=.
14094 (class tsi_range): New.
14096 2021-05-28 Richard Biener <rguenther@suse.de>
14098 PR tree-optimization/100778
14099 * tree-vect-slp.c (vect_build_slp_tree_1): Prevent possibly
14100 trapping ops in different BBs.
14102 2021-05-28 Richard Biener <rguenther@suse.de>
14105 * tree-inline.c (copy_bb): When processing __builtin_va_arg_pack
14106 copy fntype from original call.
14108 2021-05-28 Martin Liska <mliska@suse.cz>
14110 PR gcov-profile/100751
14111 * doc/gcov.texi: Revert partially a hunk that was wrong.
14113 2021-05-28 Cooper Qu <cooper.qu@linux.alibaba.com>
14115 * config/csky/csky-linux-elf.h (HAVE_sync_compare_and_swapqi):
14117 (HAVE_sync_compare_and_swaphi): Likewise.
14118 (HAVE_sync_compare_and_swapsi): Likewise.
14120 2021-05-28 Jakub Jelinek <jakub@redhat.com>
14122 PR middle-end/99928
14123 * tree.h (OMP_CLAUSE_MAP_IMPLICIT): Define.
14125 2021-05-28 Tobias Burnus <tobias@codesourcery.com>
14127 * gimplify.c (gimplify_omp_affinity): New.
14128 (gimplify_scan_omp_clauses): Call it; remove affinity clause afterwards.
14129 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_AFFINITY.
14130 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_AFFINITY.
14131 * tree.c (omp_clause_num_ops, omp_clause_code_name): Add clause.
14132 (walk_tree_1): Handle OMP_CLAUSE_AFFINITY.
14134 2021-05-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
14135 Richard Biener <rguenther@suse.de>
14137 * match.pd <popcount & / + pattern matching>:
14138 When generating popcount directly fails, try doing it in two halves.
14140 2021-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
14142 * Makefile.in (generated_files): Add gimple-match.c and
14145 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
14147 * gensupport.c (alter_predicate_for_insn): Handle MATCH_DUP.
14149 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
14151 * gensupport.c (alter_constraints): Add MATCH_SCRATCH case.
14153 2021-05-28 Kewen Lin <linkw@linux.ibm.com>
14155 PR tree-optimization/99398
14156 * tree-ssa-forwprop.c (simplify_permutation): Optimize some cases
14157 where the fed operands are CTOR/CST and propagated through
14158 VIEW_CONVERT_EXPR. Call vec_perm_indices::new_shrunk_vector.
14159 * vec-perm-indices.c (vec_perm_indices::new_shrunk_vector): New
14161 * vec-perm-indices.h (vec_perm_indices::new_shrunk_vector): New
14164 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
14166 * config/i386/mmx.md (addv2sf3): Do not call
14167 ix86_fixup_binary_operands_no_copy.
14170 (<smaxmin:code>v2sf3): Ditto.
14171 (<plusminus:insn><MMXMODEI:mode>3): Ditto.
14172 (<plusminus:insn><VI_32:mode>3): Remove expander.
14173 (<plusminus:insn><VI_32:mode>3): Rename from
14174 "*<plusminus:insn><VI_32:mode>3".
14175 (mulv4hi): Do not call ix86_fixup_binary_operands_no_copy.
14176 (mulv2hi3): Remove expander.
14177 (mulv2hi3): Rename from *mulv2hi3.
14178 (<s>mulv2hi3_highpart): Remove expander.
14179 (<s>mulv2hi3_highpart): Rename from *<s>mulv2hi3_highpart.
14180 (<smaxmin:code><MMXMODE14:mode>3): Rename from
14181 "*mmx_<smaxmin:code><MMXMODE14:mode>3".
14182 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Remove expander.
14183 (SMAXMIN_MMXMODEI): Remove mode iterator.
14184 (<smaxmin:code>v4hi3): New expander.
14185 (<smaxmin:code>v4qi3): Rename from *<smaxmin:code>v4qi3.
14186 (<smaxmin:code>v2hi3): Rename from *<smaxmin:code>v2hi3.
14187 (<smaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
14188 (SMAXMIN_VI_32): Remove mode iterator.
14189 (<umaxmin:code><MMXMODE24:mode>3): Rename from
14190 "*mmx_<umaxmin:code><MMXMODE24:mode>3".
14191 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Remove expander.
14192 (UMAXMIN_MMXMODEI): Remove mode iterator.
14193 (<umaxmin:code>v8qi3): New expander.
14194 (<umaxmin:code>v4qi3): Rename from *<umaxmin:code>v4qi3.
14195 (<umaxmin:code>v2hi3): Rename from *<umaxmin:code>v2hi3.
14196 (<umaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
14197 (UMAXMIN_VI_32): Remove mode iterator.
14198 (<any_shift:insn>v2hi3): Remove expander.
14199 (<any_shift:insn>v2hi3): Rename from *<any_shift:insn>v2hi3.
14200 (<any_logic:code><MMXMODEI:mode>3): Do not call
14201 ix86_fixup_binary_operands_no_copy.
14202 (<any_logic:code><VI_32:mode>3): Remove expander.
14203 (<any_logic:code><VI_32:mode>3): Rename from
14204 "*<any_logic:code><VI_32:mode>3".
14205 (uavg<mode>3_ceil): Do not call ix86_fixup_binary_operands_no_copy.
14206 * config/i386/sse.md (div<VF2:mode>3): Do not call
14207 ix86_fixup_binary_operands_no_copy.
14208 (div<VF1:mode>3): Ditto.
14209 (<maxmin:code><VI8_AVX2_AVX512F:mode>3): Ditto.
14210 (smulhrsv4hi3): Ditto.
14211 (smulhrsv2hi3): Ditto.
14213 2021-05-27 Martin Sebor <msebor@redhat.com>
14215 * ggc.h (gt_ggc_mx): Add overloads for all integers.
14217 * hash-map.h (class hash_map): Add pch_nx_helper overloads for all
14219 (hash_map::operator==): New function.
14221 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
14224 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
14225 For TARGET_XOP bypass SSE comparisons for all supported vector modes.
14226 * config/i386/mmx.md (*xop_maskcmp<MMXMODEI:mode>3): New insn pattern.
14227 (*xop_maskcmp<VI_32:mode>3): Ditto.
14228 (*xop_maskcmp_uns<MMXMODEI:mode>3): Ditto.
14229 (*xop_maskcmp_uns<VI_32:mode>3): Ditto.
14231 2021-05-27 Richard Earnshaw <rearnsha@arm.com>
14234 * config/arm/arm.c (arm_configure_build_target): Remove parameter
14235 opts_set, directly check opts parameters for being non-null.
14236 (arm_option_restore): Update call to arm_configure_build_target.
14237 (arm_option_override): Likewise.
14238 (arm_can_inline_p): Likewise.
14239 (arm_valid_target_attribute_tree): Likewise.
14240 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
14241 * config/arm/arm-protos.h (arm_configure_build_target): Adjust
14244 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
14246 * vr-values.c (simplify_conversion_using_ranges): Use
14247 get_range_query instead of get_global_range_query.
14249 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
14251 * gimple-range.cc (get_range_global): Move to value-query.cc.
14252 (gimple_range_global): Same.
14253 (get_global_range_query): Same.
14254 (global_range_query::range_of_expr): Same.
14255 * gimple-range.h (class global_range_query): Move to
14257 (gimple_range_global): Same.
14258 * tree-ssanames.c (get_range_info): Move to value-query.cc.
14259 (get_ptr_nonnull): Same.
14260 * tree-ssanames.h (get_range_info): Remove.
14261 (get_ptr_nonnull): Remove.
14262 * value-query.cc (get_ssa_name_range_info): Move from
14264 (get_ssa_name_ptr_info_nonnull): Same.
14265 (get_range_global): Move from gimple-range.cc.
14266 (gimple_range_global): Same.
14267 (get_global_range_query): Same.
14268 (global_range_query::range_of_expr): Same.
14269 * value-query.h (class global_range_query): Move from
14271 (gimple_range_global): Same.
14273 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
14276 * config/i386/mmx.md (uavgv4qi3_ceil): New insn pattern.
14277 (uavgv2hi3_ceil): Ditto.
14279 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
14282 * doc/extend.texi (scalar_storage_order): Rephrase slightly.
14284 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
14286 * tree-ssanames.c (get_range_info): Merge both copies of
14287 get_range_info into one that works with irange.
14288 * tree-ssanames.h (get_range_info): Remove version that works on
14291 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
14293 * builtins.c (check_nul_terminated_array): Convert to get_range_query.
14294 (expand_builtin_strnlen): Same.
14295 (determine_block_size): Same.
14296 * fold-const.c (expr_not_equal_to): Same.
14297 * gimple-fold.c (size_must_be_zero_p): Same.
14298 * gimple-match-head.c: Include gimple-range.h.
14299 * gimple-pretty-print.c (dump_ssaname_info): Convert to get_range_query.
14300 * gimple-ssa-warn-restrict.c
14301 (builtin_memref::extend_offset_range): Same.
14302 * graphite-sese-to-poly.c (add_param_constraints): Same.
14303 * internal-fn.c (get_min_precision): Same.
14304 * ipa-fnsummary.c (set_switch_stmt_execution_predicate): Same.
14305 * ipa-prop.c (ipa_compute_jump_functions_for_edge): Same.
14307 * tree-data-ref.c (split_constant_offset): Same.
14308 (dr_step_indicator): Same.
14309 * tree-dfa.c (get_ref_base_and_extent): Same.
14310 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
14311 * tree-ssa-loop-niter.c (refine_value_range_using_guard): Same.
14312 (determine_value_range): Same.
14313 (record_nonwrapping_iv): Same.
14314 (infer_loop_bounds_from_signedness): Same.
14315 (scev_var_range_cant_overflow): Same.
14316 * tree-ssa-phiopt.c (two_value_replacement): Same.
14317 * tree-ssa-pre.c (insert_into_preds_of_block): Same.
14318 * tree-ssa-reassoc.c (optimize_range_tests_to_bit_test): Same.
14319 * tree-ssa-strlen.c (handle_builtin_stxncpy_strncat): Same.
14321 (dump_strlen_info): Same.
14322 (set_strlen_range): Same.
14323 (maybe_diag_stxncpy_trunc): Same.
14324 (get_len_or_size): Same.
14325 (handle_integral_assign): Same.
14326 * tree-ssa-structalias.c (find_what_p_points_to): Same.
14327 * tree-ssa-uninit.c (find_var_cmp_const): Same.
14328 * tree-switch-conversion.c (bit_test_cluster::emit): Same.
14329 * tree-vect-patterns.c (vect_get_range_info): Same.
14330 (vect_recog_divmod_pattern): Same.
14331 * tree-vrp.c (intersect_range_with_nonzero_bits): Same.
14332 (register_edge_assert_for_2): Same.
14333 (determine_value_range_1): Same.
14334 * tree.c (get_range_pos_neg): Same.
14335 * vr-values.c (vr_values::get_lattice_entry): Same.
14336 (vr_values::update_value_range): Same.
14337 (simplify_conversion_using_ranges): Same.
14339 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
14341 * gimple-ssa-warn-alloca.c (alloca_call_type): Use
14342 get_range_query instead of query argument.
14343 (pass_walloca::execute): Enable and disable global ranger.
14345 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
14347 * gimple-ssa-evrp.c (rvrp_folder::rvrp_folder): Call
14349 (rvrp_folder::~rvrp_folder): Call disable_ranger.
14350 (hybrid_folder::hybrid_folder): Call enable_ranger.
14351 (hybrid_folder::~hybrid_folder): Call disable_ranger.
14353 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
14355 * function.c (allocate_struct_function): Set cfun->x_range_query.
14356 * function.h (struct function): Declare x_range_query.
14357 (get_range_query): New.
14358 (get_global_range_query): New.
14359 * gimple-range-cache.cc (ssa_global_cache::ssa_global_cache):
14360 Remove call to safe_grow_cleared.
14361 * gimple-range.cc (get_range_global): New.
14362 (gimple_range_global): Move from gimple-range.h.
14363 (get_global_range_query): New.
14364 (global_range_query::range_of_expr): New.
14365 (enable_ranger): New.
14366 (disable_ranger): New.
14367 * gimple-range.h (gimple_range_global): Move to gimple-range.cc.
14368 (class global_range_query): New.
14369 (enable_ranger): New.
14370 (disable_ranger): New.
14371 * gimple-ssa-evrp.c (evrp_folder::~evrp_folder): Rename
14372 dump_all_value_ranges to dump.
14373 * tree-vrp.c (vrp_prop::finalize): Same.
14374 * value-query.cc (range_query::dump): New.
14375 * value-query.h (range_query::dump): New.
14376 * vr-values.c (vr_values::dump_all_value_ranges): Rename to...
14377 (vr_values::dump): ...this.
14378 * vr-values.h (class vr_values): Rename dump_all_value_ranges to
14379 dump and make virtual.
14381 2021-05-26 Uroš Bizjak <ubizjak@gmail.com>
14383 * config/i386/i386.c (ix86_autovectorize_vector_modes):
14384 Add V4QImode and V16QImode for TARGET_SSE2.
14385 * doc/sourcebuild.texi (Vector-specific attributes):
14386 Add vect64 and vect32 description.
14388 2021-05-26 Bernd Edlinger <bernd.edlinger@hotmail.de>
14390 * gimple-range-gori.cc (range_def_chain::register_dependency):
14391 Resize m_def_chain when needed.
14393 2021-05-26 Christophe Lyon <christophe.lyon@linaro.org>
14395 * config/arm/mve.md (mve_vaddvq_<supf><mode>): Prefix with '@'.
14396 * config/arm/neon.md (reduc_plus_scal_<mode>): Move to ..
14397 * config/arm/vec-common.md: .. here. Add support for MVE.
14399 2021-05-26 Jakub Jelinek <jakub@redhat.com>
14401 * config/epiphany/epiphany.c (epiphany_print_operand_address): Remove
14403 * config/microblaze/microblaze.c (microblaze_legitimize_address,
14405 microblaze_option_override, print_operand): Likewise.
14406 * config/microblaze/microblaze.md (call_internal_plt,
14407 call_value_intern_plt, call_value_intern): Likewise.
14408 * config/arm/aout.h (ASM_OUTPUT_ALIGN): Likewise.
14409 * config/iq2000/iq2000.md (call_internal1, call_value_internal1,
14410 call_value_multiple_internal1): Likewise.
14411 * config/bfin/bfin.c (symbolic_reference_mentioned_p): Likewise.
14413 2021-05-26 Jan-Benedict Glaw <jbglaw@lug-owl.de>
14415 * config/arc/arc.c (arc_address_cost, arc_print_operand_address,
14416 arc_ccfsm_advance, symbolic_reference_mentioned_p,
14417 arc_raw_symbolic_reference_mentioned_p): Remove register
14420 2021-05-26 Jakub Jelinek <jakub@redhat.com>
14423 * omp-low.c: Include omp-offload.h.
14424 (create_omp_child_function): If current_function_decl has
14425 "omp declare target" attribute and is_gimple_omp_offloaded,
14426 remove that attribute from the copy of attribute list and
14427 add "omp target entrypoint" attribute instead.
14428 (lower_omp_target): Mark .omp_data_sizes.* and .omp_data_kinds.*
14429 variables for offloading if in omp_maybe_offloaded_ctx.
14430 * omp-offload.c (pass_omp_target_link::execute): Nullify second
14431 argument to GOMP_target_data_ext in offloaded code.
14433 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
14435 * config/csky/csky.c (csky_can_change_mode_class): Delete.
14436 For csky, HF/SF mode use the low bits of VREGS.
14438 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
14440 * gimplify.c (gimplify_decl_expr): Do not clear TREE_READONLY on a
14441 DECL which is a reference for OMP.
14443 2021-05-26 Martin Liska <mliska@suse.cz>
14445 PR gcov-profile/100751
14446 * doc/gcov.texi: Document that __gcov_dump can be called just
14447 once and that __gcov_reset resets run-time counters.
14449 2021-05-26 Martin Liska <mliska@suse.cz>
14451 * doc/install.texi: Port relevant part from install-old.texi
14452 and re-generate list of CPUs and systems.
14454 2021-05-26 Martin Liska <mliska@suse.cz>
14456 * Makefile.in: Remove it.
14457 * doc/include/fdl.texi: Update next/previous chapters.
14458 * doc/install.texi: Likewise.
14459 * doc/install-old.texi: Removed.
14461 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
14463 * config/csky/csky.c (ck810_legitimate_index_p): Support
14464 "base + index" with DF mode.
14465 * config/csky/constraints.md ("Y"): New constraint for memory operands
14466 without index register.
14467 * config/csky/csky_insn_fpuv2.md (fpuv3_movdf): Use "Y" instead of "m"
14468 when mov between memory and general registers, and lower their priority.
14469 * config/csky/csky_insn_fpuv3.md (fpuv2_movdf): Likewise.
14471 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
14473 * config/csky/csky.c (TARGET_PROMOTE_PROTOTYPES): Delete.
14475 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
14477 * config/csky/csky.md (untyped_call): Emit clobber for return
14478 registers to mark them used.
14480 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
14482 * config/csky/csky.md (cskyv2_sextend_ldbs): New.
14484 2021-05-26 Andrew Pinski <apinski@marvell.com>
14486 * match.pd (x < 0 ? ~y : y): New patterns.
14488 2021-05-26 Andrew Pinski <apinski@marvell.com>
14490 * match.pd (A?CST1:CST2): Add simplifcations for A?0:+-1, A?+-1:0,
14491 A?POW2:0 and A?0:POW2.
14493 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14495 * gimple-range-gori.cc (class logical_stmt_cache): Delete
14496 (logical_stmt_cache::logical_stmt_cache ): Delete.
14497 (logical_stmt_cache::~logical_stmt_cache): Delete.
14498 (logical_stmt_cache::cache_entry::dump): Delete.
14499 (logical_stmt_cache::get_range): Delete.
14500 (logical_stmt_cache::cached_name ): Delete.
14501 (logical_stmt_cache::same_cached_name): Delete.
14502 (logical_stmt_cache::cacheable_p): Delete.
14503 (logical_stmt_cache::slot_diagnostics ): Delete.
14504 (logical_stmt_cache::dump): Delete.
14505 (gori_compute_cache::gori_compute_cache): Delete.
14506 (gori_compute_cache::~gori_compute_cache): Delete.
14507 (gori_compute_cache::compute_operand_range): Delete.
14508 (gori_compute_cache::cache_stmt): Delete.
14509 * gimple-range-gori.h (gori_compute::compute_operand_range): Remove
14511 (class gori_compute_cache): Delete.
14513 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14515 * gimple-range.cc (fold_using_range::range_of_range_op): Use m_gori
14517 (fold_using_range::range_of_address): Adjust.
14518 (fold_using_range::range_of_phi): Adjust.
14519 * gimple-range.h (class fur_source): Adjust.
14520 (fur_source::fur_source): Adjust.
14522 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14524 * gimple-range-gori.cc (gori_compute::expr_range_at_stmt): Rename
14525 from expr_range_in_bb and adjust.
14526 (gori_compute::compute_name_range_op): Adjust.
14527 (gori_compute::optimize_logical_operands): Adjust.
14528 (gori_compute::compute_logical_operands_in_chain): Adjust.
14529 (gori_compute::compute_operand1_range): Adjust.
14530 (gori_compute::compute_operand2_range): Adjust.
14531 (ori_compute_cache::cache_stmt): Adjust.
14532 * gimple-range-gori.h (gori_compute): Rename prototype.
14534 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14536 * gimple-range.cc (gimple_ranger::range_of_expr): Non-null should be
14537 checked only after range_of_stmt, not range_on_entry.
14538 (gimple_ranger::range_on_entry): Check for non-null in any
14539 predecessor block, if it is not already non-null.
14540 (gimple_ranger::range_on_exit): DOnt check for non-null after
14541 range on entry call.
14542 (gimple_ranger::dump_bb): New. Split from dump.
14543 (gimple_ranger::dump): Adjust.
14544 * gimple-range.h (class gimple_ranger): Adjust.
14546 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14548 * gimple-range-cache.cc (struct range_timestamp): Delete.
14549 (class temporal_cache): Adjust.
14550 (temporal_cache::get_timestamp): Delete.
14551 (temporal_cache::set_dependency): Delete.
14552 (temporal_cache::temporal_value): Adjust.
14553 (temporal_cache::current_p): Take dependencies as params.
14554 (temporal_cache::set_timestamp): Adjust.
14555 (temporal_cache::set_always_current): Adjust.
14556 (ranger_cache::get_non_stale_global_range): Adjust.
14557 (ranger_cache::register_dependency): Delete.
14558 * gimple-range-cache.h (class range_cache): Adjust.
14560 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14562 * gimple-range-gori.cc (range_def_chain::range_def_chain): init
14564 (range_def_chain::~range_def_chain): Dispose of obstack rather than
14565 each individual bitmap.
14566 (range_def_chain::set_import): New.
14567 (range_def_chain::get_imports): New.
14568 (range_def_chain::chain_import_p): New.
14569 (range_def_chain::register_dependency): Rename from build_def_chain
14571 (range_def_chain::def_chain_in_bitmap_p): New.
14572 (range_def_chain::add_def_chain_to_bitmap): New.
14573 (range_def_chain::has_def_chain): Just check first depenedence.
14574 (range_def_chain::get_def_chain): Process imports, use generic
14575 register_dependency routine.
14576 (range_def_chain::dump): New.
14577 (gori_map::gori_map): Allocate import list.
14578 (gori_map::~gori_map): Release imports.
14579 (gori_map::exports): Check for past allocated block size.
14580 (gori_map::imports): New.
14581 (gori_map::def_chain_in_export_p): Delete.
14582 (gori_map::is_import_p): New.
14583 (gori_map::maybe_add_gori): Handle imports.
14584 (gori_map::dump): Adjust output, add imports.
14585 (gori_compute::has_edge_range_p): Remove def_chain_in_export call.
14586 (gori_export_iterator::gori_export_iterator): New.
14587 (gori_export_iterator::next): New.
14588 (gori_export_iterator::get_name): New.
14589 * gimple-range-gori.h (range_def_chain): Add imports and direct
14590 dependecies via struct rdc.
14591 (range_def_chain::depend1): New.
14592 (range_def_chain::depend2): New.
14593 (class gori_map): Adjust.
14594 (FOR_EACH_GORI_IMPORT_NAME): New.
14595 (FOR_EACH_GORI_EXPORT_NAME): New.
14596 (class gori_export_iterator): New.
14598 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14600 * gimple-range-cache.cc (ranger_cache::ranger_cache): Move initial
14601 export cache filling to here.
14602 * gimple-range-gori.cc (gori_compute::gori_compute) : From Here.
14604 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14606 * gimple-range-gori.cc (range_def_chain): Move to gimple-range-gori.h.
14607 (gori_map): Move to gimple-range-gori.h.
14608 (gori_compute::gori_compute): Adjust.
14609 (gori_compute::~gori_compute): Delete.
14610 (gori_compute::compute_operand_range_switch): Adjust.
14611 (gori_compute::compute_operand_range): Adjust.
14612 (gori_compute::compute_logical_operands): Adjust.
14613 (gori_compute::has_edge_range_p ): Adjust.
14614 (gori_compute::set_range_invariant): Delete.
14615 (gori_compute::dump): Adjust.
14616 (gori_compute::outgoing_edge_range_p): Adjust.
14617 * gimple-range-gori.h (class range_def_chain): Relocate here.
14618 (class gori_map): Relocate here.
14619 (class gori_compute): Inherit from gori_map, and adjust.
14621 2021-05-25 Aldy Hernandez <aldyh@redhat.com>
14623 * value-range.cc (range_tests_legacy): Use
14624 build_nonstandard_integer_type instead of int and short.
14626 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
14628 * gimplify.c (gimplify_decl_expr): Clear TREE_READONLY on the DECL
14629 when really creating an initialization statement for it.
14631 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
14633 * tree-inline.c (setup_one_parameter): Fix thinko in new condition.
14635 2021-05-25 Kito Cheng <kito.cheng@sifive.com>
14637 * config/riscv/riscv.h (ASM_SPEC): Pass -mno-relax.
14639 2021-05-25 Martin Liska <mliska@suse.cz>
14641 PR tree-optimization/92860
14643 * optc-save-gen.awk: Remove exceptions.
14645 2021-05-25 Martin Liska <mliska@suse.cz>
14647 * asan.h (sanitize_coverage_p): New function.
14648 * doc/extend.texi: Document it.
14649 * fold-const.c (fold_range_test): Use sanitize_flags_p
14650 instead of flag_sanitize_coverage.
14651 (fold_truth_andor): Likewise.
14652 * sancov.c: Likewise.
14653 * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
14654 * ipa-inline.c (sanitize_attrs_match_for_inline_p): Handle
14655 -fsanitize-coverage when inlining.
14657 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
14659 * config/csky/csky-modes.def : Fix copyright.
14661 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
14663 * config/csky/csky-modes.def : Amend copyright.
14664 * config/csky/csky_insn_fpuv2.md : Likewise.
14665 * config/csky/csky_insn_fpuv3.md : Likewise.
14667 2021-05-25 Richard Biener <rguenther@suse.de>
14669 PR middle-end/100727
14670 * calls.c (initialize_argument_information): Explicitely test
14671 for WITH_SIZE_EXPR.
14672 * gimple-expr.c (mark_addressable): Skip outer WITH_SIZE_EXPR.
14674 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
14676 * config/csky/csky.h (FRAME_POINTER_REGNUM): Use
14677 HARD_FRAME_POINTER_REGNUM and FRAME_POINTER_REGNUM instead of
14678 the signle definition. The signle definition may not work well
14679 at simplify_subreg_regno().
14680 (HARD_FRAME_POINTER_REGNUM): New.
14681 (ELIMINABLE_REGS): Add for HARD_FRAME_POINTER_REGNUM.
14682 * config/csky/csky.c (get_csky_live_regs, csky_can_eliminate,
14683 csky_initial_elimination_offset, csky_expand_prologue,
14684 csky_expand_epilogue): Add for HARD_FRAME_POINTER_REGNUM.
14686 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
14688 * config/csky/csky.c (csky_option_override):
14689 Init csky_arch_isa_features[] in advance, so TARGET_DSP
14690 and TARGET_DIV can be set well.
14692 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
14694 * config/csky/constraints.md ("l", "h"): Delete.
14695 * config/csky/csky.h (reg_class, REG_CLASS_NAMES,
14696 REG_CLASS_CONTENTS): Delete LO_REGS and HI_REGS.
14697 * config/csky/csky.c (regno_reg_classm,
14698 csky_secondary_reload, csky_register_move_cost):
14699 Use HILO_REGS instead of LO_REGS and HI_REGS.
14701 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
14703 * config/csky/constraints.md ("W"): New constriant for mem operand
14704 with base reg, index register.
14705 ("Q"): Renamed and modified "csky_valid_fpuv2_mem_operand" to
14706 "csky_valid_mem_constraint_operand" to deal with both "Q" and "W"
14708 ("Dv"): New constraint for const double value that can be used at
14710 * config/csky/csky-modes.def (HFmode): New mode.
14711 * config/csky/csky-protos.h (csky_valid_fpuv2_mem_operand): Rename
14712 to "csky_valid_mem_constraint_operand" and support new constraint
14714 (csky_get_movedouble_length): New.
14715 (fpuv3_output_move): New.
14716 (fpuv3_const_double): New.
14717 * config/csky/csky.c (csky_option_override): New arch CK860 with fpv3.
14718 (decompose_csky_address): Refine.
14719 (csky_print_operand): New "CONST_DOUBLE" operand.
14720 (csky_output_move): Support fpv3 instructions.
14721 (csky_get_movedouble_length): New.
14722 (fpuv3_output_move): New.
14723 (fpuv3_const_double): New.
14724 (csky_emit_compare): Cover float comparsion.
14725 (csky_emit_compare_float): Refine.
14726 (csky_vaild_fpuv2_mem_operand): Rename to
14727 "csky_valid_mem_constraint_operand" and support new constraint "W".
14728 (ck860_rtx_costs): New.
14729 (csky_rtx_costs): Add the cost calculation of CK860.
14730 (regno_reg_class): New vregs for fpuv3.
14731 (csky_dbx_regno): Likewise.
14732 (csky_cpu_cpp_builtins): New builtin macro for fpuv3.
14733 (csky_conditional_register_usage): Suporrot fpuv3.
14734 (csky_dwarf_register_span): Suporrot fpuv3.
14735 (csky_init_builtins, csky_mangle_type): Support "__fp16" type.
14736 (ck810_legitimate_index_p): Support fp16.
14737 * config/csky/csky.h (TARGET_TLS): ADD CK860.
14738 (CSKY_VREG_P, CSKY_VREG_LO_P, CSKY_VREG_HI_P): Support fpuv3.
14739 (TARGET_SINGLE_FPU): Support fpuv3.
14740 (TARGET_SUPPORT_FPV3): New.
14741 (FIRST_PSEUDO_REGISTER): Change to 202 to hold the new fpuv3 registers.
14742 (FIXED_REGISTERS, CALL_REALLY_USED_REGISTERS, REGISTER_NAMES,
14743 REG_CLASS_CONTENTS): Support fpuv3.
14744 * config/csky/csky.md (movsf): Move to cksy_insn_fpu.md and refine.
14745 (csky_movsf_fpv2): Likewise.
14746 (ck801_movsf): Likewise.
14747 (csky_movsf): Likewise.
14749 (csky_movdf_fpv2): Likewise.
14750 (ck801_movdf): Likewise.
14751 (csky_movdf): Likewise.
14752 (movsicc): Refine. Use "comparison_operatior" instead of
14753 "ordered_comparison_operatior".
14754 (addsicc): Likewise.
14755 (CSKY_FIRST_VFP3_REGNUM, CSKY_LAST_VFP3_REGNUM): New constant.
14756 (call_value_internal_vh): New.
14757 * config/csky/csky_cores.def (CK860): New arch and cpu.
14762 * config/csky/csky_insn_fpu.md: Refactor. Separate all float patterns
14763 into emit-patterns and match-patterns, remain the emit-patterns here,
14764 and move the match-patterns to csky_insn_fpuv2.md or
14765 csky_insn_fpuv3.md.
14766 * config/csky/csky_insn_fpuv2.md: New file for fpuv2 instructions.
14767 * config/csky/csky_insn_fpuv3.md: New file and new patterns for fpuv3
14769 * config/csky/csky_isa.def (fcr): New.
14774 (CK860): New definition for ck860.
14775 * config/csky/csky_tables.opt (ck860): New processors ck860,
14776 ck860f. And new arch ck860.
14781 * config/csky/predicates.md (csky_float_comparsion_operator): Delete
14782 "geu", "gtu", "leu", "ltu", which will never appear at float comparison.
14783 * config/csky/t-csky-elf: Support 860.
14784 * config/csky/t-csky-linux: Likewise.
14785 * doc/md.texi: Add "Q" and "W" constraints for C-SKY.
14787 2021-05-24 Aaron Sawdey <acsawdey@linux.ibm.com>
14789 * config/rs6000/genfusion.pl (gen_logical_addsubf): Refactor to
14790 add generation of logical-add and add-logical fusion pairs.
14791 * config/rs6000/rs6000-cpus.def: Add new fusion to ISA 3.1 mask
14793 * config/rs6000/rs6000.c (rs6000_option_override_internal): Turn on
14794 logical-add and add-logical fusion by default.
14795 * config/rs6000/rs6000.opt: Add -mpower10-fusion-logical-add and
14796 -mpower10-fusion-add-logical options.
14797 * config/rs6000/fusion.md: Regenerate file.
14799 2021-05-24 Aldy Hernandez <aldyh@redhat.com>
14801 * value-range.cc (irange::legacy_equal_p): Check type when
14802 comparing VR_VARYING types.
14803 (range_tests_legacy): Test comparing VARYING ranges of different
14806 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
14808 * config/aarch64/aarch64.c (neoversen1_tunings):
14809 Enable AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
14811 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
14813 * config/aarch64/aarch64.c (aarch64_classify_symbol): Use GOT for
14814 extern weak symbols. Limit symbol offsets for non-GOT symbols with
14817 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
14819 * config/arm/neon.md (vec_load_lanesxi<mode>)
14820 (vec_store_lanexoi<mode>): Move ...
14821 * config/arm/vec-common.md: here.
14823 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
14825 * config/arm/neon.md (vec_load_lanesoi<mode>)
14826 (vec_store_lanesoi<mode>): Move ...
14827 * config/arm/vec-common.md: here.
14829 2021-05-24 liuhongt <hongtao.liu@intel.com>
14832 * config/i386/i386.c (ix86_gimple_fold_builtin): Replacing
14833 stmt with GIMPLE_NOP when lhs doesn't exist.
14835 2021-05-23 Uroš Bizjak <ubizjak@gmail.com>
14838 * config/i386/mmx.md (*push<VI_32:mode>2_rex64):
14839 New instruction pattern.
14840 (*push<VI_32:mode>2): Ditto.
14841 (push splitter for SSE registers): New splitter.
14843 2021-05-23 Andrew Pinski <apinski@marvell.com>
14845 * match.pd ((A & C) != 0 ? D : 0): Limit to non pointer types.
14847 2021-05-22 Aaron Sawdey <acsawdey@linux.ibm.com>
14849 * config/rs6000/genfusion.pl (gen_addadd): Fix incorrect attr types.
14850 * config/rs6000/fusion.md: Regenerate file.
14852 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
14854 * config/rs6000/genfusion.pl (gen_addadd): New function.
14855 * config/rs6000/fusion.md: Regenerate file.
14856 * config/rs6000/rs6000-cpus.def: Add
14857 OPTION_MASK_P10_FUSION_2ADD to masks.
14858 * config/rs6000/rs6000.c (rs6000_option_override_internal):
14859 Handle default value of OPTION_MASK_P10_FUSION_2ADD.
14860 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2add.
14862 2021-05-21 Jakub Jelinek <jakub@redhat.com>
14864 PR middle-end/99928
14865 * tree.h (OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET): Define.
14866 * gimplify.c (enum gimplify_omp_var_data): Fix up
14867 GOVD_MAP_HAS_ATTACHMENTS value, add GOVD_FIRSTPRIVATE_IMPLICIT.
14868 (omp_lastprivate_for_combined_outer_constructs): If combined target
14869 has GOVD_FIRSTPRIVATE_IMPLICIT set for the decl, change it to
14870 GOVD_MAP | GOVD_SEEN.
14871 (gimplify_scan_omp_clauses): Set GOVD_FIRSTPRIVATE_IMPLICIT for
14872 firstprivate clauses with OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT.
14873 (gimplify_adjust_omp_clauses): For firstprivate clauses with
14874 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT either clear that bit and
14875 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET too, or remove it and
14876 let it be replaced by implicit map clause.
14878 2021-05-21 Jakub Jelinek <jakub@redhat.com>
14880 PR middle-end/99928
14881 * gimplify.c (omp_lastprivate_for_combined_outer_constructs): New
14883 (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LASTPRIVATE>: Use it.
14884 (gimplify_omp_for): Likewise.
14886 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14888 PR middle-end/90115
14889 * omp-low.c (oacc_privatization_candidate_p): Reject 'static',
14890 'external' in blocks.
14892 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14894 PR middle-end/90115
14895 * flag-types.h (enum openacc_privatization): New.
14896 * params.opt (-param=openacc-privatization): New.
14897 * doc/invoke.texi (openacc-privatization): Document it.
14898 * omp-general.h (get_openacc_privatization_dump_flags): New
14900 * omp-low.c (oacc_privatization_candidate_p): Add diagnostics.
14901 * omp-offload.c (execute_oacc_device_lower)
14902 <IFN_UNIQUE_OACC_PRIVATE>: Re-work diagnostics.
14903 * target.def (goacc.adjust_private_decl): Add 'location_t'
14905 * doc/tm.texi: Regenerate.
14906 * config/gcn/gcn-protos.h (gcn_goacc_adjust_private_decl): Adjust.
14907 * config/gcn/gcn-tree.c (gcn_goacc_adjust_private_decl): Likewise.
14908 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl):
14909 Likewise. Preserve it for...
14910 (nvptx_goacc_expand_var_decl): ... use here.
14912 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14914 * doc/sourcebuild.texi (Other attributes): Document '__OPTIMIZE__'
14917 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14919 PR middle-end/90115
14920 * omp-low.c (oacc_privatization_candidate_p): New function.
14921 (oacc_privatization_scan_clause_chain)
14922 (oacc_privatization_scan_decl_chain): Use it. Also
14923 'gcc_checking_assert' that we're not seeing duplicates.
14925 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14927 PR middle-end/90115
14928 * omp-offload.c (execute_oacc_device_lower): Skip processing if no
14931 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14933 PR middle-end/90115
14934 * omp-offload.c (execute_oacc_device_lower): Explain.
14936 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14938 PR middle-end/90115
14939 * omp-offload.c (execute_oacc_device_lower)
14940 <IFN_UNIQUE_OACC_PRIVATE>: Diagnose and handle for 'level == -1'
14942 * internal-fn.c (expand_UNIQUE): Don't expect
14943 'IFN_UNIQUE_OACC_PRIVATE'.
14945 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14947 PR middle-end/90115
14948 * omp-low.c (lower_omp_for): Don't evaluate OpenMP 'for' clauses.
14950 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14952 PR middle-end/90115
14953 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl)
14954 (nvptx_goacc_expand_var_decl): Tighten.
14956 2021-05-21 Julian Brown <julian@codesourcery.com>
14957 Chung-Lin Tang <cltang@codesourcery.com>
14958 Thomas Schwinge <thomas@codesourcery.com>
14960 PR middle-end/90115
14961 * doc/tm.texi.in (TARGET_GOACC_EXPAND_VAR_DECL)
14962 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Add documentation hooks.
14963 * doc/tm.texi: Regenerate.
14964 * expr.c (expand_expr_real_1): Expand decls using the
14965 expand_var_decl OpenACC hook if defined.
14966 * internal-fn.c (expand_UNIQUE): Handle IFN_UNIQUE_OACC_PRIVATE.
14967 * internal-fn.h (IFN_UNIQUE_CODES): Add OACC_PRIVATE.
14968 * omp-low.c (omp_context): Add oacc_privatization_candidates
14970 (lower_oacc_reductions): Add PRIVATE_MARKER parameter. Insert
14972 (lower_oacc_head_tail): Add PRIVATE_MARKER parameter. Modify
14973 private marker's gimple call arguments, and pass it to
14974 lower_oacc_reductions.
14975 (oacc_privatization_scan_clause_chain)
14976 (oacc_privatization_scan_decl_chain, lower_oacc_private_marker):
14978 (lower_omp_for, lower_omp_target, lower_omp_1): Use these.
14979 * omp-offload.c (convert.h): Include.
14980 (oacc_loop_xform_head_tail): Treat private-variable markers like
14981 fork/join when transforming head/tail sequences.
14982 (struct var_decl_rewrite_info): Add struct.
14983 (oacc_rewrite_var_decl, is_sync_builtin_call): New functions.
14984 (execute_oacc_device_lower): Support rewriting gang-private
14985 variables using target hook, and fix up addr_expr and var_decl
14987 * target.def (adjust_private_decl, expand_var_decl): New hooks.
14988 * config/gcn/gcn-protos.h (gcn_goacc_adjust_gangprivate_decl):
14990 (gcn_goacc_adjust_private_decl): ...this.
14991 * config/gcn/gcn-tree.c (gcn_goacc_adjust_gangprivate_decl):
14993 (gcn_goacc_adjust_private_decl): ...this. Add LEVEL parameter.
14994 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_GANGPRIVATE_DECL): Rename
14995 definition using gcn_goacc_adjust_gangprivate_decl...
14996 (TARGET_GOACC_ADJUST_PRIVATE_DECL): ...to this, using
14997 gcn_goacc_adjust_private_decl.
14998 * config/nvptx/nvptx.c (tree-pretty-print.h): Include.
14999 (gang_private_shared_size): New global variable.
15000 (gang_private_shared_align): Likewise.
15001 (gang_private_shared_sym): Likewise.
15002 (gang_private_shared_hmap): Likewise.
15003 (nvptx_option_override): Initialize these.
15004 (nvptx_file_end): Output gang_private_shared_sym.
15005 (nvptx_goacc_adjust_private_decl, nvptx_goacc_expand_var_decl):
15007 (nvptx_set_current_function): Clear gang_private_shared_hmap.
15008 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Define hook.
15009 (TARGET_GOACC_EXPAND_VAR_DECL): Likewise.
15011 2021-05-21 H.J. Lu <hjl.tools@gmail.com>
15013 * config/i386/i386-modes.def (MAX_BITSIZE_MODE_ANY_INT): Removed.
15015 2021-05-21 Richard Biener <rguenther@suse.de>
15016 H.J. Lu <hjl.tools@gmail.com>
15018 PR middle-end/90773
15019 * expr.c (expand_constructor): Elide expand_constructor if
15020 move by pieces is preferred.
15022 2021-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15024 * config/aarch64/aarch64-builtins.c (aarch64_call_properties):
15025 Take a flag and mode value as arguments.
15026 (aarch64_modifies_global_state_p): Likewise.
15027 (aarch64_reads_global_state_p): Likewise.
15028 (aarch64_could_trap_p): Likewise.
15029 (aarch64_get_attributes): Likewise.
15030 (aarch64_init_simd_builtins): Adjust callsite of above.
15031 (aarch64_init_fcmla_laneq_builtins): Use aarch64_get_attributes to get
15032 function attributes to apply to builtins.
15033 (aarch64_init_crc32_builtins): Likewise.
15034 (aarch64_init_builtin_rsqrt): Likewise.
15036 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
15038 * config/rs6000/rs6000.md (define_attr "type"): Add types for fusion.
15039 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10): Use new fusion types.
15040 (gen_2logical): Use new fusion types.
15041 * config/rs6000/fusion.md: Regenerate.
15043 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
15046 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
15047 Handle V4QI and V2HI modes.
15048 (ix86_expand_sse_movcc): Ditto.
15049 * config/i386/mmx.md (*<sat_plusminus:insn><VI_32:mode>3):
15050 New instruction pattern.
15051 (*eq<VI_32:mode>3): Ditto.
15052 (*gt<VI_32:mode>3): Ditto.
15053 (*xop_pcmov_<VI_32:mode>): Ditto.
15054 (mmx_pblendvb32): Ditto.
15055 (mmx_pblendvb64): Rename from mmx_pblendvb.
15056 (vec_cmp<VI_32:mode><VI_32:mode>): New expander.
15057 (vec_cmpu<VI_32:mode><VI_32:mode>): Ditto.
15058 (vcond<VI_32:mode><VI_32:mode>): Ditto.
15059 (vcondu<VI_32:mode><VI_32:mode>): Ditto.
15060 (vcond_mask_<VI_32:mode><VI_32:mode>): Ditto.
15062 2021-05-21 Jakub Jelinek <jakub@redhat.com>
15064 PR tree-optimization/94589
15065 * tree-ssa-phiopt.c (spaceship_replacement): For integral rhs1 and
15066 rhs2, treat x <= 4 equivalently to x < 5 etc. In cmp1 and cmp2 (if
15067 not the same as cmp3) treat <= the same as < and >= the same as >.
15068 Don't require that cond2_phi_edge is true edge, instead take
15069 false/true edges into account based on cmp1/cmp2 comparison kinds.
15071 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
15074 * config/i386/mmx.md (SMAXMIN_MMXMODEI): New mode iterator.
15075 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Macroize expander
15076 from <smaxmin:code>v4hi3> and <smaxmin:code><MMXMODE14:mode>3
15077 using SMAXMIN_MMXMODEI mode iterator.
15078 (*<smaxmin:code>v4qi3): New insn pattern.
15079 (*<smaxmin:code>v2hi3): Ditto.
15080 (SMAXMIN_VI_32): New mode iterator.
15081 (<smaxmin:code><SMAXMIN_VI_32>mode3): New expander.
15082 (UMAXMIN_MMXMODEI): New mode iterator.
15083 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Macroize expander
15084 from <umaxmin:code>v8qi3> and <umaxmin:code><MMXMODE24:mode>3
15085 using UMAXMIN_MMXMODEI mode iterator.
15086 (*<umaxmin:code>v4qi3): New insn pattern.
15087 (*<umaxmin:code>v2hi3): Ditto.
15088 (UMAXMIN_VI_32): New mode iterator.
15089 (<umaxmin:code><UMAXMIN_VI_32>mode3): New expander.
15090 (abs<VI_32:mode>2): New insn pattern.
15091 (ssse3_abs<MMXMODEI:mode>2, abs<MMXMODEI:mode>2): Move from ...
15092 * config/i386/sse.md: ... here.
15094 2021-05-20 Clement Chigot <clement.chigot@atos.net>
15095 David Edelsohn <dje.gcc@gmail.com>
15097 * collect2.c (scan_prog_file): Issue non-fatal warning for
15100 2021-05-20 Jonathan Wakely <jwakely@redhat.com>
15102 * doc/invoke.texi (-Wno-c++11-extensions)
15103 (-Wno-c++14-extensions, -Wno-c++17-extensions)
15104 (-Wno-c++20-extensions, -Wno-c++23-extensions): Document
15107 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
15109 * config/c6x/c6x.c (c6x_output_file_unwind): Use dwarf_debuginfo_p.
15110 * config/darwin.c (darwin_override_options): Likewise.
15111 * config/i386/cygming.h (DBX_REGISTER_NUMBER): Likewise.
15112 * config/i386/darwin.h (DBX_REGISTER_NUMBER): Likewise.
15113 (DWARF2_FRAME_REG_OUT): Likewise.
15114 * config/mips/mips.c (mips_output_filename): Likewise.
15115 * config/rs6000/rs6000.c (rs6000_xcoff_declare_function_name):
15117 (rs6000_dbx_register_number): Likewise.
15118 * dbxout.c: Include flags.h.
15119 * dwarf2cfi.c (cfi_label_required_p): Likewise.
15120 (dwarf2out_do_frame): Likewise.
15121 * except.c: Include flags.h.
15122 * final.c (dwarf2_debug_info_emitted_p): Likewise.
15123 (final_scan_insn_1): Likewise.
15124 * flags.h (dwarf_debuginfo_p): New function declaration.
15125 * opts.c (dwarf_debuginfo_p): New function definition.
15126 * targhooks.c (default_debug_unwind_info): Use dwarf_debuginfo_p.
15127 * toplev.c (process_options): Likewise.
15129 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
15131 * common.opt: Change type to support bitmasks.
15132 * flag-types.h (enum debug_info_type): Rename enumerator constants.
15133 (NO_DEBUG): New bitmask.
15134 (DBX_DEBUG): Likewise.
15135 (DWARF2_DEBUG): Likewise.
15136 (XCOFF_DEBUG): Likewise.
15137 (VMS_DEBUG): Likewise.
15138 (VMS_AND_DWARF2_DEBUG): Likewise.
15139 * flags.h (debug_set_to_format): New function declaration.
15140 (debug_set_count): Likewise.
15141 (debug_set_names): Likewise.
15142 * opts.c (debug_type_masks): Array of bitmasks for debug formats.
15143 (debug_set_to_format): New function definition.
15144 (debug_set_count): Likewise.
15145 (debug_set_names): Likewise.
15146 (set_debug_level): Update access to debug_type_names.
15147 * toplev.c: Likewise.
15149 2021-05-20 Martin Sebor <msebor@redhat.com>
15151 PR middle-end/100684
15152 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Handle C++ lambda.
15154 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
15157 * config/i386/i386.md (isa): Remove x64_bmi.
15158 (enabled): Remove x64_bmi.
15159 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
15160 Remove general register alternative.
15161 (*andnot<VI_32:mode>3): Ditto.
15162 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
15163 (*<any_logic:code><VI_32:mode>3): Ditto.
15165 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
15167 * config/arm/arm.c: Include head files tree-vectorizer.h and
15170 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
15173 * config/i386/mmx.md (Yv_Yw): Revert adding V4QI and V2HI modes.
15174 (*<plusminus:insn><VI_32:mode>3): Use Yw instad of <Yv_Yw> constrint.
15175 (<s>mulv4hi3_highpart): New expander.
15176 (*<s>mulv2hi3_highpart): New insn pattern.
15177 (<s>mulv2hi3_higpart): New expander.
15178 (*<any_shift:insn>v2hi3): New insn pattern.
15179 (<any_shift:insn>v2hi3): New expander.
15180 * config/i386/sse.md (smulhrsv2hi3): New expander.
15181 (*smulhrsv2hi3): New insn pattern.
15183 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
15185 * doc/invoke.texi (vect-inner-loop-cost-factor): Document new
15187 * params.opt (vect-inner-loop-cost-factor): New.
15188 * targhooks.c (default_add_stmt_cost): Replace hardcoded factor
15189 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR, include head file
15190 tree-vectorizer.h and its required ones.
15191 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Replace
15192 hardcoded factor 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR.
15193 * config/arm/arm.c (arm_add_stmt_cost): Likewise.
15194 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
15195 * config/rs6000/rs6000.c (rs6000_add_stmt_cost): Likewise.
15196 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
15198 (_loop_vec_info::_loop_vec_info): Init inner_loop_cost_factor.
15199 * tree-vectorizer.h (_loop_vec_info): Add inner_loop_cost_factor.
15200 (LOOP_VINFO_INNER_LOOP_COST_FACTOR): New macro.
15202 2021-05-20 Christophe Lyon <christophe.lyon@linaro.org>
15203 Torbjörn Svensson <torbjorn.svensson@st.com>
15206 * doc/cpp.texi (Common Predefined Macros): Document __FILE_NAME__.
15208 2021-05-20 Jakub Jelinek <jakub@redhat.com>
15210 PR middle-end/99928
15211 * gimplify.c (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LINEAR>: For
15212 explicit linear clause when combined with target, make it map(tofrom:)
15213 instead of no clause or firstprivate.
15215 2021-05-20 Jakub Jelinek <jakub@redhat.com>
15217 PR tree-optimization/94589
15218 * match.pd ((X & Y) == X -> (X & ~Y) == 0): Simplify even in presence
15219 of integral conversions.
15221 2021-05-19 Andrew MacLeod <amacleod@redhat.com>
15223 * gimple-range.cc (fur_source::get_operand): New.
15224 (gimple_range_fold): Delete.
15225 (fold_using_range::fold_stmt): Move from gimple_ranger::calc_stmt.
15226 (fold_using_range::range_of_range_op): Move from gimple_ranger.
15227 (fold_using_range::range_of_address): Ditto.
15228 (fold_using_range::range_of_phi): Ditto.
15229 (fold_using_range::range_of_call): Ditto.
15230 (fold_using_range::range_of_builtin_ubsan_call): Move from
15231 range_of_builtin_ubsan_call.
15232 (fold_using_range::range_of_builtin_call): Move from
15233 range_of_builtin_call.
15234 (gimple_ranger::range_of_builtin_call): Delete.
15235 (fold_using_range::range_of_cond_expr): Move from gimple_ranger.
15236 (gimple_ranger::fold_range_internal): New.
15237 (gimple_ranger::range_of_stmt): Use new fold_using_range API.
15238 (fold_using_range::range_of_ssa_name_with_loop_info): Move from
15239 gimple_ranger. Improve ranges of SSA_NAMES when possible.
15240 * gimple-range.h (gimple_ranger): Remove various range_of routines.
15241 (class fur_source): New.
15242 (class fold_using_range): New.
15243 (fur_source::fur_source): New.
15245 * vr-values.c (vr_values::extract_range_basic): Use fold_using_range
15246 instead of range_of_builtin_call.
15248 2021-05-19 Jonathan Wakely <jwakely@redhat.com>
15250 * doc/cpp.texi (Common Predefined Macros): Update documentation
15251 for the __GXX_EXPERIMENTAL_CXX0X__ macro.
15253 2021-05-19 Alex Coplan <alex.coplan@arm.com>
15256 * config/arm/arm.md (nonsecure_call_internal): Always ensure
15257 callee's address is in a register.
15259 2021-05-19 Geng Qi <gengqi@linux.alibaba.com>
15261 * common/config/riscv/riscv-common.c
15262 (riscv_subset_list::parsing_subset_version): Properly parse the letter
15264 (riscv_subset_list::parse_std_ext,
15265 riscv_subset_list::parse_multiletter_ext): To handle errors generated
15266 in riscv_subset_list::parsing_subset_version.
15268 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
15270 * config/aarch64/aarch64-simd.md: Use "neon_move_narrow_q"
15271 type attribute in patterns generating XTN(2).
15273 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
15275 * config/aarch64/aarch64-simd.md (aarch64_simd_vec_pack_trunc_<mode>):
15276 Remove as duplicate of...
15277 (aarch64_xtn<mode>): This.
15278 (aarch64_xtn2<mode>_le): Move position in file.
15279 (aarch64_xtn2<mode>_be): Move position in file.
15280 (aarch64_xtn2<mode>): Move position in file.
15281 (vec_pack_trunc_<mode>): Define as an expander.
15283 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
15285 * config/aarch64/aarch64-simd-builtins.def: Split builtin
15286 generation for aarch64_<sur>q<r>shr<u>n_n<mode> pattern into
15287 separate scalar and vector generators.
15288 * config/aarch64/aarch64-simd.md
15289 (aarch64_<sur>q<r>shr<u>n_n<mode>): Define as an expander and
15291 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_le): This and...
15292 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_be): This.
15293 * config/aarch64/iterators.md: Define SD_HSDI iterator.
15295 2021-05-19 Jonathn Wright <jonathan.wright@arm.com>
15297 * config/aarch64/aarch64-simd.md: Use UNSPEC_SQXTUN instead
15299 * config/aarch64/iterators.md: Remove UNSPEC_SQXTUN2.
15301 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
15303 * config/aarch64/aarch64-simd.md (aarch64_<sur>q<r>shr<u>n2_n<mode>):
15304 Implement as an expand emitting a big/little endian
15305 instruction pattern.
15306 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_le): Define.
15307 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_be): Define.
15309 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
15311 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn2<mode>):
15312 Implement as an expand emitting a big/little endian
15313 instruction pattern.
15314 (aarch64_<sur><addsub>hn2<mode>_insn_le): Define.
15315 (aarch64_<sur><addsub>hn2<mode>_insn_be): Define.
15316 * config/aarch64/iterators.md: Remove UNSPEC_[R]ADDHN2 and
15317 UNSPEC_[R]SUBHN2 unspecs and ADDSUBHN2 iterator.
15319 2021-05-19 Richard Biener <rguenther@suse.de>
15321 PR middle-end/100672
15322 * fold-const.c (fold_negate_expr_1): Use element_precision.
15323 (negate_expr_p): Likewise.
15325 2021-05-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
15327 * config/aarch64/iterators.md (SVE_PRED_LOAD): New iterator.
15328 (pred_load): New int attribute.
15329 * config/aarch64/aarch64-sve.md
15330 (aarch64_load_<ANY_EXTEND:optab><SVE_HSDI:mode><SVE_PARTIAL_I:mode>): Use
15331 SVE_PRED_LOAD enum iterator and corresponding pred_load attribute.
15332 * config/aarch64/aarch64-sve-builtins-base.cc (expand): Update call to
15333 code_for_aarch64_load.
15335 2021-05-19 Richard Biener <rguenther@suse.de>
15337 * cfgexpand.c (discover_nonconstant_array_refs_r): Make
15338 sure TARGET_MEM_REF bases are expanded as memory.
15339 * tree-ssa-operands.c (operands_scanner::get_tmr_operands):
15340 Do not mark TARGET_MEM_REF bases addressable.
15341 * tree-ssa.c (non_rewritable_mem_ref_base): Handle
15342 TARGET_MEM_REF bases as never rewritable.
15343 * gimple-walk.c (walk_stmt_load_store_addr_ops): Do not
15344 walk TARGET_MEM_REF bases as address-takens.
15345 * tree-ssa-dce.c (ref_may_be_aliased): Handle TARGET_MEM_REF.
15347 2021-05-19 Richard Biener <rguenther@suse.de>
15349 * builtins.c (get_object_alignment_1): Strip outer
15351 * tree-dfa.c (get_ref_base_and_extent): Handle outer
15352 WITH_SIZE_EXPR for size processing and process the
15354 * tree-ssa-alias.c (ao_ref_base_alias_set): Strip
15355 outer WITH_SIZE_EXPR.
15356 (ao_ref_base_alias_ptr_type): Likewise.
15357 (refs_may_alias_p_2): Allow WITH_SIZE_EXPR in ref->ref
15358 and handle that accordingly, stripping it for the
15359 core alias workers.
15360 * tree.c (get_base_address): Handle WITH_SIZE_EXPR by
15361 looking through it instead of returning NULL.
15363 2021-05-19 Jakub Jelinek <jakub@redhat.com>
15365 PR middle-end/100576
15366 * builtins.c (check_read_access): Convert bound to size_type_node if
15369 2021-05-19 Richard Biener <rguenther@suse.de>
15371 * tree-cfg.c (verify_types_in_gimple_min_lval): Inline...
15372 (verify_types_in_gimple_reference): ... here. Sanitize.
15373 (verify_gimple_call): Verify references in LHS and arguments.
15374 (verify_gimple_assign_single): Reject WITH_SIZE_EXPR.
15376 2021-05-19 Uroš Bizjak <ubizjak@gmail.com>
15378 * config/i386/i386.h (VALID_INT_MODE_P):
15379 Add V8QI, V4HI and V2SI modes for TARGET_64BIT.
15380 * config/i386/i386.md (isa): Add x64_bmi.
15381 (enabled): Handle x64_bmi.
15382 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
15383 Add alternative using 64bit general registers.
15384 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
15386 2021-05-19 Jakub Jelinek <jakub@redhat.com>
15388 PR middle-end/99928
15389 * tree.h (OMP_MASTER_COMBINED): Define.
15390 * gimplify.c (gimplify_scan_omp_clauses): Rewrite lastprivate
15391 handling for outer combined/composite constructs to a loop.
15392 Handle lastprivate on combined target.
15393 (gimplify_expr): Formatting fix.
15395 2021-05-19 Xionghu Luo <luoxhu@linux.ibm.com>
15397 * passes.def: Add sink_code pass before store_merging.
15398 * tree-ssa-sink.c (pass_sink_code:clone): New.
15400 2021-05-18 Bill Schmidt <wschmidt@linux.ibm.com>
15402 * config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Remove call to
15403 rs6000_special_adjust_field_align_p.
15404 * config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
15405 * config/rs6000/rs6000-call.c (rs6000_function_arg_boundary):
15406 Remove ABI warning.
15407 (rs6000_function_arg): Likewise.
15408 * config/rs6000/rs6000-protos.h
15409 (rs6000_special_adjust_field_align_p): Remove prototype.
15410 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
15412 * config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Remove call to
15413 rs6000_special_adjust_field_align_p.
15415 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
15418 * config/i386/i386.h (VALID_SSE2_REG_MODE):
15419 Add V4QI and V2HI modes.
15420 (VALID_INT_MODE_P): Ditto.
15421 * config/i386/mmx.md (VI_32): New mode iterator.
15422 (mmxvecsize): Handle V4QI and V2HI.
15424 (mov<VI_32:mode>): New expander.
15425 (*mov<mode>_internal): New insn pattern.
15426 (movmisalign<VI_32:mode>): New expander.
15427 (neg<VI_32:mode>): New expander.
15428 (<plusminus:insn><VI_32:mode>3): New expander.
15429 (*<plusminus:insn><VI_32:mode>3): New insn pattern.
15430 (mulv2hi3): New expander.
15431 (*mulv2hi3): New insn pattern.
15432 (one_cmpl<VI_32:mode>2): New expander.
15433 (*andnot<VI_32:mode>3): New insn pattern.
15434 (<any_logic:code><VI_32:mode>3): New expander.
15435 (*<any_logic:code><VI_32:mode>3): New insn pattern.
15437 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
15439 * config/i386/sse.md (<any_extend:insn>v4qiv4di2):
15440 Fix a mode mismatch with operand 1.
15442 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
15445 * config/i386/i386-expand.c (split_double_mode): Return
15446 temporary register when simplify_gen_subreg fails with
15447 the high half od the paradoxical subreg.
15449 2021-05-18 Richard Biener <rguenther@suse.de>
15451 * cfgexpand.c (expand_one_var): Pass in forced_stack_var
15452 and honor it when expanding.
15453 (expand_used_vars_for_block): Pass through forced_stack_var.
15454 (expand_used_vars): Likewise.
15455 (discover_nonconstant_array_refs_r): Set bits in
15456 forced_stack_vars instead of marking vars TREE_ADDRESSABLE.
15457 (avoid_type_punning_on_regs): Likewise.
15458 (discover_nonconstant_array_refs): Likewise.
15459 (pass_expand::execute): Create and pass down forced_stack_var
15460 bitmap. For parameters and returns temporarily set
15461 TREE_ADDRESSABLE when expand_function_start.
15463 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
15465 * doc/sourcebuild.texi: Document 'dg-note'.
15467 2021-05-18 Tobias Burnus <tobias@codesourcery.com>
15470 * configure: Regenerate.
15471 * configure.ac (BUILD_CFLAG, BUILD_CXXFLAGS): Add $(CFLAGS-$@).
15473 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
15475 * gimple.h (is_gimple_omp_oacc): Tighten.
15476 * omp-low.c (check_omp_nesting_restrictions): Adjust.
15478 2021-05-18 Richard Biener <rguenther@suse.de>
15480 * tree-ssa-operands.c (mark_address_taken): Simplify.
15482 2021-05-18 Martin Liska <mliska@suse.cz>
15484 * config/gcn/mkoffload.c (STR): Redefine.
15485 * config/i386/intelmic-mkoffload.c (STR): Likewise.
15486 * config/nvptx/mkoffload.c (STR): Likewise.
15488 2021-05-18 Martin Liska <mliska@suse.cz>
15490 * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
15491 Use startswith function instead of strncmp.
15492 * common/config/bfin/bfin-common.c (bfin_handle_option): Likewise.
15493 * common/config/riscv/riscv-common.c (riscv_subset_list::parse): Likewise.
15494 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_type): Likewise.
15495 * config/aarch64/aarch64.c (aarch64_process_one_target_attr): Likewise.
15496 * config/alpha/alpha.c (alpha_elf_section_type_flags): Likewise.
15497 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
15498 * config/arm/arm.c (arm_file_start): Likewise.
15499 (arm_valid_target_attribute_rec): Likewise.
15500 (thumb1_md_asm_adjust): Likewise.
15501 * config/arm/driver-arm.c (host_detect_local_cpu): Likewise.
15502 * config/avr/avr.c (STR_PREFIX_P): Likewise.
15503 (avr_set_current_function): Likewise.
15504 (avr_handle_addr_attribute): Likewise.
15505 (avr_asm_output_aligned_decl_common): Likewise.
15506 (avr_asm_named_section): Likewise.
15507 (avr_section_type_flags): Likewise.
15508 (avr_asm_select_section): Likewise.
15509 * config/c6x/c6x.c (c6x_in_small_data_p): Likewise.
15510 (c6x_section_type_flags): Likewise.
15511 * config/darwin-c.c (darwin_cfstring_ref_p): Likewise.
15512 (darwin_objc_declare_unresolved_class_reference): Likewise.
15513 (darwin_objc_declare_class_definition): Likewise.
15514 * config/darwin.c (indirect_data): Likewise.
15515 (darwin_encode_section_info): Likewise.
15516 (darwin_objc2_section): Likewise.
15517 (darwin_objc1_section): Likewise.
15518 (machopic_select_section): Likewise.
15519 (darwin_globalize_label): Likewise.
15520 (darwin_label_is_anonymous_local_objc_name): Likewise.
15521 (darwin_asm_named_section): Likewise.
15522 (darwin_asm_output_dwarf_offset): Likewise.
15523 * config/frv/frv.c (frv_string_begins_with): Likewise.
15524 (frv_in_small_data_p): Likewise.
15525 * config/gcn/mkoffload.c (STR): Likewise.
15527 * config/i386/i386-builtins.c (get_builtin_code_for_version): Likewise.
15528 * config/i386/i386-options.c (ix86_option_override_internal): Likewise.
15529 * config/i386/i386.c (x86_64_elf_section_type_flags): Likewise.
15530 (ix86_md_asm_adjust): Likewise.
15531 * config/i386/intelmic-mkoffload.c (STR): Likewise.
15532 * config/i386/winnt.c (i386_pe_asm_named_section): Likewise.
15533 (i386_pe_file_end): Likewise.
15534 * config/ia64/ia64.c (ia64_in_small_data_p): Likewise.
15535 (ia64_section_type_flags): Likewise.
15536 * config/mips/driver-native.c (host_detect_local_cpu): Likewise.
15537 * config/mips/mips.c (mips_handle_interrupt_attr): Likewise.
15538 (mips16_stub_function_p): Likewise.
15539 (mips_function_rodata_section): Likewise.
15540 * config/msp430/msp430.c (msp430_mcu_name): Likewise.
15541 (msp430_function_section): Likewise.
15542 (msp430_section_type_flags): Likewise.
15543 (msp430_expand_helper): Likewise.
15544 * config/nios2/nios2.c (nios2_small_section_name_p): Likewise.
15545 (nios2_valid_target_attribute_rec): Likewise.
15546 * config/nvptx/mkoffload.c (process): Likewise.
15548 * config/pa/som.h: Likewise.
15549 * config/pdp11/pdp11.c (pdp11_output_ident): Likewise.
15550 * config/riscv/riscv.c (riscv_elf_select_rtx_section): Likewise.
15551 * config/rs6000/rs6000.c (VTABLE_NAME_P): Likewise.
15552 (rs6000_inner_target_options): Likewise.
15553 * config/s390/driver-native.c (s390_host_detect_local_cpu): Likewise.
15554 * config/sparc/driver-sparc.c (host_detect_local_cpu): Likewise.
15555 * config/vax/vax.c (vax_output_int_move): Likewise.
15556 * config/vms/vms-ld.c (startswith): Likewise.
15557 (process_args): Likewise.
15559 * config/vms/vms.c: Likewise.
15561 2021-05-18 Jakub Jelinek <jakub@redhat.com>
15563 PR rtl-optimization/100590
15564 * regcprop.c (copyprop_hardreg_forward_1): Only DCE dead sets if
15565 they are NONJUMP_INSN_P.
15567 2021-05-18 Jakub Jelinek <jakub@redhat.com>
15570 * function.c (push_dummy_function): Set DECL_ARTIFICIAL and
15571 DECL_ASSEMBLER_NAME on the fn_decl.
15573 2021-05-18 Jakub Jelinek <jakub@redhat.com>
15575 PR tree-optimization/94589
15576 * tree-ssa-phiopt.c (spaceship_replacement): Pattern match
15577 phi result used in (res & ~1) == 0 comparison as res >= 0 as
15578 res == 2 would be UB with -ffinite-math-only.
15580 2021-05-18 Martin Liska <mliska@suse.cz>
15582 * Makefile.in: genversion.o should depend on DATESTAMP.
15584 2021-05-18 Claudiu Zissulescu <claziss@synopsys.com>
15586 * config/arc/simdext.md (negv2si2): Remove round bracket.
15588 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
15590 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Define
15591 _Bool as macro expanding to _Bool.
15593 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
15596 * tree.c (build_reference_type_for_mode)
15597 (build_pointer_type_for_mode): Pick pointer mode if MODE argument
15599 (build_reference_type, build_pointer_type): Invoke
15600 build_*_type_for_mode with VOIDmode.
15602 2021-05-17 Andrew MacLeod <amacleod@redhat.com>
15604 PR tree-optimization/100512
15605 * gimple-range-cache.cc (ranger_cache::set_global_range): Mark const
15606 and non-zero pointer ranges as invariant.
15607 * gimple-range.cc (gimple_ranger::range_of_stmt): Remove pointer
15608 processing from here.
15610 2021-05-17 Tom de Vries <tdevries@suse.de>
15613 * config/nvptx/nvptx-protos.h (nvptx_output_atomic_insn): Declare
15614 * config/nvptx/nvptx.c (nvptx_output_barrier)
15615 (nvptx_output_atomic_insn): New function.
15616 (nvptx_print_operand): Add support for 'B'.
15617 * config/nvptx/nvptx.md: Use nvptx_output_atomic_insn for atomic
15620 2021-05-17 Aldy Hernandez <aldyh@redhat.com>
15622 PR tree-optimization/100349
15623 * vr-values.c (bounds_of_var_in_loop): Bail if scev returns
15626 2021-05-17 Tamar Christina <tamar.christina@arm.com>
15628 * config/aarch64/driver-aarch64.c (DEFAULT_ARCH): New.
15629 (host_detect_local_cpu): Use it.
15631 2021-05-17 Martin Liska <mliska@suse.cz>
15633 * doc/invoke.texi: Add 2 missing dots.
15635 2021-05-17 Marius Hillenbrand <mhillen@linux.ibm.com>
15637 PR bootstrap/100552
15638 * configure.ac: Replace pattern substitution with call to sed.
15639 * configure: Regenerate.
15641 2021-05-17 Richard Biener <rguenther@suse.de>
15643 PR middle-end/100582
15644 * tree.c (array_at_struct_end_p): Get to the base of the
15645 reference before looking for the underlying decl.
15647 2021-05-17 Joern Rennecke <joern.rennecke@embecosm.com>
15649 * genoutput.c (validate_insn_alternatives) Make "wrong number of
15650 alternatives" message more specific, and remove assumption on where
15653 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
15655 * config/arm/iterators.md (V16): New iterator.
15656 (VH_cvtto): New iterator.
15657 (v_cmp_result): Added V4HF and V8HF support.
15658 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result>): Use VDQWH.
15659 (vcond<mode><mode>): Likewise.
15660 (vcond_mask_<mode><v_cmp_result>): Likewise.
15661 (vcond<VH_cvtto><mode>): New expander.
15663 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
15665 * config/arm/arm-protos.h (arm_expand_vector_compare): Update
15667 * config/arm/arm.c (arm_expand_vector_compare): Add support for
15669 (arm_expand_vcond): Likewise.
15670 * config/arm/iterators.md (supf): Remove VCMPNEQ_S, VCMPEQQ_S,
15671 VCMPEQQ_N_S, VCMPNEQ_N_S.
15672 (VCMPNEQ, VCMPEQQ, VCMPEQQ_N, VCMPNEQ_N): Remove.
15673 * config/arm/mve.md (@mve_vcmp<mve_cmp_op>q_<mode>): Add '@' prefix.
15674 (@mve_vcmp<mve_cmp_op>q_f<mode>): Likewise.
15675 (@mve_vcmp<mve_cmp_op>q_n_f<mode>): Likewise.
15676 (@mve_vpselq_<supf><mode>): Likewise.
15677 (@mve_vpselq_f<mode>"): Likewise.
15678 * config/arm/neon.md (vec_cmp<mode><v_cmp_result): Enable for MVE
15679 and move to vec-common.md.
15680 (vec_cmpu<mode><mode>): Likewise.
15681 (vcond<mode><mode>): Likewise.
15682 (vcond<V_cvtto><mode>): Likewise.
15683 (vcondu<mode><v_cmp_result>): Likewise.
15684 (vcond_mask_<mode><v_cmp_result>): Likewise.
15685 * config/arm/unspecs.md (VCMPNEQ_U, VCMPNEQ_S, VCMPEQQ_S)
15686 (VCMPEQQ_N_S, VCMPNEQ_N_S, VCMPEQQ_U, CMPEQQ_N_U, VCMPNEQ_N_U)
15687 (VCMPGEQ_N_S, VCMPGEQ_S, VCMPGTQ_N_S, VCMPGTQ_S, VCMPLEQ_N_S)
15688 (VCMPLEQ_S, VCMPLTQ_N_S, VCMPLTQ_S, VCMPCSQ_N_U, VCMPCSQ_U)
15689 (VCMPHIQ_N_U, VCMPHIQ_U): Remove.
15690 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result): Moved
15692 (vec_cmpu<mode><mode>): Likewise.
15693 (vcond<mode><mode>): Likewise.
15694 (vcond<V_cvtto><mode>): Likewise.
15695 (vcondu<mode><v_cmp_result>): Likewise.
15696 (vcond_mask_<mode><v_cmp_result>): Likewise. Added unsafe math
15699 2021-05-17 liuhongt <hongtao.liu@intel.com>
15702 * config/i386/i386.c (ix86_gimple_fold_builtin): Use
15703 gsi_insert_seq_before instead.
15705 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
15707 * doc/sourcebuild.texi (arm_qbit_ok): Rename into...
15708 (arm_sat_ok): ...this.
15710 2021-05-17 Martin Liska <mliska@suse.cz>
15712 * lto-wrapper.c (merge_flto_options): Factor out a new function.
15713 (merge_and_complain): Use it.
15714 (run_gcc): Merge also linker command line -flto=foo argument
15717 2021-05-16 Christophe Lyon <christophe.lyon@linaro.org>
15719 * config/arm/arm.h (CPP_SPEC): Remove error message about
15720 -mlittle-endian/-mbig-endian conflict.
15722 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
15724 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
15725 __ROP_PROTECT__ if -mrop-protect is selected.
15727 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
15729 * config/rs6000/rs6000-internal.h (rs6000_stack): Add
15730 rop_hash_save_offset and rop_hash_size.
15731 * config/rs6000/rs6000-logue.c (rs6000_stack_info): Compute
15732 rop_hash_size and rop_hash_save_offset.
15733 (debug_stack_info): Dump rop_hash_save_offset and rop_hash_size.
15734 (rs6000_emit_prologue): Emit hashst[p] in prologue.
15735 (rs6000_emit_epilogue): Emit hashchk[p] in epilogue.
15736 * config/rs6000/rs6000.md (unspec): Add UNSPEC_HASHST and
15738 (hashst): New define_insn.
15739 (hashchk): Likewise.
15741 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
15743 * config/rs6000/rs6000.c (rs6000_option_override_internal):
15744 Disable shrink wrap when inserting ROP-protect instructions.
15745 * config/rs6000/rs6000.opt (mrop-protect): New option.
15746 (mprivileged): Likewise.
15747 * doc/invoke.texi: Document mrop-protect and mprivileged.
15749 2021-05-15 Hans-Peter Nilsson <hp@axis.com>
15751 * reorg.c (fill_slots_from_thread): Reinstate code typoed out in
15754 2021-05-15 Martin Jambor <mjambor@suse.cz>
15757 2021-05-13 Martin Jambor <mjambor@suse.cz>
15759 PR tree-optimization/100453
15760 * tree-sra.c (sra_modify_assign): All const base accesses do not
15761 need refreshing, not just those from decl_pool.
15762 (sra_modify_assign): Do not refresh into a const base decl.
15764 2021-05-15 Jakub Jelinek <jakub@redhat.com>
15766 PR rtl-optimization/100342
15767 * regcprop.c (copy_value): When copying a source reg in a wider
15768 mode than it has recorded for the value, adjust recorded destination
15769 mode too or punt if !REG_CAN_CHANGE_MODE_P.
15771 2021-05-14 Jason Merrill <jason@redhat.com>
15773 * intl.h: Add comments.
15775 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15777 * config/aarch64/aarch64-simd.md
15778 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Split into...
15779 (aarch64_sqdmlsl2_lane<mode>_internal): ... This...
15780 (aarch64_sqdmlal2_lane<mode>_internal): ... And this.
15781 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>_internal): Split into ...
15782 (aarch64_sqdmlsl2_laneq<mode>_internal): ... This...
15783 (aarch64_sqdmlal2_laneq<mode>_internal): ... And this.
15784 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>_internal): Split into...
15785 (aarch64_sqdmlsl2_n<mode>_internal): ... This...
15786 (aarch64_sqdmlal2_n<mode>_internal): ... And this.
15788 2021-05-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15791 * config/arm/arm_neon.h (vtst_s8): Replace call to vtst builtin with it's
15792 boolean logic equivalent.
15793 (vtst_s16): Likewise.
15794 (vtst_s32): Likewise.
15795 (vtst_u8): Likewise.
15796 (vtst_u16): Likewise.
15797 (vtst_u32): Likewise.
15798 (vtst_p8): Likewise.
15799 (vtst_p16): Likewise.
15800 (vtstq_s8): Likewise.
15801 (vtstq_s16): Likewise.
15802 (vtstq_s32): Likewise.
15803 (vtstq_u8): Likewise.
15804 (vtstq_u16): Likewise.
15805 (vtstq_u32): Likewise.
15806 (vtstq_p8): Likewise.
15807 (vtstq_p16): Likewise.
15808 * config/arm/arm_neon_builtins.def: Remove entry for vtst.
15809 * config/arm/neon.md (neon_vtst<mode>): Remove pattern.
15811 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15813 * config/aarch64/aarch64-simd.md (aarch64_sqdmlal2<mode>): Merge into...
15814 (aarch64_sqdml<SBINQOPS:as>l2<mode>): ... This.
15815 (aarch64_sqdmlsl2<mode>): Delete.
15816 (aarch64_sqdmlal2_lane<mode>): Merge this...
15817 (aarch64_sqdmlsl2_lane<mode>): ... And this...
15818 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>): ... Into this.
15819 (aarch64_sqdmlal2_laneq<mode>): Merge this...
15820 (aarch64_sqdmlsl2_laneq<mode>): ... And this...
15821 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>): ... Into this.
15822 (aarch64_sqdmlal2_n<mode>): Merge this...
15823 (aarch64_sqdmlsl2_n<mode>): ... And this...
15824 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>): ... Into this.
15826 2021-05-13 Martin Sebor <msebor@redhat.com>
15828 PR middle-end/100574
15829 * builtins.c (access_ref::get_ref): Improve detection of PHIs with
15830 all null arguments.
15832 2021-05-13 Martin Sebor <msebor@redhat.com>
15834 PR tree-optimization/93100
15835 PR middle-end/98583
15836 * tree-ssa-uninit.c (check_defs): Exclude intrinsic functions that
15837 don't modify referenced objects.
15839 2021-05-13 Martin Jambor <mjambor@suse.cz>
15841 PR tree-optimization/100453
15842 * tree-sra.c (sra_modify_assign): All const base accesses do not
15843 need refreshing, not just those from decl_pool.
15844 (sra_modify_assign): Do not refresh into a const base decl.
15846 2021-05-13 Martin Liska <mliska@suse.cz>
15848 * tree-ssa-dom.c: Remove m_simplifier.
15850 2021-05-13 Richard Earnshaw <rearnsha@arm.com>
15853 * config/arm/arm.c (arm_canonicalize_comparison): Correctly
15854 canonicalize DImode inequality comparisons against the
15855 maximum integral value.
15857 2021-05-13 Jakub Jelinek <jakub@redhat.com>
15859 PR tree-optimization/98856
15860 * config/i386/i386.c (ix86_shift_rotate_cost): Add CODE argument.
15861 Expect V2DI and V4DI arithmetic right shifts to be emulated.
15862 (ix86_rtx_costs, ix86_add_stmt_cost): Adjust ix86_shift_rotate_cost
15864 * config/i386/i386-expand.c (expand_vec_perm_2perm_interleave,
15865 expand_vec_perm_2perm_pblendv): New functions.
15866 (ix86_expand_vec_perm_const_1): Use them.
15867 * config/i386/sse.md (ashr<mode>3<mask_name>): Rename to ...
15868 (<mask_codefor>ashr<mode>3<mask_name>): ... this.
15869 (ashr<mode>3): New define_expand with VI248_AVX512BW iterator.
15870 (ashrv4di3): New define_expand.
15871 (ashrv2di3): Change condition to TARGET_SSE2, handle !TARGET_XOP
15872 and !TARGET_AVX512VL expansion.
15874 2021-05-13 Uroš Bizjak <ubizjak@gmail.com>
15877 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Force mode
15878 sizes < 16 to a register when constructing vpcmov pattern.
15879 * config/i386/mmx.md (*xop_pcmov_<mode>): Use MMXMODE124 mode.
15881 2021-05-13 Martin Liska <mliska@suse.cz>
15883 * gcov-io.c (gcov_write_block): Remove.
15884 (gcov_write_words): Likewise.
15885 (gcov_read_words): Re-implement using gcov_read_bytes.
15886 (gcov_allocate): Remove.
15887 (GCOV_BLOCK_SIZE): Likewise.
15888 (struct gcov_var): Remove most of the fields.
15889 (gcov_position): Implement with ftell.
15890 (gcov_rewrite): Remove setting of start and offset fields.
15891 (from_file): Re-format.
15892 (gcov_open): Remove setbuf call. It should not be needed.
15893 (gcov_close): Remove internal buffer handling.
15894 (gcov_magic): Use __builtin_bswap32.
15895 (gcov_write_counter): Use directly gcov_write_unsigned.
15896 (gcov_write_string): Use direct fwrite and do not round
15898 (gcov_seek): Use directly fseek.
15899 (gcov_write_tag): Use gcov_write_unsigned directly.
15900 (gcov_write_length): Likewise.
15901 (gcov_write_tag_length): Likewise.
15902 (gcov_read_bytes): Use directly fread.
15903 (gcov_read_unsigned): Use gcov_read_words.
15904 (gcov_read_counter): Likewise.
15905 (gcov_read_string): Use gcov_read_bytes.
15906 * gcov-io.h (GCOV_WORD_SIZE): Adjust to reflect
15907 that size is not in bytes, but words (4B).
15908 (GCOV_TAG_FUNCTION_LENGTH): Likewise.
15909 (GCOV_TAG_ARCS_LENGTH): Likewise.
15910 (GCOV_TAG_ARCS_NUM): Likewise.
15911 (GCOV_TAG_COUNTER_LENGTH): Likewise.
15912 (GCOV_TAG_COUNTER_NUM): Likewise.
15913 (GCOV_TAG_SUMMARY_LENGTH): Likewise.
15915 2021-05-13 liuhongt <hongtao.liu@intel.com>
15918 * config/i386/sse.md (ssedoublevecmode): Add attribute for
15919 V64QI/V32HI/V16SI/V4DI.
15920 (ssehalfvecmode): Add attribute for V2DI/V2DF.
15921 (*vec_concatv4si_0): Extend to VI124_128.
15922 (*vec_concat<mode>_0): New pre-reload splitter.
15923 * config/i386/predicates.md (movq_parallel): New predicate.
15925 2021-05-13 Alexandre Oliva <oliva@adacore.com>
15927 * targhooks.c (default_zero_call_used_regs): Retry using
15928 successfully-zeroed registers as sources.
15930 2021-05-12 Tobias Burnus <tobias@codesourcery.com>
15932 * omp-low.c (finish_taskreg_scan): Use the proper detach decl.
15934 2021-05-12 Aldy Hernandez <aldyh@redhat.com>
15937 * gimple-range.cc (range_of_builtin_call): Skip out on
15938 processing __builtin_clz when varying.
15940 2021-05-12 Tom de Vries <tdevries@suse.de>
15943 * config/nvptx/nvptx-opts.h (enum ptx_version): New enum.
15944 * config/nvptx/nvptx.c (nvptx_file_start): Print .version according
15945 to ptx_version_option.
15946 * config/nvptx/nvptx.h (TARGET_PTX_6_3): Define.
15947 * config/nvptx/nvptx.md (define_insn "nvptx_shuffle<mode>")
15948 (define_insn "nvptx_vote_ballot"): Use sync variant for
15950 * config/nvptx/nvptx.opt (ptx_version): Add enum.
15951 (mptx): Add option.
15952 * doc/invoke.texi (Nvidia PTX Options): Add mptx item.
15954 2021-05-12 Richard Biener <rguenther@suse.de>
15956 PR tree-optimization/100566
15957 * tree-ssa-sccvn.c (dominated_by_p_w_unex): Properly handle
15958 allow_back for all edge queries.
15960 2021-05-12 liuhongt <hongtao.liu@intel.com>
15963 * config/i386/sse.md (<sse4_1_avx2>_pblendvb): Add
15964 splitters for pblendvb of NOT mask register.
15966 2021-05-12 Richard Biener <rguenther@suse.de>
15968 PR tree-optimization/100519
15969 * tree-ssa-reassoc.c (can_associate_p): Split into...
15970 (can_associate_op_p): ... this
15971 (can_associate_type_p): ... and this.
15972 (is_reassociable_op): Call can_associate_op_p.
15973 (break_up_subtract_bb): Call the appropriate predicates.
15974 (reassociate_bb): Likewise.
15976 2021-05-12 Martin Liska <mliska@suse.cz>
15978 * lto-wrapper.c (merge_and_complain): Merge -flto=arg options.
15979 (run_gcc): Use -flto argument detection for merged
15982 2021-05-12 Martin Liska <mliska@suse.cz>
15984 * lto-wrapper.c (print_lto_docs_link): New function.
15985 (run_gcc): Print warning about missing job server detection
15986 after we know NR of partitions. Do the same for -flto{,=1}.
15987 * opts.c (get_option_html_page): Support -flto option.
15989 2021-05-12 Martin Liska <mliska@suse.cz>
15991 * lto-wrapper.c (get_options_from_collect_gcc_options): Change
15993 (append_option): Remove.
15994 (find_option): Rework to use the vector type.
15995 (remove_option): Remove.
15996 (merge_and_complain): Use vectors for cl_decoded_option data
15998 (append_compiler_options): Likewise.
15999 (append_diag_options): Likewise.
16000 (append_linker_options): Likewise.
16001 (append_offload_options): Likewise.
16002 (compile_offload_image): Likewise.
16003 (compile_images_for_offload_targets): Likewise.
16004 (find_and_merge_options): Likewise.
16005 (run_gcc): Likewise.
16007 2021-05-12 Bernd Edlinger <bernd.edlinger@hotmail.de>
16010 * dwarf2out.c (dwarf2out_finish): Set
16011 have_multiple_function_sections with multi-range text_section.
16013 2021-05-12 Martin Liska <mliska@suse.cz>
16015 PR bootstrap/100560
16016 * Makefile.in: Remove version.h from linker command line.
16018 2021-05-12 Richard Biener <rguenther@suse.de>
16020 PR middle-end/100547
16021 * rtl.h (rtvec_alloc): Make argument size_t.
16022 * rtl.c (rtvec_alloc): Verify the count is less than INT_MAX.
16024 2021-05-12 Jakub Jelinek <jakub@redhat.com>
16026 PR middle-end/100508
16027 * cfgexpand.c (expand_debug_expr): For DEBUG_EXPR_DECL with vector
16028 type, don't reuse DECL_RTL if it has different mode, instead force
16029 creation of a new DEBUG_EXPR.
16031 2021-05-12 Jakub Jelinek <jakub@redhat.com>
16032 Marc Glisse <marc.glisse@inria.fr>
16034 PR tree-optimization/94589
16035 * match.pd ((X & Y) == X -> (X & ~Y) == 0,
16036 (X | Y) == Y -> (X & ~Y) == 0): New GIMPLE simplifications.
16038 2021-05-12 Uroš Bizjak <ubizjak@gmail.com>
16041 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Handle V2SF mode.
16042 * config/i386/mmx.md (MMXMODE124): New mode iterator.
16044 (mmxintvecmode): New mode attribute.
16045 (mmxintvecmodelower): Ditto.
16046 (*mmx_maskcmpv2sf3_comm): New insn pattern.
16047 (*mmx_maskcmpv2sf3): Ditto.
16048 (vec_cmpv2sfv2si): New expander.
16049 (vcond<V2FI:mode>v2si): Ditto.
16050 (mmx_vlendvps): New insn pattern.
16051 (vcond<MMXMODE124:mode><MMXMODEI:mode>): Also handle V2SFmode.
16052 (vcondu<MMXMODE124:mode><MMXMODEI:mode>): Ditto.
16053 (vcond_mask_<mode><mmxintvecmodelower>): Ditto.
16055 2021-05-11 Martin Sebor <msebor@redhat.com>
16057 PR middle-end/21433
16058 * expr.c (expand_expr_real_1): Replace unreachable code with an assert.
16060 2021-05-11 Richard Biener <rguenther@suse.de>
16062 * gimple-fold.c (gimple_fold_call): Do not call
16063 maybe_fold_reference on call arguments or the static chain.
16064 (fold_stmt_1): Do not call maybe_fold_reference on GIMPLE_ASM
16067 2021-05-11 Martin Liska <mliska@suse.cz>
16069 * builtins.def (DEF_HSAIL_BUILTIN): Remove.
16070 (DEF_HSAIL_ATOMIC_BUILTIN): Likewise.
16071 (DEF_HSAIL_SAT_BUILTIN): Likewise.
16072 (DEF_HSAIL_INTR_BUILTIN): Likewise.
16073 (DEF_HSAIL_CVT_ZEROI_SAT_BUILTIN): Likewise.
16074 * doc/frontends.texi: Remove BRIG.
16075 * doc/install.texi: Likewise.
16076 * doc/invoke.texi: Likewise.
16077 * doc/standards.texi: Likewise.
16078 * brig-builtins.def: Removed.
16079 * brig/ChangeLog: Removed.
16080 * brig/Make-lang.in: Removed.
16081 * brig/brig-builtins.h: Removed.
16082 * brig/brig-c.h: Removed.
16083 * brig/brig-lang.c: Removed.
16084 * brig/brigfrontend/brig-arg-block-handler.cc: Removed.
16085 * brig/brigfrontend/brig-atomic-inst-handler.cc: Removed.
16086 * brig/brigfrontend/brig-basic-inst-handler.cc: Removed.
16087 * brig/brigfrontend/brig-branch-inst-handler.cc: Removed.
16088 * brig/brigfrontend/brig-cmp-inst-handler.cc: Removed.
16089 * brig/brigfrontend/brig-code-entry-handler.cc: Removed.
16090 * brig/brigfrontend/brig-code-entry-handler.h: Removed.
16091 * brig/brigfrontend/brig-comment-handler.cc: Removed.
16092 * brig/brigfrontend/brig-control-handler.cc: Removed.
16093 * brig/brigfrontend/brig-copy-move-inst-handler.cc: Removed.
16094 * brig/brigfrontend/brig-cvt-inst-handler.cc: Removed.
16095 * brig/brigfrontend/brig-fbarrier-handler.cc: Removed.
16096 * brig/brigfrontend/brig-function-handler.cc: Removed.
16097 * brig/brigfrontend/brig-function.cc: Removed.
16098 * brig/brigfrontend/brig-function.h: Removed.
16099 * brig/brigfrontend/brig-inst-mod-handler.cc: Removed.
16100 * brig/brigfrontend/brig-label-handler.cc: Removed.
16101 * brig/brigfrontend/brig-lane-inst-handler.cc: Removed.
16102 * brig/brigfrontend/brig-machine.c: Removed.
16103 * brig/brigfrontend/brig-machine.h: Removed.
16104 * brig/brigfrontend/brig-mem-inst-handler.cc: Removed.
16105 * brig/brigfrontend/brig-module-handler.cc: Removed.
16106 * brig/brigfrontend/brig-queue-inst-handler.cc: Removed.
16107 * brig/brigfrontend/brig-seg-inst-handler.cc: Removed.
16108 * brig/brigfrontend/brig-signal-inst-handler.cc: Removed.
16109 * brig/brigfrontend/brig-to-generic.cc: Removed.
16110 * brig/brigfrontend/brig-to-generic.h: Removed.
16111 * brig/brigfrontend/brig-util.cc: Removed.
16112 * brig/brigfrontend/brig-util.h: Removed.
16113 * brig/brigfrontend/brig-variable-handler.cc: Removed.
16114 * brig/brigfrontend/hsa-brig-format.h: Removed.
16115 * brig/brigfrontend/phsa.h: Removed.
16116 * brig/brigspec.c: Removed.
16117 * brig/config-lang.in: Removed.
16118 * brig/gccbrig.texi: Removed.
16119 * brig/lang-specs.h: Removed.
16120 * brig/lang.opt: Removed.
16122 2021-05-11 Richard Biener <rguenther@suse.de>
16125 * ipa-param-manipulation.c
16126 (ipa_param_body_adjustments::modify_call_stmt): Avoid
16127 altering SSA_NAME_DEF_STMT by adjusting the calls LHS
16128 via gimple_call_lhs_ptr.
16130 2021-05-11 Alex Coplan <alex.coplan@arm.com>
16133 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear):
16134 Avoid emitting CFA adjusts on the sp if we have the fp.
16136 2021-05-11 Richard Sandiford <richard.sandiford@arm.com>
16138 * config/aarch64/iterators.md (VMUL_CHANGE_NLANES): Delete.
16139 (VMULD): New iterator.
16140 (VCOND): Handle V4HF and V8HF.
16141 (VCONQ): Fix entry for V2SF.
16142 * config/aarch64/aarch64-simd.md (mul_lane<mode>3): Use VMULD
16143 instead of VMUL. Use a 64-bit vector mode for the indexed operand.
16144 (*aarch64_mul3_elt_<vswap_width_name><mode>): Merge with...
16145 (mul_laneq<mode>3): ...this define_insn. Use VMUL instead of VDQSF.
16146 Use a 128-bit vector mode for the indexed operand. Use stype for
16147 the scheduling type.
16149 2021-05-11 Richard Biener <rguenther@suse.de>
16151 * gimple-fold.c (maybe_fold_reference): Only return
16152 is_gimple_min_invariant values.
16154 2021-05-11 Richard Biener <rguenther@suse.de>
16156 PR middle-end/100509
16157 * gimple-fold.c (fold_gimple_assign): Only call
16158 get_symbol_constant_value on register type symbols.
16160 2021-05-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
16161 Joe Ramsay <joe.ramsay@arm.com>
16164 * config/arm/arm_mve.h (__arm_vstrwq_scatter_offset): Fix wrong arguments.
16165 (__arm_vcmpneq): Remove duplicate definition.
16166 (__arm_vstrwq_scatter_offset_p): Likewise.
16167 (__arm_vmaxq_x): Likewise.
16168 (__arm_vmlsdavaq): Likewise.
16169 (__arm_vmlsdavaxq): Likewise.
16170 (__arm_vmlsdavq_p): Likewise.
16171 (__arm_vmlsdavxq_p): Likewise.
16172 (__arm_vrmlaldavhaq): Likewise.
16173 (__arm_vstrbq_p): Likewise.
16174 (__arm_vstrbq_scatter_offset): Likewise.
16175 (__arm_vstrbq_scatter_offset_p): Likewise.
16176 (__arm_vstrdq_scatter_offset): Likewise.
16177 (__arm_vstrdq_scatter_offset_p): Likewise.
16178 (__arm_vstrdq_scatter_shifted_offset): Likewise.
16179 (__arm_vstrdq_scatter_shifted_offset_p): Likewise.
16181 2021-05-11 Jakub Jelinek <jakub@redhat.com>
16183 PR middle-end/100471
16184 * omp-low.c (lower_omp_task_reductions): For OMP_TASKLOOP, if data
16185 is 0, bypass the reduction loop including
16186 GOMP_taskgroup_reduction_unregister call.
16188 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
16190 * config/rs6000/rs6000.c (struct rs6000_cost_data): New member
16191 costing_for_scalar.
16192 (rs6000_density_test): Early return if costing_for_scalar is true.
16193 (rs6000_init_cost): Init costing_for_scalar of rs6000_cost_data.
16195 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
16197 * doc/tm.texi: Regenerated.
16198 * target.def (init_cost): Add new parameter costing_for_scalar.
16199 * targhooks.c (default_init_cost): Adjust for new parameter.
16200 * targhooks.h (default_init_cost): Likewise.
16201 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
16202 (vect_compute_single_scalar_iteration_cost): Likewise.
16203 (vect_analyze_loop_2): Likewise.
16204 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
16205 (vect_bb_vectorization_profitable_p): Likewise.
16206 * tree-vectorizer.h (init_cost): Likewise.
16207 * config/aarch64/aarch64.c (aarch64_init_cost): Likewise.
16208 * config/i386/i386.c (ix86_init_cost): Likewise.
16209 * config/rs6000/rs6000.c (rs6000_init_cost): Likewise.
16211 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
16213 * config/rs6000/rs6000.c (rs6000_vect_nonmem): Renamed to
16214 vect_nonmem and moved into...
16215 (struct rs6000_cost_data): ...here.
16216 (rs6000_init_cost): Use vect_nonmem of cost_data instead.
16217 (rs6000_add_stmt_cost): Likewise.
16218 (rs6000_finish_cost): Likewise.
16220 2021-05-10 Eric Botcazou <ebotcazou@adacore.com>
16222 * range-op.cc (get_bool_state): Adjust head comment.
16223 (operator_not_equal::op1_range): Fix comment.
16224 (operator_bitwise_xor::op1_range): Remove call to gcc_unreachable.
16226 2021-05-10 Martin Sebor <msebor@redhat.com>
16228 PR middle-end/100425
16229 PR middle-end/100510
16230 * gimple-ssa-warn-alloca.c (pass_walloca::firast_time_p): Rename...
16231 (pass_walloca::xlimit_certain_p): ...to this.
16232 (pass_walloca::gate): Execute for any kind of handled warning.
16233 (pass_walloca::execute): Avoid issuing "maybe" and "unbounded"
16234 warnings when xlimit_certain_p is set.
16236 2021-05-10 Pat Haugen <pthaugen@linux.ibm.com>
16238 * config/rs6000/rs6000.c (rs6000_ira_change_pseudo_allocno_class):
16239 Return ALTIVEC_REGS if that is best_class.
16240 (rs6000_compute_pressure_classes): Add ALTIVEC_REGS.
16242 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16244 * config/arm/arm.h (CPP_SPEC): Remove error message about
16247 2021-05-10 Martin Jambor <mjambor@suse.cz>
16249 * ipa-prop.h (IPA_NODE_REF): Removed.
16250 (IPA_NODE_REF_GET_CREATE): Likewise.
16251 (IPA_EDGE_REF): Likewise.
16252 (IPA_EDGE_REF_GET_CREATE): Likewise.
16253 (IS_VALID_JUMP_FUNC_INDEX): Likewise.
16254 * ipa-cp.c (print_all_lattices): Replaced IPA_NODE_REF with a direct
16255 use of ipa_node_params_sum.
16256 (ipcp_versionable_function_p): Likewise.
16257 (push_node_to_stack): Likewise.
16258 (pop_node_from_stack): Likewise.
16259 (set_single_call_flag): Replaced two IPA_NODE_REF with one single
16260 direct use of ipa_node_params_sum.
16261 (initialize_node_lattices): Replaced IPA_NODE_REF with a direct use of
16262 ipa_node_params_sum.
16263 (ipa_context_from_jfunc): Replaced IPA_EDGE_REF with a direct use of
16265 (ipcp_verify_propagated_values): Replaced IPA_NODE_REF with a direct
16266 use of ipa_node_params_sum.
16267 (self_recursively_generated_p): Likewise.
16268 (propagate_scalar_across_jump_function): Likewise.
16269 (propagate_context_across_jump_function): Replaced IPA_EDGE_REF with a
16270 direct use of ipa_edge_args_sum, moved the lookup after the early
16271 exit. Replaced IPA_NODE_REF with a direct use of ipa_node_params_sum.
16272 (propagate_bits_across_jump_function): Replaced IPA_NODE_REF with
16273 direct uses of ipa_node_params_sum.
16274 (propagate_vr_across_jump_function): Likewise.
16275 (propagate_aggregate_lattice): Likewise.
16276 (propagate_aggs_across_jump_function): Likewise.
16277 (propagate_constants_across_call): Likewise, also replaced
16278 IPA_EDGE_REF with a direct use of ipa_edge_args_sum.
16279 (good_cloning_opportunity_p): Replaced IPA_NODE_REF with a direct use
16280 of ipa_node_params_sum.
16281 (estimate_local_effects): Likewise.
16282 (add_all_node_vals_to_toposort): Likewise.
16283 (propagate_constants_topo): Likewise.
16284 (ipcp_propagate_stage): Likewise.
16285 (ipcp_discover_new_direct_edges): Likewise.
16286 (calls_same_node_or_its_all_contexts_clone_p): Likewise.
16287 (cgraph_edge_brings_value_p): Likewise (in both overloaded functions).
16288 (get_info_about_necessary_edges): Likewise.
16289 (want_remove_some_param_p): Likewise.
16290 (create_specialized_node): Likewise.
16291 (self_recursive_pass_through_p): Likewise.
16292 (self_recursive_agg_pass_through_p): Likewise.
16293 (find_more_scalar_values_for_callers_subset): Likewise and also
16294 replaced IPA_EDGE_REF with direct uses of ipa_edge_args_sum, in one
16295 case replacing two of those with a single query.
16296 (find_more_contexts_for_caller_subset): Likewise for the
16297 ipa_polymorphic_call_context overload.
16298 (intersect_aggregates_with_edge): Replaced IPA_EDGE_REF with a direct
16299 use of ipa_edge_args_sum. Replaced IPA_NODE_REF with direct uses of
16300 ipa_node_params_sum.
16301 (find_aggregate_values_for_callers_subset): Likewise, also reusing
16302 results of ipa_edge_args_sum->get.
16303 (cgraph_edge_brings_all_scalars_for_node): Replaced IPA_NODE_REF with
16304 direct uses of ipa_node_params_sum, replaced IPA_EDGE_REF with a
16305 direct use of ipa_edge_args_sum.
16306 (cgraph_edge_brings_all_agg_vals_for_node): Likewise, moved node
16307 summary query after the early exit and reused the result later.
16308 (decide_about_value): Replaced IPA_NODE_REF with a direct use of
16309 ipa_node_params_sum.
16310 (decide_whether_version_node): Likewise. Removed re-querying for
16311 summaries after cloning.
16312 (spread_undeadness): Replaced IPA_NODE_REF with a direct use of
16313 ipa_node_params_sum.
16314 (has_undead_caller_from_outside_scc_p): Likewise, reusing results of
16316 (identify_dead_nodes): Likewise.
16317 (ipcp_store_bits_results): Replaced IPA_NODE_REF with direct uses of
16318 ipa_node_params_sum.
16319 (ipcp_store_vr_results): Likewise.
16320 * ipa-fnsummary.c (evaluate_properties_for_edge): Likewise.
16321 (ipa_fn_summary_t::duplicate): Likewise.
16322 (analyze_function_body): Likewise.
16323 (estimate_calls_size_and_time): Likewise.
16324 (ipa_cached_call_context::duplicate_from): Likewise.
16325 (ipa_call_context::equal_to): Likewise.
16326 (remap_edge_params): Likewise.
16327 (ipa_merge_fn_summary_after_inlining): Likewise.
16328 (inline_read_section): Likewise.
16329 * ipa-icf.c (sem_function::param_used_p): Likewise.
16330 * ipa-modref.c (compute_parm_map): Likewise.
16331 (compute_parm_map): Replaced IPA_EDGE_REF with a direct use of
16333 (get_access_for_fnspec): Replaced IPA_NODE_REF with a direct use of
16334 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
16336 * ipa-profile.c (check_argument_count): Likewise.
16337 * ipa-prop.c (ipa_alloc_node_params): Replaced IPA_NODE_REF_GET_CREATE
16338 with a direct use of ipa_node_params_sum.
16339 (ipa_initialize_node_params): Likewise.
16340 (ipa_print_node_jump_functions_for_edge): Replaced IPA_EDGE_REF with a
16341 direct use of ipa_edge_args_sum and reused the query result.
16342 (ipa_compute_jump_functions_for_edge): Replaced IPA_NODE_REF with a
16343 direct use of ipa_node_params_sum and replaced IPA_EDGE_REF with a
16344 direct use of ipa_edge_args_sum.
16345 (ipa_note_param_call): Replaced IPA_NODE_REF with a direct use of
16346 ipa_node_params_sum and reused the result of the query.
16347 (ipa_analyze_node): Likewise.
16348 (ipa_analyze_controlled_uses): Replaced IPA_NODE_REF with a direct use
16349 of ipa_node_params_sum.
16350 (update_jump_functions_after_inlining): Replaced IPA_EDGE_REF with
16351 direct uses of ipa_edge_args_sum.
16352 (update_indirect_edges_after_inlining): Replaced IPA_NODE_REF with
16353 direct uses of ipa_node_params_sum and replaced IPA_EDGE_REF with a
16354 direct use of ipa_edge_args_sum. Removed superficial re-querying the
16356 (propagate_controlled_uses): Replaced IPA_NODE_REF with direct uses of
16357 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
16359 (ipa_propagate_indirect_call_infos): Replaced IPA_EDGE_REF with a
16360 direct use of ipa_edge_args_sum.
16361 (ipa_edge_args_sum_t::duplicate): Replaced IPA_NODE_REF with a direct
16362 use of ipa_node_params_sum.
16363 (ipa_print_node_params): Likewise.
16364 (ipa_write_node_info): Likewise and also replaced IPA_EDGE_REF with
16365 direct uses of ipa_edge_args_sum.
16366 (ipa_read_edge_info): Replaced IPA_EDGE_REF with a direct use of
16368 (ipa_read_node_info): Replaced IPA_NODE_REF with a direct use of
16369 ipa_node_params_sum.
16370 (ipa_prop_write_jump_functions): Likewise. Move variable node to the
16371 scopes where it is used.
16373 2021-05-10 Uroš Bizjak <ubizjak@gmail.com>
16375 * config/i386/i386-expand.c (ix86_expand_sse_movcc)
16376 <case E_V2SImode>: Force op_true to register.
16378 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16380 * config/arm/iterators.md (MVE_FP_COMPARISONS): New.
16381 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_f<mode>)
16382 (mve_vcmp<mve_cmp_op>q_n_f<mode>): New, merge all vcmp_*f*
16384 (mve_vcmpeqq_f<mode>, mve_vcmpeqq_n_f<mode>, mve_vcmpgeq_f<mode>)
16385 (mve_vcmpgeq_n_f<mode>, mve_vcmpgtq_f<mode>)
16386 (mve_vcmpgtq_n_f<mode>, mve_vcmpleq_f<mode>)
16387 (mve_vcmpleq_n_f<mode>, mve_vcmpltq_f<mode>)
16388 (mve_vcmpltq_n_f<mode>, mve_vcmpneq_f<mode>)
16389 (mve_vcmpneq_n_f<mode>): Remove.
16390 * config/arm/unspecs.md (VCMPEQQ_F, VCMPEQQ_N_F, VCMPGEQ_F)
16391 (VCMPGEQ_N_F, VCMPGTQ_F, VCMPGTQ_N_F, VCMPLEQ_F, VCMPLEQ_N_F)
16392 (VCMPLTQ_F, VCMPLTQ_N_F, VCMPNEQ_F, VCMPNEQ_N_F): Remove.
16394 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16396 * config/arm/iterators.md (MVE_COMPARISONS): New.
16398 (mve_cmp_type): New.
16399 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_<mode>): New, merge all
16401 (mve_vcmpneq_<mode>, mve_vcmpcsq_n_<mode>, mve_vcmpcsq_<mode>)
16402 (mve_vcmpeqq_n_<mode>, mve_vcmpeqq_<mode>, mve_vcmpgeq_n_<mode>)
16403 (mve_vcmpgeq_<mode>, mve_vcmpgtq_n_<mode>, mve_vcmpgtq_<mode>)
16404 (mve_vcmphiq_n_<mode>, mve_vcmphiq_<mode>, mve_vcmpleq_n_<mode>)
16405 (mve_vcmpleq_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
16406 (mve_vcmpneq_n_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
16407 (mve_vcmpneq_n_<mode>): Remove.
16409 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16411 * config/arm/arm_mve.h (__arm_vcmp*): Remove 's' suffix.
16412 * config/arm/arm_mve_builtins.def (vcmp*): Remove 's' suffix.
16413 * config/arm/mve.md (mve_vcmp*): Remove 's' suffix in pattern
16416 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16418 * config/arm/arm_mve_builtins.def (vcmpneq_u): Remove.
16419 (vcmpneq_n_u): Likewise.
16420 (vcmpeqq_u,): Likewise.
16421 (vcmpeqq_n_u): Likewise.
16422 * config/arm/iterators.md (supf): Remove VCMPNEQ_U, VCMPEQQ_U,
16423 VCMPEQQ_N_U and VCMPNEQ_N_U.
16424 * config/arm/mve.md (mve_vcmpneq): Remove <supf> iteration.
16425 (mve_vcmpeqq_n): Likewise.
16426 (mve_vcmpeqq): Likewise.
16427 (mve_vcmpneq_n): Likewise.
16429 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16431 * config/arm/arm_mve.h (__arm_vcmpeq*u*, __arm_vcmpne*u*): Call
16432 the 's' version of the builtin.
16434 2021-05-10 Richard Biener <rguenther@suse.de>
16436 PR tree-optimization/100492
16437 * tree-loop-distribution.c (find_seed_stmts_for_distribution):
16438 Find nothing when the loop contains an irreducible region.
16440 2021-05-10 Richard Biener <rguenther@suse.de>
16442 PR middle-end/100464
16444 * gimple-fold.c (canonicalize_constructor_val): Do not set
16447 2021-05-10 Richard Biener <rguenther@suse.de>
16449 PR tree-optimization/100434
16450 * tree-ssa-dse.c (initialize_ao_ref_for_dse): Handle
16452 (dse_optimize_stmt): Handle call LHS by dropping the
16453 LHS or the whole call if it doesn't have other
16455 (pass_dse::execute): Adjust.
16457 2021-05-10 Martin Liska <mliska@suse.cz>
16459 * Makefile.in: Add missing genversion rule.
16461 2021-05-10 Alex Coplan <alex.coplan@arm.com>
16464 * config/arm/mve.md (*mve_mov<mode>): Simplify output code. Use
16465 vldrw.u32 and vstrw.32 for V2D[IF]mode loads and stores.
16467 2021-05-10 Martin Liska <mliska@suse.cz>
16469 * builtins.c (is_builtin_name): Use startswith
16470 function instead of strncmp.
16471 * collect2.c (main): Likewise.
16472 (has_lto_section): Likewise.
16473 (scan_libraries): Likewise.
16474 * coverage.c (coverage_checksum_string): Likewise.
16475 (coverage_init): Likewise.
16476 * dwarf2out.c (is_cxx): Likewise.
16477 (gen_compile_unit_die): Likewise.
16478 * gcc-ar.c (main): Likewise.
16479 * gcc.c (init_spec): Likewise.
16480 (read_specs): Likewise.
16481 (execute): Likewise.
16482 (check_live_switch): Likewise.
16483 * genattrtab.c (write_attr_case): Likewise.
16484 (IS_ATTR_GROUP): Likewise.
16485 * gencfn-macros.c (main): Likewise.
16486 * gengtype.c (type_for_name): Likewise.
16487 (gen_rtx_next): Likewise.
16488 (get_file_langdir): Likewise.
16489 (write_local): Likewise.
16490 * genmatch.c (get_operator): Likewise.
16491 (get_operand_type): Likewise.
16492 (expr::gen_transform): Likewise.
16493 * genoutput.c (validate_optab_operands): Likewise.
16494 * incpath.c (add_sysroot_to_chain): Likewise.
16495 * langhooks.c (lang_GNU_C): Likewise.
16496 (lang_GNU_CXX): Likewise.
16497 (lang_GNU_Fortran): Likewise.
16498 (lang_GNU_OBJC): Likewise.
16499 * lto-wrapper.c (run_gcc): Likewise.
16500 * omp-general.c (omp_max_simt_vf): Likewise.
16501 * omp-low.c (omp_runtime_api_call): Likewise.
16502 * opts-common.c (parse_options_from_collect_gcc_options): Likewise.
16503 * read-rtl-function.c (function_reader::read_rtx_operand_r): Likewise.
16504 * real.c (real_from_string): Likewise.
16505 * selftest.c (assert_str_startswith): Likewise.
16506 * timevar.c (timer::validate_phases): Likewise.
16507 * tree.c (get_file_function_name): Likewise.
16508 * ubsan.c (ubsan_use_new_style_p): Likewise.
16509 * varasm.c (default_function_rodata_section): Likewise.
16510 (incorporeal_function_p): Likewise.
16511 (default_section_type_flags): Likewise.
16512 * system.h (startswith): Define startswith.
16514 2021-05-10 Martin Liska <mliska@suse.cz>
16516 * bitmap.h (class auto_bitmap): Remove
16517 __cplusplus >= 201103.
16518 * config/aarch64/aarch64.c: Likewise.
16519 * gimple-ssa-store-merging.c (store_immediate_info::store_immediate_info):
16521 * sbitmap.h: Likewise.
16523 2021-05-10 Martin Liska <mliska@suse.cz>
16525 * Makefile.in: Rename gcov-iov to genversion and depend
16526 on version.h (instead of gcov-iov.h).
16527 * gcov-io.h: Include version.h instread of gcov-iov.h.
16528 * gengtype-state.c (read_state_version): Likewise.
16529 * gcov-iov.c: Moved to...
16530 * genversion.c: ...here.
16531 * lto-streamer.h (LTO_major_version): Define it with
16533 * version.c: Removed.
16534 * version.h: Removed.
16536 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16538 * config/arc/arc.md (UNSPEC_ARC_DMPYWH): Define.
16539 * config/arc/simdext.md (VCT): Add predicates for iterator
16542 (voptab): Likewise.
16543 (vec_widen_<V_US>mult_hi_v4hi): Change pattern predicate.
16544 (<voptab>v2si3): New patterns.
16546 (reduc_plus_scal_v4hi): Likewise.
16547 (reduc_plus_scal_v2si): Likewise.
16548 (vec_duplicatev2si): Likewise.
16549 (vec_duplicatev4hi): Likewise.
16551 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16553 * config/arc/simdext.md: Format and cleanup file.
16555 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16557 * config/arc/simdext.md (movmisalignv2hi): Allow misaligned access
16558 only when munaligned-access option is on.
16559 (movmisalign<mode>): Likewise.
16561 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16563 * common/config/arc/arc-common.c (arc_handle_option): Remove dot
16565 * config/arc/arc.c (arc_reorg): Remove underscore from string.
16567 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16569 * config/arc/arc.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
16570 (CTZ_DEFINED_VALUE_AT_ZERO): Likewise.
16571 * config/arc/arc.md (clrsbsi2): Cleanup pattern.
16572 (norm_f): Likewise.
16575 (clzsi2): Use fls instruction when available.
16576 (arc_clzsi2): Likewise.
16578 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16580 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add r26 and r27.
16582 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16584 * doc/extend.texi (__builtin_arc_sr): Swap arguments.
16586 2021-05-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
16588 PR middle-end/100467
16589 * toplev.c (compile_file): Call insn_locations_init before
16590 targetm.asm_out.code_end.
16592 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
16595 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
16597 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
16599 2021-05-07 Jakub Jelinek <jakub@redhat.com>
16600 Andrew Stubbs <amd@codesourcery.com>
16603 * builtins.c (try_store_by_multiple_pieces): Use force_operand for
16604 emit_move_insn operands.
16606 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
16608 * cfgexpand.c (expand_gimple_basic_block): Do not inherit a current
16609 location for the outgoing edges of an empty block.
16610 * dwarf2out.c (add_subscript_info): Retrieve the bounds and index
16611 type by means of the get_array_descr_info langhook, if it is set and
16612 returns true. Remove obsolete code dealing with unnamed subtypes.
16614 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16616 * gimple-range-cache.cc (ssa_block_ranges): Virtualize.
16617 (sbr_vector): Renamed from ssa_block_cache.
16618 (sbr_vector::sbr_vector): Allocate from obstack abd initialize.
16619 (ssa_block_ranges::~ssa_block_ranges): Remove.
16620 (sbr_vector::set_bb_range): Use varying and undefined cached values.
16621 (ssa_block_ranges::set_bb_varying): Remove.
16622 (sbr_vector::get_bb_range): Adjust assert.
16623 (sbr_vector::bb_range_p): Adjust assert.
16624 (~block_range_cache): No freeing loop required.
16625 (block_range_cache::get_block_ranges): Remove.
16626 (block_range_cache::set_bb_range): Inline get_block_ranges.
16627 (block_range_cache::set_bb_varying): Remove.
16628 * gimple-range-cache.h (set_bb_varying): Remove prototype.
16629 * value-range.h (irange_allocator::get_memory): New.
16631 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16633 * gimple-range-cache.cc (non_null_ref::non_null_deref_p): Search
16634 dominator tree is available and requested.
16635 (ranger_cache::ssa_range_in_bb): Don't search dom tree here.
16636 (ranger_cache::fill_block_cache): Don't search dom tree here either.
16637 * gimple-range-cache.h (non_null_deref_p): Add dom_search param.
16639 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16641 * gimple-range.cc (gimple_ranger::range_on_exit): Handle block with
16642 only PHI nodes better.
16644 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16646 * gimple-range-edge.h (gimple_outgoing_range): Rename from
16648 (gcond_edge_range): Export prototype.
16649 * gimple-range-edge.cc (gcond_edge_range): New.
16650 (gimple_outgoing_range::edge_range_p): Use gcond_edge_range.
16651 * gimple-range-gori.h (gori_compute): Use gimple_outgoing_range.
16653 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16655 * gimple-range-edge.cc (outgoing_range::calc_switch_ranges): Compute
16656 default range into a temp and allocate only what is needed.
16658 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16660 * range-op.cc (operator_trunc_mod::wi_fold): x % 0 is UNDEFINED.
16662 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16664 * gimple-range.h (gimple_range_global): Pick up parameter initial
16665 values, and use-before defined locals are UNDEFINED.
16667 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
16669 * doc/extend.texi (scalar_storage_order): Mention effect on pointer
16671 * tree.h (reverse_storage_order_for_component_p): Return false if
16672 the type is a pointer.
16674 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
16676 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
16678 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
16681 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
16682 Handle V8QI, V4HI and V2SI modes.
16683 * config/i386/mmx.md (mmx_pblendvb): New insn pattern.
16684 * config/i386/sse.md (unspec): Move UNSPEC_BLENDV ...
16685 * config/i386/i386.md (unspec): ... here.
16687 2021-05-07 Tobias Burnus <tobias@codesourcery.com>
16688 Tom de Vries <tdevries@suse.de>
16690 * omp-low.c (lower_rec_simd_input_clauses): Set max_vf = 1 if
16691 a truth_value_p reduction variable is nonintegral.
16693 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
16696 * config/i386/i386-expand.c (ix86_use_mask_cmp_p):
16697 Return false for mode sizes < 16.
16699 2021-05-07 Jakub Jelinek <jakub@redhat.com>
16702 * config/i386/mmx.md (*xop_pcmov_<mode>): New define_insn.
16704 2021-05-06 Martin Jambor <mjambor@suse.cz>
16706 * ipa-sra.c (ipa_sra_dump_all_summaries): Dump edge summaries even
16707 when there is no function summary.
16708 (ipa_sra_summarize_function): produce edge summaries even when
16711 2021-05-06 Tom Tromey <tom@tromey.com>
16713 * godump.c (string_hash_eq): Remove.
16714 (go_finish): Use htab_eq_string.
16716 2021-05-06 Tom Tromey <tom@tromey.com>
16718 * gengtype-state.c (read_state): Use htab_eq_string.
16719 (string_eq): Remove.
16721 2021-05-06 Tom Tromey <tom@tromey.com>
16723 * gensupport.c (htab_eq_string): Remove.
16725 2021-05-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
16728 * debug.h (gcc_debug_hooks): Add set_ignored_loc function pointer.
16729 * dwarf2out.h (dw_fde_node::ignored_debug): New data item.
16730 * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Add dummy
16731 set_ignored_loc callbacks.
16732 * debug.c (do_nothing_debug_hooks): Likewise.
16733 * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
16734 * dwarf2out.c (text_section_used, cold_text_section_used): Remove.
16735 (in_text_section_p, last_text_label, last_cold_label,
16736 switch_text_ranges, switch_cold_ranges): New data items.
16737 (dwarf2out_note_section_used): Remove.
16738 (dwarf2out_begin_prologue): Set fde->ignored_debug and
16740 (mark_ignored_debug_section): New helper function.
16741 (dwarf2out_end_epilogue, dwarf2out_switch_text_section): Call
16742 mark_ignored_debug_section.
16743 (dwarf2_debug_hooks): Use dwarf2out_set_ignored_loc.
16744 (dwarf2_lineno_debug_hooks): Use dummy for set_ignored_loc.
16745 (size_of_aranges): Adjust formula for multi-part text ranges size.
16746 (output_aranges): Output multi-part text ranges.
16747 (dwarf2out_set_ignored_loc): New callback function.
16748 (dwarf2out_finish): Output multi-part text ranges.
16749 (dwarf2out_c_finalize): Clear new data items.
16750 * final.c (final_start_function_1): Call set_ignored_loc callback.
16751 (final_scan_insn_1): Likewise.
16752 * ggc-page.c (gt_ggc_mx): New helper function.
16753 * stringpool.c (gt_pch_nx): Likewise.
16755 2021-05-06 Richard Biener <rguenther@suse.de>
16757 * timevar.def (TV_TREE_INSERT_PHI_NODES): Remove.
16758 (TV_TREE_SSA_REWRITE_BLOCKS): Likewise.
16759 (TV_TREE_INTO_SSA): New.
16760 * tree-into-ssa.c (insert_phi_nodes): Do not account separately.
16761 (rewrite_blocks): Likewise.
16762 (pass_data_build_ssa): Account to TV_TREE_INTO_SSA.
16764 2021-05-06 Jakub Jelinek <jakub@redhat.com>
16766 * tree-ssa-phiopt.c (value_replacement, minmax_replacement,
16767 abs_replacement, xor_replacement,
16768 cond_removal_in_popcount_clz_ctz_pattern,
16769 replace_phi_edge_with_variable): Change type of phi argument from
16770 gimple * to gphi *.
16772 2021-05-06 Richard Biener <rguenther@suse.de>
16774 * tree-ssa-loop-split.c (split_loop): Delay updating SSA form.
16775 Output an opt-info message.
16776 (do_split_loop_on_cond): Likewise.
16777 (tree_ssa_split_loops): Update SSA form here.
16779 2021-05-06 Richard Biener <rguenther@suse.de>
16781 * tree-inline.c (tree_function_versioning): Fix DECL_BY_REFERENCE
16782 return variable removal.
16784 2021-05-06 Marius Hillenbrand <mhillen@linux.ibm.com>
16786 * config/s390/s390-builtins.def (O_M5, O1_M5, ...): Remove unused macros.
16787 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
16788 (s390_vec_permi_dbl, s390_vpdi): Use the O3_U2 type for the immediate
16790 * config/s390/s390.c (s390_const_operand_ok): Remove unused
16793 2021-05-06 Jakub Jelinek <jakub@redhat.com>
16795 PR tree-optimization/94589
16796 * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Call
16797 spaceship_replacement.
16798 (cond_only_block_p, spaceship_replacement): New functions.
16800 2021-05-06 Richard Biener <rguenther@suse.de>
16803 * tree-emutls.c (gen_emutls_addr): Pass in whether we're
16804 dealing with a debug use and only query existing addresses
16806 (lower_emutls_1): Avoid splitting out addresses for debug
16807 stmts, reset the debug stmt when we fail to find existing
16809 (lower_emutls_phi_arg): Set wi.stmt.
16811 2021-05-06 Christoph Muellner <cmuellner@gcc.gnu.org>
16814 * config/riscv/riscv.c (riscv_block_move_loop): Use cbranch helper.
16815 * config/riscv/riscv.md (cbranch<mode>4): Generate helpers.
16816 (stack_protect_test): Use cbranch helper.
16818 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
16821 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
16822 always return the establisher frame for __builtin_frame_address (0).
16824 2021-05-05 Ivan Sorokin <vanyacpp@gmail.com>
16827 * config/i386/i386-builtins.c (ix86_cpu_model_type_node): New.
16828 (ix86_cpu_model_var): Likewise.
16829 (ix86_cpu_features2_type_node): Likewise.
16830 (ix86_cpu_features2_var): Likewise.
16831 (fold_builtin_cpu): Cache __cpu_model and __cpu_features2 with
16834 2021-05-05 Martin Sebor <msebor@redhat.com>
16836 * passes.def (pass_warn_printf): Run after SSA.
16838 2021-05-05 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16840 * config/arm/neon.md (neon_vtst_combine<mode>): New pattern.
16841 * config/arm/predicates.md (minus_one_operand): New predicate.
16843 2021-05-05 Jeff Law <jlaw@tachyum.com>
16845 * config/avr/avr.md: Remove references to CC_STATUS_INIT.
16847 2021-05-05 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
16849 PR rtl-optimization/100263
16850 * postreload.c (move2add_valid_value_p): Ensure register can
16853 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
16855 PR rtl-optimization/100411
16856 * cfgcleanup.c (try_crossjump_to_edge): Also skip end of prologue
16857 and beginning of function markers.
16859 2021-05-05 Jeff Law <jlaw@tachyum.com>
16861 * config/cr16/cr16.h (NOTICE_UPDATE_CC): Remove.
16862 * config/cr16/cr16.c (notice_update_cc): Remove.
16863 * config/cr16/cr16-protos.h (notice_update_cc): Remove.
16865 2021-05-05 Uroš Bizjak <ubizjak@gmail.com>
16868 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
16869 Handle V8QI, V4HI and V2SI modes.
16870 * config/i386/i386.c (ix86_build_const_vector): Handle V2SImode.
16871 (ix86_build_signbit_mask): Ditto.
16872 * config/i386/mmx.md (MMXMODE14): New mode iterator.
16873 (<smaxmin:code><MMXMODE14:mode>3): New expander.
16874 (*mmx_<smaxmin:code><MMXMODE14:mode>3): New insn pattern.
16875 (<umaxmin:code><MMXMODE24:mode>3): New expander.
16876 (*mmx_<umaxmin:code><MMXMODE24:mode>3): New insn pattern.
16877 (vec_cmp<MMXMODEI:mode><MMXMODEI:mode>): New expander.
16878 (vec_cmpu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
16879 (vcond<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
16880 (vcondu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
16881 (vcond_mask_<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
16883 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
16885 * dwarf2out.c (loc_list_from_tree_1) <DECL>: During early DWARF, do
16886 not expand the VALUE_EXPR of variables put in the non-local frame.
16887 * gimplify.c (gimplify_type_sizes) <RECORD_TYPE>: If the type is not
16888 to be ignored for debug info, ensure its variable offsets are not.
16890 2021-05-05 Richard Biener <rguenther@suse.de>
16892 PR tree-optimization/79333
16893 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt):
16894 Fold stmt following SSA edges.
16896 2021-05-05 Richard Biener <rguenther@suse.de>
16898 PR middle-end/100394
16899 * calls.c (expand_call): Preserve possibly throwing calls.
16900 * cfgexpand.c (expand_call_stmt): When a call can throw signal
16901 RTL expansion there are side-effects.
16902 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Simplify,
16903 mark all possibly throwing stmts necessary unless we can elide
16905 * tree-ssa-dse.c (pass_dse::execute): Preserve exceptions unless
16906 -fdelete-dead-exceptions.
16907 * tree.h (DECL_PURE_P): Add note about exceptions.
16909 2021-05-05 Alexandre Oliva <oliva@adacore.com>
16911 * config/i386/vxworks.h (DBX_REGISTER_NUMBER): Make it
16914 2021-05-04 David Edelsohn <dje.gcc@gmail.com>
16916 * config/rs6000/rs6000-call.c (rs6000_output_mi_thunk): Use
16917 get_fnname_from_decl for name of thunk.
16918 * config/rs6000/rs6000.c (rs6000_declare_alias): Use assemble_name
16919 and ASM_OUTPUT_LABEL.
16920 (rs6000_xcoff_declare_function_name): Use assemble_name and
16922 (rs6000_xcoff_declare_object_name): Use ASM_OUTPUT_LABEL.
16923 (rs6000_xcoff_encode_section_info): Don't add mapping class
16924 for aliases. Always add [DS] mapping class to primary
16926 (rs6000_asm_weaken_decl): Don't explicitly add [DS].
16928 2021-05-04 Martin Sebor <msebor@redhat.com>
16930 PR middle-end/100307
16931 * builtins.c (compute_objsize_r): Clear base0 for pointers.
16933 2021-05-04 Jeff Law <jlaw@tachyum.com>
16935 * config/bfin/bfin.h (NOTICE_UPDATE_CC): Remove.
16937 2021-05-04 Segher Boessenkool <segher@kernel.crashing.org>
16939 * caller-save.c: Remove CC0.
16940 * cfgcleanup.c: Remove CC0.
16941 * cfgrtl.c: Remove CC0.
16942 * combine.c: Remove CC0.
16943 * compare-elim.c: Remove CC0.
16944 * conditions.h: Remove CC0.
16945 * config/h8300/h8300.h: Remove CC0.
16946 * config/h8300/h8300-protos.h: Remove CC0.
16947 * config/h8300/peepholes.md: Remove CC0.
16948 * config/i386/x86-tune-sched.c: Remove CC0.
16949 * config/m68k/m68k.c: Remove CC0.
16950 * config/rl78/rl78.c: Remove CC0.
16951 * config/sparc/sparc.c: Remove CC0.
16952 * config/xtensa/xtensa.c: Remove CC0.
16953 (gen_conditional_move): Use pc_rtx instead of cc0_rtx in a piece of
16954 RTL where that is used as a placeholder only.
16955 * cprop.c: Remove CC0.
16956 * cse.c: Remove CC0.
16957 * cselib.c: Remove CC0.
16958 * df-problems.c: Remove CC0.
16959 * df-scan.c: Remove CC0.
16960 * doc/md.texi: Remove CC0. Adjust an example.
16961 * doc/rtl.texi: Remove CC0. Adjust an example.
16962 * doc/tm.texi: Regenerate.
16963 * doc/tm.texi.in: Remove CC0.
16964 * emit-rtl.c: Remove CC0.
16965 * final.c: Remove CC0.
16966 * fwprop.c: Remove CC0.
16967 * gcse-common.c: Remove CC0.
16968 * gcse.c: Remove CC0.
16969 * genattrtab.c: Remove CC0.
16970 * genconfig.c: Remove CC0.
16971 * genemit.c: Remove CC0.
16972 * genextract.c: Remove CC0.
16973 * gengenrtl.c: Remove CC0.
16974 * genrecog.c: Remove CC0.
16975 * haifa-sched.c: Remove CC0.
16976 * ifcvt.c: Remove CC0.
16977 * ira-costs.c: Remove CC0.
16978 * ira.c: Remove CC0.
16979 * jump.c: Remove CC0.
16980 * loop-invariant.c: Remove CC0.
16981 * lra-constraints.c: Remove CC0.
16982 * lra-eliminations.c: Remove CC0.
16983 * optabs.c: Remove CC0.
16984 * postreload-gcse.c: Remove CC0.
16985 * postreload.c: Remove CC0.
16986 * print-rtl.c: Remove CC0.
16987 * read-rtl-function.c: Remove CC0.
16988 * reg-notes.def: Remove CC0.
16989 * reg-stack.c: Remove CC0.
16990 * reginfo.c: Remove CC0.
16991 * regrename.c: Remove CC0.
16992 * reload.c: Remove CC0.
16993 * reload1.c: Remove CC0.
16994 * reorg.c: Remove CC0.
16995 * resource.c: Remove CC0.
16996 * rtl.c: Remove CC0.
16997 * rtl.def: Remove CC0.
16998 * rtl.h: Remove CC0.
16999 * rtlanal.c: Remove CC0.
17000 * sched-deps.c: Remove CC0.
17001 * sched-rgn.c: Remove CC0.
17002 * shrink-wrap.c: Remove CC0.
17003 * simplify-rtx.c: Remove CC0.
17004 * system.h: Remove CC0. Poison NOTICE_UPDATE_CC, CC_STATUS_MDEP_INIT,
17005 CC_STATUS_MDEP, and CC_STATUS.
17006 * target.def: Remove CC0.
17007 * valtrack.c: Remove CC0.
17008 * var-tracking.c: Remove CC0.
17010 2021-05-04 Richard Biener <rguenther@suse.de>
17012 PR tree-optimization/100414
17013 * tree-ssa-phiopt.c (get_non_trapping): Do not compute dominance
17015 (tree_ssa_phiopt_worker): But unconditionally here.
17017 2021-05-04 Tobias Burnus <tobias@codesourcery.com>
17019 * omp-low.c (lower_rec_input_clauses, lower_reduction_clauses): Handle
17020 && and || with floating-point and complex arguments.
17022 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
17024 * tree-inline.c (insert_debug_decl_map): Delete.
17025 (copy_debug_stmt): Minor tweak.
17026 (setup_one_parameter): Do not use a variable if the value is either
17027 a read-only DECL or a non-addressable local variable in the caller.
17028 In this case, insert the debug-only variable in the map manually.
17029 (expand_call_inline): Do not generate a CLOBBER for these values.
17030 * tree-inline.h (debug_map): Minor tweak.
17032 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
17034 * builtins.c (builtin_with_linkage_p): Return true for stp[n]cpy.
17035 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Tidy up.
17037 2021-05-04 Richard Biener <rguenther@suse.de>
17039 PR tree-optimization/100329
17040 * tree-ssa-reassoc.c (can_reassociate_p): Do not reassociate
17042 (insert_stmt_after): Assert we're not running into asm goto.
17044 2021-05-04 Richard Biener <rguenther@suse.de>
17046 PR tree-optimization/100398
17047 * tree-ssa-dse.c (pass_dse::execute): Preserve control
17050 2021-05-04 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17052 * builtins.c (try_store_by_multiple_pieces): Fix constfun's prototype.
17054 2021-05-04 Alexandre Oliva <oliva@adacore.com>
17056 * builtins.c (try_store_by_multiple_pieces): New.
17057 (expand_builtin_memset_args): Use it. If target_char_cast
17058 fails, proceed as for non-constant val. Pass len's ctz to...
17059 * expr.c (clear_storage_hints): ... this. Try store by
17060 multiple pieces after setmem.
17061 (clear_storage): Adjust.
17062 * expr.h (clear_storage_hints): Likewise.
17063 (try_store_by_multiple_pieces): Declare.
17064 * passes.def: Replace the last copy_prop with ccp.
17066 2021-05-03 Tom de Vries <tdevries@suse.de>
17069 * omp-low.c (lower_rec_input_clauses): Disable SIMT for user-defined
17072 2021-05-03 Richard Biener <rguenther@suse.de>
17074 * tree-ssa-dse.c (dse_classify_store): Track two PHI defs.
17076 2021-05-03 Richard Biener <rguenther@suse.de>
17078 * tree-ssa-dse.c: Do not include domwalk.h but cfganal.h.
17079 (dse_dom_walker): Remove.
17080 (dse_dom_walker::dse_optimize_stmt): Rename...
17081 (dse_optimize_stmt): ... to this, pass in live_bytes sbitmap.
17082 (dse_dom_walker::before_dom_children): Inline ...
17083 (pass_dse::execute): ... here. Perform a reverse program
17086 2021-05-03 H.J. Lu <hjl.tools@gmail.com>
17089 * configure: Regenerated.
17091 2021-05-03 Ilya Leoshkevich <iii@linux.ibm.com>
17094 * config/s390/s390.c (s390_hard_fp_reg_p): New function.
17095 (s390_md_asm_adjust): Handle hard registers.
17097 2021-05-03 Jakub Jelinek <jakub@redhat.com>
17099 PR tree-optimization/100382
17100 * tree-ssa-dse.c: Include tree-eh.h.
17101 (dse_dom_walker::before_dom_children): Don't remove stmts if
17102 stmt_unremovable_because_of_non_call_eh_p is true.
17104 2021-05-02 David Edelsohn <dje.gcc@gmail.com>
17106 * varasm.c (compute_reloc_for_var): Split out from...
17107 (get_variable_section): Use it.
17108 * output.h (compute_reloc_for_var): Declare.
17109 * config/rs6000/rs6000-protos.h
17110 (rs6000_xcoff_asm_output_aligned_decl_common): Change alignment to
17112 * config/rs6000/rs6000.c (rs6000_legitimize_tls_address_aix):
17113 Don't append storage mapping class to symbol.
17114 (rs6000_xcoff_asm_named_section): Add BS and UL mapping classes.
17115 Don't convert TLS BSS to common.
17116 (rs6000_xcoff_unique_section): Don't fall back to select_secton.
17117 (rs6000_xcoff_section_type_flags): Add SECTION_BSS if DECL is
17119 (rs6000_xcoff_asm_globalize_decl_name): Don't strip storage
17121 (rs6000_xcoff_asm_output_aligned_decl_common): Align is unsigned int.
17122 If align is 0 from TLS class, use the same rules as varasm.c
17123 If not common, switch to BSS section manually.
17124 If common, emit appropriate comm or lcomm directive.
17125 (rs6000_xcoff_encode_section_info): Add logic to append all
17126 storage mapping classes.
17127 (rs6000_asm_weaken_decl): Adjust for qualname symbols.
17128 * config/rs6000/xcoff.h (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
17129 rs6000_xcoff_asm_output_aligned_decl_common.
17130 (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
17131 rs6000_xcoff_asm_output_aligned_decl_common.
17132 (ASM_OUTPUT_TLS_COMMON): Use
17133 rs6000_xcoff_asm_output_aligned_decl_common.
17135 2021-05-02 Jakub Jelinek <jakub@redhat.com>
17138 * config/nvptx/nvptx.c (nvptx_sese_pseudo): Use nullptr instead of 0
17139 as first argument of pseudo_node_t constructors.
17141 2021-05-02 Jakub Jelinek <jakub@redhat.com>
17144 * config/i386/t-i386 (TM_H): Add $(srcdir)/config/i386/i386-isa.def.
17146 2021-05-01 Aldy Hernandez <aldyh@redhat.com>
17148 * value-range.cc (DEFINE_INT_RANGE_GC_STUBS): Remove.
17149 (gt_pch_nx (int_range<1> *&)): New.
17150 (gt_ggc_mx (int_range<1> *&)): New.
17151 * value-range.h (class irange): Add GTY support for
17154 2021-05-01 Geng Qi <gengqi@linux.alibaba.com>
17156 * doc/options.texi (Negative): Change either or to both and.
17158 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
17160 * config/aarch64/aarch64-simd-builtins.def: Add
17161 float_ml[as][q]_laneq builtin generator macros.
17162 * config/aarch64/aarch64-simd.md (mul_laneq<mode>3): Define.
17163 (aarch64_float_mla_laneq<mode>): Define.
17164 (aarch64_float_mls_laneq<mode>): Define.
17165 * config/aarch64/arm_neon.h (vmla_laneq_f32): Use RTL builtin
17166 instead of GCC vector extensions.
17167 (vmlaq_laneq_f32): Likewise.
17168 (vmls_laneq_f32): Likewise.
17169 (vmlsq_laneq_f32): Likewise.
17171 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
17173 * config/aarch64/aarch64-simd-builtins.def: Add
17174 float_ml[as]_lane builtin generator macros.
17175 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt<mode>):
17177 (mul_lane<mode>3): This, and re-order arguments.
17178 (aarch64_float_mla_lane<mode>): Define.
17179 (aarch64_float_mls_lane<mode>): Define.
17180 * config/aarch64/arm_neon.h (vmla_lane_f32): Use RTL builtin
17181 instead of GCC vector extensions.
17182 (vmlaq_lane_f32): Likewise.
17183 (vmls_lane_f32): Likewise.
17184 (vmlsq_lane_f32): Likewise.
17186 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
17188 * config/aarch64/aarch64-simd-builtins.def: Add float_ml[as]
17189 builtin generator macros.
17190 * config/aarch64/aarch64-simd.md (aarch64_float_mla<mode>):
17192 (aarch64_float_mls<mode>): Define.
17193 * config/aarch64/arm_neon.h (vmla_f32): Use RTL builtin
17194 instead of relying on GCC vector extensions.
17195 (vmla_f64): Likewise.
17196 (vmlaq_f32): Likewise.
17197 (vmlaq_f64): Likewise.
17198 (vmls_f32): Likewise.
17199 (vmls_f64): Likewise.
17200 (vmlsq_f32): Likewise.
17201 (vmlsq_f64): Likewise.
17202 * config/aarch64/iterators.md: Define VDQF_DF mode iterator.
17204 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
17206 * config/aarch64/aarch64-simd-builtins.def: Add
17207 float_ml[as]_n_builtin generator macros.
17208 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt_from_dup<mode>):
17210 (mul_n<mode>3): This, and re-order arguments.
17211 (aarch64_float_mla_n<mode>): Define.
17212 (aarch64_float_mls_n<mode>): Define.
17213 * config/aarch64/arm_neon.h (vmla_n_f32): Use RTL builtin
17214 instead of inline asm.
17215 (vmlaq_n_f32): Likewise.
17216 (vmls_n_f32): Likewise.
17217 (vmlsq_n_f32): Likewise.
17219 2021-04-30 Jonathan Wright <joanthan.wright@arm.com>
17221 * config/aarch64/aarch64-simd-builtins.def: Add pmull[2]
17222 builtin generator macros.
17223 * config/aarch64/aarch64-simd.md (aarch64_pmullv8qi): Define.
17224 (aarch64_pmull_hiv16qi_insn): Define.
17225 (aarch64_pmull_hiv16qi): Define.
17226 * config/aarch64/arm_neon.h (vmull_high_p8): Use RTL builtin
17227 instead of inline asm.
17228 (vmull_p8): Likewise.
17230 2021-04-30 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
17232 * config/avr/avr.md: Adjust peepholes to match and
17233 generate parallels with clobber of REG_CC.
17234 (mov<mode>_insn): Rename to mov<mode>_insn_split.
17235 (*mov<mode>_insn): Rename to mov<mode>_insn.
17237 2021-04-30 David Edelsohn <dje.gcc@gmail.com>
17239 * varasm.c (use_blocks_for_decl_p): Don't use section anchors
17240 for VAR_DECLs if -fdata-sections enabled.
17242 2021-04-30 Michael Meissner <meissner@linux.ibm.com>
17244 PR bootstrap/100327
17245 * config/rs6000/rs6000.c
17246 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Define.
17247 (rs6000_libgcc_floating_mode_supported_p): New target hook.
17249 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
17251 * tree-ssa-threadbackward.c (class thread_jumps): Split out code
17253 (class back_threader_registry): ...to here...
17254 (class back_threader_profitability): ...and here...
17255 (thread_jumps::thread_through_all_blocks): Remove argument.
17256 (back_threader_registry::back_threader_registry): New.
17257 (back_threader_registry::~back_threader_registry): New.
17258 (back_threader_registry::thread_through_all_blocks): New.
17259 (thread_jumps::profitable_jump_thread_path): Move from here...
17260 (back_threader_profitability::profitable_path_p): ...to here.
17261 (thread_jumps::find_taken_edge): New.
17262 (thread_jumps::convert_and_register_current_path): Move...
17263 (back_threader_registry::register_path): ...to here.
17264 (thread_jumps::register_jump_thread_path_if_profitable): Move...
17265 (thread_jumps::maybe_register_path): ...to here.
17266 (thread_jumps::handle_phi): Call find_taken_edge and
17267 maybe_register_path.
17268 (thread_jumps::handle_assignment): Same.
17269 (thread_jumps::fsm_find_control_statement_thread_paths): Remove
17270 tree argument to handle_phi and handle_assignment.
17271 (thread_jumps::find_jump_threads_backwards): Set m_name. Remove
17272 set of m_speed_p and m_max_threaded_paths.
17273 (pass_thread_jumps::execute): Remove second argument from
17274 find_jump_threads_backwards.
17275 (pass_early_thread_jumps::execute): Same.
17277 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
17279 * tree-ssa-dom.c (class dom_jump_threader_simplifier): New.
17280 (class dom_opt_dom_walker): Initialize some class variables.
17281 (pass_dominator::execute): Pass evrp_range_analyzer and
17282 dom_jump_threader_simplifier to dom_opt_dom_walker.
17283 Adjust for some functions moving into classes.
17284 (simplify_stmt_for_jump_threading): Adjust and move to...
17285 (jump_threader_simplifier::simplify): ...here.
17286 (dom_opt_dom_walker::before_dom_children): Adjust for
17287 m_evrp_range_analyzer.
17288 (dom_opt_dom_walker::after_dom_children): Remove x_vr_values hack.
17289 (test_for_singularity): Place in dom_opt_dom_walker class.
17290 (dom_opt_dom_walker::optimize_stmt): The argument
17291 evrp_range_analyzer is now a class field.
17292 * tree-ssa-threadbackward.c (class thread_jumps): Add m_registry.
17293 (thread_jumps::thread_through_all_blocks): New.
17294 (thread_jumps::convert_and_register_current_path): Use m_registry.
17295 (pass_thread_jumps::execute): Adjust for thread_through_all_blocks
17296 being in the threader class.
17297 (pass_early_thread_jumps::execute): Same.
17298 * tree-ssa-threadedge.c (threadedge_initialize_values): Move...
17299 (jump_threader::jump_threader): ...here.
17300 (threadedge_finalize_values): Move...
17301 (jump_threader::~jump_threader): ...here.
17302 (jump_threader::remove_jump_threads_including): New.
17303 (jump_threader::thread_through_all_blocks): New.
17304 (record_temporary_equivalences_from_phis): Move...
17305 (jump_threader::record_temporary_equivalences_from_phis): ...here.
17306 (record_temporary_equivalences_from_stmts_at_dest): Move...
17307 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
17309 (simplify_control_stmt_condition_1): Move to jump_threader class.
17310 (simplify_control_stmt_condition): Move...
17311 (jump_threader::simplify_control_stmt_condition): ...here.
17312 (thread_around_empty_blocks): Move...
17313 (jump_threader::thread_around_empty_blocks): ...here.
17314 (thread_through_normal_block): Move...
17315 (jump_threader::thread_through_normal_block): ...here.
17316 (thread_across_edge): Move...
17317 (jump_threader::thread_across_edge): ...here.
17318 (thread_outgoing_edges): Move...
17319 (jump_threader::thread_outgoing_edges): ...here.
17320 * tree-ssa-threadedge.h: Move externally facing functings...
17321 (class jump_threader): ...here...
17322 (class jump_threader_simplifier): ...and here.
17323 * tree-ssa-threadupdate.c (struct redirection_data): Remove comment.
17324 (jump_thread_path_allocator::jump_thread_path_allocator): New.
17325 (jump_thread_path_allocator::~jump_thread_path_allocator): New.
17326 (jump_thread_path_allocator::allocate_thread_edge): New.
17327 (jump_thread_path_allocator::allocate_thread_path): New.
17328 (jump_thread_path_registry::jump_thread_path_registry): New.
17329 (jump_thread_path_registry::~jump_thread_path_registry): New.
17330 (jump_thread_path_registry::allocate_thread_edge): New.
17331 (jump_thread_path_registry::allocate_thread_path): New.
17332 (dump_jump_thread_path): Make extern.
17333 (debug (const vec<jump_thread_edge *> &path)): New.
17334 (struct removed_edges): Move to tree-ssa-threadupdate.h.
17335 (struct thread_stats_d): Remove.
17336 (remove_ctrl_stmt_and_useless_edges): Make static.
17337 (lookup_redirection_data): Move...
17338 (jump_thread_path_registry::lookup_redirection_data): ...here.
17339 (ssa_redirect_edges): Make static.
17340 (thread_block_1): Move...
17341 (jump_thread_path_registry::thread_block_1): ...here.
17342 (thread_block): Move...
17343 (jump_thread_path_registry::thread_block): ...here.
17344 (thread_through_loop_header): Move...
17345 (jump_thread_path_registry::thread_through_loop_header): ...here.
17346 (mark_threaded_blocks): Move...
17347 (jump_thread_path_registry::mark_threaded_blocks): ...here.
17348 (debug_path): Move...
17349 (jump_thread_path_registry::debug_path): ...here.
17350 (debug_all_paths): Move...
17351 (jump_thread_path_registry::dump): ..here.
17352 (rewire_first_differing_edge): Move...
17353 (jump_thread_path_registry::rewire_first_differing_edge): ...here.
17354 (adjust_paths_after_duplication): Move...
17355 (jump_thread_path_registry::adjust_paths_after_duplication): ...here.
17356 (duplicate_thread_path): Move...
17357 (jump_thread_path_registry::duplicate_thread_path): ..here.
17358 (remove_jump_threads_including): Move...
17359 (jump_thread_path_registry::remove_jump_threads_including): ...here.
17360 (thread_through_all_blocks): Move to...
17361 (jump_thread_path_registry::thread_through_all_blocks): ...here.
17362 (delete_jump_thread_path): Remove.
17363 (register_jump_thread): Move...
17364 (jump_thread_path_registry::register_jump_thread): ...here.
17365 * tree-ssa-threadupdate.h: Move externally facing functions...
17366 (class jump_thread_path_allocator): ...here...
17367 (class jump_thread_path_registry): ...and here.
17368 (thread_through_all_blocks): Remove.
17369 (struct removed_edges): New.
17370 (register_jump_thread): Remove.
17371 (remove_jump_threads_including): Remove.
17372 (delete_jump_thread_path): Remove.
17373 (remove_ctrl_stmt_and_useless_edges): Remove.
17374 (free_dom_edge_info): New prototype.
17375 * tree-vrp.c: Remove x_vr_values hack.
17376 (class vrp_jump_threader_simplifier): New.
17377 (vrp_jump_threader_simplifier::simplify): New.
17378 (vrp_jump_threader::vrp_jump_threader): Adjust method signature.
17379 Remove m_dummy_cond.
17380 Instantiate m_simplifier and m_threader.
17381 (vrp_jump_threader::thread_through_all_blocks): New.
17382 (vrp_jump_threader::simplify_stmt): Remove.
17383 (vrp_jump_threader::after_dom_children): Do not set m_dummy_cond.
17384 Remove x_vr_values hack.
17385 (execute_vrp): Adjust for thread_through_all_blocks being in a
17388 2021-04-30 Christophe Lyon <christophe.lyon@linaro.org>
17390 * genflags.c (gen_insn): Print failed expansion string.
17392 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
17394 * expr.c (alignment_for_piecewise_move): Call mode_for_size
17395 without limit to MAX_FIXED_MODE_SIZE.
17397 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
17399 PR middle-end/90773
17400 * builtins.c (builtin_memset_gen_str): Don't use return from
17401 simplify_gen_subreg.
17403 2021-04-30 Uroš Bizjak <ubizjak@gmail.com>
17406 * config/i386/i386.md (*add<mode>3_carry_0r): New insn pattern.
17407 (*addsi3_carry_zext_0r): Ditto.
17408 (*sub<mode>3_carry_0): Ditto.
17409 (*subsi3_carry_zext_0r): Ditto.
17410 * config/i386/predicates.md (ix86_carry_flag_unset_operator):
17412 * config/i386/i386.c (ix86_rtx_costs) <case PLUS, case MINUS>:
17413 Also consider ix86_carry_flag_unset_operator to calculate
17414 the cost of adc/sbb insn.
17416 2021-04-30 Roman Zhuykov <zhroma@ispras.ru>
17418 PR rtl-optimization/100225
17419 PR rtl-optimization/84878
17420 * modulo-sched.c (sms_schedule): Use note_stores to skip loops
17421 where we have an instruction which touches (writes) any hard
17422 register from df->regular_block_artificial_uses set.
17423 Allow not-single-set instruction only right before basic block
17426 2021-04-30 Geng Qi <gengqi@linux.alibaba.com>
17428 * config/riscv/riscv.opt (march=,mabi=): Negative itself.
17430 2021-04-30 LevyHsu <admin@levyhsu.com>
17432 * config/riscv/riscv.c (riscv_min_arithmetic_precision): New.
17433 * config/riscv/riscv.h (TARGET_MIN_ARITHMETIC_PRECISION): New.
17434 * config/riscv/riscv.md (addv<mode>4, uaddv<mode>4): New.
17435 (subv<mode>4, usubv<mode>4, mulv<mode>4, umulv<mode>4): New.
17437 2021-04-29 Alexandre Oliva <oliva@adacore.com>
17439 * config.gcc: Merged x86 and x86_64 cpu_type-setting cases.
17441 2021-04-29 Alexandre Oliva <oliva@adacore.com>
17443 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Rename to...
17444 (ASM_OUTPUT_MAX_SKIP_ALIGN): ... this. Enclose in do/while(0).
17445 * config/i386/i386.c: Adjust.
17446 * config/i386/i386.md: Adjust.
17447 * config/i386/darwin.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Drop.
17448 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17449 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17450 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17451 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17452 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17453 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17454 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17455 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17456 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17457 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
17459 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
17461 * config/i386/i386-expand.c (ix86_expand_int_compare):
17462 Swap operands of GTU and LEU comparison to emit carry flag comparison.
17463 * config/i386/i386.md (*add<mode>3_carry_0): Change insn
17464 predicate to allow more combine opportunities with memory operands.
17465 (*sub<mode>3_carry_0): Ditto.
17467 2021-04-29 Richard Sandiford <richard.sandiford@arm.com>
17469 PR rtl-optimization/100303
17470 * rtl-ssa/accesses.cc (function_info::make_use_available): Take a
17471 boolean that indicates whether the use will only be used in
17472 debug instructions. Treat it in the same way that existing
17473 cross-EBB debug references would be handled if so.
17474 (function_info::make_uses_available): Likewise.
17475 * rtl-ssa/functions.h (function_info::make_uses_available): Update
17476 prototype accordingly.
17477 (function_info::make_uses_available): Likewise.
17478 * fwprop.c (try_fwprop_subst): Update call accordingly.
17480 2021-04-29 Jeff Law <jlaw@tachyum.com>
17482 * config/nios2/nios2-protos.h (nios2_fpu_insn_enabled): Move outside
17485 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
17486 Richard Biener <rguenther@suse.de>
17489 * config/i386/i386-builtin.def (IX86_BUILTIN_MASKLOADPD)
17490 (IX86_BUILTIN_MASKLOADPS, IX86_BUILTIN_MASKLOADPD256)
17491 (IX86_BUILTIN_MASKLOADPS256, IX86_BUILTIN_MASKLOADD)
17492 (IX86_BUILTIN_MASKLOADQ, IX86_BUILTIN_MASKLOADD256)
17493 (IX86_BUILTIN_MASKLOADQ256): Move from SPECIAL_ARGS
17494 to PURE_ARGS category.
17495 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
17496 Handle PURE_ARGS category.
17497 * config/i386/i386-expand.c (ix86_expand_builtin): Ditto.
17499 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
17501 * configure.ac: Check for the presence of sys/locking.h header and
17502 for whether _LK_LOCK is supported by _locking.
17503 * configure: Regenerate.
17504 * config.in: Likewise.
17505 * gcov-io.h: Define GCOV_LOCKED_WITH_LOCKING if HOST_HAS_LK_LOCK.
17506 * gcov-io.c (gcov_open): Add support for GCOV_LOCKED_WITH_LOCKING.
17507 * system.h: Include <sys/locking.h> if HAVE_SYS_LOCKING_H.
17509 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
17511 * config/i386/predicates.md (fcmov_comparison_operator):
17512 Do not check for trivial FP comparison operator.
17513 <case GEU, case LTU>: Allow CCGZmode.
17514 <case GTU, case LEU>: Do not allow CCCmode.
17515 (ix86_comparison_operator) <case GTU, case LEU>: Allow only CCmode.
17516 (ix86_carry_flag_operator): Match only LTU and UNLT code.
17517 Do not check for trivial FP comparison operator. Allow CCGZmode.
17519 2021-04-29 Tom de Vries <tdevries@suse.de>
17521 * omp-expand.c (expand_omp_simd): Add step_orig, and replace uses of
17522 fd->loop.step by either step or orig_step.
17524 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
17526 * config/sparc/sparc.c (gen_load_pcrel_sym): Delete.
17527 (load_got_register): Do the PIC dance here.
17528 (sparc_legitimize_tls_address): Simplify.
17529 (sparc_emit_probe_stack_range): Likewise.
17530 (sparc32_initialize_trampoline): Likewise.
17531 (sparc64_initialize_trampoline): Likewise.
17532 * config/sparc/sparc.md (load_pcrel_sym<P:mode>): Add @ marker.
17533 (probe_stack_range<P:mode>): Likewise.
17534 (flush<P:mode>): Likewise.
17535 (tgd_hi22<P:mode>): Likewise.
17536 (tgd_lo10<P:mode>): Likewise.
17537 (tgd_add<P:mode>): Likewise.
17538 (tgd_call<P:mode>): Likewise.
17539 (tldm_hi22<P:mode>): Likewise.
17540 (tldm_lo10<P:mode>): Likewise.
17541 (tldm_add<P:mode>): Likewise.
17542 (tldm_call<P:mode>): Likewise.
17543 (tldo_hix22<P:mode>): Likewise.
17544 (tldo_lox10<P:mode>): Likewise.
17545 (tldo_add<P:mode>): Likewise.
17546 (tie_hi22<P:mode>): Likewise.
17547 (tie_lo10<P:mode>): Likewise.
17548 (tie_add<P:mode>): Likewise.
17549 (tle_hix22<P:mode>): Likewise.
17550 (tle_lox10<P:mode>): Likewise.
17551 (stack_protect_setsi): Rename to...
17552 (stack_protect_set32): ...this.
17553 (stack_protect_setdi): Rename to...
17554 (stack_protect_set64): ...this.
17555 (stack_protect_set): Adjust calls to above.
17556 (stack_protect_testsi): Rename to...
17557 (stack_protect_test32): ...this.
17558 (stack_protect_testdi): Rename to...
17559 (stack_protect_test64): ...this.
17560 (stack_protect_test): Adjust calls to above.
17562 2021-04-29 H.J. Lu <hjl.tools@gmail.com>
17564 PR middle-end/90773
17565 * builtins.c (builtin_memcpy_read_str): Add a dummy argument.
17566 (builtin_strncpy_read_str): Likewise.
17567 (builtin_memset_read_str): Add an argument for the previous RTL
17568 information and generate the new RTL from the previous RTL info.
17569 (builtin_memset_gen_str): Likewise.
17570 * builtins.h (builtin_strncpy_read_str): Update the prototype.
17571 (builtin_memset_read_str): Likewise.
17572 * expr.c (by_pieces_ninsns): If targetm.overlap_op_by_pieces_p()
17573 returns true, round up size and alignment to the widest integer
17574 mode for maximum size.
17575 (pieces_addr::adjust): Add a pointer to by_pieces_prev argument
17576 and pass it to m_constfn.
17577 (op_by_pieces_d): Add m_push and m_overlap_op_by_pieces.
17578 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
17579 initialize m_push. Initialize m_overlap_op_by_pieces with
17580 targetm.overlap_op_by_pieces_p ().
17581 (op_by_pieces_d::run): Pass the previous RTL information to
17582 pieces_addr::adjust and generate overlapping operations if
17583 m_overlap_op_by_pieces is true.
17585 (move_by_pieces_d::move_by_pieces_d): Updated for op_by_pieces_d
17587 (store_by_pieces_d::store_by_pieces_d): Updated for op_by_pieces_d
17589 (can_store_by_pieces): Use by_pieces_constfn on constfun.
17590 (store_by_pieces): Use by_pieces_constfn on constfun. Updated
17591 for op_by_pieces_d change.
17592 (clear_by_pieces_1): Add a dummy argument.
17593 (clear_by_pieces): Updated for op_by_pieces_d change.
17594 (compare_by_pieces_d::compare_by_pieces_d): Likewise.
17595 (string_cst_read_str): Add a dummy argument.
17596 * expr.h (by_pieces_constfn): Add a dummy argument.
17597 (by_pieces_prev): New.
17598 * target.def (overlap_op_by_pieces_p): New target hook.
17599 * config/i386/i386.c (TARGET_OVERLAP_OP_BY_PIECES_P): New.
17600 * doc/tm.texi.in: Add TARGET_OVERLAP_OP_BY_PIECES_P.
17601 * doc/tm.texi: Regenerated.
17603 2021-04-29 Richard Biener <rguenther@suse.de>
17605 PR tree-optimization/100253
17606 * tree-vect-stmts.c (vectorizable_load): Do not assume
17607 element alignment when DR_MISALIGNMENT is -1.
17608 (vectorizable_store): Likewise.
17610 2021-04-29 Jakub Jelinek <jakub@redhat.com>
17613 * config/aarch64/aarch64.c (aarch64_add_offset_1_temporaries): Use
17614 absu_hwi instead of abs_hwi.
17616 2021-04-29 Richard Biener <rguenther@suse.de>
17618 PR middle-end/38474
17619 * tree-ssa-structalias.c (add_graph_edge): Avoid direct
17620 forwarding when indirect forwarding through ESCAPED
17623 2021-04-29 Tom de Vries <tdevries@suse.de>
17626 * internal-fn.c (expand_GOMP_SIMT_ENTER_ALLOC)
17627 (expand_GOMP_SIMT_LAST_LANE, expand_GOMP_SIMT_ORDERED_PRED)
17628 (expand_GOMP_SIMT_VOTE_ANY, expand_GOMP_SIMT_XCHG_BFLY)
17629 (expand_GOMP_SIMT_XCHG_IDX): Ensure target is assigned to.
17631 2021-04-29 Richard Biener <rguenther@suse.de>
17633 PR tree-optimization/99912
17634 * tree-ssa-dse.c (dse_dom_walker::m_need_cfg_cleanup): New.
17635 (dse_dom_walker::todo): Likewise.
17636 (dse_dom_walker::dse_optimize_stmt): Move VDEF check to the
17638 (dse_dom_walker::before_dom_children): Remove trivially
17639 dead SSA defs and schedule CFG cleanup if we removed all
17641 (pass_dse::execute): Get TODO as computed by the DOM walker
17642 and return it. Wipe dominator info earlier.
17644 2021-04-29 Richard Biener <rguenther@suse.de>
17647 * ipa-prop.c (ipcp_modif_dom_walker::before_dom_children):
17648 Track blocks to cleanup EH in new m_need_eh_cleanup.
17649 (ipcp_modif_dom_walker::cleanup_eh): New.
17650 (ipcp_transform_function): Release dominator info before
17653 2021-04-29 Martin Sebor <msebor@redhat.com>
17655 PR middle-end/100250
17656 * attribs.c (attr_access::array_as_string): Avoid dereferencing
17657 a pointer when it's null.
17659 2021-04-29 Martin Sebor <msebor@redhat.com>
17661 * Makefile.in (OBJS): Add ipa-free-lang-data.o.
17662 * ipa-free-lang-data.cc: New file.
17663 * tree.c: Move pass free_lang_data to file above.
17664 (build_array_type_1): Declare extern.
17665 * tree.h (build_array_type_1): Declare.
17667 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17669 * config/aarch64/aarch64-simd-builtins.def: Modify comment to
17670 make consistent with updated RTL pattern.
17671 * config/aarch64/aarch64-simd.md (aarch64_<sur>qmovn<mode>):
17672 Implement using ss_truncate and us_truncate rather than
17674 * config/aarch64/iterators.md: Remove redundant unspecs and
17675 iterator: UNSPEC_[SU]QXTN and SUQMOVN respectively.
17677 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17679 * config/aarch64/arm_acle.h (__attribute__): Make intrinsic
17680 attributes consistent with those defined in arm_neon.h.
17682 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17684 * config/aarch64/arm_fp16.h (__attribute__): Make intrinsic
17685 attributes consistent with those defined in arm_neon.h.
17687 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17689 * config/aarch64/aarch64-simd-builtins.def: Add
17690 float_trunc_rodd builtin generator macros.
17691 * config/aarch64/aarch64-simd.md (aarch64_float_trunc_rodd_df):
17693 (aarch64_float_trunc_rodd_lo_v2sf): Define.
17694 (aarch64_float_trunc_rodd_hi_v4sf_le): Define.
17695 (aarch64_float_trunc_rodd_hi_v4sf_be): Define.
17696 (aarch64_float_trunc_rodd_hi_v4sf): Define.
17697 * config/aarch64/arm_neon.h (vcvtx_f32_f64): Use RTL builtin
17698 instead of inline asm.
17699 (vcvtx_high_f32_f64): Likewise.
17700 (vcvtxd_f32_f64): Likewise.
17701 * config/aarch64/iterators.md: Add FCVTXN unspec.
17703 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17705 * config/aarch64/aarch64-simd-builtins.def: Add tbx1 builtin
17707 * config/aarch64/aarch64-simd.md (aarch64_tbx1<mode>):
17709 * config/aarch64/arm_neon.h (vqtbx1_s8): USE RTL builtin
17710 instead of inline asm.
17711 (vqtbx1_u8): Likewise.
17712 (vqtbx1_p8): Likewise.
17713 (vqtbx1q_s8): Likewise.
17714 (vqtbx1q_u8): Likewise.
17715 (vqtbx1q_p8): Likewise.
17716 (vtbx2_s8): Likewise.
17717 (vtbx2_u8): Likewise.
17718 (vtbx2_p8): Likewise.
17720 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17722 * config/aarch64/aarch64-simd-builtins.def: Add tbl1 builtin
17724 * config/aarch64/arm_neon.h (vqtbl1_p8): Use RTL builtin
17725 instead of inline asm.
17726 (vqtbl1_s8): Likewise.
17727 (vqtbl1_u8): Likewise.
17728 (vqtbl1q_p8): Likewise.
17729 (vqtbl1q_s8): Likewise.
17730 (vqtbl1q_u8): Likewise.
17731 (vtbl1_s8): Likewise.
17732 (vtbl1_u8): Likewise.
17733 (vtbl1_p8): Likewise.
17734 (vtbl2_s8): Likewise.
17735 (vtbl2_u8): Likewise.
17736 (vtbl2_p8): Likewise.
17738 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17740 * config/aarch64/aarch64-simd-builtins.def: Add polynomial
17741 ssri_n buitin generator macro.
17742 * config/aarch64/arm_neon.h (vsri_n_p8): Use RTL builtin
17743 instead of inline asm.
17744 (vsri_n_p16): Likewise.
17745 (vsri_n_p64): Likewise.
17746 (vsriq_n_p8): Likewise.
17747 (vsriq_n_p16): Likewise.
17748 (vsriq_n_p64): Likewise.
17750 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17752 * config/aarch64/aarch64-simd-builtins.def: Use VALLP mode
17753 iterator for polynomial ssli_n builtin generator macro.
17754 * config/aarch64/arm_neon.h (vsli_n_p8): Use RTL builtin
17755 instead of inline asm.
17756 (vsli_n_p16): Likewise.
17757 (vsliq_n_p8): Likewise.
17758 (vsliq_n_p16): Likewise.
17759 * config/aarch64/iterators.md: Define VALLP mode iterator.
17761 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17763 * config/aarch64/aarch64-simd-builtins.def: Use VDQV_L
17764 iterator to generate [su]adalp RTL builtins.
17765 * config/aarch64/aarch64-simd.md: Use VDQV_L iterator in
17766 [su]adalp RTL pattern.
17767 * config/aarch64/arm_neon.h (vpadal_s32): Use RTL builtin
17768 instead of inline asm.
17769 (vpadal_u32): Likewise.
17771 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17773 * config/aarch64/aarch64-simd-builtins.def: Add [su]addlp
17774 builtin generator macros.
17775 * config/aarch64/aarch64-simd.md (aarch64_<su>addlp<mode>):
17777 * config/aarch64/arm_neon.h (vpaddl_s8): Use RTL builtin
17778 instead of inline asm.
17779 (vpaddl_s16): Likewise.
17780 (vpaddl_s32): Likewise.
17781 (vpaddl_u8): Likewise.
17782 (vpaddl_u16): Likewise.
17783 (vpaddl_u32): Likewise.
17784 (vpaddlq_s8): Likewise.
17785 (vpaddlq_s16): Likewise.
17786 (vpaddlq_s32): Likewise.
17787 (vpaddlq_u8): Likewise.
17788 (vpaddlq_u16): Likewise.
17789 (vpaddlq_u32): Liwewise.
17790 * config/aarch64/iterators.md: Define [SU]ADDLP unspecs with
17791 appropriate attributes.
17793 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17795 * config/aarch64/aarch64-simd-builtins.def: Use VDQ_I iterator
17796 for aarch64_addp<mode> builtin macro generator.
17797 * config/aarch64/aarch64-simd.md: Use VDQ_I iterator in
17798 aarch64_addp<mode> RTL pattern.
17799 * config/aarch64/arm_neon.h (vpaddq_s8): Use RTL builtin
17800 instead of inline asm.
17801 (vpaddq_s16): Likewise.
17802 (vpaddq_s32): Likewise.
17803 (vpaddq_s64): Likewise.
17804 (vpaddq_u8): Likewise.
17805 (vpaddq_u16): Likewise.
17806 (vpaddq_u32): Likewise.
17807 (vpaddq_u64): Likewise.
17809 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17811 * config/aarch64/aarch64-simd-builtins.def: Add sq[r]dmulh_n
17812 builtin generator macros.
17813 * config/aarch64/aarch64-simd.md (aarch64_sq<r>dmulh_n<mode>):
17815 * config/aarch64/arm_neon.h (vqdmulh_n_s16): Use RTL builtin
17816 instead of inline asm.
17817 (vqdmulh_n_s32): Likewise.
17818 (vqdmulhq_n_s16): Likewise.
17819 (vqdmulhq_n_s32): Likewise.
17820 (vqrdmulh_n_s16): Likewise.
17821 (vqrdmulh_n_s32): Likewise.
17822 (vqrdmulhq_n_s16): Likewise.
17823 (vqrdmulhq_n_s32): Likewise.
17825 2021-04-28 Tobias Burnus <tobias@codesourcery.com>
17827 * doc/install.texi (--enable-offload-defaulted): Document.
17829 2021-04-28 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
17831 * config/avr/avr-dimode.md: Turn existing patterns into
17832 define_insn_and_split style patterns where the splitter
17833 adds a clobber of the condition code register. Drop "cc"
17834 attribute. Add new patterns to match output of
17836 * config/avr/avr-fixed.md: Likewise.
17837 * config/avr/avr.c (cc_reg_rtx): New.
17838 (avr_parallel_insn_from_insns): Adjust insn count
17839 for removal of set of cc0.
17840 (avr_is_casesi_sequence): Likewise.
17841 (avr_casei_sequence_check_operands): Likewise.
17842 (avr_optimize_casesi): Likewise. Also insert
17843 new insns after jump_insn.
17844 (avr_pass_casesi::avr_rest_of_handle_casesi): Adjust
17845 for removal of set of cc0.
17846 (avr_init_expanders): Initialize cc_reg_rtx.
17847 (avr_regno_reg_class): Handle REG_CC.
17848 (cond_string): Remove usage of CC_OVERFLOW_UNUSABLE.
17849 (avr_notice_update_cc): Remove function.
17850 (ret_cond_branch): Remove usage of CC_OVERFLOW_UNUSABLE.
17851 (compare_condition): Adjust for PARALLEL with
17853 (out_shift_with_cnt): Likewise.
17854 (ashlhi3_out): Likewise.
17855 (ashrhi3_out): Likewise.
17856 (lshrhi3_out): Likewise.
17857 (avr_class_max_nregs): Return single reg for REG_CC.
17858 (avr_compare_pattern): Check for REG_CC instead
17860 (avr_reorg_remove_redundant_compare): Likewise.
17861 (avr_reorg):Adjust for PARALLEL with REG_CC clobber.
17862 (avr_hard_regno_nregs): Return single reg for REG_CC.
17863 (avr_hard_regno_mode_ok): Allow only CCmode for REG_CC.
17864 (avr_md_asm_adjust): Clobber REG_CC.
17865 (TARGET_HARD_REGNO_NREGS): Define.
17866 (TARGET_CLASS_MAX_NREGS): Define.
17867 (TARGET_MD_ASM_ADJUST): Define.
17868 * config/avr/avr.h (FIRST_PSEUDO_REGISTER): Adjust
17870 (enum reg_class): Add CC_REG class.
17871 (NOTICE_UPDATE_CC): Remove.
17872 (CC_OVERFLOW_UNUSABLE): Remove.
17873 (CC_NO_CARRY): Remove.
17874 * config/avr/avr.md: Turn existing patterns into
17875 define_insn_and_split style patterns where the splitter
17876 adds a clobber of the condition code register. Drop "cc"
17877 attribute. Add new patterns to match output of
17879 (sez): Remove unused pattern.
17881 2021-04-28 Richard Earnshaw <rearnsha@arm.com>
17884 * config/arm/arm.c (arm_hard_regno_mode_ok): Only allow VPR to be
17887 2021-04-28 Richard Sandiford <richard.sandiford@arm.com>
17890 * config/aarch64/constraints.md (Utq): Require the address to
17891 be valid for both the element mode and for V2DImode.
17893 2021-04-28 Jakub Jelinek <jakub@redhat.com>
17894 Tobias Burnus <tobias@codesourcery.com>
17896 * configure.ac (OFFLOAD_DEFAULTED): AC_DEFINE if offload-defaulted.
17897 * gcc.c (process_command): New variable.
17898 (driver::maybe_putenv_OFFLOAD_TARGETS): If OFFLOAD_DEFAULTED,
17899 set it if -foffload is defaulted.
17900 * lto-wrapper.c (OFFLOAD_TARGET_DEFAULT_ENV): Define.
17901 (compile_offload_image): If OFFLOAD_DEFAULTED and
17902 OFFLOAD_TARGET_DEFAULT is in the environment, don't fail
17903 if corresponding mkoffload can't be found.
17904 (compile_images_for_offload_targets): Likewise. Free and clear
17905 offload_names if no valid offload is found.
17906 * config.in: Regenerate.
17907 * configure: Regenerate.
17909 2021-04-28 Richard Biener <rguenther@suse.de>
17911 PR tree-optimization/100292
17912 * tree-vect-generic.c (expand_vector_condition): Do not fold
17915 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
17917 * config/rs6000/aix.h (SUBTARGET_DRIVER_SELF_SPECS): New.
17918 * config/rs6000/aix64.opt (m64): New.
17921 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
17923 * config/vax/vax.c (print_operand_address, vax_address_cost_1)
17924 (index_term_p): Handle ASHIFT too.
17926 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
17928 * config/vax/builtins.md (jbb<ccss>i<mode>): Remove operand #3.
17929 (sync_lock_test_and_set<mode>): Adjust accordingly.
17930 (sync_lock_release<mode>): Likewise.
17932 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
17934 * config/vax/vax-protos.h (adjacent_operands_p): Remove
17936 * config/vax/vax.c (adjacent_operands_p): Remove.
17938 2021-04-27 Maciej W. Rozycki <macro@linux-mips.org>
17940 * ifcvt.c (dead_or_predicable) [!IFCVT_MODIFY_TESTS]: Fall
17941 through to the non-conditional execution case if getting the
17942 condition for conditional execution has failed.
17944 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
17946 PR middle-end/100284
17947 * gimple.c (gimple_could_trap_p_1): Remove VEC_COND_EXPR test.
17948 * tree-eh.c (operation_could_trap_p): Handle VEC_COND_EXPR rather
17949 than asserting on it.
17951 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
17953 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): Protect
17954 with TARGET_AIX_OS.
17956 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
17959 * calls.c (precompute_register_parameters): Additionally test
17960 targetm.precompute_tls_p to pre-compute argument.
17961 * config/rs6000/aix.h (TARGET_PRECOMPUTE_TLS_P): Define.
17962 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): New.
17963 * target.def (precompute_tls_p): New.
17964 * doc/tm.texi.in (TARGET_PRECOMPUTE_TLS_P): Add hook documentation.
17965 * doc/tm.texi: Regenerated.
17967 2021-04-27 Jakub Jelinek <jakub@redhat.com>
17970 * config/aarch64/aarch64.c (aarch64_print_operand): Cast -UINTVAL
17971 back to HOST_WIDE_INT.
17973 2021-04-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
17976 * simplify-rtx.c (simplify_context::simplify_subreg): Check the
17977 memory alignment for the outer mode.
17979 2021-04-27 H.J. Lu <hjl.tools@gmail.com>
17981 PR middle-end/90773
17982 * expr.c (op_by_pieces_d::get_usable_mode): New member function.
17983 (op_by_pieces_d::run): Cange a while loop to a do-while loop.
17985 2021-04-27 Alex Coplan <alex.coplan@arm.com>
17988 * config/arm/arm.c (arm_split_compare_and_swap): Fix up codegen
17989 with negative immediates: ensure we expand cbranchsi4_scratch
17990 correctly and ensure we satisfy its constraints.
17991 * config/arm/sync.md
17992 (@atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1): Don't
17993 attempt to tie two output operands together with constraints;
17994 collapse two alternatives.
17995 (@atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): Likewise.
17996 * config/arm/thumb1.md (cbranchsi4_neg_late): New.
17998 2021-04-27 Jakub Jelinek <jakub@redhat.com>
18001 * config/aarch64/predicates.md (aarch64_sub_immediate,
18002 aarch64_plus_immediate): Use -UINTVAL instead of -INTVAL.
18003 * config/aarch64/aarch64.md (casesi, rotl<mode>3): Likewise.
18004 * config/aarch64/aarch64.c (aarch64_print_operand,
18005 aarch64_split_atomic_op, aarch64_expand_subvti): Likewise.
18007 2021-04-27 Jakub Jelinek <jakub@redhat.com>
18009 PR tree-optimization/100239
18010 * tree-vect-generic.c (lower_vec_perm): Don't accept constant
18011 permutations with all indices from the first zero element as vec_shl.
18013 2021-04-27 Jakub Jelinek <jakub@redhat.com>
18015 PR rtl-optimization/100254
18016 * cfgcleanup.c (outgoing_edges_match): Check REG_EH_REGION on
18017 last1 and last2 insns rather than BB_END (bb1) and BB_END (bb2) insns.
18019 2021-04-27 Richard Biener <rguenther@suse.de>
18021 PR tree-optimization/99912
18022 * passes.def: Add comment about new TODO_remove_unused_locals.
18023 * tree-stdarg.c (pass_data_stdarg): Run TODO_remove_unused_locals
18026 2021-04-27 Richard Biener <rguenther@suse.de>
18028 PR tree-optimization/99912
18029 * passes.def (pass_all_optimizations): Add pass_dse before
18030 the first pass_dce, move the first pass_dse before the
18031 pass_dce following pass_pre.
18033 2021-04-27 Jakub Jelinek <jakub@redhat.com>
18035 PR tree-optimization/95527
18036 * generic-match-head.c: Include tm.h.
18037 * gimple-match-head.c: Include tm.h.
18038 * match.pd (CLZ == INTEGER_CST): Don't use
18039 #ifdef CLZ_DEFINED_VALUE_AT_ZERO, only test CLZ_DEFINED_VALUE_AT_ZERO
18040 if clz == CFN_CLZ. Add missing val declaration.
18041 (CTZ cmp CST): New simplifications.
18043 2021-04-27 Jakub Jelinek <jakub@redhat.com>
18045 PR tree-optimization/96696
18046 * expr.c (expand_expr_divmod): New function.
18047 (expand_expr_real_2) <case TRUNC_DIV_EXPR>: Use it for truncations and
18048 divisions. Formatting fixes.
18049 <case MULT_EXPR>: Optimize x / y * y as x - x % y if the latter is
18052 2021-04-27 Martin Jambor <mjambor@suse.cz>
18055 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call):
18056 If removing a call statement LHS SSA name, release it.
18058 2021-04-27 Richard Earnshaw <rearnsha@arm.com>
18061 * config/arm/arm.c (THUMB2_WORK_REGS): Check PIC_OFFSET_TABLE_REGNUM
18062 is valid before including it in the mask.
18064 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
18067 * config/aarch64/aarch64.c (aarch64_comp_type_attributes): Handle
18070 2021-04-27 Richard Biener <rguenther@suse.de>
18072 PR tree-optimization/100051
18073 * tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Add
18074 disambiguator based on access size vs. decl size.
18076 2021-04-27 Richard Biener <rguenther@suse.de>
18078 PR tree-optimization/100278
18079 * tree-ssa-pre.c (compute_avail): Give up when we cannot
18080 adjust TBAA beacuse of mismatching bases.
18082 2021-04-27 Jakub Jelinek <jakub@redhat.com>
18085 * config/i386/i386.md (*<insn><mode>3_mask, *<insn><mode>3_mask_1):
18086 For any_rotate define_insn_split and following splitters, use
18087 SWI iterator instead of SWI48.
18089 2021-04-27 Richard Biener <rguenther@suse.de>
18091 PR tree-optimization/99776
18092 * match.pd (bit_field_ref (ctor)): Relax element extract
18093 type compatibility checks.
18095 2021-04-27 Cui,Lili <lili.cui@intel.com>
18097 * common/config/i386/i386-common.c (processor_names):
18098 Sync processor_names with processor_type.
18099 * config/i386/i386-options.c (processor_cost_table):
18100 Sync processor_cost_table with processor_type.
18102 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
18104 * value-range.cc (irange::irange_set_1bit_anti_range): Add assert.
18105 (irange::set): Call irange_set_1bit_anti_range for handling all
18106 1-bit ranges. Fall through on ~[MIN,MAX].
18108 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
18110 * value-range.cc (irange::legacy_num_pairs): Remove.
18111 (irange::invert): Change gcc_assert to gcc_checking_assert.
18112 * value-range.h (irange::num_pairs): Adjust for a cached
18113 num_pairs(). Also, rename all gcc_assert's to
18114 gcc_checking_assert's.
18116 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
18118 * value-range.cc (irange::operator=): Set m_kind.
18119 (irange::copy_to_legacy): Handle varying and undefined sources
18120 as a legacy copy since they can be easily copied.
18121 (irange::irange_set): Set m_kind.
18122 (irange::irange_set_anti_range): Same.
18123 (irange::set): Rename normalize_min_max to normalize_kind.
18124 (irange::verify_range): Adjust for multi-ranges having the
18126 (irange::irange_union): Set m_kind.
18127 (irange::irange_intersect): Same.
18128 (irange::invert): Same.
18129 * value-range.h (irange::kind): Always return m_kind.
18130 (irange::varying_p): Rename to...
18131 (irange::varying_comptaible_p): ...this.
18132 (irange::undefined_p): Only look at m_kind.
18133 (irange::irange): Always set VR_UNDEFINED if applicable.
18134 (irange::set_undefined): Always set VR_UNDEFINED.
18135 (irange::set_varying): Always set m_kind to VR_VARYING.
18136 (irange::normalize_min_max): Rename to...
18137 (irange::normalize_kind): ...this.
18139 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
18141 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::set_ssa_range_info):
18142 Adjust for constant_p including varying_p.
18143 * tree-vrp.c (vrp_prop::finalize): Same.
18144 (determine_value_range): Same.
18145 * vr-values.c (vr_values::range_of_expr): Same.
18146 * value-range.cc (irange::symbolic_p): Do not check varying_p.
18147 (irange::constant_p): Same.
18149 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
18151 * value-range.cc (irange::legacy_lower_bound): Replace
18152 !undefined_p check with num_ranges > 0.
18153 (irange::legacy_upper_bound): Same.
18154 * value-range.h (irange::type): Same.
18155 (irange::lower_bound): Same.
18156 (irange::upper_bound): Same.
18158 2021-04-26 Richard Biener <rguenther@suse.de>
18160 PR tree-optimization/99956
18161 * gimple-loop-interchange.cc (compute_access_stride):
18162 Try instantiating the access in a shallower loop nest
18163 if instantiating failed.
18164 (compute_access_strides): Pass adjustable loop_nest
18165 to compute_access_stride.
18167 2021-04-26 Christophe Lyon <christophe.lyon@linaro.org>
18169 * doc/sourcebuild.texi (arm_cmse_hw): Document.
18171 2021-04-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18173 * config/aarch64/iterators.md (vwcore): Handle V4BF, V8BF.
18175 2021-04-26 Thomas Schwinge <thomas@codesourcery.com>
18176 Nathan Sidwell <nathan@codesourcery.com>
18177 Tom de Vries <vries@codesourcery.com>
18178 Julian Brown <julian@codesourcery.com>
18179 Kwok Cheung Yeung <kcy@codesourcery.com>
18181 * omp-offload.c (oacc_validate_dims): Implement
18182 '-Wopenacc-parallelism'.
18183 * doc/invoke.texi (-Wopenacc-parallelism): Document.
18185 2021-04-26 Richard Biener <rguenther@suse.de>
18187 * tree-cfg.h (gimplify_build1): Remove.
18188 (gimplify_build2): Likewise.
18189 (gimplify_build3): Likewise.
18190 * tree-cfg.c (gimplify_build1): Move to tree-vect-generic.c.
18191 (gimplify_build2): Likewise.
18192 (gimplify_build3): Likewise.
18193 * tree-vect-generic.c (gimplify_build1): Move from tree-cfg.c.
18195 (gimplify_build2): Likewise.
18196 (gimplify_build3): Likewise.
18197 (tree_vec_extract): Use resimplify with following SSA edges.
18198 (expand_vector_parallel): Avoid passing NULL size/bitpos
18199 to tree_vec_extract.
18200 * expr.c (store_constructor): Deal with zero-element CTORs.
18201 * match.pd (bit_field_ref <vector CTOR>): Make sure to
18202 produce vector constants when possible.
18204 2021-04-26 Richard Biener <rguenther@suse.de>
18206 * tree-complex.c: Include gimple-fold.h.
18207 (expand_complex_addition): Use gimple_build.
18208 (expand_complex_multiplication_components): Likewise.
18209 (expand_complex_multiplication): Likewise.
18210 (expand_complex_div_straight): Likewise.
18211 (expand_complex_div_wide): Likewise.
18212 (expand_complex_division): Likewise.
18213 (expand_complex_conjugate): Likewise.
18214 (expand_complex_comparison): Likewise.
18216 2021-04-26 Richard Biener <rguenther@suse.de>
18218 * tree-ssa-phiopt.c (two_value_replacement): Remove use
18219 of legacy gimplify_buildN API.
18221 2021-04-26 Richard Biener <rguenther@suse.de>
18223 PR tree-optimization/99473
18224 * tree-ssa-phiopt.c (cond_store_replacement): Handle all
18227 2021-04-26 Richard Biener <rguenther@suse.de>
18229 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_builtin):
18230 Use replace_call_with_value.
18232 2021-04-26 Richard Biener <rguenther@suse.de>
18234 * tree-ssa-propagate.h (valid_gimple_rhs_p): Remove.
18235 (update_gimple_call): Likewise.
18236 (update_call_from_tree): Likewise.
18237 * tree-ssa-propagate.c (valid_gimple_rhs_p): Remove.
18238 (valid_gimple_call_p): Likewise.
18239 (move_ssa_defining_stmt_for_defs): Likewise.
18240 (finish_update_gimple_call): Likewise.
18241 (update_gimple_call): Likewise.
18242 (update_call_from_tree): Likewise.
18243 (propagate_tree_value_into_stmt): Use replace_call_with_value.
18244 * gimple-fold.h (update_gimple_call): Declare.
18245 * gimple-fold.c (valid_gimple_rhs_p): Move here from
18246 tree-ssa-propagate.c.
18247 (update_gimple_call): Likewise.
18248 (valid_gimple_call_p): Likewise.
18249 (finish_update_gimple_call): Likewise, and simplify.
18250 (gimplify_and_update_call_from_tree): Implement
18251 update_call_from_tree functionality, avoid excessive
18252 push/pop_gimplify_context.
18253 (gimple_fold_builtin): Use only gimplify_and_update_call_from_tree.
18254 (gimple_fold_call): Likewise.
18255 * gimple-ssa-sprintf.c (try_substitute_return_value): Likewise.
18256 * tree-ssa-ccp.c (ccp_folder::fold_stmt): Likewise.
18257 (pass_fold_builtins::execute): Likewise.
18258 (optimize_stack_restore): Use replace_call_with_value.
18259 * tree-cfg.c (fold_loop_internal_call): Likewise.
18260 * tree-ssa-dce.c (maybe_optimize_arith_overflow): Use
18261 only gimplify_and_update_call_from_tree.
18262 * tree-ssa-strlen.c (handle_builtin_strlen): Likewise.
18263 (handle_builtin_strchr): Likewise.
18264 * tsan.c: Include gimple-fold.h instead of tree-ssa-propagate.h.
18266 2021-04-26 Jakub Jelinek <jakub@redhat.com>
18269 * vmsdbgout.c (ASM_OUTPUT_DEBUG_STRING, vmsdbgout_begin_block,
18270 vmsdbgout_end_block, lookup_filename, vmsdbgout_source_line): Remove
18273 2021-04-25 liuhongt <hongtao.liu@intel.com>
18276 * config/i386/i386-builtin.def (BDESC): Change the icode of
18277 the following builtins to CODE_FOR_nothing.
18278 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
18279 IX86_BUILTIN_PCMPEQB128, IX86_BUILTIN_PCMPEQW128,
18280 IX86_BUILTIN_PCMPEQD128, IX86_BUILTIN_PCMPEQQ,
18281 IX86_BUILTIN_PCMPEQB256, IX86_BUILTIN_PCMPEQW256,
18282 IX86_BUILTIN_PCMPEQD256, IX86_BUILTIN_PCMPEQQ256,
18283 IX86_BUILTIN_PCMPGTB128, IX86_BUILTIN_PCMPGTW128,
18284 IX86_BUILTIN_PCMPGTD128, IX86_BUILTIN_PCMPGTQ,
18285 IX86_BUILTIN_PCMPGTB256, IX86_BUILTIN_PCMPGTW256,
18286 IX86_BUILTIN_PCMPGTD256, IX86_BUILTIN_PCMPGTQ256.
18287 * config/i386/sse.md (avx2_eq<mode>3): Deleted.
18288 (sse2_eq<mode>3): Ditto.
18289 (sse4_1_eqv2di3): Ditto.
18290 (sse2_gt<mode>3): Rename to ..
18291 (*sse2_gt<mode>3): .. this.
18293 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
18296 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
18299 * config/darwin.c (darwin_binds_local_p): Assume that any
18300 public symbol might be interposed for PIC code. Update function
18301 header comment to reflect current Darwin capability.
18303 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
18306 * config/darwin.c (darwin_binds_local_p): Assume that any
18307 public symbol might be interposed for PIC code. Update function
18308 header comment to reflect current Darwin capability.
18310 2021-04-24 Richard Sandiford <richard.sandiford@arm.com>
18312 * doc/sourcebuild.texi: Document no-opts and any-opts target
18315 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
18317 * config/bpf/bpf.h (ASM_OUTPUT_ALIGNED_BSS): Use .type and .lcomm.
18319 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
18321 * config/bpf/bpf.h (FUNCTION_BOUNDARY): Set to 64.
18323 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
18326 * config/i386/i386-options.c (ix86_option_override_internal):
18327 Error out when -m96bit-long-double is used with 64bit targets.
18328 * config/i386/i386.md (*pushxf_rounded): Remove pattern.
18330 2021-04-23 Martin Liska <mliska@suse.cz>
18332 * lto-wrapper.c: Remove FIXME about usage of
18333 hardware_concurrency. The function is not on par with
18336 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
18339 * config/i386/sync.md (FILD_ATOMIC/FIST_ATOMIC FP load peephole2):
18340 Copy operand 3 to operand 4. Use sse_reg_operand
18341 as operand 3 predicate.
18342 (FILD_ATOMIC/FIST_ATOMIC FP load peephole2 with mem blockage): Ditto.
18343 (LDX_ATOMIC/STX_ATOMIC FP load peephole2): Ditto.
18344 (LDX_ATOMIC/LDX_ATOMIC FP load peephole2 with mem blockage): Ditto.
18345 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2):
18346 Copy operand 1 to operand 0.
18347 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2 with mem blockage): Ditto.
18348 (LDX_ATOMIC/STX_ATOMIC FP store peephole2): Ditto.
18349 (LDX_ATOMIC/LDX_ATOMIC FP store peephole2 with mem blockage): Ditto.
18351 2021-04-23 Alex Coplan <alex.coplan@arm.com>
18353 PR rtl-optimization/100230
18354 * early-remat.c (early_remat::sort_candidates): Use delete[]
18355 instead of delete for array allocated with new[].
18357 2021-04-23 Richard Biener <rguenther@suse.de>
18359 * genmatch.c (lower_cond): Remove VEC_COND_EXPR special-casing.
18360 (capture_info::capture_info): Likewise.
18361 (capture_info::walk_match): Likewise.
18362 (expr::gen_transform): Likewise.
18363 (dt_simplify::gen_1): Likewise.
18364 * gimple-match-head.c (maybe_resimplify_conditional_op):
18365 Remove VEC_COND_EXPR special-casing.
18366 (gimple_simplify): Likewise.
18367 * gimple.c (gimple_could_trap_p_1): Adjust.
18368 * tree-ssa-pre.c (compute_avail): Allow VEC_COND_EXPR
18369 to participate in PRE.
18371 2021-04-23 Richard Biener <rguenther@suse.de>
18373 * cfganal.c (connect_infinite_loops_to_exit): First call
18374 add_noreturn_fake_exit_edges.
18375 * ipa-sra.c (process_scan_results): Do not call the now redundant
18376 add_noreturn_fake_exit_edges.
18377 * predict.c (tree_estimate_probability): Likewise.
18378 (rebuild_frequencies): Likewise.
18379 * store-motion.c (one_store_motion_pass): Likewise.
18381 2021-04-23 Richard Biener <rguenther@suse.de>
18383 PR tree-optimization/100222
18384 * predict.c (pass_profile::execute): Remove redundant call to
18385 mark_irreducible_loops.
18386 (report_predictor_hitrates): Likewise.
18388 2021-04-23 Richard Biener <rguenther@suse.de>
18390 * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Avoid
18391 valid_gimple_rhs_p by instead gimplifying to one.
18393 2021-04-23 Richard Biener <rguenther@suse.de>
18395 PR tree-optimization/99971
18396 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
18397 Always use TBAA for loads.
18399 2021-04-23 liuhongt <hongtao.liu@intel.com>
18402 * config/i386/i386-options.c (ix86_option_override_internal):
18403 Clear MASK_AVX256_SPLIT_UNALIGNED_LOAD/STORE in x_target_flags
18404 when X86_TUNE_AVX256_UNALIGNED_LOAD/STORE_OPTIMAL is enabled
18405 by target attribute.
18407 2021-04-23 David Edelsohn <dje.gcc@gmail.com>
18409 * config/rs6000/aix71.h (PREFERRED_DEBUGGING_TYPE): Change to
18411 * config/rs6000/aix72.h (PREFERRED_DEBUGGING_TYPE): Same.
18413 2021-04-22 David Edelsohn <dje.gcc@gmail.com>
18415 * config.gcc (powerpc-ibm-aix6.*): Remove.
18416 * config/rs6000/aix61.h: Delete.
18418 2021-04-22 Martin Liska <mliska@suse.cz>
18420 PR testsuite/100159
18421 PR testsuite/100192
18422 * builtins.c (expand_builtin): Fix typos and missing comments.
18423 * dwarf2out.c (gen_subprogram_die): Likewise.
18424 (gen_struct_or_union_type_die): Likewise.
18426 2021-04-22 Uroš Bizjak <ubizjak@gmail.com>
18429 * config/i386/i386-expand.c (ix86_expand_convert_uns_sidf_sse):
18430 Remove the sign with FE_DOWNWARD, where x - x = -0.0.
18432 2021-04-21 Iain Sandoe <iain@sandoe.co.uk>
18434 * config/i386/darwin.h (TARGET_64BIT): Remove definition
18435 based on TARGET_ISA_64BIT.
18436 (TARGET_64BIT_P): Remove definition based on
18437 TARGET_ISA_64BIT_P().
18439 2021-04-21 Martin Liska <mliska@suse.cz>
18442 2021-04-21 Martin Liska <mliska@suse.cz>
18444 * lto-wrapper.c (cpuset_popcount): Remove.
18445 (init_num_threads): Remove and use hardware_concurrency.
18447 2021-04-21 Martin Liska <mliska@suse.cz>
18450 * main.c (main): Call toplev::finalize in CHECKING_P mode.
18451 * ipa-modref.c (ipa_modref_c_finalize): summaries are NULL
18452 when incremental LTO linking happens.
18454 2021-04-21 Martin Liska <mliska@suse.cz>
18456 * lto-wrapper.c (run_gcc): When -flto=jobserver is used, but the
18457 makeserver cannot be detected, then use -flto=N fallback.
18459 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
18461 * acinclude.m4 (gcc_AC_INITFINI_ARRAY): When cross-compiling,
18462 default to yes for aarch64-linux-gnu.
18463 * configure: Regenerate.
18465 2021-04-21 Martin Liska <mliska@suse.cz>
18467 * lto-wrapper.c (cpuset_popcount): Remove.
18468 (init_num_threads): Remove and use hardware_concurrency.
18470 2021-04-21 Martin Liska <mliska@suse.cz>
18472 * config/i386/i386.c: Remove superfluous || TARGET_MACHO
18473 which remains to be '(... || 0)' and clang complains about it.
18474 * dwarf2out.c (AT_vms_delta): Declare conditionally.
18475 (add_AT_vms_delta): Likewise.
18476 * tree.c (fld_simplified_type): Use rather more common pattern
18477 for disabling of something (#if 0).
18478 (get_tree_code_name): Likewise.
18479 (verify_type_variant): Likewise.
18481 2021-04-21 Martin Liska <mliska@suse.cz>
18483 * config/i386/i386-expand.c (decide_alignment): Use newly named
18484 macro TARGET_CPU_P.
18485 * config/i386/i386.c (ix86_decompose_address): Likewise.
18486 (ix86_address_cost): Likewise.
18487 (ix86_lea_outperforms): Likewise.
18488 (ix86_avoid_lea_for_addr): Likewise.
18489 (ix86_add_stmt_cost): Likewise.
18490 * config/i386/i386.h (TARGET_*): Remove.
18491 (TARGET_CPU_P): New macro.
18492 * config/i386/i386.md: Use newly named macro TARGET_CPU_P.
18493 * config/i386/x86-tune-sched-atom.c (do_reorder_for_imul): Likewise.
18494 (swap_top_of_ready_list): Likewise.
18495 (ix86_atom_sched_reorder): Likewise.
18496 * config/i386/x86-tune-sched-bd.c (ix86_bd_has_dispatch): Likewise.
18497 * config/i386/x86-tune-sched.c (ix86_adjust_cost): Likewise.
18499 2021-04-21 Martin Liska <mliska@suse.cz>
18501 * config/i386/i386-options.c (TARGET_EXPLICIT_NO_SAHF_P):
18503 (SET_TARGET_NO_SAHF): Likewise.
18504 (TARGET_EXPLICIT_PREFETCH_SSE_P): Likewise.
18505 (SET_TARGET_PREFETCH_SSE): Likewise.
18506 (TARGET_EXPLICIT_NO_TUNE_P): Likewise.
18507 (SET_TARGET_NO_TUNE): Likewise.
18508 (TARGET_EXPLICIT_NO_80387_P): Likewise.
18509 (SET_TARGET_NO_80387): Likewise.
18511 * config/i386/i386.h (TARGET_*): Remove.
18512 * opth-gen.awk: Generate new used macros.
18514 2021-04-21 Martin Liska <mliska@suse.cz>
18516 * config/i386/i386.h (PTA_*): Remove.
18517 (enum pta_flag): New.
18518 (DEF_PTA): Generate PTA_* values from i386-isa.def.
18519 * config/i386/i386-isa.def: New file.
18521 2021-04-21 Alex Coplan <alex.coplan@arm.com>
18524 * config/aarch64/aarch64-bti-insert.c (aarch64_bti_j_insn_p): New.
18525 (rest_of_insert_bti): Avoid inserting duplicate bti j insns for
18526 jump table targets.
18528 2021-04-21 H.J. Lu <hjl.tools@gmail.com>
18530 * config.gcc: Install mwaitintrin.h for i[34567]86-*-* and
18531 x86_64-*-* targets.
18532 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_MWAIT_SET):
18534 (OPTION_MASK_ISA2_MWAIT_UNSET): Likewise.
18535 (ix86_handle_option): Handle -mmwait.
18536 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
18537 Replace OPTION_MASK_ISA_SSE3 with OPTION_MASK_ISA2_MWAIT on
18538 __builtin_ia32_monitor and __builtin_ia32_mwait.
18539 * config/i386/i386-options.c (isa2_opts): Add -mmwait.
18540 (ix86_valid_target_attribute_inner_p): Likewise.
18541 (ix86_option_override_internal): Enable mwait/monitor
18542 instructions for -msse3.
18543 * config/i386/i386.h (TARGET_MWAIT): New.
18544 (TARGET_MWAIT_P): Likewise.
18545 * config/i386/i386.opt: Add -mmwait.
18546 * config/i386/mwaitintrin.h: New file.
18547 * config/i386/pmmintrin.h: Include <mwaitintrin.h>.
18548 * config/i386/sse.md (sse3_mwait): Replace TARGET_SSE3 with
18550 (@sse3_monitor_<mode>): Likewise.
18551 * config/i386/x86gprintrin.h: Include <mwaitintrin.h>.
18552 * doc/extend.texi: Document mwait target attribute.
18553 * doc/invoke.texi: Document -mmwait.
18555 2021-04-21 Martin Liska <mliska@suse.cz>
18557 * config/i386/i386-options.c (DEF_ENUM): Remove it.
18558 * config/i386/i386-opts.h (DEF_ENUM): Likewise.
18559 * config/i386/stringop.def (DEF_ENUM): Likewise.
18561 2021-04-21 Martin Liska <mliska@suse.cz>
18563 * tree-cfg.c (gimple_verify_flow_info): Use qD instead
18564 of print_generic_expr.
18566 2021-04-21 Jakub Jelinek <jakub@redhat.com>
18568 PR rtl-optimization/100148
18569 * cprop.c (constprop_register): Use next_nondebug_insn instead of
18572 2021-04-21 Martin Liska <mliska@suse.cz>
18575 * cgraphunit.c (cgraph_node::analyze): Remove duplicate
18576 free_dominance_info calls.
18578 2021-04-21 Richard Biener <rguenther@suse.de>
18580 * gimple-fold.c (maybe_fold_reference): Remove is_lhs
18581 parameter (and assume it to be false).
18582 (fold_gimple_assign): Adjust, remove all callers of
18583 maybe_fold_reference calling it with is_lhs true.
18584 (gimple_fold_call): Likewise.
18585 (fold_stmt_1): Likewise.
18587 2021-04-21 Richard Biener <rguenther@suse.de>
18589 * fold-const.c (pedantic_non_lvalue_loc): Remove.
18590 (fold_binary_loc): Adjust.
18591 (fold_ternary_loc): Likewise.
18593 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
18595 PR middle-end/100130
18596 * varasm.c (get_block_for_decl): Make sure that any use of the
18597 retain attribute matches the section's retain flag.
18598 (switch_to_section): Check for retain mismatches even when
18599 changing sections, but do not warn if the given decl is the
18600 section's named.decl.
18601 (output_object_block): Pass the first decl in the block (if any)
18602 to switch_to_section.
18604 2021-04-20 H.J. Lu <hjl.tools@gmail.com>
18606 * config/i386/i386-c.c (ix86_target_macros_internal): Define
18607 __CRC32__ for -mcrc32.
18608 * config/i386/i386-options.c (ix86_option_override_internal):
18609 Enable crc32 instruction for -msse4.2.
18610 * config/i386/i386.md (sse4_2_crc32<mode>): Remove TARGET_SSE4_2
18612 (sse4_2_crc32di): Likewise.
18613 * config/i386/ia32intrin.h: Use crc32 target option for CRC32
18616 2021-04-20 Segher Boessenkool <segher@kernel.crashing.org>
18619 * config/rs6000/rs6000.c (rs6000_machine_from_flags): Do not consider
18622 2021-04-20 Martin Liska <mliska@suse.cz>
18624 * doc/invoke.texi: Fix typo.
18625 * params.opt: Likewise.
18627 2021-04-20 Martin Liska <mliska@suse.cz>
18629 * doc/invoke.texi: Document new param.
18631 2021-04-19 Andrew MacLeod <amacleod@redhat.com>
18633 PR tree-optimization/100081
18634 * gimple-range-cache.h (ranger_cache): Inherit from gori_compute
18635 rather than gori_compute_cache.
18636 * gimple-range-gori.cc (is_gimple_logical_p): Move to top of file.
18637 (range_def_chain::m_logical_depth): New member.
18638 (range_def_chain::range_def_chain): Initialize m_logical_depth.
18639 (range_def_chain::get_def_chain): Don't build defchains through more
18640 than LOGICAL_LIMIT logical expressions.
18641 * params.opt (param_ranger_logical_depth): New.
18643 2021-04-19 Richard Earnshaw <rearnsha@arm.com>
18646 * config/arm/arm.c (arm_configure_build_target): Do not strip
18647 extended FPU/SIMD feature bits from the target ISA when -mfpu
18648 is specified (partial revert of r11-8168).
18650 2021-04-19 Thomas Schwinge <thomas@codesourcery.com>
18652 * params.opt (-param=openacc-kernels=): Add.
18653 * omp-oacc-kernels-decompose.cc
18654 (pass_omp_oacc_kernels_decompose::gate): Use it.
18655 * doc/invoke.texi (-fopenacc-kernels=@var{mode}): Move...
18656 (--param): ... here, 'openacc-kernels'.
18658 2021-04-19 Martin Liska <mliska@suse.cz>
18661 * gengtype.c (finish_root_table): Align function arguments
18662 in between declaration and definition.
18664 2021-04-19 Eric Botcazou <ebotcazou@adacore.com>
18666 * config/i386/winnt.c (i386_pe_seh_cold_init): Properly deal with
18667 frames larger than the SEH maximum frame size.
18669 2021-04-18 Segher Boessenkool <segher@kernel.crashing.org>
18671 PR rtl-optimization/99927
18672 * combine.c (distribute_notes) [REG_UNUSED]: If the register already
18673 is dead, just drop it.
18675 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
18678 * config/i386/winnt-d.c (TARGET_D_TEMPLATES_ALWAYS_COMDAT): Define.
18679 * doc/tm.texi: Regenerate.
18680 * doc/tm.texi.in (D language and ABI): Add @hook for
18681 TARGET_D_TEMPLATES_ALWAYS_COMDAT.
18683 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
18685 * config/darwin-d.c (darwin_d_handle_target_object_format): New
18687 (darwin_d_register_target_info): New function.
18688 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18689 * config/dragonfly-d.c (dragonfly_d_handle_target_object_format): New
18691 (dragonfly_d_register_target_info): New function.
18692 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18693 * config/freebsd-d.c (freebsd_d_handle_target_object_format): New
18695 (freebsd_d_register_target_info): New function.
18696 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18697 * config/glibc-d.c (glibc_d_handle_target_object_format): New
18699 (glibc_d_register_target_info): New function.
18700 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18701 * config/i386/i386-d.c (ix86_d_handle_target_object_format): New
18703 (ix86_d_register_target_info): Add ix86_d_handle_target_object_format
18704 as handler for objectFormat key.
18705 * config/i386/winnt-d.c (winnt_d_handle_target_object_format): New
18707 (winnt_d_register_target_info): New function.
18708 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18709 * config/netbsd-d.c (netbsd_d_handle_target_object_format): New
18711 (netbsd_d_register_target_info): New function.
18712 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18713 * config/openbsd-d.c (openbsd_d_handle_target_object_format): New
18715 (openbsd_d_register_target_info): New function.
18716 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18717 * config/pa/pa-d.c (pa_d_handle_target_object_format): New function.
18718 (pa_d_register_target_info): Add pa_d_handle_target_object_format as
18719 handler for objectFormat key.
18720 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_object_format): New
18722 (rs6000_d_register_target_info): Add
18723 rs6000_d_handle_target_object_format as handler for objectFormat key.
18724 * config/sol2-d.c (solaris_d_handle_target_object_format): New
18726 (solaris_d_register_target_info): New function.
18727 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18729 2021-04-16 Jakub Jelinek <jakub@redhat.com>
18732 * config/aarch64/aarch64.c (aarch64_function_arg_alignment): Change
18733 abi_break argument from bool * to unsigned *, store there the pre-GCC 9
18735 (aarch64_layout_arg, aarch64_gimplify_va_arg_expr): Adjust callers.
18736 (aarch64_function_arg_regno_p): Likewise. Only emit -Wpsabi note if
18737 the old and new alignment after applying MIN/MAX to it is different.
18739 2021-04-16 Tamar Christina <tamar.christina@arm.com>
18742 * config/aarch64/aarch64-sve.md (@aarch64_sve_trn1_conv<mode>): New.
18743 * config/aarch64/aarch64.c (aarch64_expand_sve_const_pred_trn): Use new
18745 * config/aarch64/iterators.md (UNSPEC_TRN1_CONV): New.
18747 2021-04-16 Bill Schmidt <wschmidt@linux.ibm.com>
18749 * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions): Revise
18750 this section and its subsections.
18752 2021-04-16 Jakub Jelinek <jakub@redhat.com>
18755 * config/aarch64/aarch64.md (*neg_asr_si2_extr, *extrsi5_insn_di): New
18756 define_insn patterns.
18758 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
18760 PR rtl-optimization/98689
18761 * reg-notes.def (UNTYPED_CALL): New note.
18762 * combine.c (distribute_notes): Handle it.
18763 * emit-rtl.c (try_split): Likewise.
18764 * rtlanal.c (rtx_properties::try_to_add_insn): Likewise. Assume
18765 that calls with the note implicitly set all return value registers.
18766 * builtins.c (expand_builtin_apply): Add a REG_UNTYPED_CALL
18769 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
18771 PR rtl-optimization/99596
18772 * rtlanal.c (rtx_properties::try_to_add_insn): Don't add global
18773 register accesses for const calls. Assume that pure functions
18774 can only read from global registers. Ignore cases in which
18775 the stack pointer has been marked global.
18777 2021-04-16 Jakub Jelinek <jakub@redhat.com>
18780 * tree-vect-loop.c (vect_transform_loop): Don't remove just
18781 dead scalar .MASK_LOAD calls, but also dead .COND_* calls - replace
18782 them by their last argument.
18784 2021-04-15 Martin Liska <mliska@suse.cz>
18786 * doc/invoke.texi: Other params don't use it, remove it.
18788 2021-04-15 Richard Biener <rguenther@suse.de>
18790 * gimple-builder.h: Add deprecation note.
18792 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
18795 * attribs.h (restrict_type_identity_attributes_to): Declare.
18796 * attribs.c (restrict_type_identity_attributes_to): New function.
18798 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
18801 * attribs.h (affects_type_identity_attributes): Declare.
18802 * attribs.c (remove_attributes_matching): New function.
18803 (affects_type_identity_attributes): Likewise.
18805 2021-04-15 Jakub Jelinek <jakub@redhat.com>
18808 * config/aarch64/aarch64.md (*<LOGICAL:optab>_<SHIFT:optab><mode>3):
18809 Add combine splitters for *<LOGICAL:optab>_ashl<mode>3 with
18810 ZERO_EXTEND, SIGN_EXTEND or AND.
18812 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
18814 PR rtl-optimization/99929
18815 * rtl.h (same_vector_encodings_p): New function.
18816 * cse.c (exp_equiv_p): Check that CONST_VECTORs have the same encoding.
18817 * cselib.c (rtx_equal_for_cselib_1): Likewise.
18818 * jump.c (rtx_renumbered_equal_p): Likewise.
18819 * lra-constraints.c (operands_match_p): Likewise.
18820 * reload.c (operands_match_p): Likewise.
18821 * rtl.c (rtx_equal_p_cb, rtx_equal_p): Likewise.
18823 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
18825 * print-rtl.c (rtx_writer::print_rtx_operand_codes_E_and_V): Print
18826 more information about variable-length CONST_VECTORs.
18828 2021-04-14 Vladimir N. Makarov <vmakarov@redhat.com>
18830 PR rtl-optimization/100066
18831 * lra-constraints.c (split_reg): Check paradoxical_subreg_p for
18832 ordered modes when choosing splitting mode for hard reg.
18834 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
18837 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector_sel):
18839 (aarch64_expand_sve_const_vector): Use it for nelts_per_pattern==2.
18841 2021-04-14 Andreas Krebbel <krebbel@linux.ibm.com>
18843 * config/s390/s390-builtins.def (O_M5, O_M12, ...): Add new macros
18844 for mask operand types.
18845 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
18846 (s390_vec_permi_dbl, s390_vpdi): Use the M5 type for the immediate
18848 (s390_vec_msum_u128, s390_vmslg): Use the M12 type for the
18850 * config/s390/s390.c (s390_const_operand_ok): Check the new
18851 operand types and generate a list of valid values.
18853 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
18855 * doc/tm.texi: Regenerate.
18856 * doc/tm.texi.in (D language and ABI): Add @hook for
18857 TARGET_D_REGISTER_OS_TARGET_INFO.
18859 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
18861 * config/aarch64/aarch64-d.c (aarch64_d_handle_target_float_abi): New
18863 (aarch64_d_register_target_info): New function.
18864 * config/aarch64/aarch64-protos.h (aarch64_d_register_target_info):
18866 * config/aarch64/aarch64.h (TARGET_D_REGISTER_CPU_TARGET_INFO):
18868 * config/arm/arm-d.c (arm_d_handle_target_float_abi): New function.
18869 (arm_d_register_target_info): New function.
18870 * config/arm/arm-protos.h (arm_d_register_target_info): Declare.
18871 * config/arm/arm.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18872 * config/i386/i386-d.c (ix86_d_handle_target_float_abi): New function.
18873 (ix86_d_register_target_info): New function.
18874 * config/i386/i386-protos.h (ix86_d_register_target_info): Declare.
18875 * config/i386/i386.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18876 * config/mips/mips-d.c (mips_d_handle_target_float_abi): New function.
18877 (mips_d_register_target_info): New function.
18878 * config/mips/mips-protos.h (mips_d_register_target_info): Declare.
18879 * config/mips/mips.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18880 * config/pa/pa-d.c (pa_d_handle_target_float_abi): New function.
18881 (pa_d_register_target_info): New function.
18882 * config/pa/pa-protos.h (pa_d_register_target_info): Declare.
18883 * config/pa/pa.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18884 * config/riscv/riscv-d.c (riscv_d_handle_target_float_abi): New
18886 (riscv_d_register_target_info): New function.
18887 * config/riscv/riscv-protos.h (riscv_d_register_target_info): Declare.
18888 * config/riscv/riscv.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18889 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_float_abi): New
18891 (rs6000_d_register_target_info): New function.
18892 * config/rs6000/rs6000-protos.h (rs6000_d_register_target_info):
18894 * config/rs6000/rs6000.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18895 * config/s390/s390-d.c (s390_d_handle_target_float_abi): New function.
18896 (s390_d_register_target_info): New function.
18897 * config/s390/s390-protos.h (s390_d_register_target_info): Declare.
18898 * config/s390/s390.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18899 * config/sparc/sparc-d.c (sparc_d_handle_target_float_abi): New
18901 (sparc_d_register_target_info): New function.
18902 * config/sparc/sparc-protos.h (sparc_d_register_target_info): Declare.
18903 * config/sparc/sparc.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18904 * doc/tm.texi: Regenerate.
18905 * doc/tm.texi.in (D language and ABI): Add @hook for
18906 TARGET_D_REGISTER_CPU_TARGET_INFO.
18908 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
18910 * config/i386/i386-d.c (ix86_d_has_stdcall_convention): New function.
18911 * config/i386/i386-protos.h (ix86_d_has_stdcall_convention): Declare.
18912 * config/i386/i386.h (TARGET_D_HAS_STDCALL_CONVENTION): Define.
18913 * doc/tm.texi: Regenerate.
18914 * doc/tm.texi.in (D language and ABI): Add @hook for
18915 TARGET_D_HAS_STDCALL_CONVENTION.
18917 2021-04-14 Richard Biener <rguenther@suse.de>
18919 * tree-cfg.c (verify_gimple_assign_ternary): Verify that
18920 VEC_COND_EXPRs have a gimple_val condition.
18921 * tree-ssa-propagate.c (valid_gimple_rhs_p): VEC_COND_EXPR
18922 can no longer have a GENERIC condition.
18924 2021-04-14 Richard Earnshaw <rearnsha@arm.com>
18927 * config/arm/arm.c (arm_configure_build_target): Strip isa_all_fpbits
18928 from the isa_delta when -mfpu has been used.
18929 (arm_options_perform_arch_sanity_checks): It's the architecture that
18930 lacks an FPU not the processor.
18932 2021-04-13 Richard Biener <rguenther@suse.de>
18934 PR tree-optimization/100053
18935 * tree-ssa-sccvn.c (vn_nary_op_get_predicated_value): Do
18936 not use optimistic dominance queries for backedges to validate
18938 (dominated_by_p_w_unex): Add parameter to ignore executable
18939 state on backedges.
18940 (rpo_elim::eliminate_avail): Adjust.
18942 2021-04-13 Jakub Jelinek <jakub@redhat.com>
18945 * config/aarch64/aarch64.md (*aarch64_bfxil<mode>_extr,
18946 *aarch64_bfxilsi_extrdi): New define_insn patterns.
18948 2021-04-13 Jakub Jelinek <jakub@redhat.com>
18951 * simplify-rtx.c (simplify_immed_subreg): For MODE_COMPOSITE_P
18952 outermode, return NULL if the result doesn't encode back to the
18953 original byte sequence.
18954 (simplify_gen_subreg): Don't create SUBREGs from constants to
18955 MODE_COMPOSITE_P outermode.
18957 2021-04-12 Jakub Jelinek <jakub@redhat.com>
18959 PR rtl-optimization/99905
18960 * combine.c (expand_compound_operation): If pos + len > modewidth,
18961 perform the right shift by pos in inner_mode and then convert to mode,
18962 instead of trying to simplify a shift of rtx with inner_mode by pos
18963 as if it was a shift in mode.
18965 2021-04-12 Jakub Jelinek <jakub@redhat.com>
18968 * combine.c (simplify_and_const_int_1): Don't optimize varop
18969 away if it has side-effects.
18971 2021-04-12 Martin Liska <mliska@suse.cz>
18973 * doc/extend.texi: Escape @smallexample content.
18975 2021-04-12 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
18977 * config/s390/s390.md ("*movdi_31", "*movdi_64"): Add
18978 alternative in order to load a DFP zero.
18980 2021-04-12 Martin Liska <mliska@suse.cz>
18982 * doc/extend.texi: Be more precise in documentation
18983 of symver attribute.
18985 2021-04-12 Martin Liska <mliska@suse.cz>
18988 * gimplify.c (gimplify_expr): Right now, we unpoison all
18989 variables before a goto <dest>. We should not do it if we are
18992 2021-04-12 Cui,Lili <lili.cui@intel.com>
18994 * common/config/i386/cpuinfo.h (get_intel_cpu): Handle
18996 * common/config/i386/i386-common.c (processor_names): Add
18998 (processor_alias_table): Add rocketlake.
18999 * common/config/i386/i386-cpuinfo.h (processor_subtypes): Add
19000 INTEL_COREI7_ROCKETLAKE.
19001 * config.gcc: Add -march=rocketlake.
19002 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19004 * config/i386/i386-options.c (m_ROCKETLAKE) : Define.
19005 (processor_cost_table): Add rocketlake cost.
19006 * config/i386/i386.h (ix86_size_cost) : Define
19008 (processor_type) : Add PROCESSOR_ROCKETLAKE.
19009 (PTA_ROCKETLAKE): Ditto.
19010 * doc/extend.texi: Add rocketlake.
19011 * doc/invoke.texi: Add rocketlake.
19013 2021-04-12 Cui,Lili <lili.cui@intel.com>
19015 * config/i386/i386.h (PTA_ALDERLAKE): Change alderlake ISA list.
19016 * config/i386/i386-options.c (m_CORE_AVX2): Add m_ALDERLAKE.
19017 * common/config/i386/cpuinfo.h (get_intel_cpu): Add AlderLake model.
19018 * doc/invoke.texi: Change alderlake ISA list.
19020 2021-04-11 Hafiz Abid Qadeer <abidh@codesourcery.com>
19022 PR middle-end/98088
19023 * omp-expand.c (expand_oacc_collapse_init): Update condition in
19026 2021-04-10 H.J. Lu <hjl.tools@gmail.com>
19029 * config/i386/serializeintrin.h (_serialize): Defined as macro.
19031 2021-04-10 Jakub Jelinek <jakub@redhat.com>
19034 * expr.c (expand_expr_addr_expr_1): Test is_global_var rather than
19035 just TREE_STATIC on COMPOUND_LITERAL_EXPR_DECLs.
19037 2021-04-10 Jakub Jelinek <jakub@redhat.com>
19039 PR middle-end/99989
19040 * gimple-ssa-warn-alloca.c
19041 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit to
19042 0 with integer precision unconditionally.
19044 2021-04-10 Jakub Jelinek <jakub@redhat.com>
19046 PR rtl-optimization/98601
19047 * rtlanal.c (rtx_addr_can_trap_p_1): Allow in assert unknown size
19048 not just for BLKmode, but also for VOIDmode. For STRICT_ALIGNMENT
19049 unaligned_mems handle VOIDmode like BLKmode.
19051 2021-04-10 Jan Hubicka <hubicka@ucw.cz>
19054 * tree.c (free_lang_data_in_decl): Do not release body of
19055 declare_variant_alt.
19057 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
19059 * config/aarch64/aarch64.c (aarch64_option_restore): If the
19060 architecture was specified explicitly and the tuning wasn't,
19061 tune for the architecture rather than the configured default CPU.
19063 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
19065 * config/aarch64/aarch64.md (tlsdesc_small_sve_<mode>): Use X30
19066 as the temporary register.
19068 2021-04-09 Martin Liska <mliska@suse.cz>
19070 * doc/extend.texi: Move non-target attributes on the top level.
19072 2021-04-09 Martin Liska <mliska@suse.cz>
19074 * doc/invoke.texi: Document minimum and maximum value of the
19075 argument for both supported compression algorithms.
19077 2021-04-08 David Edelsohn <dje.gcc@gmail.com>
19079 * config/rs6000/rs6000.c (rs6000_xcoff_select_section): Select
19080 TLS BSS before TLS data.
19081 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .comm.
19083 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
19085 * doc/sourcebuild.texi (stdint_types_mbig_endian): Document.
19087 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
19089 * match.pd: Extend vec_cond folds to handle shifts.
19091 2021-04-08 Maciej W. Rozycki <macro@orcam.me.uk>
19093 * config/vax/vax.md: Fix comment for `*bit<mode>' pattern's
19096 2021-04-08 Alex Coplan <alex.coplan@arm.com>
19099 * config/arm/iterators.md (MVE_vecs): New.
19100 (V_elem): Also handle V2DF.
19101 * config/arm/mve.md (*mve_mov<mode>): Rename to ...
19102 (*mve_vdup<mode>): ... this. Remove second alternative since
19103 vec_duplicate of const_int is not canonical RTL, and we don't
19104 want to match symbol_refs.
19105 (*mve_vec_duplicate<mode>): Delete (pattern is redundant).
19107 2021-04-08 Xionghu Luo <luoxhu@linux.ibm.com>
19109 * fold-const.c (fold_single_bit_test): Fix typo.
19110 * print-rtl.c (print_rtx_insn_vec): Call print_rtl_single
19113 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
19115 PR tree-optimization/97513
19116 * tree-vect-slp.c (vect_add_slp_permutation): New function,
19118 (vectorizable_slp_permutation): ...here. Detect cases in which
19119 all VEC_PERM_EXPRs are guaranteed to have the same stepped
19120 permute vector and only generate one permute vector for that case.
19121 Extend that case to handle variable-length vectors.
19123 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
19125 PR tree-optimization/99873
19126 * tree-vect-slp.c (vect_slp_prefer_store_lanes_p): New function.
19127 (vect_build_slp_instance): Don't split store groups that could
19128 use IFN_STORE_LANES.
19130 2021-04-07 Jakub Jelinek <jakub@redhat.com>
19133 * varasm.c (output_constant_pool_contents): Don't strip name encoding
19134 from XSTR (desc->sym, 0) or from label before passing those to
19137 2021-04-07 Richard Biener <rguenther@suse.de>
19139 PR tree-optimization/99954
19140 * tree-loop-distribution.c: Include tree-affine.h.
19141 (generate_memcpy_builtin): Try using tree-affine to prove
19143 (loop_distribution::classify_builtin_ldst): Always classify
19146 2021-04-07 Richard Biener <rguenther@suse.de>
19148 PR tree-optimization/99947
19149 * tree-vect-loop.c (vectorizable_induction): Pre-allocate
19150 steps vector to avoid pushing elements from the reallocated
19153 2021-04-07 Richard Biener <rguenther@suse.de>
19155 * tree-ssa-sccvn.h (print_vn_reference_ops): Declare.
19156 * tree-ssa-pre.c (print_pre_expr): Factor out VN reference operand
19158 * tree-ssa-sccvn.c (print_vn_reference_ops): ... into this new
19160 (debug_vn_reference_ops): New.
19162 2021-04-07 Bin Cheng <bin.cheng@linux.alibaba.com>
19164 PR tree-optimization/98736
19165 * tree-loop-distribution.c
19166 * (loop_distribution::bb_top_order_init):
19167 Compute RPO with programing order preserved by calling function
19168 rev_post_order_and_mark_dfs_back_seme.
19170 2021-04-06 Vladimir N. Makarov <vmakarov@redhat.com>
19173 * lra-constraints.c (split_reg): Don't check paradoxical_subreg_p.
19174 * lra-lives.c (clear_sparseset_regnos, regnos_in_sparseset_p): New
19176 (process_bb_lives): Don't update biggest mode of hard reg for
19177 implicit in multi-register group. Use the new functions for
19178 updating dead_set and unused_set by register notes.
19180 2021-04-06 Xianmiao Qu <xianmiao_qu@c-sky.com>
19182 * config/csky/csky_pipeline_ck802.md : Use insn reservation name
19185 2021-04-06 H.J. Lu <hjl.tools@gmail.com>
19187 * config/i386/x86-tune-costs.h (skylake_memcpy): Updated.
19188 (skylake_memset): Likewise.
19189 (skylake_cost): Change CLEAR_RATIO to 17.
19190 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
19191 Replace m_CANNONLAKE, m_ICELAKE_CLIENT, m_ICELAKE_SERVER,
19192 m_TIGERLAKE and m_SAPPHIRERAPIDS with m_SKYLAKE and m_CORE_AVX512.
19194 2021-04-06 Richard Biener <rguenther@suse.de>
19196 PR tree-optimization/99880
19197 * tree-vect-loop.c (maybe_set_vectorized_backedge_value): Only
19198 set vectorized defs of relevant PHIs.
19200 2021-04-06 Richard Biener <rguenther@suse.de>
19202 PR tree-optimization/99924
19203 * tree-vect-slp.c (vect_bb_partition_graph_r): Do not mark
19204 nodes w/o scalar stmts as visited.
19206 2021-04-06 Alex Coplan <alex.coplan@arm.com>
19209 * config/arm/arm.c (arm_libcall_uses_aapcs_base): Also use base
19210 PCS for [su]fix_optab.
19212 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
19214 * config/darwin.c (machopic_legitimize_pic_address): Check
19215 that the current pic register is one of the hard reg set
19216 before setting liveness.
19218 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
19220 * config/darwin.c (machopic_legitimize_pic_address): Fix
19221 whitespace, remove unused code.
19223 2021-04-03 Jakub Jelinek <jakub@redhat.com>
19225 PR tree-optimization/99882
19226 * gimple-ssa-store-merging.c (bswap_view_convert): Handle val with
19229 2021-04-03 Jakub Jelinek <jakub@redhat.com>
19231 PR rtl-optimization/99863
19232 * dse.c (replace_read): Drop regs_live argument. Instead of
19233 regs_live, use store_insn->fixed_regs_live if non-NULL,
19234 otherwise punt if insns sequence clobbers or sets any hard
19237 2021-04-03 Jakub Jelinek <jakub@redhat.com>
19240 * targhooks.h (default_print_patchable_function_entry_1): Declare.
19241 * targhooks.c (default_print_patchable_function_entry_1): New function,
19242 copied from default_print_patchable_function_entry with an added flags
19244 (default_print_patchable_function_entry): Rewritten into a small
19245 wrapper around default_print_patchable_function_entry_1.
19246 * config/rs6000/rs6000.c (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY):
19248 (rs6000_print_patchable_function_entry): New function.
19250 2021-04-02 Eric Botcazou <ebotcazou@adacore.com>
19252 * doc/invoke.texi (fdelete-dead-exceptions): Minor tweak.
19254 2021-04-01 Jason Merrill <jason@redhat.com>
19257 * common.opt: Document v15 and v16.
19259 2021-04-01 Richard Biener <rguenther@suse.de>
19261 PR tree-optimization/99863
19262 * gimplify.c (gimplify_init_constructor): Recompute vector
19265 2021-04-01 Jakub Jelinek <jakub@redhat.com>
19267 * doc/extend.texi (symver attribute): Fix up syntax errors
19270 2021-04-01 Jakub Jelinek <jakub@redhat.com>
19272 PR tree-optimization/96573
19273 * gimple-ssa-store-merging.c (init_symbolic_number): Handle
19274 also pointer types.
19276 2021-04-01 Richard Biener <rguenther@suse.de>
19278 PR tree-optimization/99856
19279 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Promote
19280 precision to vector element precision.
19282 2021-04-01 Martin Jambor <mjambor@suse.cz>
19284 PR tree-optimization/97009
19285 * tree-sra.c (access_or_its_child_written): New function.
19286 (propagate_subaccesses_from_rhs): Use it instead of a simple grp_write
19289 2021-03-31 Jan Hubicka <hubicka@ucw.cz>
19292 * cif-code.def (USES_COMDAT_LOCAL): Make CIF_FINAL_NORMAL.
19294 2021-03-31 Pat Haugen <pthaugen@linux.ibm.com>
19297 * config/rs6000/altivec.md (xxspltiw_v4si, xxspltiw_v4sf_inst,
19298 xxspltidp_v2df_inst, xxsplti32dx_v4si_inst, xxsplti32dx_v4sf_inst,
19299 xxblend_<mode>, xxpermx_inst, xxeval): Mark prefixed.
19300 * config/rs6000/mma.md (mma_<vvi4i4i8>, mma_<avvi4i4i8>,
19301 mma_<vvi4i4i2>, mma_<avvi4i4i2>, mma_<vvi4i4>, mma_<avvi4i4>,
19302 mma_<pvi4i2>, mma_<apvi4i2>, mma_<vvi4i4i4>, mma_<avvi4i4i4>):
19304 * config/rs6000/rs6000.c (rs6000_final_prescan_insn): Adjust test.
19305 * config/rs6000/rs6000.md (define_attr "maybe_prefixed"): New.
19306 (define_attr "prefixed"): Update initializer.
19308 2021-03-31 Jakub Jelinek <jakub@redhat.com>
19311 * dwarf2out.c (debug_ranges_dwo_section): New variable.
19312 (DW_RANGES_IDX_SKELETON): Define.
19313 (struct dw_ranges): Add begin_entry and end_entry members.
19314 (DEBUG_DWO_RNGLISTS_SECTION): Define.
19315 (add_ranges_num): Adjust r initializer for addition of *_entry
19317 (add_ranges_by_labels): For -gsplit-dwarf and force_direct,
19318 set idx to DW_RANGES_IDX_SKELETON.
19319 (use_distinct_base_address_for_range): New function.
19320 (index_rnglists): Don't set r->idx if it is equal to
19321 DW_RANGES_IDX_SKELETON. Initialize r->begin_entry and
19322 r->end_entry for -gsplit-dwarf if those will be needed by
19324 (output_rnglists): Add DWO argument. If true, switch to
19325 debug_ranges_dwo_section rather than debug_ranges_section.
19326 Adjust l1/l2 label indexes. Only output the offset table when
19327 dwo is true and don't include in there the skeleton range
19328 entry if present. For -gsplit-dwarf, skip ranges that belong
19329 to the other rnglists section. Change return type from void
19330 to bool and return true if there are any range entries for
19331 the other section. For dwarf_split_debug_info use
19332 DW_RLE_startx_endx, DW_RLE_startx_length and DW_RLE_base_addressx
19333 entries instead of DW_RLE_start_end, DW_RLE_start_length and
19334 DW_RLE_base_address. Use use_distinct_base_address_for_range.
19335 (init_sections_and_labels): Initialize debug_ranges_dwo_section
19336 if -gsplit-dwarf and DWARF >= 5. Adjust ranges_section_label
19337 and range_base_label indexes.
19338 (dwarf2out_finish): Call index_rnglists earlier before finalizing
19339 .debug_addr. Never emit DW_AT_rnglists_base attribute. For
19340 -gsplit-dwarf and DWARF >= 5 call output_rnglists up to twice
19341 with different dwo arguments.
19342 (dwarf2out_c_finalize): Clear debug_ranges_dwo_section.
19344 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
19346 PR tree-optimization/98268
19347 * gimple-fold.c (maybe_canonicalize_mem_ref_addr): Call
19348 recompute_tree_invariant_for_addr_expr after successfully
19349 folding a TARGET_MEM_REF that occurs inside an ADDR_EXPR.
19351 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
19353 PR tree-optimization/99726
19354 * tree-data-ref.c (create_intersect_range_checks_index): Bail
19355 out if there is more than one access function SCEV for the loop
19358 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
19360 PR rtl-optimization/97141
19361 PR rtl-optimization/98726
19362 * emit-rtl.c (valid_for_const_vector_p): Return true for
19364 * rtx-vector-builder.h (rtx_vector_builder::step): Return a
19365 poly_wide_int instead of a wide_int.
19366 (rtx_vector_builder::apply_set): Take a poly_wide_int instead
19368 * rtx-vector-builder.c (rtx_vector_builder::apply_set): Likewise.
19369 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Return
19370 false for CONST_VECTORs that cannot be forced to memory.
19371 * config/aarch64/aarch64-simd.md (mov<mode>): If a CONST_VECTOR
19372 is too complex to force to memory, build it up from individual
19375 2021-03-31 Jan Hubicka <jh@suse.cz>
19378 * cgraph.c (cgraph_node::release_body): Fix overactive check.
19380 2021-03-31 Christophe Lyon <christophe.lyon@linaro.org>
19383 * config/arm/vec-common.md (mul<mode>3): Disable on iwMMXT, expect
19386 2021-03-31 H.J. Lu <hjl.tools@gmail.com>
19388 * config/i386/i386-expand.c (expand_set_or_cpymem_via_rep):
19389 For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, don't convert QImode
19391 (decide_alg): For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, use
19392 "rep movsb/stosb" only for known sizes.
19393 * config/i386/i386-options.c (processor_cost_table): Use Ice
19394 Lake cost for Cannon Lake, Ice Lake, Tiger Lake, Sapphire
19395 Rapids and Alder Lake.
19396 * config/i386/i386.h (TARGET_PREFER_KNOWN_REP_MOVSB_STOSB): New.
19397 * config/i386/x86-tune-costs.h (icelake_memcpy): New.
19398 (icelake_memset): Likewise.
19399 (icelake_cost): Likewise.
19400 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
19403 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
19406 * config/aarch64/aarch64.c
19407 (aarch64_vectorize_preferred_vector_alignment): Query the size
19408 of the provided SVE vector; do not assume that all SVE vectors
19409 have the same size.
19411 2021-03-31 Jan Hubicka <jh@suse.cz>
19414 * cgraph.c (cgraph_node::release_body): Remove all callers and
19416 * cgraphclones.c (cgraph_node::materialize_clone): Do not do it here.
19417 * cgraphunit.c (cgraph_node::expand): And here.
19419 2021-03-31 Martin Liska <mliska@suse.cz>
19421 * ipa-modref.c (analyze_ssa_name_flags): Fix coding style
19422 and one negated condition.
19424 2021-03-31 Jakub Jelinek <jakub@redhat.com>
19425 Richard Sandiford <richard.sandiford@arm.com>
19428 * config/aarch64/aarch64.md (*add<mode>3_poly_1): Swap Uai and Uav
19429 constraints on operands[2] and similarly 0 and rk constraints
19430 on operands[1] corresponding to that.
19432 2021-03-31 Jakub Jelinek <jakub@redhat.com>
19435 * configure.ac (HAVE_LD_BROKEN_PE_DWARF5): New AC_DEFINE if PECOFF
19436 linker doesn't support DWARF sections new in DWARF5.
19437 * config/i386/i386-options.c (ix86_option_override_internal): Default
19438 to dwarf_version 4 if HAVE_LD_BROKEN_PE_DWARF5 for TARGET_PECOFF
19440 * config.in: Regenerated.
19441 * configure: Regenerated.
19443 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19446 * config/aarch64/aarch64.c (aarch64_analyze_loop_vinfo): Check for
19447 available issue_info before using it.
19449 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19452 * config/aarch64/aarch64.md (sub<mode>3_compare1_imm): Do not allow zero
19455 2021-03-30 Xionghu Luo <luoxhu@linux.ibm.com>
19458 * config/rs6000/altivec.md (altivec_lvsl_reg): Change to ...
19459 (altivec_lvsl_reg_<mode>): ... this.
19460 (altivec_lvsr_reg): Change to ...
19461 (altivec_lvsr_reg_<mode>): ... this.
19462 * config/rs6000/predicates.md (vec_set_index_operand): New.
19463 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
19464 Enable 32bit variable vec_insert for all TARGET_VSX.
19465 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
19466 Enable 32bit variable vec_insert for p9 and above.
19467 (rs6000_expand_vector_set_var_p8): Rename to ...
19468 (rs6000_expand_vector_set_var_p7): ... this.
19469 (rs6000_expand_vector_set): Use TARGET_VSX and adjust assert
19471 * config/rs6000/vector.md (vec_set<mode>): Use vec_set_index_operand.
19472 * config/rs6000/vsx.md (xl_len_r): Use gen_altivec_lvsl_reg_di and
19473 gen_altivec_lvsr_reg_di.
19475 2021-03-30 H.J. Lu <hjl.tools@gmail.com>
19478 * config/i386/ia32intrin.h (__rdtsc): Defined as macro.
19479 (__rdtscp): Likewise.
19481 2021-03-30 Tamar Christina <tamar.christina@arm.com>
19483 PR tree-optimization/99825
19484 * tree-vect-slp-patterns.c (vect_check_evenodd_blend):
19485 Reject non-mult 2 lanes.
19487 2021-03-30 Richard Earnshaw <rearnsha@arm.com>
19490 * config/arm/arm.c (arm_file_start): Fix emission of
19491 Tag_ABI_VFP_args attribute.
19493 2021-03-30 Richard Biener <rguenther@suse.de>
19495 PR tree-optimization/99824
19496 * stor-layout.c (set_min_and_max_values_for_integral_type):
19497 Assert the precision is within the bounds of
19498 WIDE_INT_MAX_PRECISION.
19499 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
19500 the outermost component ref only to lower the access size
19501 and initialize that from the access type.
19503 2021-03-30 Richard Sandiford <richard.sandiford@arm.com>
19506 * config/aarch64/aarch64.md (mov<mode>): Pass multi-instruction
19507 CONST_INTs to aarch64_expand_mov_immediate when called after RA.
19509 2021-03-30 Mihailo Stojanovic <mihailo.stojanovic@typhoon-hil.com>
19511 * config/aarch64/aarch64.md
19512 (<optab>_trunc<fcvt_target><GPI:mode>2): Set the "arch"
19513 attribute to disambiguate between SIMD and FP variants of the
19516 2021-03-29 Jan Hubicka <hubicka@ucw.cz>
19518 * ipa-modref.c (merge_call_lhs_flags): Correct handling of deref.
19519 (analyze_ssa_name_flags): Fix typo in comment.
19521 2021-03-29 Alex Coplan <alex.coplan@arm.com>
19524 * config/aarch64/aarch64-sve-builtins.cc
19525 (function_builder::add_function): Add placeholder_p argument, use
19526 placeholder decls if this is set.
19527 (function_builder::add_unique_function): Instead of conditionally adding
19528 direct overloads, unconditionally add either a direct overload or a
19530 (function_builder::add_overloaded_function): Set placeholder_p if we're
19531 using C++ overloads. Use the obstack for string storage instead
19532 of relying on the tree nodes.
19533 (function_builder::add_overloaded_functions): Don't return early for
19534 m_direct_overloads: we need to add placeholders.
19535 * config/aarch64/aarch64-sve-builtins.h
19536 (function_builder::add_function): Add placeholder_p argument.
19538 2021-03-29 Richard Biener <rguenther@suse.de>
19540 PR tree-optimization/99807
19541 * tree-vect-slp.c (vect_slp_analyze_node_operations_1): Move
19542 assert below VEC_PERM handling.
19544 2021-03-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19547 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>): Use
19548 aarch64_simd_or_scalar_imm_zero to match zeroes. Remove pattern
19549 matching const_int 0.
19550 (move_lo_quad_internal_be_<mode>): Likewise.
19551 (move_lo_quad_<mode>): Update for the above.
19552 * config/aarch64/iterators.md (VQ_2E): Delete.
19554 2021-03-29 Jakub Jelinek <jakub@redhat.com>
19556 PR tree-optimization/99777
19557 * fold-const.c (extract_muldiv_1): For conversions, punt on casts from
19558 types other than scalar integral types.
19560 2021-03-28 David Edelsohn <dje.gcc@gmail.com>
19562 * config/rs6000/rs6000.c (rs6000_output_dwarf_dtprel): Do not add
19563 XCOFF TLS reloc decorations.
19565 2021-03-28 Gerald Pfeifer <gerald@pfeifer.com>
19567 * doc/analyzer.texi (Analyzer Internals): Update link to
19568 "A Memory Model for Static Analysis of C Programs".
19570 2021-03-26 David Edelsohn <dje.gcc@gmail.com>
19572 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Call function.
19573 * config/rs6000/rs6000-protos.h (rs6000_special_adjust_field_align):
19575 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align): New.
19576 (rs6000_special_round_type_align): Recursively check innermost first
19579 2021-03-26 Jakub Jelinek <jakub@redhat.com>
19582 * dwarf2out.h (struct dw_fde_node): Add rule18 member.
19583 * dwarf2cfi.c (dwarf2out_frame_debug_expr): When handling (set hfp sp)
19584 assignment with drap_reg active, queue reg save for hfp with offset 0
19585 and flush queued reg saves. When handling a push with rule18,
19586 defer queueing reg save for hfp and just assert the offset is 0.
19587 (scan_trace): Assert that fde->rule18 is false.
19589 2021-03-26 Vladimir Makarov <vmakarov@redhat.com>
19592 * ira-costs.c (record_reg_classes): Put case with
19593 CT_RELAXED_MEMORY adjacent to one with CT_MEMORY.
19594 * ira.c (ira_setup_alts): Ditto.
19595 * lra-constraints.c (process_alt_operands): Ditto.
19596 * recog.c (asm_operand_ok): Ditto.
19597 * reload.c (find_reloads): Ditto.
19599 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19601 * config/aarch64/aarch64-protos.h
19602 (cpu_addrcost_table::post_modify_ld3_st3): New member variable.
19603 (cpu_addrcost_table::post_modify_ld4_st4): Likewise.
19604 * config/aarch64/aarch64.c (generic_addrcost_table): Update
19605 accordingly, using the same costs as for post_modify.
19606 (exynosm1_addrcost_table, xgene1_addrcost_table): Likewise.
19607 (thunderx2t99_addrcost_table, thunderx3t110_addrcost_table):
19608 (tsv110_addrcost_table, qdf24xx_addrcost_table): Likewise.
19609 (a64fx_addrcost_table): Likewise.
19610 (neoversev1_addrcost_table): New.
19611 (neoversev1_tunings): Use neoversev1_addrcost_table.
19612 (aarch64_address_cost): Use the new post_modify costs for CImode
19615 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19617 * config/aarch64/aarch64.opt
19618 (-param=aarch64-loop-vect-issue-rate-niters=): New parameter.
19619 * doc/invoke.texi: Document it.
19620 * config/aarch64/aarch64-protos.h (aarch64_base_vec_issue_info)
19621 (aarch64_scalar_vec_issue_info, aarch64_simd_vec_issue_info)
19622 (aarch64_advsimd_vec_issue_info, aarch64_sve_vec_issue_info)
19623 (aarch64_vec_issue_info): New structures.
19624 (cpu_vector_cost): Write comments above the variables rather
19626 (cpu_vector_cost::issue_info): New member variable.
19627 * config/aarch64/aarch64.c: Include gimple-pretty-print.h
19628 and tree-ssa-loop-niter.h.
19629 (generic_vector_cost, a64fx_vector_cost, qdf24xx_vector_cost)
19630 (thunderx_vector_cost, tsv110_vector_cost, cortexa57_vector_cost)
19631 (exynosm1_vector_cost, xgene1_vector_cost, thunderx2t99_vector_cost)
19632 (thunderx3t110_vector_cost): Initialize issue_info to null.
19633 (neoversev1_scalar_issue_info, neoversev1_advsimd_issue_info)
19634 (neoversev1_sve_issue_info, neoversev1_vec_issue_info): New structures.
19635 (neoversev1_vector_cost): Use them.
19636 (aarch64_vec_op_count, aarch64_sve_op_count): New structures.
19637 (aarch64_vector_costs::saw_sve_only_op): New member variable.
19638 (aarch64_vector_costs::num_vector_iterations): Likewise.
19639 (aarch64_vector_costs::scalar_ops): Likewise.
19640 (aarch64_vector_costs::advsimd_ops): Likewise.
19641 (aarch64_vector_costs::sve_ops): Likewise.
19642 (aarch64_vector_costs::seen_loads): Likewise.
19643 (aarch64_simd_vec_costs_for_flags): New function.
19644 (aarch64_analyze_loop_vinfo): Initialize num_vector_iterations.
19645 Count the number of predicate operations required by SVE WHILE
19647 (aarch64_comparison_type, aarch64_multiply_add_p): New functions.
19648 (aarch64_sve_only_stmt_p, aarch64_in_loop_reduction_latency): Likewise.
19649 (aarch64_count_ops): Likewise.
19650 (aarch64_add_stmt_cost): Record whether see an SVE operation
19651 that cannot currently be implementing using Advanced SIMD.
19652 Record issue information about the scalar, Advanced SIMD
19653 and (where relevant) SVE versions of a loop.
19654 (aarch64_vec_op_count::dump): New function.
19655 (aarch64_sve_op_count::dump): Likewise.
19656 (aarch64_estimate_min_cycles_per_iter): Likewise.
19657 (aarch64_adjust_body_cost): If issue information is available,
19658 try to compare the issue rates of the various loop implementations
19659 and increase or decrease the vector body cost accordingly.
19661 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19663 * config/aarch64/aarch64.c (aarch64_detect_vector_stmt_subtype):
19664 Assume a zero cost for induction phis.
19666 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19668 * config/aarch64/aarch64.c (aarch64_embedded_comparison_type): New
19670 (aarch64_adjust_stmt_cost): Add the costs of embedded scalar and
19671 vector comparisons.
19673 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19675 * config/aarch64/aarch64.c (aarch64_detect_scalar_stmt_subtype):
19677 (aarch64_add_stmt_cost): Call it.
19679 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19681 * config/aarch64/aarch64-tuning-flags.def (matched_vector_throughput):
19682 New tuning parameter.
19683 * config/aarch64/aarch64.c (neoversev1_tunings): Use it.
19684 (aarch64_estimated_sve_vq): New function.
19685 (aarch64_vector_costs::analyzed_vinfo): New member variable.
19686 (aarch64_vector_costs::is_loop): Likewise.
19687 (aarch64_vector_costs::unrolled_advsimd_niters): Likewise.
19688 (aarch64_vector_costs::unrolled_advsimd_stmts): Likewise.
19689 (aarch64_record_potential_advsimd_unrolling): New function.
19690 (aarch64_analyze_loop_vinfo, aarch64_analyze_bb_vinfo): Likewise.
19691 (aarch64_add_stmt_cost): Call aarch64_analyze_loop_vinfo or
19692 aarch64_analyze_bb_vinfo on the first use of a costs structure.
19693 Detect whether we're vectorizing a loop for SVE that might be
19694 completely unrolled if it used Advanced SIMD instead.
19695 (aarch64_adjust_body_cost_for_latency): New function.
19696 (aarch64_finish_cost): Call it.
19698 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19700 * config/aarch64/aarch64.c (aarch64_vector_costs): New structure.
19701 (aarch64_init_cost): New function.
19702 (aarch64_add_stmt_cost): Use aarch64_vector_costs instead of
19703 the default unsigned[3].
19704 (aarch64_finish_cost, aarch64_destroy_cost_data): New functions.
19705 (TARGET_VECTORIZE_INIT_COST): Override.
19706 (TARGET_VECTORIZE_FINISH_COST): Likewise.
19707 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
19709 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19711 * config/aarch64/aarch64.c (neoversev1_advsimd_vector_cost)
19712 (neoversev1_sve_vector_cost): New cost structures.
19713 (neoversev1_vector_cost): Likewise.
19714 (neoversev1_tunings): Use them. Enable use_new_vector_costs.
19716 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19718 * config/aarch64/aarch64-protos.h
19719 (sve_vec_cost::scatter_store_elt_cost): New member variable.
19720 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
19721 accordingly, taking the cost from the cost of a scalar_store.
19722 (a64fx_sve_vector_cost): Likewise.
19723 (aarch64_detect_vector_stmt_subtype): Detect scatter stores.
19725 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19727 * config/aarch64/aarch64-protos.h
19728 (simd_vec_cost::store_elt_extra_cost): New member variable.
19729 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
19730 accordingly, using the vec_to_scalar cost for the new field.
19731 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
19732 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
19733 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
19734 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
19735 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
19736 (thunderx3t110_advsimd_vector_cost): Likewise.
19737 (aarch64_detect_vector_stmt_subtype): Detect single-element stores.
19739 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19741 * config/aarch64/aarch64-protos.h (simd_vec_cost::ld2_st2_permute_cost)
19742 (simd_vec_cost::ld3_st3_permute_cost): New member variables.
19743 (simd_vec_cost::ld4_st4_permute_cost): Likewise.
19744 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
19745 accordingly, using zero for the new costs.
19746 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
19747 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
19748 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
19749 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
19750 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
19751 (thunderx3t110_advsimd_vector_cost): Likewise.
19752 (aarch64_ld234_st234_vectors): New function.
19753 (aarch64_adjust_stmt_cost): Likewise.
19754 (aarch64_add_stmt_cost): Call aarch64_adjust_stmt_cost if using
19755 the new vector costs.
19757 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19759 * config/aarch64/aarch64-protos.h (sve_vec_cost): Turn into a
19760 derived class of simd_vec_cost. Add information about CLAST[AB]
19761 and FADDA instructions.
19762 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
19763 accordingly, using the vec_to_scalar costs for the new fields.
19764 (a64fx_sve_vector_cost): Likewise.
19765 (aarch64_reduc_type): New function.
19766 (aarch64_sve_in_loop_reduction_latency): Likewise.
19767 (aarch64_detect_vector_stmt_subtype): Take a vinfo parameter.
19768 Use aarch64_sve_in_loop_reduction_latency to handle SVE reductions
19769 that occur in the loop body.
19770 (aarch64_add_stmt_cost): Update call accordingly.
19772 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19774 * config/aarch64/aarch64-tuning-flags.def (use_new_vector_costs):
19776 * config/aarch64/aarch64-protos.h (simd_vec_cost): Put comments
19777 above the fields rather than to the right.
19778 (simd_vec_cost::reduc_i8_cost): New member variable.
19779 (simd_vec_cost::reduc_i16_cost): Likewise.
19780 (simd_vec_cost::reduc_i32_cost): Likewise.
19781 (simd_vec_cost::reduc_i64_cost): Likewise.
19782 (simd_vec_cost::reduc_f16_cost): Likewise.
19783 (simd_vec_cost::reduc_f32_cost): Likewise.
19784 (simd_vec_cost::reduc_f64_cost): Likewise.
19785 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
19786 accordingly, using the vec_to_scalar_cost for the new fields.
19787 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
19788 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
19789 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
19790 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
19791 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
19792 (thunderx3t110_advsimd_vector_cost): Likewise.
19793 (aarch64_use_new_vector_costs_p): New function.
19794 (aarch64_simd_vec_costs): New function, split out from...
19795 (aarch64_builtin_vectorization_cost): ...here.
19796 (aarch64_is_reduction): New function.
19797 (aarch64_detect_vector_stmt_subtype): Likewise.
19798 (aarch64_add_stmt_cost): Call aarch64_detect_vector_stmt_subtype if
19799 using the new vector costs.
19801 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
19804 * tree-emutls.c (get_emutls_init_templ_addr): Mark initializer of weak
19805 TLS declarations as public.
19807 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
19809 * config/aarch64/aarch64-d.c (IN_TARGET_CODE): Define.
19810 * config/arm/arm-d.c (IN_TARGET_CODE): Likewise.
19811 * config/i386/i386-d.c (IN_TARGET_CODE): Likewise.
19812 * config/mips/mips-d.c (IN_TARGET_CODE): Likewise.
19813 * config/pa/pa-d.c (IN_TARGET_CODE): Likewise.
19814 * config/riscv/riscv-d.c (IN_TARGET_CODE): Likewise.
19815 * config/rs6000/rs6000-d.c (IN_TARGET_CODE): Likewise.
19816 * config/s390/s390-d.c (IN_TARGET_CODE): Likewise.
19817 * config/sparc/sparc-d.c (IN_TARGET_CODE): Likewise.
19819 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
19822 * config.gcc (*-*-cygwin*): Add winnt-d.o
19823 (*-*-mingw*): Likewise.
19824 * config/i386/cygwin.h (EXTRA_TARGET_D_OS_VERSIONS): New macro.
19825 * config/i386/mingw32.h (EXTRA_TARGET_D_OS_VERSIONS): Likewise.
19826 * config/i386/t-cygming: Add winnt-d.o.
19827 * config/i386/winnt-d.c: New file.
19829 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
19831 * config/freebsd-d.c: Include memmodel.h.
19833 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
19836 * config.gcc (*-*-openbsd*): Add openbsd-d.o.
19837 * config/t-openbsd: Add openbsd-d.o.
19838 * config/openbsd-d.c: New file.
19840 2021-03-25 Stam Markianos-Wright <stam.markianos-wright@arm.com>
19842 PR tree-optimization/96974
19843 * tree-vect-stmts.c (vect_get_vector_types_for_stmt): Replace assert
19844 with graceful exit.
19846 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
19849 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
19853 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
19854 always_inline in system headers.
19856 2021-03-25 Kewen Lin <linkw@linux.ibm.com>
19858 * tree-vect-loop.c (vect_model_reduction_cost): Init inside_cost.
19860 2021-03-25 Jakub Jelinek <jakub@redhat.com>
19863 * tree-core.h (enum operand_equal_flag): Add OEP_ADDRESS_OF_SAME_FIELD.
19864 * fold-const.c (operand_compare::operand_equal_p): Don't compare
19865 field offsets if OEP_ADDRESS_OF_SAME_FIELD.
19867 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
19871 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
19872 always_inline in system headers.
19874 2021-03-25 Richard Biener <rguenther@suse.de>
19876 PR tree-optimization/99746
19877 * tree-vect-slp-patterns.c (complex_pattern::build): Do not mark
19878 the scalar stmt as patterned. Instead set up required things
19881 2021-03-25 Xionghu Luo <luoxhu@linux.ibm.com>
19883 * config/rs6000/rs6000.c (power8_costs): Change l2 cache
19886 2021-03-24 Martin Liska <mliska@suse.cz>
19889 * common/config/i386/i386-common.c (ARRAY_SIZE): Fix off-by-one
19891 * config/i386/i386-options.c (ix86_option_override_internal):
19892 Add run-time assert.
19894 2021-03-24 Martin Jambor <mjambor@suse.cz>
19897 * ipa-cp.c (initialize_node_lattices): Mark as bottom all
19898 parameters with unknown type.
19899 (ipacp_value_safe_for_type): New function.
19900 (propagate_vals_across_arith_jfunc): Verify that the constant type
19901 can be used for a type of the formal parameter.
19902 (propagate_vals_across_ancestor): Likewise.
19903 (propagate_scalar_across_jump_function): Likewise. Pass the type
19904 also to propagate_vals_across_ancestor.
19906 2021-03-24 Christophe Lyon <christophe.lyon@linaro.org>
19909 * config/arm/mve.md (movmisalign<mode>_mve_store): Use Ux
19911 (movmisalign<mode>_mve_load): Likewise.
19913 2021-03-24 Jakub Jelinek <jakub@redhat.com>
19916 * config/arm/vec-common.md (one_cmpl<mode>2, neg<mode>2,
19917 movmisalign<mode>): Disable expanders for TARGET_REALLY_IWMMXT.
19919 2021-03-24 Alexandre Oliva <oliva@adacore.com>
19921 * doc/sourcebuild.texi (sysconf): New effective target.
19923 2021-03-24 Alexandre Oliva <oliva@adacore.com>
19925 * config/i386/predicates.md (reg_or_const_vec_operand): New.
19926 * config/i386/sse.md (ssse3_pshufbv8qi3): Add an expander for
19927 the now *-prefixed insn_and_split, turn the splitter const vec
19928 into an input for the insn, making it an ignored immediate for
19929 non-split cases, and loaded into the scratch register
19932 2021-03-23 Vladimir N. Makarov <vmakarov@redhat.com>
19935 * config/aarch64/constraints.md (Utq, UOb, UOh, UOw, UOd, UOty):
19936 Use define_relaxed_memory_constraint for them.
19938 2021-03-23 Iain Sandoe <iain@sandoe.co.uk>
19941 * config/host-darwin.c (darwin_gt_pch_use_address): Add a
19942 colon to the diagnostic message.
19944 2021-03-23 Ilya Leoshkevich <iii@linux.ibm.com>
19946 * fwprop.c (fwprop_propagation::fwprop_propagation): Look at
19948 (try_fwprop_subst_note): Use set_info instead of insn_info.
19949 (try_fwprop_subst_pattern): Likewise.
19950 (try_fwprop_subst_notes): Likewise.
19951 (try_fwprop_subst): Likewise.
19952 (forward_propagate_subreg): Likewise.
19953 (forward_propagate_and_simplify): Likewise.
19954 (forward_propagate_into): Likewise.
19955 * rtl-ssa/accesses.h (set_info::single_nondebug_use) New
19957 (set_info::single_nondebug_insn_use): Likewise.
19958 (set_info::single_phi_use): Likewise.
19959 * rtl-ssa/member-fns.inl (set_info::single_nondebug_use) New
19961 (set_info::single_nondebug_insn_use): Likewise.
19962 (set_info::single_phi_use): Likewise.
19964 2021-03-23 Christophe Lyon <christophe.lyon@linaro.org>
19966 * doc/sourcebuild.texi (arm_dsp_ok, arm_dsp): Document.
19968 2021-03-23 Jakub Jelinek <jakub@redhat.com>
19971 * config/aarch64/aarch64.c (aarch64_add_offset): Tell
19972 expand_mult to perform an unsigned rather than a signed
19975 2021-03-23 H.J. Lu <hjl.tools@gmail.com>
19978 * config/i386/cpuid.h (__cpuid): Add __volatile__.
19979 (__cpuid_count): Likewise.
19981 2021-03-23 Richard Biener <rguenther@suse.de>
19983 PR tree-optimization/99721
19984 * tree-vect-slp.c (vect_slp_analyze_node_operations):
19985 Make sure we can schedule the node.
19987 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
19989 * config/riscv/riscv.c (riscv_subword): Take endianness into
19990 account when calculating the byte offset.
19992 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
19994 * config/riscv/predicates.md (subreg_lowpart_operator): New predicate
19995 * config/riscv/riscv.md (*addsi3_extended2, *subsi3_extended2)
19996 (*negsi2_extended2, *mulsi3_extended2, *<optab>si3_mask)
19997 (*<optab>si3_mask_1, *<optab>di3_mask, *<optab>di3_mask_1)
19998 (*<optab>si3_extend_mask, *<optab>si3_extend_mask_1): Use
19999 new predicate "subreg_lowpart_operator"
20001 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
20003 * config/riscv/riscv.c (riscv_swap_instruction): New function
20004 to byteswap an SImode rtx containing an instruction.
20005 (riscv_trampoline_init): Byteswap the generated instructions
20008 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
20010 * common/config/riscv/riscv-common.c
20011 (TARGET_DEFAULT_TARGET_FLAGS): Set default endianness.
20012 * config.gcc (riscv32be-*, riscv64be-*): Set
20013 TARGET_BIG_ENDIAN_DEFAULT to 1.
20014 * config/riscv/elf.h (LINK_SPEC): Change -melf* value
20015 depending on default endianness.
20016 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
20017 * config/riscv/linux.h (LINK_SPEC): Likewise.
20018 * config/riscv/riscv.c (TARGET_DEFAULT_TARGET_FLAGS): Set
20019 default endianness.
20020 * config/riscv/riscv.h (DEFAULT_ENDIAN_SPEC): New macro.
20022 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
20024 * config/riscv/elf.h (LINK_SPEC): Pass linker endianness flag.
20025 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
20026 * config/riscv/linux.h (LINK_SPEC): Likewise.
20027 * config/riscv/riscv.h (ASM_SPEC): Pass -mbig-endian and
20029 (BYTES_BIG_ENDIAN): Handle big endian.
20030 (WORDS_BIG_ENDIAN): Define to BYTES_BIG_ENDIAN.
20031 * config/riscv/riscv.opt (-mbig-endian, -mlittle-endian): New
20033 * doc/invoke.texi (-mbig-endian, -mlittle-endian): Document.
20035 2021-03-23 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
20037 * regcprop.c (find_oldest_value_reg): Ask target whether
20038 different mode is fine for replacement register.
20040 2021-03-23 Aldy Hernandez <aldyh@redhat.com>
20042 PR tree-optimization/99296
20043 * value-range.cc (irange::irange_set_1bit_anti_range): New.
20044 (irange::irange_set_anti_range): Call irange_set_1bit_anti_range
20045 * value-range.h (irange::irange_set_1bit_anti_range): New.
20047 2021-03-22 Vladimir N. Makarov <vmakarov@redhat.com>
20050 * config/aarch64/constraints.md (UtQ): Use
20051 define_relaxed_memory_constraint for it.
20052 * doc/md.texi (define_relaxed_memory_constraint): Describe it.
20053 * genoutput.c (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
20054 * genpreds.c (constraint_data): Add bitfield is_relaxed_memory.
20055 (have_relaxed_memory_constraints): New static var.
20056 (relaxed_memory_start, relaxed_memory_end): Ditto.
20057 (add_constraint): Add arg is_relaxed_memory. Check name for
20058 relaxed memory. Set up is_relaxed_memory in constraint_data and
20059 have_relaxed_memory_constraints. Adjust calls.
20060 (choose_enum_order): Process relaxed memory.
20061 (write_tm_preds_h): Ditto.
20062 (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
20063 * gensupport.c (process_rtx): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
20064 * ira-costs.c (record_reg_classes): Process CT_RELAXED_MEMORY.
20065 * ira-lives.c (single_reg_class): Use
20066 insn_extra_relaxed_memory_constraint.
20067 * ira.c (ira_setup_alts): CT_RELAXED_MEMORY.
20068 * lra-constraints.c (valid_address_p): Use
20069 insn_extra_relaxed_memory_constraint instead of other memory
20071 (process_alt_operands): Process CT_RELAXED_MEMORY.
20072 (curr_insn_transform): Use insn_extra_relaxed_memory_constraint.
20073 * recog.c (asm_operand_ok, preprocess_constraints): Process
20075 * reload.c (find_reloads): Ditto.
20076 * rtl.def (DEFINE_RELAXED_MEMORY_CONSTRAINT): New.
20077 * stmt.c (parse_input_constraint): Use
20078 insn_extra_relaxed_memory_constraint.
20080 2021-03-22 Segher Boessenkool <segher@kernel.crashing.org>
20083 * ubsan.c (ubsan_instrument_float_cast): Don't test for unordered if
20086 2021-03-22 Alex Coplan <alex.coplan@arm.com>
20089 * config/arm/arm-protos.h (neon_make_constant): Add generate
20090 argument to guard emitting insns, default to true.
20091 * config/arm/arm.c (arm_legitimate_constant_p_1): Reject
20092 CONST_VECTORs which neon_make_constant can't handle.
20093 (neon_vdup_constant): Add generate argument, avoid emitting
20094 insns if it's not set.
20095 (neon_make_constant): Plumb new generate argument through.
20096 * config/arm/constraints.md (Ui): New. Use it...
20097 * config/arm/mve.md (*mve_mov<mode>): ... here.
20098 * config/arm/vec-common.md (movv8hf): Use neon_make_constant to
20099 synthesize constants.
20101 2021-03-22 Richard Biener <rguenther@suse.de>
20103 * debug.h: Add deprecation warning.
20105 2021-03-22 Richard Biener <rguenther@suse.de>
20107 PR tree-optimization/99694
20108 * tree-ssa-sccvn.c (visit_phi): Ignore edges with the
20111 2021-03-22 Kito Cheng <kito.cheng@sifive.com>
20114 * config/riscv/riscv.c (riscv_expand_block_move): Get RTL value
20115 after type checking.
20117 2021-03-22 Jakub Jelinek <jakub@redhat.com>
20121 * dwarf2out.c (get_full_len): Use get_precision rather than
20123 (add_const_value_attribute): Make sure add_AT_wide argument has
20124 precision prec rather than some very wide one.
20126 2021-03-22 Kewen Lin <linkw@linux.ibm.com>
20128 * config/rs6000/rs6000.md (*rotldi3_insert_sf,
20129 *mov<SFDF:mode><SFDF2:mode>cc_p9, floatsi<mode>2_lfiwax,
20130 floatsi<mode>2_lfiwax_mem, floatunssi<mode>2_lfiwzx,
20131 floatunssi<mode>2_lfiwzx_mem, *floatsidf2_internal,
20132 *floatunssidf2_internal, fix_trunc<mode>si2_stfiwx,
20133 fix_trunc<mode>si2_internal, fixuns_trunc<mode>si2_stfiwx,
20134 *round32<mode>2_fprs, *roundu32<mode>2_fprs,
20135 *fix_trunc<mode>si2_internal): Fix empty split condition.
20136 * config/rs6000/vsx.md (*vsx_le_undo_permute_<mode>,
20137 vsx_reduc_<VEC_reduc_name>_v2df, vsx_reduc_<VEC_reduc_name>_v4sf,
20138 *vsx_reduc_<VEC_reduc_name>_v2df_scalar,
20139 *vsx_reduc_<VEC_reduc_name>_v4sf_scalar): Likewise.
20141 2021-03-22 Xionghu Luo <luoxhu@linux.ibm.com>
20144 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
20145 Convert idx to DImode.
20146 (rs6000_expand_vector_set_var_p8): Likewise.
20148 2021-03-21 Jakub Jelinek <jakub@redhat.com>
20151 * dwarf2out.c (insert_float): Change return type from void to
20152 unsigned, handle GET_MODE_SIZE (mode) == 2 and return element size.
20153 (mem_loc_descriptor, loc_descriptor, add_const_value_attribute):
20156 2021-03-20 H.J. Lu <hjl.tools@gmail.com>
20159 * config/i386/i386.c (construct_container): Check cfun != NULL
20160 before accessing silent_p.
20162 2021-03-20 Ahamed Husni <ahamedhusni73@gmail.com>
20164 * asan.c: Fix typos in comments.
20166 2021-03-20 Vladimir N. Makarov <vmakarov@redhat.com>
20168 PR rtl-optimization/99680
20169 * lra-constraints.c (skip_contraint_modifiers): Rename to skip_constraint_modifiers.
20170 (process_address_1): Check empty constraint before using
20173 2021-03-19 Pat Haugen <pthaugen@linux.ibm.com>
20175 * config/rs6000/rs6000.c (power10_cost): New.
20176 (rs6000_option_override_internal): Set Power10 costs.
20177 (rs6000_issue_rate): Set Power10 issue rate.
20178 * config/rs6000/power10.md: Rewrite for Power10.
20180 2021-03-19 Vladimir N. Makarov <vmakarov@redhat.com>
20183 * lra-constraints.c (process_address_1): Don't use unknown
20184 constraint for address constraint.
20186 2021-03-19 Iain Sandoe <iain@sandoe.co.uk>
20189 * config.gcc (powerpc-*-darwin8): Delete the reference to
20190 the now removed darwin8.h.
20192 2021-03-19 Olivier Hainque <hainque@adacore.com>
20195 * config/vxworksae.h (VX_CPU_PREFIX): Define.
20197 2021-03-19 John David Anglin <danglin@gcc.gnu.org>
20199 * config/pa/pa.c (import_milli): Use memcpy instead of strncpy.
20201 2021-03-19 Tamar Christina <tamar.christina@arm.com>
20203 PR tree-optimization/99656
20204 * tree-vect-slp-patterns.c (linear_loads_p,
20205 complex_add_pattern::matches, is_eq_or_top,
20206 vect_validate_multiplication, complex_mul_pattern::matches,
20207 complex_fms_pattern::matches): Remove complex_perm_kinds_t.
20208 * tree-vectorizer.h: (complex_load_perm_t): Removed.
20209 (slp_tree_to_load_perm_map_t): Use complex_perm_kinds_t instead of
20210 complex_load_perm_t.
20212 2021-03-19 H.J. Lu <hjl.tools@gmail.com>
20215 * config/i386/i386-options.c (ix86_init_machine_status): Set
20217 * config/i386/i386.c (init_cumulative_args): Set silent_p to
20219 (construct_container): Return early for return and argument
20220 errors if silent_p is true.
20221 * config/i386/i386.h (machine_function): Add silent_p.
20223 2021-03-19 Jakub Jelinek <jakub@redhat.com>
20226 * config/arm/constraints.md (Ds): New constraint.
20227 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): Use w,Ds
20228 constraint instead of w,Dm.
20230 2021-03-19 Andrew Stubbs <ams@codesourcery.com>
20232 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Fix quotes
20235 2021-03-19 Eric Botcazou <ebotcazou@adacore.com>
20237 PR middle-end/99641
20238 * fold-const.c (native_encode_initializer) <CONSTRUCTOR>: For an
20239 array type, do the computation of the current position in sizetype.
20241 2021-03-18 Vladimir N. Makarov <vmakarov@redhat.com>
20244 * lra-constraints.c (process_address_1): Use lookup_constraint
20245 only for a single constraint.
20247 2021-03-18 Martin Sebor <msebor@redhat.com>
20249 PR middle-end/99502
20250 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): Rename...
20251 (inbounds_memaccess_p): ...to this. Check the ending offset of
20252 the accessed member.
20254 2021-03-18 Andrew Stubbs <ams@codesourcery.com>
20256 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Add %< and
20257 %> quote markers to error messages.
20258 (gcn_goacc_validate_dims): Likewise.
20259 (gcn_conditional_register_usage): Remove exclaimation mark from error
20261 (gcn_vectorize_vec_perm_const): Ensure perm is fully uninitialized.
20263 2021-03-18 Jan Hubicka <hubicka@ucw.cz>
20265 * config/i386/x86-tune-costs.h (struct processor_costs): Fix costs of
20268 2021-03-18 Sinan Lin <sinan@isrc.iscas.ac.cn>
20269 Kito Cheng <kito.cheng@sifive.com>
20271 * config/riscv/riscv.c (riscv_block_move_straight): Change type
20272 to unsigned HOST_WIDE_INT for parameter and local variable with
20273 HOST_WIDE_INT type.
20274 (riscv_adjust_block_mem): Ditto.
20275 (riscv_block_move_loop): Ditto.
20276 (riscv_expand_block_move): Ditto.
20278 2021-03-18 Nick Clifton <nickc@redhat.com>
20280 * config/v850/v850.c (construct_restore_jr): Increase static
20282 (construct_save_jarl): Likewise.
20283 * config/v850/v850.h (DWARF2_DEBUGGING_INFO): Define.
20285 2021-03-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20287 * config/aarch64/aarch64.c (aarch64_adjust_generic_arch_tuning): Define.
20288 (aarch64_override_options_internal): Use it.
20289 (generic_tunings): Add AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS to
20292 2021-03-17 Sandra Loosemore <sandra@codesourcery.com>
20294 * config/nios2/nios2.c (nios2_custom_check_insns): Clean up
20295 error message format issues.
20296 (nios2_option_override): Likewise.
20297 (nios2_expand_fpu_builtin): Likewise.
20298 (nios2_init_custom_builtins): Adjust to avoid bogus strncpy
20299 truncation warning.
20300 (nios2_expand_custom_builtin): More error message format fixes.
20301 (nios2_expand_rdwrctl_builtin): Likewise.
20302 (nios2_expand_rdprs_builtin): Likewise.
20303 (nios2_expand_eni_builtin): Likewise.
20304 (nios2_expand_builtin): Likewise.
20305 (nios2_register_custom_code): Likewise.
20306 (nios2_valid_target_attribute_rec): Likewise.
20307 (nios2_add_insn_asm): Fix uninitialized variable warning.
20309 2021-03-17 Jan Hubicka <jh@suse.cz>
20311 * config/i386/x86-tune-costs.h (struct processor_costs): Update costs
20312 of gather to match reality.
20313 * config/i386/x86-tune.def (X86_TUNE_USE_GATHER): Enable for znver3.
20315 2021-03-17 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20317 * config/aarch64/aarch64-builtins.c (aarch64_expand_rng_builtin): Use EQ
20318 to compare against CC_REG rather than NE.
20320 2021-03-17 H.J. Lu <hjl.tools@gmail.com>
20323 * config/i386/i386.c (ix86_force_load_from_GOT_p): Support
20324 inline assembly statements.
20325 (ix86_print_operand): Update 'P' handling for -fno-plt.
20327 2021-03-17 Tamar Christina <tamar.christina@arm.com>
20330 * config/aarch64/aarch64.c
20331 (aarch64_simd_clone_compute_vecsize_and_simdlen): Remove unused var.
20333 2021-03-16 Segher Boessenkool <segher@kernel.crashing.org>
20336 * config/rs6000/predicates.md (branch_comparison_operator): Allow
20337 ordered and unordered for CCFPmode, if flag_finite_math_only.
20339 2021-03-16 Jakub Jelinek <jakub@redhat.com>
20342 * config/i386/i386-expand.c (ix86_split_lea_for_addr): Emit a MULT
20343 rather than ASHIFT.
20344 * config/i386/i386.md (mult by 1248 into ashift): New splitter.
20346 2021-03-16 Martin Liska <mliska@suse.cz>
20349 * optc-save-gen.awk: Add flag_ipa_ra to exceptions for
20350 cl_optimization_compare function.
20352 2021-03-16 Ilya Leoshkevich <iii@linux.ibm.com>
20354 * config/s390/s390.c (f_constraint_p): Treat "fv" constraints
20357 2021-03-16 Jakub Jelinek <jakub@redhat.com>
20360 * config/i386/i386.h (struct machine_function): Add
20361 has_explicit_vzeroupper bitfield.
20362 * config/i386/i386-expand.c (ix86_expand_builtin): Set
20363 cfun->machine->has_explicit_vzeroupper when expanding
20364 IX86_BUILTIN_VZEROUPPER.
20365 * config/i386/i386-features.c (rest_of_handle_insert_vzeroupper):
20366 Do the mode switching only when TARGET_VZEROUPPER, expensive
20367 optimizations turned on and not optimizing for size.
20368 (pass_insert_vzeroupper::gate): Enable even when
20369 cfun->machine->has_explicit_vzeroupper is set.
20371 2021-03-16 Jakub Jelinek <jakub@redhat.com>
20374 * config/aarch64/aarch64.c
20375 (aarch64_simd_clone_compute_vecsize_and_simdlen): If not a function
20376 definition, walk TYPE_ARG_TYPES list if non-NULL for argument types
20377 instead of DECL_ARGUMENTS. Ignore types for uniform arguments.
20379 2021-03-15 Richard Biener <rguenther@suse.de>
20381 PR tree-optimization/98834
20382 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle missing
20383 subsetting by truncating the access size.
20385 2021-03-15 Jan Hubicka <hubicka@ucw.cz>
20387 * config/i386/i386-options.c (processor_cost_table): Add znver3_cost.
20388 * config/i386/x86-tune-costs.h (znver3_cost): New gobal variable; copy
20391 2021-03-15 Martin Liska <mliska@suse.cz>
20393 * spellcheck.c: Add missing comma in initialization.
20395 2021-03-14 Uroš Bizjak <ubizjak@gmail.com>
20397 * config/i386/sse.md (*vec_extract<mode>): Merge alternative 0 with
20398 alternative 2 and alternative 1 with alternative 3 using
20399 YW register constraint.
20400 (*vec_extract<PEXTR_MODE12:mode>_zext): Merge alternatives
20401 using YW register constraint.
20402 (*vec_extractv16qi_zext): Ditto.
20403 (*vec_extractv4si): Merge alternatives 4 and 5
20404 using Yw register constraint.
20405 (*ssse3_palignr<mode>_perm): Use Yw instead of v for alternative 3.
20407 2021-03-13 Martin Sebor <msebor@redhat.com>
20409 PR tree-optimization/99489
20410 * builtins.c (gimple_call_alloc_size): Fail gracefully when argument
20411 is not a call statement.
20413 2021-03-13 Jakub Jelinek <jakub@redhat.com>
20415 PR tree-optimization/99544
20416 * match.pd (X + (X << C) -> X * (1 + (1 << C))): Don't simplify
20417 if for vector types multiplication can't be done in type's mode.
20419 2021-03-12 Eric Botcazou <ebotcazou@adacore.com>
20422 * config/sparc/constraints.md (w): Rename to...
20423 (W): ... this and ditch previous implementation.
20424 * config/sparc/sparc.md (*movdi_insn_sp64): Replace W with m.
20425 (*movdf_insn_sp64): Likewise.
20426 (*mov<VM64:mode>_insn_sp64): Likewise.
20427 * config/sparc/sync.md (*atomic_compare_and_swap<mode>_1): Replace
20429 (atomic_compare_and_swap_leon3_1): Likewise.
20430 (*atomic_compare_and_swapdi_v8plus): Likewise.
20431 * config/sparc/sparc.c (memory_ok_for_ldd): Remove useless test on
20432 architecture and add missing address validity check during LRA.
20434 2021-03-12 Tobias Burnus <tobias@codesourcery.com>
20437 * gimplify.c (omp_add_variable): Handle NULL_TREE as size
20438 occuring for assumed-size arrays in use_device_{ptr,addr}.
20440 2021-03-12 Jakub Jelinek <jakub@redhat.com>
20443 * config/i386/constraints.md (YW): New internal constraint.
20444 * config/i386/sse.md (v_Yw): Add V4TI, V2TI, V1TI and TI cases.
20445 (*<sse2_avx2>_<insn><mode>3<mask_name>,
20446 *<sse2_avx2>_uavg<mode>3<mask_name>, *abs<mode>2,
20447 *<s>mul<mode>3_highpart<mask_name>): Use <v_Yw> instead of v in
20449 (<sse2_avx2>_psadbw): Use YW instead of v in constraints.
20450 (*avx2_pmaddwd, *sse2_pmaddwd, *<code>v8hi3, *<code>v16qi3,
20451 avx2_pmaddubsw256, ssse3_pmaddubsw128): Merge last two alternatives
20452 into one, use Yw instead of former x,v.
20453 (ashr<mode>3, <insn><mode>3): Use <v_Yw> instead of x in constraints of
20454 the last alternative.
20455 (<sse2_avx2>_packsswb<mask_name>, <sse2_avx2>_packssdw<mask_name>,
20456 <sse2_avx2>_packuswb<mask_name>, <sse4_1_avx2>_packusdw<mask_name>,
20457 *<ssse3_avx2>_pmulhrsw<mode>3<mask_name>, <ssse3_avx2>_palignr<mode>,
20458 <ssse3_avx2>_pshufb<mode>3<mask_name>): Merge last two alternatives
20459 into one, use <v_Yw> instead of former x,v.
20460 (avx2_interleave_highv32qi<mask_name>,
20461 vec_interleave_highv16qi<mask_name>): Use Yw instead of v in
20462 constraints. Add && <mask_avx512bw_condition> to condition.
20463 (avx2_interleave_lowv32qi<mask_name>,
20464 vec_interleave_lowv16qi<mask_name>,
20465 avx2_interleave_highv16hi<mask_name>,
20466 vec_interleave_highv8hi<mask_name>,
20467 avx2_interleave_lowv16hi<mask_name>, vec_interleave_lowv8hi<mask_name>,
20468 avx2_pshuflw_1<mask_name>, sse2_pshuflw_1<mask_name>,
20469 avx2_pshufhw_1<mask_name>, sse2_pshufhw_1<mask_name>,
20470 avx2_<code>v16qiv16hi2<mask_name>, sse4_1_<code>v8qiv8hi2<mask_name>,
20471 *sse4_1_<code>v8qiv8hi2<mask_name>_1, <sse2_avx2>_<insn><mode>3): Use
20472 Yw instead of v in constraints.
20473 * config/i386/mmx.md (Yv_Yw): New define_mode_attr.
20474 (*mmx_<insn><mode>3, mmx_ashr<mode>3, mmx_<insn><mode>3): Use <Yv_Yw>
20475 instead of Yv in constraints.
20476 (*mmx_<insn><mode>3, *mmx_mulv4hi3, *mmx_smulv4hi3_highpart,
20477 *mmx_umulv4hi3_highpart, *mmx_pmaddwd, *mmx_<code>v4hi3,
20478 *mmx_<code>v8qi3, mmx_pack<s_trunsuffix>swb, mmx_packssdw,
20479 mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
20480 *mmx_uavgv8qi3, *mmx_uavgv4hi3, mmx_psadbw): Use Yw instead of Yv in
20482 (*mmx_pinsrw, *mmx_pinsrb, *mmx_pextrw, *mmx_pextrw_zext, *mmx_pextrb,
20483 *mmx_pextrb_zext): Use YW instead of Yv in constraints.
20484 (*mmx_eq<mode>3, mmx_gt<mode>3): Use x instead of Yv in constraints.
20485 (mmx_andnot<mode>3, *mmx_<code><mode>3): Split last alternative into
20486 two, one with just x, another isa avx512vl with v.
20488 2021-03-12 Martin Liska <mliska@suse.cz>
20490 * doc/invoke.texi: Add missing param documentation.
20492 2021-03-11 David Malcolm <dmalcolm@redhat.com>
20495 * Makefile.in (ANALYZER_OBJS): Add analyzer/feasible-graph.o and
20496 analyzer/trimmed-graph.o.
20497 * doc/analyzer.texi (Analyzer Paths): Rewrite description of
20498 feasibility checking to reflect new implementation.
20499 * doc/invoke.texi (-fdump-analyzer-feasibility): Document new
20501 * shortest-paths.h (shortest_paths::get_shortest_distance): New.
20503 2021-03-11 David Malcolm <dmalcolm@redhat.com>
20505 * digraph.cc (selftest::test_shortest_paths): Update
20506 shortest_paths init for new param. Add test of
20507 SPS_TO_GIVEN_TARGET.
20508 * shortest-paths.h (enum shortest_path_sense): New.
20509 (shortest_paths::shortest_paths): Add "sense" param.
20510 Update for renamings. Generalize to use "sense" param.
20511 (shortest_paths::get_shortest_path): Rename param.
20512 (shortest_paths::m_sense): New field.
20513 (shortest_paths::m_prev): Rename...
20514 (shortest_paths::m_best_edge): ...to this.
20515 (shortest_paths::get_shortest_path): Update for renamings.
20516 Conditionalize flipping of path on sense of traversal.
20518 2021-03-11 David Malcolm <dmalcolm@redhat.com>
20520 * digraph.cc (selftest::test_shortest_paths): Add test coverage
20521 for paths from B and C.
20522 * shortest-paths.h (shortest_paths::shortest_paths): Handle
20523 unreachable nodes, rather than asserting.
20525 2021-03-11 David Edelsohn <dje.gcc@gmail.com>
20528 * config/rs6000/rs6000.c (rs6000_xcoff_file_start): Don't create
20529 xcoff_tbss_section_name.
20530 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .lcomm.
20531 * xcoffout.c (xcoff_tbss_section_name): Delete.
20532 * xcoffout.h (xcoff_tbss_section_name): Delete.
20534 2021-03-11 Richard Biener <rguenther@suse.de>
20536 PR tree-optimization/99523
20537 * tree-cfg.c (dump_function_to_file): Dump SSA names
20538 w/o identifier to the decls section as well, not only those
20539 without a VAR_DECL.
20541 2021-03-11 Jakub Jelinek <jakub@redhat.com>
20544 * ipa-icf-gimple.c (func_checker::compare_gimple_call): For internal
20545 function calls with lhs fail if the lhs don't have compatible types.
20547 2021-03-11 Hans-Peter Nilsson <hp@axis.com>
20549 * config/cris/cris.h (HARD_FRAME_POINTER_REGNUM): Define.
20550 Change FRAME_POINTER_REGNUM to correspond to a new faked
20551 register faked_fp, part of GENNONACR_REGS like faked_ap.
20552 (CRIS_FAKED_REGS_CONTENTS): New helper macro.
20553 (FIRST_PSEUDO_REGISTER, FIXED_REGISTERS, CALL_USED_REGISTERS):
20554 (REG_ALLOC_ORDER, REG_CLASS_CONTENTS, REGNO_OK_FOR_BASE_P)
20555 (ELIMINABLE_REGS, REGISTER_NAMES): Adjust accordingly.
20556 * config/cris/cris.md (CRIS_FP_REGNUM): Renumber to new faked
20558 (CRIS_REAL_FP_REGNUM): New constant.
20559 * config/cris/cris.c (cris_reg_saved_in_regsave_area): Check
20560 for HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM.
20561 (cris_initial_elimination_offset): Handle elimination changes
20562 to HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM
20563 and add one from FRAME_POINTER_REGNUM to
20564 HARD_FRAME_POINTER_REGNUM.
20565 (cris_expand_prologue, cris_expand_epilogue): Emit code for
20566 hard_frame_pointer_rtx instead of frame_pointer_rtx.
20568 2021-03-10 David Edelsohn <dje.gcc@gmail.com>
20571 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Add check for DCmode.
20572 * config/rs6000/rs6000.c (rs6000_special_round_type_align): Same.
20574 2021-03-10 Vladimir N. Makarov <vmakarov@redhat.com>
20577 * lra-constraints.c (process_address_1): Don't check unknown
20578 constraint, use X for empty constraint.
20580 2021-03-10 Alex Coplan <alex.coplan@arm.com>
20582 * config/aarch64/aarch64.c (aarch64_vfp_is_call_or_return_candidate):
20583 Fix typo in comment describing "is_ha" argument.
20585 2021-03-10 John David Anglin <danglin@gcc.gnu.org>
20587 * doc/sourcebuild.texi: Document LRA target selector.
20589 2021-03-10 David Malcolm <dmalcolm@redhat.com>
20591 * doc/ux.texi: Add subsection contrasting interactive versus
20592 batch usage of GCC.
20594 2021-03-10 Joel Hutton <joel.hutton@arm.com>
20597 * tree-vect-stmts.c (vectorizable_store): Fix scatter store mask
20599 (vectorizable_load): Fix gather load mask check condition.
20601 2021-03-10 Richard Biener <rguenther@suse.de>
20603 PR tree-optimization/99510
20604 * tree.c (check_aligned_type): Check that the candidate
20605 has TYPE_USER_ALIGN set instead of matching with the
20608 2021-03-10 Eric Botcazou <ebotcazou@adacore.com>
20610 * config/sparc/sparc.c (sparc_regmode_natural_size): Return 4 for
20611 float and vector integer modes only if the mode is not larger.
20613 2021-03-10 Hans-Peter Nilsson <hp@axis.com>
20615 * config/cris/cris.h (DWARF_FRAME_REGISTERS): Define.
20617 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
20619 * ira.c (ira_setup_alts, ira_get_dup_out_num): Process digital
20621 * ira-lives.c (single_reg_class): Ditto.
20623 2021-03-09 Sebastian Huber <sebastian.huber@embedded-brains.de>
20625 * config.gcc (aarch64-*-rtems*): Include general rtems.h after
20626 the architecture-specific rtems.h.
20627 (aarch64-*-rtems*): Likewise.
20628 (arm*-*-rtems*): Likewise.
20629 (epiphany-*-rtems*): Likewise.
20630 (riscv*-*-rtems*): Likewise.
20632 2021-03-09 Jakub Jelinek <jakub@redhat.com>
20634 PR tree-optimization/99305
20635 * tree-ssa-phiopt.c (conditional_replacement): Test integer_pow2p
20636 before integer_all_onesp instead of vice versa.
20638 2021-03-09 Richard Earnshaw <rearnsha@arm.com>
20640 * common/config/arm/arm-common.c (arm_config_default): Change type
20641 of 'i' to unsigned.
20643 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
20646 * lra-constraints.c (process_address_1): Process constraint 'g'
20647 separately and digital constraints containing more one digit.
20649 2021-03-09 Nick Clifton <nickc@redhat.com>
20651 * config/rx/rx.h (DBX_DEBUGGING_INFO): Define.
20652 (DWARF"_DEBUGGING_INFO): Define.
20654 2021-03-09 Eric Botcazou <ebotcazou@adacore.com>
20657 * calls.c (initialize_argument_information): When the argument
20658 is passed by reference, do not make a copy in a thunk only if
20659 the argument is already in memory. Remove redundant test for
20660 the case of callee copy.
20662 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
20665 * lra-constraints.c (process_address_1): Process 0..9 constraints
20666 in process_address_1.
20668 2021-03-09 Andreas Krebbel <krebbel@linux.ibm.com>
20670 * config/s390/s390.c (struct s390_processor processor_table):
20671 Binutils name string must not be empty.
20673 2021-03-09 Claudiu Zissulescu <claziss@synopsys.com>
20675 * config/arc/arc.c (arc_attr_type): Remove function.
20677 2021-03-09 Martin Liska <mliska@suse.cz>
20680 * config/i386/i386-options.c (ix86_option_override_internal):
20681 Set isa_flags for OPTS argument and not for the global
20684 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
20686 * config/rs6000/predicates.md (ds_form_mem_operand): Check
20689 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
20692 * config/rs6000/predicates.md (ds_form_mem_operand) New
20694 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10) Use
20695 ds_form_mem_operand in ld/lwa patterns.
20696 * config/rs6000/fusion.md: Regenerate file.
20698 2021-03-08 Martin Sebor <msebor@redhat.com>
20700 PR middle-end/98266
20701 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): New function.
20702 (array_bounds_checker::check_array_bounds): Call it.
20704 2021-03-08 Martin Sebor <msebor@redhat.com>
20706 PR middle-end/97631
20707 * tree-ssa-strlen.c (maybe_warn_overflow): Test rawmem.
20708 (handle_builtin_stxncpy_strncat): Rename locals. Determine
20709 destination size from allocation calls. Issue a more appropriate
20711 (handle_builtin_memcpy): Pass true as rawmem to maybe_warn_overflow.
20712 (handle_builtin_memset): Same.
20714 2021-03-08 Peter Bergner <bergner@linux.ibm.com>
20717 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_permute): Add an assert
20718 to ensure we do not have an Altivec style address.
20719 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): Disable if passed
20720 an Altivec style address.
20721 (*vsx_le_perm_store_<mode>): Likewise.
20722 (splitters after *vsx_le_perm_store_<mode>): Likewise.
20723 (vsx_load_<mode>): Disable special expander if passed an Altivec
20725 (vsx_store_<mode>): Likewise.
20727 2021-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20730 * config/aarch64/predicates.md (aarch64_simd_shift_imm_vec_qi): Define.
20731 (aarch64_simd_shift_imm_vec_hi): Likewise.
20732 (aarch64_simd_shift_imm_vec_si): Likewise.
20733 (aarch64_simd_shift_imm_vec_di): Likewise.
20734 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le): Use
20735 predicate from above.
20736 (aarch64_shrn<mode>_insn_be): Likewise.
20737 (aarch64_rshrn<mode>_insn_le): Likewise.
20738 (aarch64_rshrn<mode>_insn_be): Likewise.
20739 (aarch64_shrn2<mode>_insn_le): Likewise.
20740 (aarch64_shrn2<mode>_insn_be): Likewise.
20741 (aarch64_rshrn2<mode>_insn_le): Likewise.
20742 (aarch64_rshrn2<mode>_insn_be): Likewise.
20744 2021-03-08 Vladimir N. Makarov <vmakarov@redhat.com>
20747 * lra-constraints.c (skip_contraint_modifiers): New function.
20748 (process_address_1): Use it before lookup_constraint call.
20750 2021-03-08 Martin Liska <mliska@suse.cz>
20753 * config/i386/i386-options.c (ix86_option_override_internal):
20754 Enable UINTR and HRESET for -march that supports it.
20756 2021-03-08 Ilya Leoshkevich <iii@linux.ibm.com>
20758 * config/s390/s390.c (f_constraint_p): New function.
20759 (s390_md_asm_adjust): Implement TARGET_MD_ASM_ADJUST.
20760 (TARGET_MD_ASM_ADJUST): Likewise.
20762 2021-03-08 Tobias Burnus <tobias@codesourcery.com>
20765 * tree-nested.c (convert_local_reference_stmt): Avoid calling
20766 lookup_field_for_decl for Fortran module (= namespace context).
20768 2021-03-08 Andreas Krebbel <krebbel@linux.ibm.com>
20770 * config/s390/s390.c (s390_expand_vec_compare): Implement <0
20771 comparison with arithmetic right shift.
20772 (s390_expand_vcond): No need for a force_reg anymore.
20773 s390_vec_compare will do it.
20774 * config/s390/vector.md ("vec_cmp<mode><tointvec>"): Accept also
20775 immediate operands.
20777 2021-03-07 Jakub Jelinek <jakub@redhat.com>
20780 * config/i386/constraints.md (Yw): Use SSE_REGS if TARGET_SSE
20781 but TARGET_AVX512BW or TARGET_AVX512VL is not set. Adjust description
20783 * config/i386/sse.md (v_Yw): New define_mode_attr.
20784 (*<insn><mode>3, *mul<mode>3<mask_name>, *avx2_<code><mode>3,
20785 *sse4_1_<code><mode>3<mask_name>): Use <v_Yw> instead of v
20787 * config/i386/mmx.md (mmx_pshufw_1, *vec_dupv4hi): Use Yw instead of
20788 xYw in constraints.
20790 2021-03-06 Julian Brown <julian@codesourcery.com>
20792 * tree-pretty-print.c (dump_generic_node): Emit non-generic
20793 address space info for aggregates.
20795 2021-03-06 Hans-Peter Nilsson <hp@axis.com>
20797 * config/cris/cris.h (MAX_FIXED_MODE_SIZE): Don't define.
20799 2021-03-05 Jakub Jelinek <jakub@redhat.com>
20801 PR middle-end/99322
20802 * tree-cfg.c (bb_to_omp_idx): New variable.
20803 (execute_build_cfg): Release the bb_to_omp_idx vector after
20804 cleanup_tree_cfg returns.
20805 (handle_abnormal_edges): Remove bb_to_omp_idx argument, adjust
20806 for bb_to_omp_idx being a vec<int> instead of pointer to array
20808 (make_edges): Remove bb_to_omp_idx local variable, don't pass
20809 it to handle_abnormal_edges, adjust for bb_to_omp_idx being a
20810 vec<int> instead of pointer to array of ints and don't free/release
20812 (remove_bb): When removing a bb and placing forced label somewhere
20813 else, ensure it is put into the same OpenMP region during cfg
20814 pass if possible or to entry successor as fallback. Unregister
20815 bb from bb_to_omp_idx.
20817 2021-03-05 Vladimir N. Makarov <vmakarov@redhat.com>
20820 * lra-constraints.c (process_address_1): Skip decomposing address
20821 for asm insn operand with unknown constraint.
20823 2021-03-05 Martin Jambor <mjambor@suse.cz>
20826 * cgraph.c (cgraph_edge::set_call_stmt): Do not update all
20827 corresponding speculative edges if we are about to resolve
20828 sepculation. Make edge direct (and so resolve speculations) before
20829 removing it from call_site_hash.
20830 (cgraph_edge::make_direct): Relax the initial assert to allow calling
20831 the function on speculative direct edges.
20833 2021-03-05 Eric Botcazou <ebotcazou@adacore.com>
20835 PR rtl-optimization/99376
20836 * rtlanal.c (nonzero_bits1) <arithmetic operators>: If the number
20837 of low-order zero bits is too large, set the result to 0 directly.
20839 2021-03-04 Jakub Jelinek <jakub@redhat.com>
20841 PR middle-end/93235
20842 * expmed.c (store_bit_field_using_insv): Return false of xop0 is a
20843 SUBREG and a SUBREG to op_mode can't be created.
20845 2021-03-04 Alex Coplan <alex.coplan@arm.com>
20848 * config/aarch64/aarch64-sve-builtins.cc
20849 (function_resolver::require_vector_type): Handle error_mark_node.
20851 2021-03-04 Ilya Leoshkevich <iii@linux.ibm.com>
20853 * cfgexpand.c (expand_asm_loc): Pass new parameter.
20854 (expand_asm_stmt): Likewise.
20855 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Add new
20857 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
20858 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
20859 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
20860 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
20861 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
20862 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
20863 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
20864 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
20865 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
20866 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
20867 * doc/tm.texi (md_asm_adjust): Likewise.
20868 * target.def (md_asm_adjust): Likewise.
20870 2021-03-04 Richard Biener <rguenther@suse.de>
20872 PR middle-end/97855
20873 * tree-pretty-print.c: Poison pp_printf.
20874 (dump_decl_name): Avoid use of pp_printf.
20875 (dump_block_node): Likewise.
20876 (dump_generic_node): Likewise.
20878 2021-03-04 Martin Sebor <msebor@redhat.com>
20880 PR middle-end/96963
20881 PR middle-end/94655
20882 * builtins.c (handle_array_ref): New helper.
20883 (handle_mem_ref): New helper.
20884 (compute_objsize_r): Factor out ARRAY_REF and MEM_REF handling
20885 into new helper functions. Correct a workaround for vectorized
20888 2021-03-03 Pat Haugen <pthaugen@linux.ibm.com>
20890 * config/rs6000/dfp.md (extendddtd2, trunctddd2, *cmp<mode>_internal1,
20891 floatditd2, ftrunc<mode>2, fix<mode>di2, dfp_ddedpd_<mode>,
20892 dfp_denbcd_<mode>, dfp_dxex_<mode>, dfp_diex_<mode>,
20893 *dfp_sgnfcnc_<mode>, dfp_dscli_<mode>, dfp_dscri_<mode>): Update size
20894 attribute for Power10.
20895 * config/rs6000/mma.md (*movoo): Likewise.
20896 * config/rs6000/rs6000.md (define_attr "size"): Add 256.
20897 (define_mode_attr bits): Add DD/TD modes.
20898 * config/rs6000/sync.md (load_quadpti, store_quadpti, load_lockedpti,
20899 store_conditionalpti): Update size attribute for Power10.
20901 2021-03-03 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20904 * config/sparc/t-sparc (tree-ssanames.o-warn): Don't error for
20905 -Wuninitialized, -Wmaybe-uninitialized.
20906 (wide-int.o-warn): Likewise.
20908 2021-03-03 Richard Earnshaw <rearnsha@arm.com>
20910 * common/config/arm/arm-common.c: Include configargs.h.
20911 (arm_config_default): New function.
20912 (arm_target_mode): Renamed from arm_target_thumb_only. Handle
20913 processors that do not support Thumb. Take into account the
20914 --with-mode configuration setting for selecting the default.
20915 * config/arm/arm.h (OPTION_DEFAULT_SPECS): Remove entry for 'mode'.
20916 (TARGET_MODE_SPEC_FUNCTIONS): Update for function name change.
20918 2021-03-03 Martin Liska <mliska@suse.cz>
20920 PR gcov-profile/97461
20921 * gcov-io.h (GCOV_PREALLOCATED_KVP): Remove.
20923 2021-03-03 Eric Botcazou <ebotcazou@adacore.com>
20926 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
20927 point back the hard frame pointer to its default location when the
20928 frame is larger than SEH_MAX_FRAME_SIZE.
20930 2021-03-03 Jakub Jelinek <jakub@redhat.com>
20933 * config/i386/predicates.md (logic_operator): New define_predicate.
20934 * config/i386/i386.md (mov + mem using comm arith peephole2):
20935 Punt if operands[1] is EXT_REX_SSE_REGNO_P, AVX512BW is not enabled
20936 and the inner mode is [QH]Imode.
20938 2021-03-03 Jakub Jelinek <jakub@redhat.com>
20941 * dwarf2out.c (dw_loc_list_struct): Add end_entry member.
20942 (new_loc_list): Clear end_entry.
20943 (output_loc_list): Only use DW_LLE_startx_length for -gsplit-dwarf
20944 if HAVE_AS_LEB128, otherwise use DW_LLE_startx_endx. Fix comment
20946 (index_location_lists): For dwarf_version >= 5 without HAVE_AS_LEB128,
20947 initialize also end_entry.
20949 2021-03-03 Jakub Jelinek <jakub@redhat.com>
20952 * cfgrtl.c (fixup_partitions): When changing some bbs from hot to cold
20953 partitions, if in non-layout mode after reorder_blocks also move
20954 affected blocks to ensure a single partition transition.
20956 2021-03-03 Jason Merrill <jason@redhat.com>
20959 * cgraphunit.c (process_function_and_variable_attributes): Don't
20960 warn about flatten on an alias if the target also has it.
20961 * cgraph.h (symtab_node::get_alias_target_tree): New.
20963 2021-03-02 David Edelsohn <dje.gcc@gmail.com>
20965 * config/rs6000/rs6000.md (tls_get_tpointer_internal): Prepend
20966 period to symbol name.
20967 (tls_get_addr_internal<mode>): Same.
20969 2021-03-02 David Malcolm <dmalcolm@redhat.com>
20972 * diagnostic-show-locus.c
20973 (selftest::test_one_liner_many_fixits_2): Fix accidental usage of
20976 2021-03-02 Martin Sebor <msebor@redhat.com>
20978 PR middle-end/99276
20979 * builtins.c (warn_for_access): Remove stray warning text.
20981 2021-03-02 Martin Sebor <msebor@redhat.com>
20983 PR middle-end/99295
20984 * doc/extend.texi (attribute malloc): Reword and clarify nonaliasing
20987 2021-03-02 Jakub Jelinek <jakub@redhat.com>
20990 * dwarf2out.c (output_macinfo_op): Use DW_MACRO_*_str* even with
20991 -gdwarf-5 -gstrict-dwarf. For -gsplit-dwarf -gdwarf-5 use
20992 DW_MACRO_*_strx instead of DW_MACRO_*_strp. Handle
20993 DW_MACRO_define_strx and DW_MACRO_undef_strx.
20994 (save_macinfo_strings): Use DW_MACRO_*_str* even with
20995 -gdwarf-5 -gstrict-dwarf. Handle DW_MACRO_define_strx and
20996 DW_MACRO_undef_strx.
20998 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
21000 * config/s390/s390-builtin-types.def (BT_FN_V4SF_V8HI_UINT): New
21002 (BT_FN_V8HI_V8HI_UINT): Likewise.
21003 (BT_FN_V8HI_V4SF_V4SF_UINT): Likewise.
21004 * config/s390/s390-builtins.def (B_NNPA): New macro definition.
21005 (s390_vclfnhs, s390_vclfnls, s390_vcrnfs, s390_vcfn, s390_vcnf):
21006 New builtin definitions.
21007 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Bump
21008 vector extension version.
21009 * config/s390/s390.c (s390_expand_builtin): Check if builtins are
21010 available with current -march level.
21011 * config/s390/s390.md (UNSPEC_NNPA_VCLFNHS_V8HI)
21012 (UNSPEC_NNPA_VCLFNLS_V8HI, UNSPEC_NNPA_VCRNFS_V8HI)
21013 (UNSPEC_NNPA_VCFN_V8HI, UNSPEC_NNPA_VCNF_V8HI): New constants.
21014 * config/s390/vecintrin.h (vec_extend_to_fp32_hi): New macro.
21015 (vec_extend_to_fp32_lo): Likewise.
21016 (vec_round_from_fp32): Likewise.
21017 (vec_convert_to_fp16): Likewise.
21018 (vec_convert_from_fp16): Likewise.
21019 * config/s390/vx-builtins.md (vclfnhs_v8hi): New insn pattern.
21020 (vclfnls_v8hi): Likewise.
21021 (vcrnfs_v8hi): Likewise.
21022 (vcfn_v8hi): Likewise.
21023 (vcnf_v8hi): Likewise.
21025 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
21027 * common/config/s390/s390-common.c (processor_flags_table): New entry.
21028 * config.gcc: Enable arch14 for --with-arch and --with-tune.
21029 * config/s390/driver-native.c (s390_host_detect_local_cpu): Pick
21030 arch14 for unknown CPU models.
21031 * config/s390/s390-opts.h (enum processor_type): Add PROCESSOR_ARCH14.
21032 * config/s390/s390.c (s390_issue_rate): Add case for PROCESSOR_ARCH14.
21033 (s390_get_sched_attrmask): Likewise.
21034 (s390_get_unit_mask): Likewise.
21035 * config/s390/s390.h (enum processor_flags): Add PF_NNPA and PF_ARCH14.
21036 (TARGET_CPU_ARCH14, TARGET_CPU_ARCH14_P, TARGET_CPU_NNPA)
21037 (TARGET_CPU_NNPA_P, TARGET_ARCH14, TARGET_ARCH14_P, TARGET_NNPA)
21038 (TARGET_NNPA_P): New macro definitions.
21039 * config/s390/s390.md ("cpu_facility", "enabled"): Add arch14 and nnpa.
21040 * config/s390/s390.opt: Add PROCESSOR_ARCH14.
21042 2021-03-02 Jakub Jelinek <jakub@redhat.com>
21044 PR middle-end/95757
21045 * tree-vrp.c (register_edge_assert_for): Remove superfluous ()s around
21046 condition. Call register_edge_assert_for_1 for == 0, != 0, == 1 and
21047 != 1 comparisons if name is lhs of a comparison.
21049 2021-03-01 Iain Sandoe <iain@sandoe.co.uk>
21053 * config/darwin-protos.h (darwin_should_restore_cfa_state): New.
21054 * config/darwin.c (darwin_should_restore_cfa_state): New.
21055 * config/darwin.h (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New.
21056 * doc/tm.texi: Regenerated.
21057 * doc/tm.texi.in: Document TARGET_ASM_SHOULD_RESTORE_CFA_STATE.
21058 * dwarf2cfi.c (connect_traces): If the target requests, restore
21059 the CFA expression after a DW_CFA_restore.
21060 * target.def (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New hook.
21062 2021-03-01 Martin Liska <mliska@suse.cz>
21065 * optc-save-gen.awk: Add 4 more exceptions.
21067 2021-03-01 Nathan Sidwell <nathan@acm.org>
21070 * tree.h (TYPE_ALIGN_RAW): New accessor.
21071 (TYPE_ALIGN): Use it.
21073 2021-03-01 Jan Hubicka <jh@suse.cz>
21076 * ipa-fnsummary.c (compute_fn_summary): Fix sanity check.
21078 2021-03-01 Eric Botcazou <ebotcazou@adacore.com>
21081 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
21082 point the hard frame pointer to the SSE register save area instead
21083 of the general register save area. Perform only minimal adjustment
21084 for small frames if it is initially not correctly aligned.
21085 (ix86_expand_prologue): Remove early saves for a SEH target.
21086 * config/i386/winnt.c (struct seh_frame_state): Document constraint.
21088 2021-02-28 Jakub Jelinek <jakub@redhat.com>
21091 * ipa.c (symbol_table::remove_unreachable_nodes): Fix a comment
21092 typo - referneced -> referenced.
21093 * tree.c (component_ref_size): Fix comment typo -
21094 refernce -> reference.
21095 * tree-ssa-alias.c (access_path_may_continue_p): Fix comment typo -
21096 traling -> trailing.
21097 (aliasing_component_refs_p): Fix comment typos -
21098 refernce -> reference and refernece -> reference and
21099 traling -> trailing.
21100 (nonoverlapping_refs_since_match_p): Fix comment typo -
21101 referneces -> references.
21102 * doc/invoke.texi (--param modref-max-bases): Fix a typo -
21103 referneces -> references.
21105 2021-02-27 Iain Sandoe <iain@sandoe.co.uk>
21107 * config/host-darwin.c (darwin_gt_pch_use_address): Modify
21108 diagnostic message to avoid use of a contraction and format
21111 2021-02-27 Jakub Jelinek <jakub@redhat.com>
21114 * gcse.c (gcse_or_cprop_is_too_expensive): Use %wu instead of
21115 HOST_WIDE_INT_PRINT_UNSIGNED in warning format string.
21116 * ipa-devirt.c (ipa_odr_read_section): Use %wd instead of
21117 HOST_WIDE_INT_PRINT_DEC in inform format string. Fix comment
21120 2021-02-26 Richard Biener <rguenther@suse.de>
21122 PR middle-end/99281
21123 * expr.c (store_field): For calls with return-slot optimization
21124 and addressable return type expand the store directly.
21126 2021-02-26 Richard Biener <rguenther@suse.de>
21129 * builtins.c (warn_string_no_nul): Fix diagnostic formatting.
21131 2021-02-26 Peter Bergner <bergner@linux.ibm.com>
21134 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Replace assert
21137 2021-02-26 Aaron Sawdey <acsawdey@linux.ibm.com>
21139 * config.gcc: Add rs6000-pcrel-opt.o.
21140 * config/rs6000/rs6000-pcrel-opt.c: New file.
21141 * config/rs6000/pcrel-opt.md: New file.
21142 * config/rs6000/predicates.md: Add d_form_memory predicate.
21143 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_PCREL_OPT.
21144 * config/rs6000/rs6000-passes.def: Add pass_pcrel_opt.
21145 * config/rs6000/rs6000-protos.h: Add reg_to_non_prefixed(),
21146 pcrel_opt_valid_mem_p(), output_pcrel_opt_reloc(),
21147 and make_pass_pcrel_opt().
21148 * config/rs6000/rs6000.c (reg_to_non_prefixed): Make global.
21149 (rs6000_option_override_internal): Add pcrel-opt.
21150 (rs6000_delegitimize_address): Support pcrel-opt.
21151 (rs6000_opt_masks): Add pcrel-opt.
21152 (pcrel_opt_valid_mem_p): New function.
21153 (reg_to_non_prefixed): Make global.
21154 (rs6000_asm_output_opcode): Reset prepend_p_to_next_insn.
21155 (output_pcrel_opt_reloc): New function.
21156 * config/rs6000/rs6000.md (loads_extern_addr): New attr.
21157 (pcrel_extern_addr): Set loads_extern_addr.
21158 Add include for pcrel-opt.md.
21159 * config/rs6000/rs6000.opt: Add -mpcrel-opt.
21160 * config/rs6000/t-rs6000: Add rules for pcrel-opt.c and
21163 2021-02-26 YunQiang Su <yunqiang.su@cipunited.com>
21166 * config/mips/mips.c (mips_expand_ext_as_unaligned_load):
21167 If TARGET_64BIT and dest is SUBREG, we check the width, if it
21168 equal to SImode, we use SImode operation, just like what we are
21171 2021-02-26 Marek Polacek <polacek@redhat.com>
21173 * builtins.c (warn_for_access): Fix typos.
21175 2021-02-25 Iain Sandoe <iain@sandoe.co.uk>
21177 * config/aarch64/aarch64.md (<optab>_rol<mode>3): Add a '#'
21178 mark in front of the immediate quantity.
21179 (<optab>_rolsi3_uxtw): Likewise.
21181 2021-02-25 Richard Earnshaw <rearnsha@arm.com>
21184 * config/arm/thumb2.md (nonsecure_call_reg_thumb2_fpcxt): New pattern.
21185 (nonsecure_call_value_reg_thumb2_fpcxt): Likewise.
21186 (nonsecure_call_reg_thumb2): Restrict to using r4 for the callee
21187 address and disable when the FPCXT is not available.
21188 (nonsecure_call_value_reg_thumb2): Likewise.
21190 2021-02-25 Nathan Sidwell <nathan@acm.org>
21193 * doc/invoke.texi (flang-info-module-cmi): Renamed option.
21195 2021-02-25 Tamar Christina <tamar.christina@arm.com>
21197 * tree-vect-slp.c (optimize_load_redistribution_1): Abort on NULL nodes.
21199 2021-02-25 Richard Biener <rguenther@suse.de>
21201 PR tree-optimization/99253
21202 * tree-vect-loop.c (check_reduction_path): First compute
21203 code, then verify out-of-loop uses.
21205 2021-02-25 Jakub Jelinek <jakub@redhat.com>
21208 * match.pd ((T)(A) + CST -> (T)(A + CST)): Add :s to convert.
21210 2021-02-25 Jakub Jelinek <jakub@redhat.com>
21212 PR tree-optimization/80635
21213 * tree-vrp.c (vrp_simplify_cond_using_ranges): Also handle
21214 VIEW_CONVERT_EXPR if modes are the same, innerop is integral and
21215 has mode precision.
21217 2021-02-25 Richard Biener <rguenther@suse.de>
21219 * tree-vect-slp.c (optimize_load_redistribution_1): Delay
21220 load_map population.
21221 (vect_match_slp_patterns_2): Revert part of last change.
21222 (vect_analyze_slp): Do not interleave optimize_load_redistribution
21223 with pattern detection but do it afterwards. Dump the
21224 whole SLP graph after pattern recognition and load
21225 redistribution optimization finished.
21227 2021-02-24 Jakub Jelinek <jakub@redhat.com>
21230 * omp-low.c (struct omp_context): Add teams_nested_p and
21231 nonteams_nested_p members.
21232 (scan_omp_target): Diagnose teams nested inside of target with other
21233 directives strictly nested inside of the same target.
21234 (check_omp_nesting_restrictions): Set ctx->teams_nested_p or
21235 ctx->nonteams_nested_p as needed.
21237 2021-02-24 Vladimir N. Makarov <vmakarov@redhat.com>
21239 PR inline-asm/99123
21240 * lra-constraints.c (uses_hard_regs_p): Don't use decompose_mem_address.
21242 2021-02-24 Hans-Peter Nilsson <hp@axis.com>
21244 * config/cris/cris.c (cris_expand_prologue): Set
21245 current_function_static_stack_size, if flag_stack_usage_info.
21247 2021-02-24 Pat Haugen <pthaugen@linux.ibm.com>
21249 * config/rs6000/rs6000.c (next_insn_prefixed_p): Rename.
21250 (rs6000_final_prescan_insn): Adjust.
21251 (rs6000_asm_output_opcode): Likewise.
21253 2021-02-24 Martin Sebor <msebor@redhat.com>
21255 PR middle-end/97172
21256 * attribs.c (attr_access::free_lang_data): Clear attribute arg spec
21257 from function arguments.
21259 2021-02-24 Tamar Christina <tamar.christina@arm.com>
21261 PR tree-optimization/99220
21262 * tree-vect-slp.c (optimize_load_redistribution_1): Remove
21263 node from cache when it's about to be deleted.
21265 2021-02-24 Jakub Jelinek <jakub@redhat.com>
21267 PR tree-optimization/99225
21268 * fold-const.c (fold_binary_loc) <case NE_EXPR>: In (x & (1 << y)) != 0
21269 to ((x >> y) & 1) != 0 simplifications use build_one_cst instead of
21270 build_int_cst (..., 1). Formatting fixes.
21272 2021-02-24 Tamar Christina <tamar.christina@arm.com>
21274 PR tree-optimization/99149
21275 * tree-vect-slp-patterns.c (vect_detect_pair_op): Don't recreate the
21277 (vect_slp_reset_pattern): Remove.
21278 (complex_fma_pattern::matches): Remove call to vect_slp_reset_pattern.
21279 (complex_mul_pattern::build, complex_fma_pattern::build,
21280 complex_fms_pattern::build): Fix ref counts.
21281 * tree-vect-slp.c (vect_free_slp_tree): Undo SLP only pattern relevancy
21282 when node is being deleted.
21283 (vect_match_slp_patterns_2): Correct result of cache hit on patterns.
21284 (vect_schedule_slp): Invalidate SLP_TREE_REPRESENTATIVE of removed
21286 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize value.
21288 2021-02-24 Matthias Klose <doko@ubuntu.com>
21291 2020-12-07 Matthias Klose <doko@ubuntu.com>
21293 * genextract.c (print_header): Undefine ENABLE_RTL_CHECKING
21294 and ENABLE_RTL_FLAG_CHECKING.
21296 2021-02-24 Richard Biener <rguenther@suse.de>
21299 * builtins.c (fold_builtin_next_arg): Avoid NULL arg.
21301 2021-02-23 Peter Bergner <bergner@linux.ibm.com>
21303 * config/rs6000/mma.md (mma_assemble_pair): Rename from this...
21304 (vsx_assemble_pair): ...to this.
21305 (*mma_assemble_pair): Rename from this...
21306 (*vsx_assemble_pair): ...to this.
21307 (mma_disassemble_pair): Rename from this...
21308 (vsx_disassemble_pair): ...to this.
21309 (*mma_disassemble_pair): Rename from this...
21310 (*vsx_disassemble_pair): ...to this.
21311 * config/rs6000/rs6000-builtin.def (BU_MMA_V2, BU_MMA_V3,
21312 BU_COMPAT): New macros.
21313 (mma_assemble_pair): Rename from this...
21314 (vsx_assemble_pair): ...to this.
21315 (mma_disassemble_pair): Rename from this...
21316 (vsx_disassemble_pair): ...to this.
21317 (mma_assemble_pair): New compatibility built-in.
21318 (mma_disassemble_pair): Likewise.
21319 * config/rs6000/rs6000-call.c (struct builtin_compatibility): New.
21320 (RS6000_BUILTIN_COMPAT): Define.
21321 (bdesc_compat): New.
21322 (mma_expand_builtin): Use VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
21323 (rs6000_gimple_fold_mma_builtin): Use MMA_BUILTIN_DISASSEMBLE_PAIR
21324 and VSX_BUILTIN_ASSEMBLE_PAIR.
21325 (rs6000_init_builtins): Register compatibility built-ins.
21326 (mma_init_builtins): Use VSX_BUILTIN_ASSEMBLE_PAIR,
21327 VSX_BUILTIN_ASSEMBLE_PAIR_INTERNAL, VSX_BUILTIN_DISASSEMBLE_PAIR and
21328 VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
21329 * doc/extend.texi (__builtin_mma_assemble_pair): Rename from this...
21330 (__builtin_vsx_assemble_pair): ...to this.
21331 (__builtin_mma_disassemble_pair): Rename from this...
21332 (__builtin_vsx_disassemble_pair): ...to this.
21334 2021-02-23 Martin Liska <mliska@suse.cz>
21337 * ipa-icf.c (sem_variable::merge): Do not merge 2 variables
21338 with different alignment. That leads to an invalid red zone
21339 size allocated in runtime.
21341 2021-02-23 Jakub Jelinek <jakub@redhat.com>
21343 PR tree-optimization/99204
21344 * fold-const.c (fold_read_from_constant_string): Check that
21345 tree_fits_uhwi_p (index) rather than just that index is INTEGER_CST.
21347 2021-02-23 Segher Boessenkool <segher@kernel.crashing.org>
21348 Kewen Lin <linkw@gcc.gnu.org>
21350 * config/rs6000/rs6000.md (*rotl<mode>3_insert_3): Renamed to...
21351 (rotl<mode>3_insert_3): ...this.
21352 (plus_ior_xor): New code_iterator.
21353 (define_split for GPR rl*imi): New splitter.
21354 * config/rs6000/vsx.md (vsx_init_v4si): Use gen_rotldi3_insert_3
21355 for integer merging.
21357 2021-02-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21359 * config/aarch64/aarch64-tuning-flags.def (cse_sve_vl_constants):
21361 * config/aarch64/aarch64.md (add<mode>3): Force CONST_POLY_INT immediates
21362 into a register when the above is enabled.
21363 * config/aarch64/aarch64.c (neoversev1_tunings):
21364 AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
21365 (aarch64_rtx_costs): Use AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
21367 2021-02-22 Hans-Peter Nilsson <hp@axis.com>
21369 * config/cris/cris.c (cris_print_operand) <'T'>: Change
21370 valid operand from is now an addi mult-value to shift-value.
21371 * config/cris/cris.md (*addi): Change expression of scaled
21372 operand from mult to ashift.
21373 * config/cris/cris.md (*addi_reload): New insn_and_split.
21375 2021-02-22 John David Anglin <danglin@gcc.gnu.org>
21378 * config/pa/pa.c (TARGET_ASM_CAN_OUTPUT_MI_THUNK): Define as
21379 hook_bool_const_tree_hwi_hwi_const_tree_true.
21380 (pa_asm_output_mi_thunk): Add support for nonzero vcall_offset.
21382 2021-02-22 Andre Vieira <andre.simoesdiasvieira@arm.com>
21384 PR rtl-optimization/98791
21385 * ira-conflicts.c (process_regs_for_copy): Don't create allocno copies
21386 for unordered modes.
21388 2021-02-22 Martin Liska <mliska@suse.cz>
21390 * tree-inline.c (inline_forbidden_p): Set
21391 inline_forbidden_reason.
21393 2021-02-22 Richard Biener <rguenther@suse.de>
21395 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Dump
21398 2021-02-22 Richard Biener <rguenther@suse.de>
21400 PR tree-optimization/99165
21401 * gimple-ssa-store-merging.c (pass_store_merging::process_store):
21402 Accumulate changed to ret.
21404 2021-02-21 Uros Bizjak <ubizjak@gmail.com>
21407 2020-12-09 Uroš Bizjak <ubizjak@gmail.com>
21409 * config/i386/i386.h (REG_ALLOC_ORDER): Remove
21411 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
21414 * config/s390/vector.md (trunctf<DFP_ALL:mode>2_vr): New
21416 (trunctf<DFP_ALL:mode>2): Likewise.
21417 (trunctdtf2_vr): Likewise.
21418 (trunctdtf2): Likewise.
21419 (extend<DFP_ALL:mode>tf2_vr): Likewise.
21420 (extend<DFP_ALL:mode>tf2): Likewise.
21421 (extendtftd2_vr): Likewise.
21422 (extendtftd2): Likewise.
21424 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
21426 * config/s390/vector.md (*fprx2_to_tf): Rename to fprx2_to_tf,
21427 add memory alternative.
21428 (tf_to_fprx2): New pattern.
21430 2021-02-19 Martin Sebor <msebor@redhat.com>
21433 * attribs.c (init_attr_rdwr_indices): Guard vblist use.
21434 (attr_access::free_lang_data): Remove a spurious test.
21436 2021-02-19 Nathan Sidwell <nathan@acm.org>
21438 * doc/invoke.texi (flang-info-module-read): Document.
21440 2021-02-19 Martin Liska <mliska@suse.cz>
21442 PR translation/99167
21443 * params.opt: Fix typo.
21445 2021-02-19 Richard Biener <rguenther@suse.de>
21447 PR middle-end/99122
21448 * tree-inline.c (inline_forbidden_p): Do not inline functions
21449 with VLA arguments or return value.
21451 2021-02-19 Jakub Jelinek <jakub@redhat.com>
21454 * config/arm/arm.md (*stack_protect_combined_set_insn,
21455 *stack_protect_combined_test_insn): If force_const_mem result
21456 is not valid general operand, force its address into the destination
21459 2021-02-19 Jakub Jelinek <jakub@redhat.com>
21462 * tree-cfg.c (gimple_merge_blocks): If bb a starts with eh landing
21463 pad or non-local label, put FORCED_LABELs from bb b after that label
21464 rather than before it.
21466 2021-02-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
21469 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3): Use
21470 expand_vector_broadcast' to emit the vec_duplicate operand.
21472 2021-02-18 Vladimir N. Makarov <vmakarov@redhat.com>
21474 PR rtl-optimization/96264
21475 * lra-remat.c (reg_overlap_for_remat_p): Check also output insn
21478 2021-02-18 H.J. Lu <hjl.tools@gmail.com>
21481 * varasm.c (get_section): Replace SUPPORTS_SHF_GNU_RETAIN with
21482 looking up the retain attribute.
21483 (resolve_unique_section): Likewise.
21484 (get_variable_section): Likewise.
21485 (switch_to_section): Likewise. Warn when a symbol without the
21486 retain attribute and a symbol with the retain attribute are
21487 placed in the section with the same name, instead of the used
21489 * doc/extend.texi: Document the "retain" attribute.
21491 2021-02-18 Nathan Sidwell <nathan@acm.org>
21494 * doc/invoke.texi (flang-info-include-translate): Document header
21497 2021-02-18 Richard Biener <rguenther@suse.de>
21499 PR middle-end/99122
21500 * ipa-fnsummary.c (analyze_function_body): Set
21501 CIF_FUNCTION_NOT_INLINABLE for VLA parameter calls.
21502 * tree-inline.c (insert_init_debug_bind): Pass NULL for
21503 error_mark_node values.
21504 (force_value_to_type): Do not build V_C_Es for WITH_SIZE_EXPR
21506 (setup_one_parameter): Delay force_value_to_type until when
21509 2021-02-18 Hans-Peter Nilsson <hp@axis.com>
21511 PR tree-optimization/99142
21512 * match.pd (clz cmp 0): Gate replacement on single_use of clz result.
21514 2021-02-18 Jakub Jelinek <jakub@redhat.com>
21516 * wide-int-bitmask.h (wide_int_bitmask::wide_int_bitmask (),
21517 wide_int_bitmask::wide_int_bitmask (uint64_t),
21518 wide_int_bitmask::wide_int_bitmask (uint64_t, uint64_t),
21519 wide_int_bitmask::operator ~ () const,
21520 wide_int_bitmask::operator | (wide_int_bitmask) const,
21521 wide_int_bitmask::operator & (wide_int_bitmask) const): Use constexpr
21523 * config/i386/i386.h (PTA_3DNOW, PTA_3DNOW_A, PTA_64BIT, PTA_ABM,
21524 PTA_AES, PTA_AVX, PTA_BMI, PTA_CX16, PTA_F16C, PTA_FMA, PTA_FMA4,
21525 PTA_FSGSBASE, PTA_LWP, PTA_LZCNT, PTA_MMX, PTA_MOVBE, PTA_NO_SAHF,
21526 PTA_PCLMUL, PTA_POPCNT, PTA_PREFETCH_SSE, PTA_RDRND, PTA_SSE, PTA_SSE2,
21527 PTA_SSE3, PTA_SSE4_1, PTA_SSE4_2, PTA_SSE4A, PTA_SSSE3, PTA_TBM,
21528 PTA_XOP, PTA_AVX2, PTA_BMI2, PTA_RTM, PTA_HLE, PTA_PRFCHW, PTA_RDSEED,
21529 PTA_ADX, PTA_FXSR, PTA_XSAVE, PTA_XSAVEOPT, PTA_AVX512F, PTA_AVX512ER,
21530 PTA_AVX512PF, PTA_AVX512CD, PTA_NO_TUNE, PTA_SHA, PTA_PREFETCHWT1,
21531 PTA_CLFLUSHOPT, PTA_XSAVEC, PTA_XSAVES, PTA_AVX512DQ, PTA_AVX512BW,
21532 PTA_AVX512VL, PTA_AVX512IFMA, PTA_AVX512VBMI, PTA_CLWB, PTA_MWAITX,
21533 PTA_CLZERO, PTA_NO_80387, PTA_PKU, PTA_AVX5124VNNIW, PTA_AVX5124FMAPS,
21534 PTA_AVX512VPOPCNTDQ, PTA_SGX, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES,
21535 PTA_AVX512VBMI2, PTA_VPCLMULQDQ, PTA_AVX512BITALG, PTA_RDPID,
21536 PTA_PCONFIG, PTA_WBNOINVD, PTA_AVX512VP2INTERSECT, PTA_PTWRITE,
21537 PTA_AVX512BF16, PTA_WAITPKG, PTA_MOVDIRI, PTA_MOVDIR64B, PTA_ENQCMD,
21538 PTA_CLDEMOTE, PTA_SERIALIZE, PTA_TSXLDTRK, PTA_AMX_TILE, PTA_AMX_INT8,
21539 PTA_AMX_BF16, PTA_UINTR, PTA_HRESET, PTA_KL, PTA_WIDEKL, PTA_AVXVNNI,
21540 PTA_X86_64_BASELINE, PTA_X86_64_V2, PTA_X86_64_V3, PTA_X86_64_V4,
21541 PTA_CORE2, PTA_NEHALEM, PTA_WESTMERE, PTA_SANDYBRIDGE, PTA_IVYBRIDGE,
21542 PTA_HASWELL, PTA_BROADWELL, PTA_SKYLAKE, PTA_SKYLAKE_AVX512,
21543 PTA_CASCADELAKE, PTA_COOPERLAKE, PTA_CANNONLAKE, PTA_ICELAKE_CLIENT,
21544 PTA_ICELAKE_SERVER, PTA_TIGERLAKE, PTA_SAPPHIRERAPIDS, PTA_ALDERLAKE,
21545 PTA_KNL, PTA_BONNELL, PTA_SILVERMONT, PTA_GOLDMONT, PTA_GOLDMONT_PLUS,
21546 PTA_TREMONT, PTA_KNM): Use constexpr instead of const.
21548 2021-02-18 Jakub Jelinek <jakub@redhat.com>
21550 PR middle-end/99109
21551 * gimple-array-bounds.cc (build_zero_elt_array_type): Rename to ...
21552 (build_printable_array_type): ... this. Add nelts argument. For
21553 overaligned eltype, use TYPE_MAIN_VARIANT (eltype) instead. If
21554 nelts, call build_array_type_nelts.
21555 (array_bounds_checker::check_mem_ref): Use build_printable_array_type
21556 instead of build_zero_elt_array_type and build_array_type_nelts.
21558 2021-02-18 Jakub Jelinek <jakub@redhat.com>
21561 * config/i386/i386.c (distance_non_agu_define): Don't call
21562 extract_insn_cached here.
21563 (ix86_lea_outperforms): Save and restore recog_data around call
21564 to distance_non_agu_define and distance_agu_use.
21565 (ix86_ok_to_clobber_flags): Remove.
21566 (ix86_avoid_lea_for_add): Don't call ix86_ok_to_clobber_flags.
21567 (ix86_avoid_lea_for_addr): Likewise. Adjust function comment.
21568 * config/i386/i386.md (*lea<mode>): Change from define_insn_and_split
21569 into define_insn. Move the splitting to define_peephole2 and
21570 check there using peep2_regno_dead_p if FLAGS_REG is dead.
21572 2021-02-17 Julian Brown <julian@codesourcery.com>
21574 * gimplify.c (gimplify_scan_omp_clauses): Handle ATTACH_DETACH
21577 2021-02-17 Xi Ruoyao <xry111@mengyan1223.wang>
21580 * config/mips/mips.c (mips_symbol_insns): Do not use
21581 MSA_SUPPORTED_MODE_P if mode is MAX_MACHINE_MODE.
21583 2021-02-16 Vladimir N. Makarov <vmakarov@redhat.com>
21585 PR inline-asm/98096
21586 * stmt.c (resolve_operand_name_1): Take inout operands into account
21587 for access to labels by names.
21588 * doc/extend.texi: Describe counting operands for accessing labels.
21590 2021-02-16 Richard Biener <rguenther@suse.de>
21592 PR tree-optimization/38474
21593 * tree-ssa-structalias.c (variable_info::address_taken): New.
21594 (new_var_info): Initialize address_taken.
21595 (process_constraint): Set address_taken.
21596 (solve_constraints): Use the new address_taken flag rather
21597 than is_reg_var for sorting variables.
21598 (dump_constraint): Dump the variable number if the name
21601 2021-02-16 Jakub Jelinek <jakub@redhat.com>
21604 * tree-vect-stmts.c (vectorizable_simd_clone_call): For num_calls != 1
21605 multiply by 4096 and for inbranch by 8192.
21606 * config/i386/i386.c (ix86_simd_clone_usable): For TARGET_AVX512F,
21607 return 3, 2 or 1 for mangle letters 'b', 'c' or 'd'.
21609 2021-02-15 Maya Rashish <coypu@sdf.org>
21611 * config/aarch64/aarch64.c (aarch64_init_builtins):
21612 Call SUBTARGET_INIT_BUILTINS.
21614 2021-02-15 Peter Bergner <bergner@linux.ibm.com>
21616 PR rtl-optimization/98872
21617 * init-regs.c (initialize_uninitialized_regs): Skip initialization
21618 if CONST0_RTX is NULL.
21620 2021-02-15 Richard Sandiford <richard.sandiford@arm.com>
21622 PR rtl-optimization/98863
21623 * rtl-ssa/functions.h (function_info::bb_live_out_info): Delete.
21624 (function_info::build_info): Turn into a declaration, moving the
21625 definition to internals.h.
21626 (function_info::bb_walker): Declare.
21627 (function_info::create_reg_use): Likewise.
21628 (function_info::calculate_potential_phi_regs): Take a build_info
21630 (function_info::place_phis, function_info::create_ebbs): Declare.
21631 (function_info::calculate_ebb_live_in_for_debug): Likewise.
21632 (function_info::populate_backedge_phis): Delete.
21633 (function_info::start_block, function_info::end_block): Declare.
21634 (function_info::populate_phi_inputs): Delete.
21635 (function_info::m_potential_phi_regs): Move information to build_info.
21636 * rtl-ssa/internals.h: New file.
21637 (function_info::bb_phi_info): New class.
21638 (function_info::build_info): Moved from functions.h.
21639 Add a constructor and destructor.
21640 (function_info::build_info::ebb_use): Delete.
21641 (function_info::build_info::ebb_def): Likewise.
21642 (function_info::build_info::bb_live_out): Likewise.
21643 (function_info::build_info::tmp_ebb_live_in_for_debug): New variable.
21644 (function_info::build_info::potential_phi_regs): Likewise.
21645 (function_info::build_info::potential_phi_regs_for_debug): Likewise.
21646 (function_info::build_info::ebb_def_regs): Likewise.
21647 (function_info::build_info::bb_phis): Likewise.
21648 (function_info::build_info::bb_mem_live_out): Likewise.
21649 (function_info::build_info::bb_to_rpo): Likewise.
21650 (function_info::build_info::def_stack): Likewise.
21651 (function_info::build_info::old_def_stack_limit): Likewise.
21652 * rtl-ssa/internals.inl (function_info::build_info::record_reg_def):
21653 Remove the regno argument. Push the previous definition onto the
21654 definition stack where necessary.
21655 * rtl-ssa/accesses.cc: Include internals.h.
21656 * rtl-ssa/changes.cc: Likewise.
21657 * rtl-ssa/blocks.cc: Likewise.
21658 (function_info::build_info::build_info): Define.
21659 (function_info::build_info::~build_info): Likewise.
21660 (function_info::bb_walker): New class.
21661 (function_info::bb_walker::bb_walker): Define.
21662 (function_info::add_live_out_use): Convert a logarithmic-complexity
21663 test into a linear one. Allow the same definition to be passed
21665 (function_info::calculate_potential_phi_regs): Moved from
21666 functions.cc. Take a build_info parameter and store the
21667 information there instead.
21668 (function_info::place_phis): New function.
21669 (function_info::add_entry_block_defs): Update call to record_reg_def.
21670 (function_info::calculate_ebb_live_in_for_debug): New function.
21671 (function_info::add_phi_nodes): Use bb_phis to decide which
21672 registers need phi nodes and initialize ebb_def_regs accordingly.
21673 Do not add degenerate phis here.
21674 (function_info::add_artificial_accesses): Use create_reg_use.
21675 Assert that all definitions are listed in the DF LR sets.
21676 Update call to record_reg_def.
21677 (function_info::record_block_live_out): Record live-out register
21678 values in the phis of successor blocks. Use the live-out set
21679 when processing the last block in an EBB, instead of always
21680 using the live-in sets of successor blocks. AND the live sets
21681 with the set of registers that have been defined in the EBB,
21682 rather than with all potential phi registers. Cope correctly
21683 with branches back to the start of the current EBB.
21684 (function_info::start_block): New function.
21685 (function_info::end_block): Likewise.
21686 (function_info::populate_phi_inputs): Likewise.
21687 (function_info::create_ebbs): Likewise.
21688 (function_info::process_all_blocks): Rewrite into a multi-phase
21690 * rtl-ssa/functions.cc: Include internals.h.
21691 (function_info::calculate_potential_phi_regs): Move to blocks.cc.
21692 (function_info::init_function_data): Remove caller.
21693 * rtl-ssa/insns.cc: Include internals.h
21694 (function_info::create_reg_use): New function. Lazily any
21695 degenerate phis needed by the linear RPO view.
21696 (function_info::record_use): Use create_reg_use. When processing
21697 debug uses, use potential_phi_regs and test it before checking
21698 whether the register is live on entry to the current EBB. Lazily
21699 calculate ebb_live_in_for_debug.
21700 (function_info::record_call_clobbers): Update call to record_reg_def.
21701 (function_info::record_def): Likewise.
21703 2021-02-15 Martin Liska <mliska@suse.cz>
21705 * toplev.c (init_asm_output): Free output of
21706 gen_command_line_string function.
21707 (process_options): Likewise.
21709 2021-02-15 Martin Liska <mliska@suse.cz>
21711 * params.opt: Add 2 missing Param keywords.
21713 2021-02-15 Eric Botcazou <ebotcazou@adacore.com>
21715 * df-core.c (df_worklist_dataflow_doublequeue): Use proper cast.
21717 2021-02-15 Jakub Jelinek <jakub@redhat.com>
21719 PR tree-optimization/99079
21720 * match.pd (A % (pow2pcst << N) -> A & ((pow2pcst << N) - 1)): Remove
21721 useless tree_nop_conversion_p (type, TREE_TYPE (@3)) check. Instead
21722 require both type and TREE_TYPE (@1) to be integral types and either
21723 type having smaller or equal precision, or TREE_TYPE (@1) being
21724 unsigned type, or type being signed type. If TREE_TYPE (@1)
21725 doesn't have wrapping overflow, perform the subtraction of one in
21728 2021-02-14 Jan Hubicka <hubicka@ucw.cz>
21729 Richard Biener <rguether@suse.de>
21732 * ipa-reference.c (ipa_init): Only conditinally initialize
21733 reference_vars_to_consider.
21734 (propagate): Conditionally deninitialize reference_vars_to_consider.
21735 (ipa_reference_write_optimization_summary): Sanity check that
21736 reference_vars_to_consider is not allocated.
21738 2021-02-13 Levy Hsu <admin@levyhsu.com>
21741 * config/riscv/riscv-shorten-memrefs.c (pass_shorten_memrefs): Add
21742 extend parameter to get_si_mem_base_reg declaration.
21743 (get_si_mem_base_reg): Add extend parameter. Set it.
21744 (analyze): Pass extend arg to get_si_mem_base_reg.
21745 (transform): Likewise. Use it when rewriting mems.
21746 * config/riscv/riscv.c (riscv_legitimize_move): Check for subword
21747 loads and emit sign/zero extending load followed by subreg move.
21749 2021-02-13 Jim Wilson <jimw@sifive.com>
21752 * config/riscv/riscv.c (riscv_compressed_lw_address_p): Drop early
21753 exit when !reload_completed. Only perform check for compressed reg
21754 if reload_completed.
21755 (riscv_rtx_costs): In MEM case, when optimizing for size and
21756 shorten memrefs, if not compressible, then increase cost.
21758 2021-02-13 Jakub Jelinek <jakub@redhat.com>
21760 PR rtl-optimization/98439
21761 * recog.c (pass_split_before_regstack::gate): Enable even when
21762 pass_split_before_sched2 is enabled if -fselective-scheduling2 is
21765 2021-02-13 Jakub Jelinek <jakub@redhat.com>
21768 * config/i386/mmx.md (*mmx_pshufd_1): Add a combine splitter for
21769 swap of V2SImode elements in memory into DImode memory rotate by 32.
21771 2021-02-12 Martin Sebor <msebor@redhat.com>
21773 * tree-pretty-print.c (print_generic_expr_to_str): Update comment.
21775 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
21777 * rtl-ssa/accesses.cc (function_info::make_use_available): Use
21778 m_temp_obstack rather than m_obstack to allocate the temporary use.
21780 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
21782 * df-problems.c (df_lr_bb_local_compute): Treat partial definitions
21783 as read-modify operations.
21785 2021-02-12 Richard Biener <rguenther@suse.de>
21787 PR middle-end/38474
21788 * ipa-fnsummary.c (unmodified_parm_1): Only walk when
21789 fbi->aa_walk_budget is bigger than zero. Update
21790 fbi->aa_walk_budget.
21791 (param_change_prob): Likewise.
21792 * ipa-prop.c (detect_type_change_from_memory_writes):
21793 Properly account walk_aliased_vdefs.
21794 (parm_preserved_before_stmt_p): Canonicalize updates.
21795 (parm_ref_data_preserved_p): Likewise.
21796 (parm_ref_data_pass_through_p): Likewise.
21797 (determine_known_aggregate_parts): Account own alias queries.
21799 2021-02-12 Martin Liska <mliska@suse.cz>
21801 * opts-common.c (decode_cmdline_option): Release werror_arg.
21802 * opts.c (gen_producer_string): Release output of
21803 gen_command_line_string.
21805 2021-02-12 Richard Biener <rguenther@suse.de>
21807 PR tree-optimization/38474
21808 * params.opt (-param=max-store-chains-to-track=): New param.
21809 (-param=max-stores-to-track=): Likewise.
21810 * doc/invoke.texi (max-store-chains-to-track): Document.
21811 (max-stores-to-track): Likewise.
21812 * gimple-ssa-store-merging.c (pass_store_merging::m_n_chains):
21814 (pass_store_merging::m_n_stores): Likewise.
21815 (pass_store_merging::terminate_and_process_chain): Update
21816 m_n_stores and m_n_chains.
21817 (pass_store_merging::process_store): Likewise. Terminate
21818 oldest chains if the number of stores or chains get too large.
21819 (imm_store_chain_info::terminate_and_process_chain): Dump
21822 2021-02-11 Eric Botcazou <ebotcazou@adacore.com>
21824 * config/i386/winnt.c (i386_pe_seh_unwind_emit): When switching to
21825 the cold section, emit a nop before the directive if the previous
21826 active instruction can throw.
21828 2021-02-11 Peter Bergner <bergner@linux.ibm.com>
21831 * config/rs6000/predicates.md (mma_assemble_input_operand): Restrict
21832 memory addresses that are legal for quad word accesses.
21834 2021-02-11 Andrea Corallo <andrea.corallo@arm.com>
21837 * config/arm/thumb2.md (*doloop_end_internal): Generate
21838 alternative sequence to handle long range branches.
21840 2021-02-11 Joel Hutton <joel.hutton@arm.com>
21842 PR tree-optimization/98772
21843 * optabs-tree.c (supportable_half_widening_operation): New function
21844 to check for supportable V8QI->V8HI widening patterns.
21845 * optabs-tree.h (supportable_half_widening_operation): New function.
21846 * tree-vect-stmts.c (vect_create_half_widening_stmts): New function
21847 to create promotion stmts for V8QI->V8HI widening patterns.
21848 (vectorizable_conversion): Add case for V8QI->V8HI.
21850 2021-02-11 Richard Biener <rguenther@suse.de>
21852 * sparseset.h (SPARSESET_ELT_BITS): Remove.
21853 (SPARSESET_ELT_TYPE): Use unsigned int.
21854 * fwprop.c: Do not include sparseset.h.
21856 2021-02-10 Jakub Jelinek <jakub@redhat.com>
21859 * varasm.c (declare_weak): For -fsyntax-only, allow even
21860 TREE_ASM_WRITTEN function decls.
21862 2021-02-10 Jakub Jelinek <jakub@redhat.com>
21865 * config/i386/sse.md (fix<fixunssuffix>_truncv2sfv2di2,
21866 <insn>v8qiv8hi2, <insn>v8qiv8si2, <insn>v4qiv4si2, <insn>v4hiv4si2,
21867 <insn>v8qiv8di2, <insn>v4qiv4di2, <insn>v2qiv2di2, <insn>v4hiv4di2,
21868 <insn>v2hiv2di2, <insn>v2siv2di2): Force operands[1] into REG before
21869 calling simplify_gen_subreg on it.
21871 2021-02-10 Martin Liska <mliska@suse.cz>
21873 * config/nvptx/nvptx.c (nvptx_option_override): Use
21874 flag_patchable_function_entry instead of the removed
21875 function_entry_patch_area_size.
21877 2021-02-10 Martin Liska <mliska@suse.cz>
21879 PR tree-optimization/99002
21880 PR tree-optimization/99026
21881 * gimple-if-to-switch.cc (if_chain::is_beneficial): Fix memory
21882 leak when adjacent cases are merged.
21883 * tree-switch-conversion.c (switch_decision_tree::analyze_switch_statement): Use
21885 (make_pass_lower_switch): Remove trailing whitespace.
21886 * tree-switch-conversion.h (release_clusters): New.
21888 2021-02-10 Richard Biener <rguenther@suse.de>
21890 PR rtl-optimization/99054
21891 * cfgrtl.c (rtl-optimization/99054): Return an auto_vec.
21892 (fixup_partitions): Adjust.
21893 (rtl_verify_edges): Likewise.
21895 2021-02-10 Jakub Jelinek <jakub@redhat.com>
21897 PR middle-end/99007
21898 * gimplify.c (gimplify_scan_omp_clauses): For MEM_REF on reductions,
21899 temporarily disable gimplify_ctxp->into_ssa around gimplify_expr
21902 2021-02-10 Richard Biener <rguenther@suse.de>
21905 * ipa-pure-const.c (propagate_malloc): Use an auto_vec<>
21908 2021-02-10 Richard Biener <rguenther@suse.de>
21910 PR tree-optimization/99024
21911 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Only
21912 clear loop->aux if it is associated with the destroyed loop_vinfo.
21914 2021-02-10 Martin Liska <mliska@suse.cz>
21916 PR tree-optimization/99002
21917 * gimple-if-to-switch.cc (find_conditions): Fix memory leak
21920 2021-02-10 Martin Liska <mliska@suse.cz>
21923 * ipa-icf.c (sem_item::add_reference): Fix memory leak when
21924 a reference exists.
21926 2021-02-10 Jakub Jelinek <jakub@redhat.com>
21929 * dwarf2out.c (prune_unused_types_walk): Mark DW_TAG_variable DIEs
21930 at class scope for DWARF5+.
21932 2021-02-09 Eric Botcazou <ebotcazou@adacore.com>
21934 PR rtl-optimization/96015
21935 * reorg.c (skip_consecutive_labels): Minor comment tweaks.
21936 (relax_delay_slots): When deleting a jump to the next active
21937 instruction over a barrier, first delete the barrier if the
21938 jump is the only way to reach the target label.
21940 2021-02-09 Andre Vieira <andre.simoesdiasvieira@arm.com>
21942 * config/aarch64/aarch64-cost-tables.h: Add entries for vect.mul.
21943 * config/aarch64/aarch64.c (aarch64_rtx_mult_cost): Use vect.mul for
21944 vector multiplies and vect.alu for SSRA.
21945 * config/arm/aarch-common-protos.h (struct vector_cost_table): Define
21946 vect.mul cost field.
21947 * config/arm/aarch-cost-tables.h: Add entries for vect.mul.
21948 * config/arm/arm.c: Likewise.
21950 2021-02-09 Richard Biener <rguenther@suse.de>
21952 PR tree-optimization/98863
21953 * tree-ssa-sccvn.h (vn_avail::next_undo): Add.
21954 * tree-ssa-sccvn.c (last_pushed_avail): New global.
21955 (rpo_elim::eliminate_push_avail): Chain pushed avails.
21956 (unwind_state::avail_top): Add.
21957 (do_unwind): Rewrite unwinding of avail entries.
21958 (do_rpo_vn): Initialize last_pushed_avail and
21959 avail_top of the undo state.
21961 2021-02-09 Jakub Jelinek <jakub@redhat.com>
21963 PR middle-end/99004
21964 * calls.c (maybe_warn_rdwr_sizes): Change s0 and s1 type from
21965 const char * to char * and free those pointers after use.
21967 2021-02-09 Richard Biener <rguenther@suse.de>
21969 PR tree-optimization/99017
21970 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Allow
21971 zero vector cost entries.
21973 2021-02-08 Andre Vieira <andre.simoesdiasvieira@arm.com>
21975 PR middle-end/98974
21976 * tree-vect-stmts.c (vectorizable_condition): Remove shadow vec_num
21977 parameter in vectorizable_condition.
21979 2021-02-08 Richard Biener <rguenther@suse.de>
21982 * tree.c (walk_tree_1): Walk VECTOR_CST elements.
21984 2021-02-08 Martin Liska <mliska@suse.cz>
21987 * cfgexpand.c (pass_expand::execute): Parse per-function option
21988 flag_patchable_function_entry and use it.
21989 * common.opt: Remove function_entry_patch_area_size and
21990 function_entry_patch_area_start global variables.
21991 * opts.c (parse_and_check_patch_area): New function.
21992 (common_handle_option): Use it.
21993 * opts.h (parse_and_check_patch_area): New function.
21994 * toplev.c (process_options): Parse and use
21995 function_entry_patch_area_size.
21997 2021-02-08 Martin Sebor <msebor@redhat.com>
21999 * doc/extend.texi (attribute malloc): Correct typos.
22001 2021-02-05 Nathan Sidwell <nathan@acm.org>
22004 * gcc.c (driver::maybe_run_linker): Check for input file
22005 accessibility if not linking.
22007 2021-02-05 Richard Biener <rguenther@suse.de>
22009 PR tree-optimization/98855
22010 * tree-vectorizer.h (add_stmt_cost): New overload.
22011 * tree-vect-slp.c (li_cost_vec_cmp): New.
22012 (vect_bb_slp_scalar_cost): Cost individual loop regions
22013 separately. Account for the scalar instance root stmt.
22015 2021-02-05 Tom de Vries <tdevries@suse.de>
22018 * tree-switch-conversion.c (jump_table_cluster::emit): Add loc
22020 (bit_test_cluster::emit): Reuse location_t for newly created
22022 (switch_decision_tree::try_switch_expansion): Preserve
22024 * tree-switch-conversion.h: Change function signatures.
22026 2021-02-05 Jakub Jelinek <jakub@redhat.com>
22029 * config/i386/i386-options.c (m_NONE, m_ALL): Define.
22030 * config/i386/x86-tune.def (X86_TUNE_BRANCH_PREDICTION_HINTS,
22031 X86_TUNE_PROMOTE_QI_REGS): Use m_NONE instead of 0U.
22032 (X86_TUNE_QIMODE_MATH): Use m_ALL instead of ~0U.
22034 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22036 * config/aarch64/aarch64-simd-builtins.def (get_high): Define builtin.
22037 * config/aarch64/aarch64-simd.md (aarch64_get_high<mode>): Define.
22038 * config/aarch64/arm_neon.h (__GET_HIGH): Delete.
22039 (vget_high_f16): Reimplement using new builtin.
22040 (vget_high_f32): Likewise.
22041 (vget_high_f64): Likewise.
22042 (vget_high_p8): Likewise.
22043 (vget_high_p16): Likewise.
22044 (vget_high_p64): Likewise.
22045 (vget_high_s8): Likewise.
22046 (vget_high_s16): Likewise.
22047 (vget_high_s32): Likewise.
22048 (vget_high_s64): Likewise.
22049 (vget_high_u8): Likewise.
22050 (vget_high_u16): Likewise.
22051 (vget_high_u32): Likewise.
22052 (vget_high_u64): Likewise.
22054 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22056 * config/aarch64/aarch64-simd-builtins.def (get_low): Define builtin.
22057 * config/aarch64/aarch64-simd.md (aarch64_get_low<mode>): Define.
22058 * config/aarch64/arm_neon.h (__GET_LOW): Delete.
22059 (vget_low_f16): Reimplement using new builtin.
22060 (vget_low_f32): Likewise.
22061 (vget_low_f64): Likewise.
22062 (vget_low_p8): Likewise.
22063 (vget_low_p16): Likewise.
22064 (vget_low_p64): Likewise.
22065 (vget_low_s8): Likewise.
22066 (vget_low_s16): Likewise.
22067 (vget_low_s32): Likewise.
22068 (vget_low_s64): Likewise.
22069 (vget_low_u8): Likewise.
22070 (vget_low_u16): Likewise.
22071 (vget_low_u32): Likewise.
22072 (vget_low_u64): Likewise.
22074 2021-02-05 Kito Cheng <kito.cheng@sifive.com>
22076 * gcc.c (print_multilib_info): Check all required argument is provided
22079 2021-02-05 liuhongt <hongtao.liu@intel.com>
22082 * config/i386/i386-expand.c (ix86_expand_sse_cmp): Don't
22083 generate integer mask comparison for 128/256-bits vector when
22084 op_true/op_false is NULL_RTX or CONSTM1_RTX/CONST0_RTX. Also
22085 delete redundant !maskcmp condition.
22086 (ix86_expand_int_vec_cmp): Ditto but no redundant deletion
22088 (ix86_expand_sse_movcc): Delete definition of maskcmp, add the
22089 condition directly to if (maskcmp), add extra check for
22090 cmpmode, it should be MODE_INT.
22091 (ix86_expand_fp_vec_cmp): Pass NULL to ix86_expand_sse_cmp's
22092 parameters op_true/op_false.
22093 (ix86_use_mask_cmp_p): New.
22095 2021-02-05 liuhongt <hongtao.liu@intel.com>
22098 * config/i386/x86-tune.def (X86_TUNE_AVX256_UNALIGNED_LOAD_OPTIMAL):
22099 Remove m_GENERIC from ~list.
22100 (X86_TUNE_AVX256_UNALIGNED_STORE_OPTIMAL): Ditto.
22102 2021-02-04 David Malcolm <dmalcolm@redhat.com>
22105 * diagnostic-show-locus.c (compatible_locations_p): Require
22106 locations in the same macro map to be either both from the
22107 macro definition, or both from the macro arguments.
22109 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
22111 * config/aarch64/aarch64-simd-builtins.def: Add
22112 [su]mull_hi_lane[q] builtin generator macros.
22113 * config/aarch64/aarch64-simd.md
22114 (aarch64_<su>mull_hi_lane<mode>_insn): Define.
22115 (aarch64_<su>mull_hi_lane<mode>): Define.
22116 (aarch64_<su>mull_hi_laneq<mode>_insn): Define.
22117 (aarch64_<su>mull_hi_laneq<mode>): Define.
22118 * config/aarch64/arm_neon.h (vmull_high_lane_s16): Use RTL
22119 builtin instead of inline asm.
22120 (vmull_high_lane_s32): Likewise.
22121 (vmull_high_lane_u16): Likewise.
22122 (vmull_high_lane_u32): Likewise.
22123 (vmull_high_laneq_s16): Likewise.
22124 (vmull_high_laneq_s32): Likewise.
22125 (vmull_high_laneq_u16): Likewise.
22126 (vmull_high_laneq_u32): Liekwise.
22128 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
22130 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_hi_n
22131 builtin generator macros.
22132 * config/aarch64/aarch64-simd.md
22133 (aarch64_<su>mull_hi_n<mode>_insn): Define.
22134 (aarch64_<su>mull_hi_n<mode>): Define.
22135 * config/aarch64/arm_neon.h (vmull_high_n_s16): Use RTL builtin
22136 instead of inline asm.
22137 (vmull_high_n_s32): Likewise.
22138 (vmull_high_n_u16): Likewise.
22139 (vmull_high_n_u32): Likewise.
22141 2021-02-04 Richard Biener <rguenther@suse.de>
22143 PR tree-optimization/98855
22144 * tree-vect-loop.c (vectorizable_phi): Do not cost
22145 single-argument PHIs.
22146 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Likewise.
22147 * tree-vect-stmts.c (vectorizable_bswap): Also perform
22148 costing for SLP operation.
22150 2021-02-04 Martin Liska <mliska@suse.cz>
22152 * doc/extend.texi: Mention -mprefer-vector-width in target
22155 2021-02-03 Martin Sebor <msebor@redhat.com>
22157 PR tree-optimization/98937
22158 * tree-ssa-strlen.c (strlen_dom_walker::~strlen_dom_walker): Define.
22159 Flush pointer_query cache.
22161 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
22163 * config/rs6000/genfusion.pl (gen_2logical): Add missing
22164 fixes based on patch review.
22165 * config/rs6000/fusion.md: Regenerate file.
22167 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
22169 * config/rs6000/t-rs6000: Comment out auto generation of
22172 2021-02-03 Andrew Stubbs <ams@codesourcery.com>
22174 * config/gcn/gcn-opts.h (enum processor_type): Add PROCESSOR_GFX908.
22175 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Add gfx908.
22176 (output_file_start): Add gfx908.
22177 * config/gcn/gcn.opt (gpu_type): Add gfx908.
22178 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Add march=gfx908.
22179 (MULTILIB_DIRNAMES): Add gfx908.
22180 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): New define.
22181 (main): Recognize gfx908.
22182 * config/gcn/t-omp-device: Add gfx908.
22184 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
22186 * config/aarch64/aarch64-simd-builtins.def: Add
22187 [su]mlsl_hi_lane[q] builtin macro generators.
22188 * config/aarch64/aarch64-simd.md
22189 (aarch64_<su>mlsl_hi_lane<mode>_insn): Define.
22190 (aarch64_<su>mlsl_hi_lane<mode>): Define.
22191 (aarch64_<su>mlsl_hi_laneq<mode>_insn): Define.
22192 (aarch64_<su>mlsl_hi_laneq<mode>): Define.
22193 * config/aarch64/arm_neon.h (vmlsl_high_lane_s16): Use RTL
22194 builtin instead of inline asm.
22195 (vmlsl_high_lane_s32): Likewise.
22196 (vmlsl_high_lane_u16): Likewise.
22197 (vmlsl_high_lane_u32): Likewise.
22198 (vmlsl_high_laneq_s16): Likewise.
22199 (vmlsl_high_laneq_s32): Likewise.
22200 (vmlsl_high_laneq_u16): Likewise.
22201 (vmlsl_high_laneq_u32): Likewise.
22202 (vmlal_high_laneq_u32): Likewise.
22204 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
22206 * config/aarch64/aarch64-simd-builtins.def: Add
22207 [su]mlal_hi_lane[q] builtin generator macros.
22208 * config/aarch64/aarch64-simd.md
22209 (aarch64_<su>mlal_hi_lane<mode>_insn): Define.
22210 (aarch64_<su>mlal_hi_lane<mode>): Define.
22211 (aarch64_<su>mlal_hi_laneq<mode>_insn): Define.
22212 (aarch64_<su>mlal_hi_laneq<mode>): Define.
22213 * config/aarch64/arm_neon.h (vmlal_high_lane_s16): Use RTL
22214 builtin instead of inline asm.
22215 (vmlal_high_lane_s32): Likewise.
22216 (vmlal_high_lane_u16): Likewise.
22217 (vmlal_high_lane_u32): Likewise.
22218 (vmlal_high_laneq_s16): Likewise.
22219 (vmlal_high_laneq_s32): Likewise.
22220 (vmlal_high_laneq_u16): Likewise.
22221 (vmlal_high_laneq_u32): Likewise.
22223 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
22225 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_hi_n
22226 builtin generator macros.
22227 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_hi_n<mode>_insn):
22229 (aarch64_<su>mlsl_hi_n<mode>): Define.
22230 * config/aarch64/arm_neon.h (vmlsl_high_n_s16): Use RTL builtin
22231 instead of inline asm.
22232 (vmlsl_high_n_s32): Likewise.
22233 (vmlsl_high_n_u16): Likewise.
22234 (vmlsl_high_n_u32): Likewise.
22236 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
22238 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_hi_n
22239 builtin generator macros.
22240 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_hi_n<mode>_insn):
22242 (aarch64_<su>mlal_hi_n<mode>): Define.
22243 * config/aarch64/arm_neon.h (vmlal_high_n_s16): Use RTL builtin
22244 instead of inline asm.
22245 (vmlal_high_n_s32): Likewise.
22246 (vmlal_high_n_u16): Likewise.
22247 (vmlal_high_n_u32): Likewise.
22249 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
22251 * config/aarch64/aarch64-simd-builtins.def: Add RTL builtin
22253 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal_hi<mode>):
22255 (aarch64_<su>mlal_hi<mode>_insn): This.
22256 (aarch64_<su>mlal_hi<mode>): Define.
22257 * config/aarch64/arm_neon.h (vmlal_high_s8): Use RTL builtin
22258 instead of inline asm.
22259 (vmlal_high_s16): Likewise.
22260 (vmlal_high_s32): Likewise.
22261 (vmlal_high_u8): Likewise.
22262 (vmlal_high_u16): Likewise.
22263 (vmlal_high_u32): Likewise.
22265 2021-02-03 Ilya Leoshkevich <iii@linux.ibm.com>
22267 * lra-spills.c (remove_pseudos): Call lra_update_insn_recog_data()
22268 after calling alter_subreg() on a (mem).
22270 2021-02-03 Martin Liska <mliska@suse.cz>
22273 * lto-streamer-out.c (produce_lto_section): Fill up missing
22275 * lto-streamer.h (struct lto_section): Add _padding field.
22277 2021-02-03 Richard Biener <rguenther@suse.de>
22279 * lto-streamer.c (lto_get_section_name): Free temporary
22281 * tree-loop-distribution.c
22282 (loop_distribution::merge_dep_scc_partitions): Free edge data.
22284 2021-02-03 Jakub Jelinek <jakub@redhat.com>
22286 PR middle-end/97487
22287 * ifcvt.c (noce_can_force_operand): New function.
22288 (noce_emit_move_insn): Use it.
22289 (noce_try_sign_mask): Likewise. Formatting fix.
22291 2021-02-03 Jakub Jelinek <jakub@redhat.com>
22293 PR middle-end/97971
22294 * lra-constraints.c (process_alt_operands): For inline asm, don't call
22295 fatal_insn, but instead return false.
22297 2021-02-03 Jakub Jelinek <jakub@redhat.com>
22299 PR tree-optimization/98287
22300 * config/i386/mmx.md (<insn><mode>3): For shifts don't enable expander
22303 2021-02-03 Tamar Christina <tamar.christina@arm.com>
22305 PR tree-optimization/98928
22306 * tree-vect-loop.c (vect_analyze_loop_2): Change
22307 STMT_VINFO_SLP_VECT_ONLY to STMT_VINFO_SLP_VECT_ONLY_PATTERN.
22308 * tree-vect-slp-patterns.c (complex_pattern::build): Likewise.
22309 * tree-vectorizer.h (STMT_VINFO_SLP_VECT_ONLY_PATTERN): New.
22310 (class _stmt_vec_info): Add slp_vect_pattern_only_p.
22312 2021-02-02 Richard Biener <rguenther@suse.de>
22314 * gimple-loop-interchange.cc (prepare_data_references):
22316 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
22317 * tree-ssa-loop-im.c (hoist_memory_references): Likewise.
22318 * tree-vect-stmts.c (vectorizable_condition): Do not
22320 (vectorizable_comparison): Likewise.
22322 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22324 * config/aarch64/aarch64-simd-builtins.def (ursqrte): Define builtin.
22325 * config/aarch64/aarch64-simd.md (aarch64_ursqrte<mode>): New pattern.
22326 * config/aarch64/arm_neon.h (vrsqrte_u32): Reimplement using builtin.
22327 (vrsqrteq_u32): Likewise.
22329 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22331 * config/aarch64/aarch64-simd-builtins.def (sqxtun2): Define builtin.
22332 * config/aarch64/aarch64-simd.md (aarch64_sqxtun2<mode>_le): Define.
22333 (aarch64_sqxtun2<mode>_be): Likewise.
22334 (aarch64_sqxtun2<mode>): Likewise.
22335 * config/aarch64/arm_neon.h (vqmovun_high_s16): Reimplement using builtin.
22336 (vqmovun_high_s32): Likewise.
22337 (vqmovun_high_s64): Likewise.
22338 * config/aarch64/iterators.md (UNSPEC_SQXTUN2): Define.
22340 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22342 * config/aarch64/aarch64-simd-builtins.def (bfdot_lane, bfdot_laneq): Use
22344 (bfmlalb_lane, bfmlalt_lane, bfmlalb_lane_q, bfmlalt_lane_q): Use FP flags.
22346 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22348 * config/aarch64/aarch64-simd-builtins.def (fcmla_lane0, fcmla_lane90,
22349 fcmla_lane180, fcmla_lane270, fcmlaq_lane0, fcmlaq_lane90, fcmlaq_lane180,
22350 fcmlaq_lane270, scvtf, ucvtf, fcvtzs, fcvtzu, scvtfsi, scvtfdi, ucvtfsi,
22351 ucvtfdi, fcvtzshf, fcvtzuhf, fmlal_lane_low, fmlsl_lane_low,
22352 fmlal_laneq_low, fmlsl_laneq_low, fmlalq_lane_low, fmlslq_lane_low,
22353 fmlalq_laneq_low, fmlslq_laneq_low, fmlal_lane_high, fmlsl_lane_high,
22354 fmlal_laneq_high, fmlsl_laneq_high, fmlalq_lane_high, fmlslq_lane_high,
22355 fmlalq_laneq_high, fmlslq_laneq_high): Use FP flags.
22357 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22359 * config/aarch64/aarch64-builtins.c (FLAG_LOAD): Define.
22360 * config/aarch64/aarch64-simd-builtins.def (ld1x2, ld2, ld3, ld4, ld2r,
22361 ld3r, ld4r, ld1, ld1x3, ld1x4): Use LOAD flags.
22363 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22365 * config/aarch64/aarch64-simd-builtins.def (combine, zip1, zip2,
22366 uzp1, uzp2, trn1, trn2, simd_bsl): Use AUTO_FP flags.
22368 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22370 * config/aarch64/aarch64-simd-builtins.def (clrsb, clz, ctz, popcount,
22371 vec_smult_lane_, vec_smlal_lane_, vec_smult_laneq_, vec_smlal_laneq_,
22372 vec_umult_lane_, vec_umlal_lane_, vec_umult_laneq_, vec_umlal_laneq_,
22373 ashl, sshl, ushl, srshl, urshl, sdot_lane, udot_lane, sdot_laneq,
22374 udot_laneq, usdot_lane, usdot_laneq, sudot_lane, sudot_laneq, ashr,
22375 ashr_simd, lshr, lshr_simd, srshr_n, urshr_n, ssra_n, usra_n, srsra_n,
22376 ursra_n, sshll_n, ushll_n, sshll2_n, ushll2_n, ssri_n, usri_n, ssli_n,
22377 ssli_n, usli_n, bswap, rbit, simd_bsl, eor3q, rax1q, xarq, bcaxq): Use
22378 NONE builtin flags.
22380 2021-02-02 Jakub Jelinek <jakub@redhat.com>
22382 PR tree-optimization/98848
22383 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Punt if
22384 STMT_VINFO_DEF_TYPE (last_stmt_info) is vect_reduction_def.
22386 2021-02-02 Kito Cheng <kito.cheng@sifive.com>
22389 * expr.c: Check mode before calling store_expr.
22391 2021-02-02 Christophe Lyon <christophe.lyon@linaro.org>
22393 * config/arm/iterators.md (supf): Remove VORNQ_S and VORNQ_U.
22395 * config/arm/mve.md (mve_vornq_s<mode>): New entry for vorn
22396 instruction using expression ior.
22397 (mve_vornq_u<mode>): New expander.
22398 (mve_vornq_f<mode>): Use ior code instead of unspec.
22399 * config/arm/unspecs.md (VORNQ_S, VORNQ_U, VORNQ_F): Remove.
22401 2021-02-02 Alexandre Oliva <oliva@adacore.com>
22403 * tree-nested.c (convert_nonlocal_reference_op): Move
22404 current_function_decl restore after re-gimplification.
22405 (convert_local_reference_op): Likewise.
22407 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22409 * config/aarch64/aarch64-simd-builtins.def (rshrn, rshrn2):
22411 * config/aarch64/aarch64-simd.md (aarch64_rshrn<mode>_insn_le):
22413 (aarch64_rshrn<mode>_insn_be): Likewise.
22414 (aarch64_rshrn<mode>): Likewise.
22415 (aarch64_rshrn2<mode>_insn_le): Likewise.
22416 (aarch64_rshrn2<mode>_insn_be): Likewise.
22417 (aarch64_rshrn2<mode>): Likewise.
22418 * config/aarch64/aarch64.md (unspec): Add UNSPEC_RSHRN.
22419 * config/aarch64/arm_neon.h (vrshrn_high_n_s16): Reimplement
22421 (vrshrn_high_n_s32): Likewise.
22422 (vrshrn_high_n_s64): Likewise.
22423 (vrshrn_high_n_u16): Likewise.
22424 (vrshrn_high_n_u32): Likewise.
22425 (vrshrn_high_n_u64): Likewise.
22426 (vrshrn_n_s16): Likewise.
22427 (vrshrn_n_s32): Likewise.
22428 (vrshrn_n_s64): Likewise.
22429 (vrshrn_n_u16): Likewise.
22430 (vrshrn_n_u32): Likewise.
22431 (vrshrn_n_u64): Likewise.
22433 2021-02-01 Sergei Trofimovich <siarheit@google.com>
22435 PR tree-optimization/98499
22436 * ipa-modref.c (analyze_ssa_name_flags): treat RVO
22437 conservatively and assume all possible side-effects.
22439 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22441 * config/aarch64/aarch64-simd-builtins.def (vec_unpacks_hi,
22442 vec_unpacku_hi_): Define builtins.
22443 * config/aarch64/arm_neon.h (vmovl_high_s8): Reimplement using
22445 (vmovl_high_s16): Likewise.
22446 (vmovl_high_s32): Likewise.
22447 (vmovl_high_u8): Likewise.
22448 (vmovl_high_u16): Likewise.
22449 (vmovl_high_u32): Likewise.
22451 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22453 * config/aarch64/aarch64-simd-builtins.def (sabdl, uabdl):
22455 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl<mode>): New
22457 * config/aarch64/aarch64.md (unspec): Define UNSPEC_SABDL,
22459 * config/aarch64/arm_neon.h (vabdl_s8): Reimplemet using
22461 (vabdl_s16): Likewise.
22462 (vabdl_s32): Likewise.
22463 (vabdl_u8): Likewise.
22464 (vabdl_u16): Likewise.
22465 (vabdl_u32): Likewise.
22466 * config/aarch64/iterators.md (ABDL): New int iterator.
22467 (sur): Handle UNSPEC_SABDL, UNSPEC_UABDL.
22469 2021-02-01 Martin Sebor <msebor@redhat.com>
22471 * tree.h (BLOCK_VARS): Add comment.
22472 (BLOCK_SUBBLOCKS): Same.
22473 (BLOCK_SUPERCONTEXT): Same.
22474 (BLOCK_ABSTRACT_ORIGIN): Same.
22475 (inlined_function_outer_scope_p): Same.
22477 2021-02-01 Martin Sebor <msebor@redhat.com>
22479 PR middle-end/97172
22480 * attribs.c (attr_access::free_lang_data): Define new function.
22481 * attribs.h (attr_access::free_lang_data): Declare new function.
22483 2021-02-01 Richard Biener <rguenther@suse.de>
22485 * vec.h (auto_vec::auto_vec): Add memory stat parameters
22487 * bitmap.h (auto_bitmap::auto_bitmap): Likewise.
22489 2021-02-01 Tamar Christina <tamar.christina@arm.com>
22491 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>,
22492 aarch64_<su>mlsl<mode>, aarch64_<su>mlsl_n<mode>): Flip mult operands.
22494 2021-02-01 Richard Biener <rguenther@suse.de>
22496 PR rtl-optimization/98863
22497 * config/i386/i386-features.c (convert_scalars_to_vector):
22498 Set DF_RD_PRUNE_DEAD_DEFS.
22500 2021-01-31 Eric Botcazou <ebotcazou@adacore.com>
22502 * system.h (SIZE_MAX): Define if not already defined.
22504 2021-01-30 Aaron Sawdey <acsawdey@linux.ibm.com>
22506 * config/rs6000/genfusion.pl (gen_2logical): New function to
22507 generate patterns for logical-logical fusion.
22508 * config/rs6000/fusion.md: Regenerated patterns.
22509 * config/rs6000/rs6000-cpus.def: Add
22510 OPTION_MASK_P10_FUSION_2LOGICAL.
22511 * config/rs6000/rs6000.c (rs6000_option_override_internal):
22512 Enable logical-logical fusion for p10.
22513 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2logical.
22515 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
22517 * config/rs6000/rs6000.opt: Add periods to new AIX options.
22519 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
22521 * config/rs6000/rs6000.opt (mabi=vec-extabi): New.
22522 (mabi=vec-default): New.
22523 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
22524 __EXTABI__ for AIX Vector extended ABI.
22525 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Print AIX Vector
22527 (conditional_register_usage): If AIX vec_extabi enabled, vs20-vs31
22529 * doc/invoke.texi (PowerPC mabi): Add AIX vec-extabi and vec-default.
22531 2021-01-30 Jakub Jelinek <jakub@redhat.com>
22533 * config/i386/i386-features.c (remove_partial_avx_dependency): Clear
22534 DF_DEFER_INSN_RESCAN after calling df_process_deferred_rescans.
22536 2021-01-29 Vladimir N. Makarov <vmakarov@redhat.com>
22539 * lra-constraints.c (in_class_p): Don't narrow class only for REG
22542 2021-01-29 Will Schmidt <will_schmidt@vnet.ibm.com>
22544 * config/rs6000/rs6000-call.c (rs6000_expand_binup_builtin): Add
22545 clauses for CODE_FOR_vsx_xvcvuxddp_scale and
22546 CODE_FOR_vsx_xvcvsxddp_scale to the parameter checking code.
22548 2021-01-29 Andrew MacLeod <amacleod@redhat.com>
22550 PR tree-optimization/98866
22551 * gimple-range-gori.h (gori_compute:set_range_invariant): New.
22552 * gimple-range-gori.cc (gori_map::set_range_invariant): New.
22553 (gori_map::m_maybe_invariant): Rename from all_outgoing.
22554 (gori_map::gori_map): Rename all_outgoing to m_maybe_invariant.
22555 (gori_map::is_export_p): Ditto.
22556 (gori_map::calculate_gori): Ditto.
22557 (gori_compute::set_range_invariant): New.
22558 * gimple-range.cc (gimple_ranger::range_of_stmt): Set range
22559 invariant for pointers evaluating to [1, +INF].
22561 2021-01-29 Richard Biener <rguenther@suse.de>
22563 PR rtl-optimization/98863
22564 * config/i386/i386-features.c (remove_partial_avx_dependency):
22565 Do not perform DF analysis.
22566 (pass_data_remove_partial_avx_dependency): Remove
22569 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
22571 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_n
22572 builtin generator macros.
22573 * config/aarch64/aarch64-simd.md (aarch64_<su>mull_n<mode>):
22575 * config/aarch64/arm_neon.h (vmull_n_s16): Use RTL builtin
22576 instead of inline asm.
22577 (vmull_n_s32): Likewise.
22578 (vmull_n_u16): Likewise.
22579 (vmull_n_u32): Likewise.
22581 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22583 * config/aarch64/aarch64-simd-builtins.def (sabdl2, uabdl2):
22585 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
22587 (aarch64_<sur>abdl2<mode>): ... This.
22588 (<sur>sadv16qi): Adjust use of above.
22589 * config/aarch64/arm_neon.h (vabdl_high_s8): Reimplement using
22591 (vabdl_high_s16): Likewise.
22592 (vabdl_high_s32): Likewise.
22593 (vabdl_high_u8): Likewise.
22594 (vabdl_high_u16): Likewise.
22595 (vabdl_high_u32): Likewise.
22597 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22599 * config/aarch64/aarch64-simd-builtins.def (sabal2): Define
22601 (uabal2): Likewise.
22602 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal2<mode>): New
22604 * config/aarch64/aarch64.md (unspec): Add UNSPEC_SABAL2 and
22606 * config/aarch64/arm_neon.h (vabal_high_s8): Reimplement using
22608 (vabal_high_s16): Likewise.
22609 (vabal_high_s32): Likewise.
22610 (vabal_high_u8): Likewise.
22611 (vabal_high_u16): Likewise.
22612 (vabal_high_u32): Likewise.
22613 * config/aarch64/iterators.md (ABAL2): New mode iterator.
22614 (sur): Handle UNSPEC_SABAL2, UNSPEC_UABAL2.
22616 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22618 * config/aarch64/aarch64-simd-builtins.def (sabal): Define
22621 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal<mode>_4):
22623 (aarch64_<sur>abal<mode>): ... This
22624 (<sur>sadv16qi): Adust use of the above.
22625 * config/aarch64/arm_neon.h (vabal_s8): Reimplement using
22627 (vabal_s16): Likewise.
22628 (vabal_s32): Likewise.
22629 (vabal_u8): Likewise.
22630 (vabal_u16): Likewise.
22631 (vabal_u32): Likewise.
22633 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22635 * config/aarch64/aarch64-simd-builtins.def (saddlv, uaddlv):
22637 * config/aarch64/aarch64-simd.md (aarch64_<su>addlv<mode>):
22639 * config/aarch64/arm_neon.h (vaddlv_s8): Reimplement using
22641 (vaddlv_s16): Likewise.
22642 (vaddlv_u8): Likewise.
22643 (vaddlv_u16): Likewise.
22644 (vaddlvq_s8): Likewise.
22645 (vaddlvq_s16): Likewise.
22646 (vaddlvq_s32): Likewise.
22647 (vaddlvq_u8): Likewise.
22648 (vaddlvq_u16): Likewise.
22649 (vaddlvq_u32): Likewise.
22650 (vaddlv_s32): Likewise.
22651 (vaddlv_u32): Likewise.
22652 * config/aarch64/iterators.md (VDQV_L): New mode iterator.
22653 (unspec): Add UNSPEC_SADDLV, UNSPEC_UADDLV.
22654 (Vwstype): New mode attribute.
22656 (VWIDE_S): Likewise.
22657 (USADDLV): New int iterator.
22658 (su): Handle UNSPEC_SADDLV, UNSPEC_UADDLV.
22660 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
22662 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_lane[q]
22663 builtin generator macros.
22664 * config/aarch64/aarch64-simd.md (aarch64_vec_<su>mlsl_lane<Qlane>):
22666 * config/aarch64/arm_neon.h (vmlsl_lane_s16): Use RTL builtin
22667 instead of inline asm.
22668 (vmlsl_lane_s32): Likewise.
22669 (vmlsl_lane_u16): Likewise.
22670 (vmlsl_lane_u32): Likewise.
22671 (vmlsl_laneq_s16): Likewise.
22672 (vmlsl_laneq_s32): Likewise.
22673 (vmlsl_laneq_u16): Likewise.
22674 (vmlsl_laneq_u32): Likewise.
22676 2021-01-29 Richard Biener <rguenther@suse.de>
22678 * doc/invoke.texi (--param max-gcse-memory): Document unit
22680 * gcse.c (gcse_or_cprop_is_too_expensive): Adjust.
22681 * params.opt (--param max-gcse-memory): Adjust default and
22682 document unit of size.
22684 2021-01-29 Richard Biener <rguenther@suse.de>
22686 PR rtl-optimization/98863
22687 * gcse.c (gcse_or_cprop_is_too_expensive): Use unsigned
22688 HOST_WIDE_INT for the memory estimate.
22690 2021-01-29 Bin Cheng <bin.cheng@linux.alibaba.com>
22691 Richard Biener <rguenther@suse.de>
22693 PR tree-optimization/97627
22694 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
22695 Do not analyze fake edges.
22697 2021-01-29 Richard Biener <rguenther@suse.de>
22699 PR rtl-optimization/98144
22700 * df.h (df_mir_bb_info): Add con_visited member.
22701 * df-problems.c (df_mir_alloc): Initialize con_visited,
22702 do not fully populate IN and OUT.
22703 (df_mir_reset): Likewise.
22704 (df_mir_confluence_0): Set con_visited.
22705 (df_mir_confluence_n): Properly handle implicitely
22706 fully populated IN and OUT as designated by con_visited
22707 and update con_visited accordingly.
22709 2021-01-29 Jakub Jelinek <jakub@redhat.com>
22712 * config/arm/vec-common.md (mve_vshlq_<supf><mode>,
22713 vashl<mode>3, vashr<mode>3, vlshr<mode>3): Add
22714 && !TARGET_REALLY_IWMMXT to conditions.
22716 2021-01-29 Jakub Jelinek <jakub@redhat.com>
22719 * cfgbuild.c (find_bb_boundaries): Reset debug_insn when seeing
22722 2021-01-28 Marek Polacek <polacek@redhat.com>
22725 * stor-layout.c (finalize_type_size): If we reset TYPE_USER_ALIGN in
22726 the main variant, maybe reset it in its variants too.
22727 * tree.c (check_base_type): Return true only if TYPE_USER_ALIGN match.
22728 (check_aligned_type): Check if TYPE_USER_ALIGN match.
22730 2021-01-28 Christophe Lyon <christophe.lyon@linaro.org>
22733 * config/arm/arm.c (arm_rtx_costs_internal): Adjust cost of vector
22734 of constant zero for comparisons.
22736 2021-01-28 Michael Meissner <meissner@linux.ibm.com>
22738 * config/rs6000/rs6000.c (rs6000_mangle_decl_assembler_name): Add
22739 support for mapping built-in function names for long double
22740 built-in functions if long double is IEEE 128-bit.
22742 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
22744 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_n
22745 builtin generator macros.
22746 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_n<mode>):
22748 * config/aarch64/arm_neon.h (vmlsl_n_s16): Use RTL builtin
22749 instead of inline asm.
22750 (vmlsl_n_s32): Likewise.
22751 (vmlsl_n_u16): Likewise.
22752 (vmlsl_n_u32): Likewise.
22754 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
22756 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_n
22757 builtin generator macros.
22758 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>):
22760 * config/aarch64/arm_neon.h (vmlal_n_s16): Use RTL builtin
22761 instead of inline asm.
22762 (vmlal_n_s32): Likewise.
22763 (vmlal_n_u16): Likewise.
22764 (vmlal_n_u32): Likewise.
22766 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22768 * config/aarch64/aarch64-simd-builtins.def (shrn2): Define
22770 * config/aarch64/aarch64-simd.md (aarch64_shrn2<mode>_insn_le):
22772 (aarch64_shrn2<mode>_insn_be): Likewise.
22773 (aarch64_shrn2<mode>): Likewise.
22774 * config/aarch64/arm_neon.h (vshrn_high_n_s16): Reimlplement
22776 (vshrn_high_n_s32): Likewise.
22777 (vshrn_high_n_s64): Likewise.
22778 (vshrn_high_n_u16): Likewise.
22779 (vshrn_high_n_u32): Likewise.
22780 (vshrn_high_n_u64): Likewise.
22782 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22784 * config/aarch64/aarch64-simd-builtins.def (shrn): Define
22786 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le):
22788 (aarch64_shrn<mode>_insn_be): Likewise.
22789 (aarch64_shrn<mode>): Likewise.
22790 * config/aarch64/arm_neon.h (vshrn_n_s16): Reimplement using
22792 (vshrn_n_s32): Likewise.
22793 (vshrn_n_s64): Likewise.
22794 (vshrn_n_u16): Likewise.
22795 (vshrn_n_u32): Likewise.
22796 (vshrn_n_u64): Likewise.
22797 * config/aarch64/iterators.md (vn_mode): New mode attribute.
22799 2021-01-28 Richard Biener <rguenther@suse.de>
22801 PR rtl-optimization/80960
22802 * dse.c (check_mem_read_rtx): Call get_addr on the
22805 2021-01-28 Xionghu Luo <luoxhu@linux.ibm.com>
22806 David Edelsohn <dje.gcc@gmail.com>
22809 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
22810 Don't generate VIEW_CONVERT_EXPR for fcode ALTIVEC_BUILTIN_VEC_INSERT
22812 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
22814 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Remove the
22815 wrapper call rs6000_expand_vector_set_var for cleanup. Call
22816 rs6000_expand_vector_set_var_p9 and rs6000_expand_vector_set_var_p8
22818 (rs6000_expand_vector_set_var): Delete.
22819 (rs6000_expand_vector_set_var_p9): Make static.
22820 (rs6000_expand_vector_set_var_p8): Make static.
22822 2021-01-28 Xing GUO <higuoxing@gmail.com>
22824 * common/config/riscv/riscv-common.c
22825 (riscv_subset_list::parsing_subset_version): Fix -march option parsing
22826 when `p` extension exists.
22828 2021-01-27 Vladimir N. Makarov <vmakarov@redhat.com>
22830 PR rtl-optimization/97684
22831 * ira.c (ira): Call ira_set_pseudo_classes before
22832 update_equiv_regs when it is necessary.
22834 2021-01-27 Jakub Jelinek <jakub@redhat.com>
22837 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Use
22838 %w0, %w1 and %2 instead of %0, %1 and %2.
22840 2021-01-27 Aaron Sawdey <acsawdey@linux.ibm.com>
22842 * config/rs6000/genfusion.pl: New script to generate
22843 define_insn_and_split patterns so combine can arrange fused
22844 instructions next to each other.
22845 * config/rs6000/fusion.md: New file, generated fused instruction
22846 patterns for combine.
22847 * config/rs6000/predicates.md (const_m1_to_1_operand): New predicate.
22848 (non_update_memory_operand): New predicate.
22849 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_P10_FUSION and
22850 OPTION_MASK_P10_FUSION_LD_CMPI to ISA_3_1_MASKS_SERVER and
22852 * config/rs6000/rs6000-protos.h (address_is_non_pfx_d_or_x): Add
22854 * config/rs6000/rs6000.c (rs6000_option_override_internal):
22855 Automatically set OPTION_MASK_P10_FUSION and
22856 OPTION_MASK_P10_FUSION_LD_CMPI if target is power10.
22857 (rs600_opt_masks): Allow -mpower10-fusion
22858 in function attributes.
22859 (address_is_non_pfx_d_or_x): New function.
22860 * config/rs6000/rs6000.h: Add MASK_P10_FUSION.
22861 * config/rs6000/rs6000.md: Include fusion.md.
22862 * config/rs6000/rs6000.opt: Add -mpower10-fusion
22863 and -mpower10-fusion-ld-cmpi.
22864 * config/rs6000/t-rs6000: Add dependencies involving fusion.md.
22866 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
22868 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal
22869 builtin generator macros.
22870 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal<mode>):
22872 (aarch64_<su>mlal<mode>): This.
22873 * config/aarch64/arm_neon.h (vmlal_s8): Use RTL builtin
22874 instead of inline asm.
22875 (vmlal_s16): Likewise.
22876 (vmlal_s32): Likewise.
22877 (vmlal_u8): Likewise.
22878 (vmlal_u16): Likewise.
22879 (vmlal_u32): Likewise.
22881 2021-01-27 Richard Biener <rguenther@suse.de>
22883 PR tree-optimization/98854
22884 * tree-vect-slp.c (vect_build_slp_tree_2): Also build
22885 PHIs from scalars when the number of CTORs matches the
22886 number of children.
22888 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
22890 * config/aarch64/aarch64-simd-builtins.def: Add mls_n builtin
22892 * config/aarch64/aarch64-simd.md (*aarch64_mls_elt_merge<mode>):
22894 (aarch64_mls_n<mode>): This.
22895 * config/aarch64/arm_neon.h (vmls_n_s16): Use RTL builtin
22897 (vmls_n_s32): Likewise.
22898 (vmls_n_u16): Likewise.
22899 (vmls_n_u32): Likewise.
22900 (vmlsq_n_s16): Likewise.
22901 (vmlsq_n_s32): Likewise.
22902 (vmlsq_n_u16): Likewise.
22903 (vmlsq_n_u32): Likewise.
22905 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
22907 * config/aarch64/aarch64-simd-builtins.def: Add mls builtin
22909 * config/aarch64/arm_neon.h (vmls_s8): Use RTL builtin rather
22911 (vmls_s16): Likewise.
22912 (vmls_s32): Likewise.
22913 (vmls_u8): Likewise.
22914 (vmls_u16): Likewise.
22915 (vmls_u32): Likewise.
22916 (vmlsq_s8): Likewise.
22917 (vmlsq_s16): Likewise.
22918 (vmlsq_s32): Likewise.
22919 (vmlsq_u8): Likewise.
22920 (vmlsq_u16): Likewise.
22921 (vmlsq_u32): Likewise.
22923 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
22925 * config/aarch64/aarch64-simd-builtins.def: Add mla_n builtin
22927 * config/aarch64/aarch64-simd.md (*aarch64_mla_elt_merge<mode>):
22929 (aarch64_mla_n<mode>): This.
22930 * config/aarch64/arm_neon.h (vmla_n_s16): Use RTL builtin
22932 (vmla_n_s32): Likewise.
22933 (vmla_n_u16): Likewise.
22934 (vmla_n_u32): Likewise.
22935 (vmlaq_n_s16): Likewise.
22936 (vmlaq_n_s32): Likewise.
22937 (vmlaq_n_u16): Likewise.
22938 (vmlaq_n_u32): Likewise.
22940 2021-01-27 liuhongt <hongtao.liu@intel.com>
22943 * config/i386/sse.md (sse2_gt<mode>3): Drop !TARGET_XOP in condition.
22944 (*sse2_eq<mode>3): Ditto.
22946 2021-01-27 Jakub Jelinek <jakub@redhat.com>
22948 * tree-pass.h (PROP_trees): Rename to ...
22949 (PROP_gimple): ... this.
22950 * cfgexpand.c (pass_data_expand): Replace PROP_trees with PROP_gimple.
22951 * passes.c (execute_function_dump, execute_function_todo,
22952 execute_one_ipa_transform_pass, execute_one_pass): Likewise.
22953 * varpool.c (ctor_for_folding): Likewise.
22955 2021-01-27 Jakub Jelinek <jakub@redhat.com>
22957 PR tree-optimization/97260
22958 * varpool.c: Include tree-pass.h.
22959 (ctor_for_folding): In GENERIC return DECL_INITIAL for TREE_READONLY
22960 non-TREE_SIDE_EFFECTS automatic variables.
22962 2021-01-26 Paul Fee <paul.f.fee@gmail.com>
22964 * doc/cpp.texi (__cplusplus): Document value for -std=c++23
22966 * doc/invoke.texi: Document -std=c++23 and -std=gnu++23.
22967 * dwarf2out.c (highest_c_language): Recognise C++20 and C++23.
22968 (gen_compile_unit_die): Recognise C++23.
22970 2021-01-26 Jakub Jelinek <jakub@redhat.com>
22973 * dwarf2asm.c (dw2_assemble_integer): Cast DWARF2_ADDR_SIZE to int
22976 2021-01-26 Jakub Jelinek <jakub@redhat.com>
22979 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
22980 Use UINTVAL (shft_amnt) and UINTVAL (mask) instead of INTVAL (shft_amnt)
22981 and INTVAL (mask). Add && INTVAL (mask) > 0 condition.
22983 2021-01-26 Richard Biener <rguenther@suse.de>
22985 * gimple-pretty-print.c (dump_binary_rhs): Handle
22986 VEC_WIDEN_{PLUS,MINUS}_{LO,HI}_EXPR.
22988 2021-01-26 Richard Biener <rguenther@suse.de>
22990 PR middle-end/98726
22991 * tree.h (vector_cst_int_elt): Remove.
22992 * tree.c (vector_cst_int_elt): Use poly_wide_int for computations,
22995 2021-01-26 Andrew Stubbs <ams@codesourcery.com>
22997 * config/gcn/gcn.c (gcn_expand_reduc_scalar): Use move instructions
22998 for V64DFmode min/max reductions.
23000 2021-01-26 Jakub Jelinek <jakub@redhat.com>
23002 * dwarf2asm.c (dw2_assemble_integer): Handle size twice as large
23003 as DWARF2_ADDR_SIZE if x is not a scalar int by emitting it as
23004 two halves, one with x and the other with const0_rtx, ordered
23005 depending on endianity.
23007 2021-01-26 Alexandre Oliva <oliva@adacore.com>
23009 * gimplify.c (gimplify_decl_expr): Skip asan marking calls for
23010 temporaries not seen in binding block, and not about to be
23011 added as gimple variables.
23013 2021-01-25 Martin Sebor <msebor@redhat.com>
23016 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust warning text.
23018 2021-01-25 Martin Liska <mliska@suse.cz>
23020 * value-prof.c (get_nth_most_common_value): Use %s instead
23023 2021-01-25 Jakub Jelinek <jakub@redhat.com>
23026 * configure.ac (HAVE_AS_GDWARF_5_DEBUG_FLAG): Only define if
23027 readelf -wi is able to read the emitted .debug_info back.
23028 * configure: Regenerated.
23030 2021-01-25 Martin Liska <mliska@suse.cz>
23032 PR gcov-profile/98739
23033 * common.opt: Add missing sign symbol.
23034 * value-prof.c (get_nth_most_common_value): Restore handling
23035 of PROFILE_REPRODUCIBILITY_PARALLEL_RUNS and
23036 PROFILE_REPRODUCIBILITY_MULTITHREADED.
23038 2021-01-25 Richard Biener <rguenther@suse.de>
23040 PR middle-end/98807
23041 * tree.c (vector_element_bits): Always use precision of
23042 the element type for boolean vectors.
23044 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
23046 * config/rtems.h (STARTFILE_SPEC): Remove qnolinkcmds.
23047 (ENDFILE_SPEC): Evaluate qnolinkcmds.
23049 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
23051 * config/rtems.h (STARTFILE_SPEC): Remove nostdlib and
23052 nostartfiles handling since this is already done by
23053 LINK_COMMAND_SPEC. Evaluate qnolinkcmds.
23054 (ENDFILE_SPEC): Remove nostdlib and nostartfiles handling since this
23055 is already done by LINK_COMMAND_SPEC.
23056 (LIB_SPECS): Remove nostdlib and nodefaultlibs handling since
23057 this is already done by LINK_COMMAND_SPEC. Remove qnolinkcmds
23060 2021-01-25 Jakub Jelinek <jakub@redhat.com>
23063 * fold-const-call.c (host_size_t_cst_p): Renamed to ...
23064 (size_t_cst_p): ... this. Check and store unsigned HOST_WIDE_INT
23065 value rather than host size_t.
23066 (fold_const_call): Change type of s2 from size_t to
23067 unsigned HOST_WIDE_INT. Use size_t_cst_p instead of
23068 host_size_t_cst_p. For strncmp calls, pass MIN (s2, SIZE_MAX)
23069 instead of s2 as last argument.
23071 2021-01-25 Tamar Christina <tamar.christina@arm.com>
23073 * config/arm/iterators.md (rotsplit1, rotsplit2, conj_op, fcmac1,
23074 VCMLA_OP, VCMUL_OP): New.
23075 * config/arm/mve.md (mve_vcmlaq<mve_rot><mode>): Support vec_dup 0.
23076 * config/arm/neon.md (cmul<conj_op><mode>3): New.
23077 * config/arm/unspecs.md (UNSPEC_VCMLA_CONJ, UNSPEC_VCMLA180_CONJ,
23078 UNSPEC_VCMUL_CONJ): New.
23079 * config/arm/vec-common.md (cmul<conj_op><mode>3, arm_vcmla<rot><mode>,
23080 cml<fcmac1><conj_op><mode>4): New.
23082 2021-01-23 Jakub Jelinek <jakub@redhat.com>
23085 * config/rs6000/mmintrin.h (__m64): Add __may_alias__ attribute.
23087 2021-01-22 Jonathan Wright <jonathan.wright@arm.com>
23089 * config/aarch64/aarch64-simd-builtins.def: Add mla builtin
23091 * config/aarch64/arm_neon.h (vmla_s8): Use RTL builtin rather
23093 (vmla_s16): Likewise.
23094 (vmla_s32): Likewise.
23095 (vmla_u8): Likewise.
23096 (vmla_u16): Likewise.
23097 (vmla_u32): Likewise.
23098 (vmlaq_s8): Likewise.
23099 (vmlaq_s16): Likewise.
23100 (vmlaq_s32): Likewise.
23101 (vmlaq_u8): Likewise.
23102 (vmlaq_u16): Likewise.
23103 (vmlaq_u32): Likewise.
23105 2021-01-22 David Malcolm <dmalcolm@redhat.com>
23107 * doc/invoke.texi (GCC_EXTRA_DIAGNOSTIC_OUTPUT): Add @findex
23110 2021-01-22 Jakub Jelinek <jakub@redhat.com>
23113 * dwarf2out.c (output_file_names): For -gdwarf-5, if there are no
23114 filenames to emit, still emit the required 0 index directory and
23115 filename entries that match DW_AT_comp_dir and DW_AT_name of the
23118 2021-01-22 Marek Polacek <polacek@redhat.com>
23121 * doc/invoke.texi: Update C++ ABI Version 15 description.
23123 2021-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23125 PR tree-optimization/98766
23126 * tree-ssa-math-opts.c (convert_mult_to_fma): Use maybe_le when
23127 comparing against type size with param_avoid_fma_max_bits.
23129 2021-01-22 Richard Biener <rguenther@suse.de>
23131 PR middle-end/98793
23132 * tree.c (vector_element_bits): Key single-bit bool vector on
23133 integer mode rather than not vector mode.
23135 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
23138 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
23139 Generate ARRAY_REF(VIEW_CONVERT_EXPR) for P8 and later
23141 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): Update
23142 to call different path for P8 and P9.
23143 (rs6000_expand_vector_set_var_p9): New function.
23144 (rs6000_expand_vector_set_var_p8): New function.
23146 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
23150 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
23151 Ajdust variable index vec_insert from address dereference to
23152 ARRAY_REF(VIEW_CONVERT_EXPR) tree expression.
23153 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
23155 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): New function.
23157 2021-01-22 Martin Liska <mliska@suse.cz>
23159 PR gcov-profile/98739
23160 * profile.c (compute_value_histograms): Drop time profile for
23161 -fprofile-reproducible=multithreaded.
23163 2021-01-22 Nathan Sidwell <nathan@acm.org>
23165 * gcc.c (process_command): Don't check OPT_SPECIAL_input_file
23168 2021-01-22 Richard Biener <rguenther@suse.de>
23170 PR middle-end/98773
23171 * tree-data-ref.c (initalize_matrix_A): Revert previous
23172 change, retaining failing on HOST_WIDE_INT_MIN CHREC_RIGHT.
23174 2021-01-22 Jakub Jelinek <jakub@redhat.com>
23176 PR tree-optimization/90248
23177 * match.pd (X cmp 0.0 ? 1.0 : -1.0 -> copysign(1, +-X),
23178 X cmp 0.0 ? -1.0 : +1.0 -> copysign(1, -+X)): Remove
23180 (X * (X cmp 0.0 ? 1.0 : -1.0) -> +-abs(X),
23181 X * (X cmp 0.0 ? -1.0 : 1.0) -> +-abs(X)): New simplifications.
23183 2021-01-22 Jakub Jelinek <jakub@redhat.com>
23185 PR tree-optimization/98255
23186 * tree-dfa.c (get_ref_base_and_extent): For ARRAY_REFs, sign
23187 extend index - low_bound from sizetype's precision rather than index
23189 (get_addr_base_and_unit_offset_1): Likewise.
23190 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Likewise.
23191 * gimple-fold.c (fold_const_aggregate_ref_1): Likewise.
23193 2021-01-22 Richard Biener <rguenther@suse.de>
23195 PR tree-optimization/98786
23196 * tree-ssa-phiopt.c (factor_out_conditional_conversion): Avoid
23197 adding new uses of abnormals. Verify we deal with a conditional
23200 2021-01-22 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23203 * optc-save-gen.awk: Add arm_fp16_format to checked_options.
23205 2021-01-22 liuhongt <hongtao.liu@intel.com>
23209 * config/i386/sse.md (VI_128_256): New mode iterator.
23210 (*avx_cmp<mode>3_1, *avx_cmp<mode>3_2, *avx_cmp<mode>3_3,
23211 *avx_cmp<mode>3_4, *avx2_eq<mode>3, *avx2_pcmp<mode>3_1,
23212 *avx2_pcmp<mode>3_2, *avx2_gt<mode>3): New
23213 define_insn_and_split to lower avx512 vector comparison to avx
23214 version when dest is vector.
23215 (*<avx512>_cmp<mode>3,*<avx512>_cmp<mode>3,*<avx512>_ucmp<mode>3):
23216 define_insn_and_split for negating the comparison result.
23217 * config/i386/predicates.md (float_vector_all_ones_operand):
23219 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Use
23220 general NOT operator without UNSPEC_MASKOP.
23222 2021-01-21 Vladimir N. Makarov <vmakarov@redhat.com>
23224 PR rtl-optimization/98777
23225 * lra-int.h (lra_pmode_pseudo): New extern.
23226 * lra.c (lra_pmode_pseudo): New global.
23228 * lra-eliminations.c (eliminate_regs_in_insn): Use it.
23230 2021-01-21 Ilya Leoshkevich <iii@linux.ibm.com>
23232 * fwprop.c (fwprop_propagation::classify_result): Allow
23233 (subreg (mem)) simplifications.
23235 2021-01-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23237 * config/aarch64/aarch64-simd.md (aarch64_sqdml<SBINQOPS:as>l<mode>):
23239 (aarch64_sqdmlal<mode>): ... This...
23240 (aarch64_sqdmlsl<mode>): ... And this.
23241 (aarch64_sqdml<SBINQOPS:as>l_lane<mode>): Split into...
23242 (aarch64_sqdmlal_lane<mode>): ... This...
23243 (aarch64_sqdmlsl_lane<mode>): ... And this.
23244 (aarch64_sqdml<SBINQOPS:as>l_laneq<mode>): Split into...
23245 (aarch64_sqdmlsl_laneq<mode>): ... This...
23246 (aarch64_sqdmlal_laneq<mode>): ... And this.
23247 (aarch64_sqdml<SBINQOPS:as>l_n<mode>): Split into...
23248 (aarch64_sqdmlsl_n<mode>): ... This...
23249 (aarch64_sqdmlal_n<mode>): ... And this.
23250 (aarch64_sqdml<SBINQOPS:as>l2<mode>_internal): Split into...
23251 (aarch64_sqdmlal2<mode>_internal): ... This...
23252 (aarch64_sqdmlsl2<mode>_internal): ... And this.
23254 2021-01-21 Christophe Lyon <christophe.lyon@linaro.org>
23256 * config/arm/arm_mve.h (__arm_vcmpneq_s8): Fix return type.
23258 2021-01-21 Andrea Corallo <andrea.corallo@arm.com>
23261 * doc/sourcebuild.texi (arm_thumb2_no_arm_v8_1_lob): Document.
23263 2021-01-21 liuhongt <hongtao.liu@intel.com>
23265 PR rtl-optimization/98694
23266 * regcprop.c (copy_value): If SRC had been assigned a mode
23267 narrower than the copy, we can't link DEST into the chain even
23268 they have same hard_regno_nregs(i.e. HImode/SImode in i386
23271 2021-01-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23273 * config/aarch64/aarch64-simd.md (aarch64_get_lane<mode>):
23274 Convert to define_insn_and_split. Split into simple move when moving
23277 2021-01-20 Segher Boessenkool <segher@kernel.crashing.org>
23279 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_store): Change assert.
23280 Adjust comment. Simplify code.
23282 2021-01-20 Jakub Jelinek <jakub@redhat.com>
23285 * dwarf2out.c (reset_indirect_string): Also reset indirect strings
23286 with DW_FORM_line_strp form.
23287 (prune_unused_types_update_strings): Don't add into debug_str_hash
23288 indirect strings with DW_FORM_line_strp form.
23289 (adjust_name_comp_dir): New function.
23290 (dwarf2out_finish): Call it on CU DIEs after resetting
23291 debug_line_str_hash.
23293 2021-01-20 Vladimir N. Makarov <vmakarov@redhat.com>
23295 PR rtl-optimization/98722
23296 * lra-eliminations.c (eliminate_regs_in_insn): Check that target
23297 has no 3-op add insn to transform insns containing two pluses.
23299 2021-01-20 Richard Biener <rguenther@suse.de>
23301 * hwint.h (add_hwi): New function.
23302 (mul_hwi): Likewise.
23303 * tree-data-ref.c (initialize_matrix_A): Properly translate
23304 tree constants and avoid HOST_WIDE_INT_MIN.
23305 (lambda_matrix_row_add): Avoid undefined integer overflow
23306 and return true on such overflow.
23307 (lambda_matrix_right_hermite): Handle overflow from
23308 lambda_matrix_row_add gracefully. Simplify previous fix.
23309 (analyze_subscript_affine_affine): Likewise.
23311 2021-01-20 Eugene Rozenfeld <erozen@microsoft.com>
23313 PR tree-optimization/96674
23314 * match.pd: New patterns: x < y || y == XXX_MIN --> x <= y - 1
23315 x >= y && y != XXX_MIN --> x > y - 1
23317 2021-01-20 Richard Sandiford <richard.sandiford@arm.com>
23319 PR tree-optimization/98535
23320 * tree-vect-slp.c (duplicate_and_interleave): Use quick_grow_cleared.
23321 If the high and low permutes are the same, remove the high permutes
23322 from the working set and only continue with the low ones.
23324 2021-01-20 Jakub Jelinek <jakub@redhat.com>
23326 PR tree-optimization/98721
23327 * builtins.c (access_ref::inform_access): Don't assume
23328 SSA_NAME_IDENTIFIER must be non-NULL. Print messages about
23329 object whenever allocfn is NULL, rather than only when DECL_P
23330 is true. Use %qE instead of %qD for that. Formatting fixes.
23332 2021-01-20 Richard Biener <rguenther@suse.de>
23334 PR tree-optimization/98758
23335 * tree-data-ref.c (int_divides_p): Use lambda_int arguments.
23336 (lambda_matrix_right_hermite): Avoid undefinedness with
23337 signed integer abs and multiplication.
23338 (analyze_subscript_affine_affine): Use lambda_int.
23340 2021-01-20 David Malcolm <dmalcolm@redhat.com>
23343 * dwarf2out.c (output_line_info): Rename static variable
23344 "generation", moving it out of the function to...
23345 (output_line_info_generation): New.
23346 (init_sections_and_labels): Likewise, renaming the variable to...
23347 (init_sections_and_labels_generation): New.
23348 (dwarf2out_c_finalize): Reset the new variables.
23350 2021-01-19 Martin Sebor <msebor@redhat.com>
23352 PR middle-end/98664
23353 * tree-ssa-live.c (remove_unused_scope_block_p): Keep scopes for
23354 all functions, even if they're not declared artificial or inline.
23355 * tree.c (tree_inlined_location): Use macro expansion location
23356 only if scope traversal fails to expose one.
23358 2021-01-19 Richard Sandiford <richard.sandiford@arm.com>
23360 PR rtl-optimization/92294
23361 * alias.c (compare_base_symbol_refs): Take an extra parameter
23362 and add the distance between two symbols to it. Enshrine in
23363 comments that -1 means "either 0 or 1, but we can't tell
23364 which at compile time".
23365 (memrefs_conflict_p): Update call accordingly.
23366 (rtx_equal_for_memref_p): Likewise. Take the distance between symbols
23369 2021-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23371 * config/aarch64/aarch64-simd-builtins.def (sqshl, uqshl,
23372 sqrshl, uqrshl, sqadd, uqadd, sqsub, uqsub, suqadd, usqadd, sqmovn,
23373 uqmovn, sqxtn2, uqxtn2, sqabs, sqneg, sqdmlal, sqdmlsl, sqdmlal_lane,
23374 sqdmlsl_lane, sqdmlal_laneq, sqdmlsl_laneq, sqdmlal_n, sqdmlsl_n,
23375 sqdmlal2, sqdmlsl2, sqdmlal2_lane, sqdmlsl2_lane, sqdmlal2_laneq,
23376 sqdmlsl2_laneq, sqdmlal2_n, sqdmlsl2_n, sqdmull, sqdmull_lane,
23377 sqdmull_laneq, sqdmull_n, sqdmull2, sqdmull2_lane, sqdmull2_laneq,
23378 sqdmull2_n, sqdmulh, sqrdmulh, sqdmulh_lane, sqdmulh_laneq,
23379 sqrdmulh_lane, sqrdmulh_laneq, sqshrun_n, sqrshrun_n, sqshrn_n,
23380 uqshrn_n, sqrshrn_n, uqrshrn_n, sqshlu_n, sqshl_n, uqshl_n, sqrdmlah,
23381 sqrdmlsh, sqrdmlah_lane, sqrdmlsh_lane, sqrdmlah_laneq, sqrdmlsh_laneq,
23382 sqmovun): Use NONE flags.
23384 2021-01-19 Richard Biener <rguenther@suse.de>
23387 * ipa-modref.c (analyze_stmt): Only record a summary for a
23390 2021-01-19 Richard Biener <rguenther@suse.de>
23392 PR middle-end/98638
23393 * tree-ssanames.c (fini_ssanames): Zero SSA_NAME_DEF_STMT.
23395 2021-01-19 Daniel Hellstrom <daniel@gaisler.com>
23397 * config/sparc/rtemself.h (TARGET_OS_CPP_BUILTINS): Add
23398 built-in define __FIX_LEON3FT_TN0018.
23400 2021-01-19 Richard Biener <rguenther@suse.de>
23403 * tree-inline.c (tree_function_versioning): Set input_location
23404 to UNKNOWN_LOCATION throughout the function.
23406 2021-01-19 Tobias Burnus <tobias@codesourcery.com>
23409 * omp-low.c (lower_omp_target): Handle nonpointer is_device_ptr.
23411 2021-01-19 Martin Jambor <mjambor@suse.cz>
23414 * ipa-sra.c (ssa_name_only_returned_p): New parameter fun. Check
23415 whether non-call exceptions allow removal of a statement.
23416 (isra_analyze_call): Pass the appropriate function to
23417 ssa_name_only_returned_p.
23419 2021-01-19 Geng Qi <gengqi@linux.alibaba.com>
23421 * config/riscv/arch-canonicalize (longext_sort): New function for
23422 sorting 'multi-letter'.
23423 * config/riscv/multilib-generator: Adjusting the loop of 'alt' in
23424 'alts'. The 'arch' may not be the first of 'alts'.
23425 (_expand_combination): Add underline for the 'ext' without '*'.
23426 This is because, a single-letter extension can always be treated well
23427 with a '_' prefix, but it cannot be separated out if it is appended
23430 2021-01-18 Vladimir N. Makarov <vmakarov@redhat.com>
23433 * ira.c (ira): Skip abnormal critical edge splitting.
23435 2021-01-18 Jakub Jelinek <jakub@redhat.com>
23437 PR tree-optimization/98727
23438 * tree-ssa-math-opts.c (match_arith_overflow): Fix up computation of
23439 second .MUL_OVERFLOW operand for signed multiplication with overflow
23440 checking if the second operand of multiplication is not constant.
23442 2021-01-18 David Edelsohn <dje.gcc@gmail.com>
23444 * doc/invoke.texi (-gdwarf): TPF defaults to version 2 and AIX
23445 defaults to version 4.
23447 2021-01-18 David Malcolm <dmalcolm@redhat.com>
23449 * attribs.h (fndecl_dealloc_argno): New decl.
23450 * builtins.c (call_dealloc_argno): Split out second half of
23452 (fndecl_dealloc_argno): New.
23453 * doc/extend.texi (Common Function Attributes): Document the
23454 interaction between the analyzer and the malloc attribute.
23455 * doc/invoke.texi (Static Analyzer Options): Likewise.
23457 2021-01-17 David Edelsohn <dje.gcc@gmail.com>
23459 * config/rs6000/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Override
23460 dwarf_version to 4.
23461 * config/rs6000/aix72.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
23463 2021-01-17 Martin Jambor <mjambor@suse.cz>
23466 * cgraph.c (clone_of_p): Check also former_clone_of as we climb
23469 2021-01-17 Mark Wielaard <mark@klomp.org>
23471 * common.opt (gdwarf-): Init(5).
23472 * doc/invoke.texi (-gdwarf): Document default to 5.
23474 2021-01-16 Kwok Cheung Yeung <kcy@codesourcery.com>
23476 * builtin-types.def
23477 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT): Rename
23479 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR):
23480 ...this. Add extra argument.
23481 * gimplify.c (omp_default_clause): Ensure that event handle is
23482 firstprivate in a task region.
23483 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_DETACH.
23484 (gimplify_adjust_omp_clauses): Likewise.
23485 * omp-builtins.def (BUILT_IN_GOMP_TASK): Change function type to
23486 BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR.
23487 * omp-expand.c (expand_task_call): Add GOMP_TASK_FLAG_DETACH to flags
23488 if detach clause specified. Add detach argument when generating
23490 * omp-low.c (scan_sharing_clauses): Setup data environment for detach
23492 (finish_taskreg_scan): Move field for variable containing the event
23493 handle to the front of the struct.
23494 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_DETACH. Fix
23496 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
23497 OMP_CLAUSE_DETACH clause.
23498 (convert_local_omp_clauses): Handle OMP_CLAUSE_DETACH clause.
23499 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_DETACH.
23500 * tree.c (omp_clause_num_ops): Add entry for OMP_CLAUSE_DETACH.
23502 (omp_clause_code_name): Add entry for OMP_CLAUSE_DETACH. Fix
23504 (walk_tree_1): Handle OMP_CLAUSE_DETACH.
23506 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
23508 * config/nios2/t-rtems: Reset all MULTILIB_* variables. Shorten
23509 multilib directory names. Use MULTILIB_REQUIRED instead of
23510 MULTILIB_EXCEPTIONS. Add -mhw-mul -mhw-mulx -mhw-div
23511 -mcustom-fpu-cfg=fph2 multilib.
23513 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
23515 * config/nios2/nios2.c (NIOS2_FPU_CONFIG_NUM): Adjust value.
23516 (nios2_init_fpu_configs): Provide register values for new
23517 -mcustom-fpu-cfg=fph2 option variant.
23518 * doc/invoke.texi (-mcustom-fpu-cfg=fph2): Document new option
23521 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
23523 * config/nios2/nios2.c (nios2_custom_check_insns): Remove
23524 custom instruction warnings.
23526 2021-01-16 Jakub Jelinek <jakub@redhat.com>
23528 PR tree-optimization/96669
23529 * match.pd ((CST << x) & 1 -> x == 0): New simplification.
23531 2021-01-16 Jakub Jelinek <jakub@redhat.com>
23533 PR tree-optimization/96271
23534 * passes.def: Pass false argument to first two pass_cd_dce
23535 instances and true to last instance. Add comment that
23536 last instance rewrites no longer addressed locals.
23537 * tree-ssa-dce.c (pass_cd_dce): Add update_address_taken_p member and
23539 (pass_cd_dce::set_pass_param): New method.
23540 (pass_cd_dce::execute): Return TODO_update_address_taken from
23541 last cd_dce instance.
23543 2021-01-15 Carl Love <cel@us.ibm.com>
23545 * config/rs6000/altivec.h (vec_mulh, vec_div, vec_dive, vec_mod):
23547 * config/rs6000/altivec.md (VIlong): Move define to file vsx.md.
23548 * config/rs6000/rs6000-builtin.def (DIVES_V4SI, DIVES_V2DI,
23549 DIVEU_V4SI, DIVEU_V2DI, DIVS_V4SI, DIVS_V2DI, DIVU_V4SI,
23550 DIVU_V2DI, MODS_V2DI, MODS_V4SI, MODU_V2DI, MODU_V4SI,
23551 MULHS_V2DI, MULHS_V4SI, MULHU_V2DI, MULHU_V4SI, MULLD_V2DI):
23552 Add builtin define.
23553 (MULH, DIVE, MOD): Add new BU_P10_OVERLOAD_2 definitions.
23554 * config/rs6000/rs6000-call.c (VSX_BUILTIN_VEC_DIV,
23555 VSX_BUILTIN_VEC_DIVE, P10_BUILTIN_VEC_MOD, P10_BUILTIN_VEC_MULH):
23556 New overloaded definitions.
23557 (builtin_function_type) [P10V_BUILTIN_DIVEU_V4SI,
23558 P10V_BUILTIN_DIVEU_V2DI, P10V_BUILTIN_DIVU_V4SI,
23559 P10V_BUILTIN_DIVU_V2DI, P10V_BUILTIN_MODU_V2DI,
23560 P10V_BUILTIN_MODU_V4SI, P10V_BUILTIN_MULHU_V2DI,
23561 P10V_BUILTIN_MULHU_V4SI]: Add case
23562 statement for builtins.
23563 * config/rs6000/rs6000.md (bits): Add new attribute sizes V4SI, V2DI.
23564 * config/rs6000/vsx.md (VIlong): Moved from config/rs6000/altivec.md.
23565 (UNSPEC_VDIVES, UNSPEC_VDIVEU): New unspec definitions.
23566 (vsx_mul_v2di): Add if TARGET_POWER10 statement.
23567 (vsx_udiv_v2di): Add if TARGET_POWER10 statement.
23568 (dives_<mode>, diveu_<mode>, div<mode>3, uvdiv<mode>3,
23569 mods_<mode>, modu_<mode>, mulhs_<mode>, mulhu_<mode>, mulv2di3):
23570 Add define_insn, mode is VIlong.
23571 * doc/extend.texi (vec_mulh, vec_mul, vec_div, vec_dive, vec_mod):
23572 Add builtin descriptions.
23574 2021-01-15 Eric Botcazou <ebotcazou@adacore.com>
23576 * final.c (final_start_function_1): Reset force_source_line.
23578 2021-01-15 Jakub Jelinek <jakub@redhat.com>
23580 PR tree-optimization/96669
23581 * match.pd (((1 << A) & 1) != 0 -> A == 0,
23582 ((1 << A) & 1) == 0 -> A != 0): Generalize for 1s replaced by
23583 possibly different power of two constants and to right shift too.
23585 2021-01-15 Jakub Jelinek <jakub@redhat.com>
23587 PR tree-optimization/96681
23588 * match.pd ((x < 0) ^ (y < 0) to (x ^ y) < 0): New simplification.
23589 ((x >= 0) ^ (y >= 0) to (x ^ y) < 0): Likewise.
23590 ((x < 0) ^ (y >= 0) to (x ^ y) >= 0): Likewise.
23591 ((x >= 0) ^ (y < 0) to (x ^ y) >= 0): Likewise.
23593 2021-01-15 Alexandre Oliva <oliva@adacore.com>
23595 * opts.c (gen_command_line_string): Exclude -dumpbase-ext.
23597 2021-01-15 Tamar Christina <tamar.christina@arm.com>
23599 * config/aarch64/aarch64-simd.md (cml<fcmac1><conj_op><mode>4,
23600 cmul<conj_op><mode>3): New.
23601 * config/aarch64/iterators.md (UNSPEC_FCMUL,
23602 UNSPEC_FCMUL180, UNSPEC_FCMLA_CONJ, UNSPEC_FCMLA180_CONJ,
23603 UNSPEC_CMLA_CONJ, UNSPEC_CMLA180_CONJ, UNSPEC_CMUL, UNSPEC_CMUL180,
23604 FCMLA_OP, FCMUL_OP, conj_op, rotsplit1, rotsplit2, fcmac1, sve_rot1,
23605 sve_rot2, SVE2_INT_CMLA_OP, SVE2_INT_CMUL_OP, SVE2_INT_CADD_OP): New.
23606 (rot): Add UNSPEC_FCMUL, UNSPEC_FCMUL180.
23607 (rot_op): Renamed to conj_op.
23608 * config/aarch64/aarch64-sve.md (cml<fcmac1><conj_op><mode>4,
23609 cmul<conj_op><mode>3): New.
23610 * config/aarch64/aarch64-sve2.md (cml<fcmac1><conj_op><mode>4,
23611 cmul<conj_op><mode>3): New.
23613 2021-01-15 David Malcolm <dmalcolm@redhat.com>
23617 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
23618 Escape the tempfile name when constructing the expected output.
23620 2021-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23622 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlsl_hi<mode>):
23624 (aarch64_<su>mlsl_hi<mode>): ... This.
23625 (aarch64_<su>mlsl_hi<mode>): Define.
23626 (*aarch64_<su>mlsl<mode): Rename to...
23627 (aarch64_<su>mlsl<mode): ... This.
23628 * config/aarch64/aarch64-simd-builtins.def (smlsl, umlsl,
23629 smlsl_hi, umlsl_hi): Define builtins.
23630 * config/aarch64/arm_neon.h (vmlsl_high_s8, vmlsl_high_s16,
23631 vmlsl_high_s32, vmlsl_high_u8, vmlsl_high_u16, vmlsl_high_u32,
23632 vmlsl_s8, vmlsl_s16, vmlsl_s32, vmlsl_u8,
23633 vmlsl_u16, vmlsl_u32): Reimplement with builtins.
23635 2021-01-15 Uroš Bizjak <ubizjak@gmail.com>
23637 * config/i386/i386-c.c (ix86_target_macros):
23638 Use cpp_define_formatted for __SIZEOF_FLOAT80__ definition.
23640 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
23643 * config.gcc (aarch64*-*-*): Add aarch64-cc-fusion.o to extra_objs.
23644 * Makefile.in (RTL_SSA_H): New variable.
23645 * config/aarch64/t-aarch64 (aarch64-cc-fusion.o): New rule.
23646 * config/aarch64/aarch64-protos.h (make_pass_cc_fusion): Declare.
23647 * config/aarch64/aarch64-passes.def: Add pass_cc_fusion after
23649 * config/aarch64/aarch64-cc-fusion.cc: New file.
23651 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
23653 * recog.h (insn_change_watermark::~insn_change_watermark): Avoid
23654 calling cancel_changes for changes that no longer exist.
23656 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
23658 * rtl-ssa/functions.h (function_info::ref_defs): Rename to...
23659 (function_info::reg_defs): ...this.
23660 * rtl-ssa/member-fns.inl (function_info::ref_defs): Rename to...
23661 (function_info::reg_defs): ...this.
23663 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23666 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
23668 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23671 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23674 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
23676 2021-01-15 Richard Biener <rguenther@suse.de>
23678 PR tree-optimization/96376
23679 * tree-vect-stmts.c (get_load_store_type): Disregard alignment
23680 for VMAT_INVARIANT.
23682 2021-01-15 Martin Liska <mliska@suse.cz>
23684 * doc/install.texi: Document that some tests need pytest module.
23685 * doc/sourcebuild.texi: Likewise.
23687 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23690 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
23692 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23694 * config/arm/mve.md (mve_vshrq_n_s<mode>_imm): New entry.
23695 (mve_vshrq_n_u<mode>_imm): Likewise.
23696 * config/arm/neon.md (vashr<mode>3, vlshr<mode>3): Move to ...
23697 * config/arm/vec-common.md: ... here.
23699 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23701 * config/arm/mve.md (mve_vshlq_<supf><mode>): Move to
23703 * config/arm/neon.md (vashl<mode>3): Delete.
23704 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): New.
23705 (vasl<mode>3): New expander.
23707 2021-01-15 Richard Biener <rguenther@suse.de>
23709 PR tree-optimization/98685
23710 * tree-vect-slp.c (vect_schedule_slp_node): Refactor handling
23711 of vector extern defs.
23713 2021-01-14 David Malcolm <dmalcolm@redhat.com>
23716 * diagnostic.c (diagnostic_kind_text): Break out this array
23718 (diagnostic_build_prefix): ...here.
23719 (fancy_abort): Detect when diagnostic_initialize has not yet been
23720 called and fall back to a minimal implementation of printing the
23721 ICE, rather than segfaulting in internal_error.
23723 2021-01-14 David Malcolm <dmalcolm@redhat.com>
23725 * diagnostic.c (diagnostic_initialize): Eliminate
23726 parseable_fixits_p in favor of initializing extra_output_kind from
23727 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
23728 (convert_column_unit): New function, split out from...
23729 (diagnostic_converted_column): ...this.
23730 (print_parseable_fixits): Add "column_unit" and "tabstop" params.
23731 Use them to call convert_column_unit on the column values.
23732 (diagnostic_report_diagnostic): Eliminate conditional on
23733 parseable_fixits_p in favor of a switch statement on
23734 extra_output_kind, passing the appropriate values to the new
23735 params of print_parseable_fixits.
23736 (selftest::test_print_parseable_fixits_none): Update for new
23737 params of print_parseable_fixits.
23738 (selftest::test_print_parseable_fixits_insert): Likewise.
23739 (selftest::test_print_parseable_fixits_remove): Likewise.
23740 (selftest::test_print_parseable_fixits_replace): Likewise.
23741 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
23743 (selftest::diagnostic_c_tests): Call it.
23744 * diagnostic.h (enum diagnostics_extra_output_kind): New.
23745 (diagnostic_context::parseable_fixits_p): Delete field in favor
23747 (diagnostic_context::extra_output_kind): ...this new field.
23748 * doc/invoke.texi (Environment Variables): Add
23749 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
23750 * opts.c (common_handle_option): Update handling of
23751 OPT_fdiagnostics_parseable_fixits for change to diagnostic_context
23754 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23756 * tree-vect-slp-patterns.c (class complex_operations_pattern,
23757 complex_operations_pattern::matches,
23758 complex_operations_pattern::recognize,
23759 complex_operations_pattern::build): New.
23760 (slp_patterns): Use it.
23762 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23764 * internal-fn.def (COMPLEX_FMS, COMPLEX_FMS_CONJ): New.
23765 * optabs.def (cmls_optab, cmls_conj_optab): New.
23766 * doc/md.texi: Document them.
23767 * tree-vect-slp-patterns.c (class complex_fms_pattern,
23768 complex_fms_pattern::matches, complex_fms_pattern::recognize,
23769 complex_fms_pattern::build): New.
23771 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23773 * internal-fn.def (COMPLEX_FMA, COMPLEX_FMA_CONJ): New.
23774 * optabs.def (cmla_optab, cmla_conj_optab): New.
23775 * doc/md.texi: Document them.
23776 * tree-vect-slp-patterns.c (vect_match_call_p,
23777 class complex_fma_pattern, vect_slp_reset_pattern,
23778 complex_fma_pattern::matches, complex_fma_pattern::recognize,
23779 complex_fma_pattern::build): New.
23781 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23783 * internal-fn.def (COMPLEX_MUL, COMPLEX_MUL_CONJ): New.
23784 * optabs.def (cmul_optab, cmul_conj_optab): New.
23785 * doc/md.texi: Document them.
23786 * tree-vect-slp-patterns.c (vect_match_call_complex_mla,
23787 vect_normalize_conj_loc, is_eq_or_top, vect_validate_multiplication,
23788 vect_build_combine_node, class complex_mul_pattern,
23789 complex_mul_pattern::matches, complex_mul_pattern::recognize,
23790 complex_mul_pattern::build): New.
23792 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23794 * tree-vect-slp.c (optimize_load_redistribution_1): New.
23795 (optimize_load_redistribution, vect_is_slp_load_node): New.
23796 (vect_match_slp_patterns): Use it.
23798 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23800 * tree-vect-slp-patterns.c (complex_add_pattern::build):
23803 2021-01-14 Thomas Schwinge <thomas@codesourcery.com>
23805 * config/gcn/mkoffload.c (main): Create an offload image only in
23806 64-bit configurations.
23808 2021-01-14 H.J. Lu <hjl.tools@gmail.com>
23811 * config/i386/i386-options.c (ix86_option_override_internal):
23812 Issue an error for -fcf-protection with CF_BRANCH when compiling
23813 for 32-bit non-TARGET_CMOV targets.
23815 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
23818 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p):
23819 Remove declaration and initialization of shadow variable "ret".
23820 (ix86_option_override_internal): Remove delcaration of
23821 shadow variable "i". Redeclare shadowed variable to unsigned.
23822 * common/config/i386/i386-common.c (pta_size): Redeclare to unsigned.
23823 * config/i386/i386-builtins.c (get_builtin_code_for_version):
23824 Update for redeclaration.
23825 * config/i386/i386.h (pta_size): Ditto.
23827 2021-01-14 Richard Biener <rguenther@suse.de>
23829 PR tree-optimization/98674
23830 * tree-data-ref.c (base_supports_access_fn_components_p): New.
23831 (initialize_data_dependence_relation): For two bases without
23832 possible access fns resort to type size equality when determining
23833 shape compatibility.
23835 2021-01-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23838 * config/arm/arm_neon.h: Replace calls to __builtin_vcge* by
23839 <=, >= operators in vcle and vcge intrinsics respectively.
23840 * config/arm/arm_neon_builtins.def: Remove entry for
23843 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
23846 * config/i386/i386-options.c (ix86_function_specific_save):
23847 Remove redundant assignment to opts->x_ix86_branch_cost.
23848 * config/i386/i386.c (ix86_prefetch_sse):
23849 Rename from x86_prefetch_sse. Update all uses.
23850 * config/i386/i386.h: Update for rename.
23851 * config/i386/i386-options.h: Ditto.
23853 2021-01-14 Jakub Jelinek <jakub@redhat.com>
23856 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3,
23857 *sse4_1_zero_extendv4hiv4si2_3, *sse4_1_zero_extendv2siv2di2_3):
23858 Use Bm instead of m for non-avx. Add isa attribute.
23860 2021-01-14 Jakub Jelinek <jakub@redhat.com>
23862 PR tree-optimization/96688
23863 * match.pd (~(X >> Y) -> ~X >> Y): New simplification if
23864 ~X can be simplified.
23866 2021-01-14 Richard Sandiford <richard.sandiford@arm.com>
23868 * tree-vect-stmts.c (vect_model_load_cost): Account for unused
23869 IFN_LOAD_LANES results.
23871 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23873 * config/aarch64/aarch64-simd.md (aarch64_<su>xtl<mode>):
23875 (aarch64_xtn<mode>): Likewise.
23876 * config/aarch64/aarch64-simd-builtins.def (sxtl, uxtl, xtn):
23879 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
23881 (vmovl_s16): Likewise.
23882 (vmovl_s32): Likewise.
23883 (vmovl_u8): Likewise.
23884 (vmovl_u16): Likewise.
23885 (vmovl_u32): Likewise.
23886 (vmovn_s16): Likewise.
23887 (vmovn_s32): Likewise.
23888 (vmovn_s64): Likewise.
23889 (vmovn_u16): Likewise.
23890 (vmovn_u32): Likewise.
23891 (vmovn_u64): Likewise.
23893 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23895 * config/aarch64/aarch64-simd.md (aarch64_<su>qxtn2<mode>_le):
23897 (aarch64_<su>qxtn2<mode>_be): Likewise.
23898 (aarch64_<su>qxtn2<mode>): Likewise.
23899 * config/aarch64/aarch64-simd-builtins.def (sqxtn2, uqxtn2):
23901 * config/aarch64/iterators.md (SAT_TRUNC): Define code_iterator.
23902 (su): Handle ss_truncate and us_truncate.
23903 * config/aarch64/arm_neon.h (vqmovn_high_s16): Reimplement using
23905 (vqmovn_high_s32): Likewise.
23906 (vqmovn_high_s64): Likewise.
23907 (vqmovn_high_u16): Likewise.
23908 (vqmovn_high_u32): Likewise.
23909 (vqmovn_high_u64): Likewise.
23911 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23913 * config/aarch64/aarch64-simd.md (aarch64_xtn2<mode>_le):
23915 (aarch64_xtn2<mode>_be): Likewise.
23916 (aarch64_xtn2<mode>): Likewise.
23917 * config/aarch64/aarch64-simd-builtins.def (xtn2): Define
23919 * config/aarch64/arm_neon.h (vmovn_high_s16): Reimplement using
23921 (vmovn_high_s32): Likewise.
23922 (vmovn_high_s64): Likewise.
23923 (vmovn_high_u16): Likewise.
23924 (vmovn_high_u32): Likewise.
23925 (vmovn_high_u64): Likewise.
23927 2021-01-13 Stafford Horne <shorne@gmail.com>
23929 * config/or1k/or1k.h (ASM_PREFERRED_EH_DATA_FORMAT): New macro.
23931 2021-01-13 Stafford Horne <shorne@gmail.com>
23933 * config/or1k/linux.h (TARGET_ASM_FILE_END): Define macro.
23935 2021-01-13 Stafford Horne <shorne@gmail.com>
23937 * config/or1k/or1k.h (TARGET_CPU_CPP_BUILTINS): Add builtin
23938 define for __or1k_hard_float__.
23940 2021-01-13 Stafford Horne <shorne@gmail.com>
23942 * config/or1k/or1k.h (NO_PROFILE_COUNTERS): Define as 1.
23943 (PROFILE_HOOK): Define to call _mcount.
23944 (FUNCTION_PROFILER): Change from abort to no-op.
23946 2021-01-13 Jakub Jelinek <jakub@redhat.com>
23948 PR tree-optimization/96691
23949 * match.pd ((~X | C) ^ D -> (X | C) ^ (~D ^ C),
23950 (~X & C) ^ D -> (X & C) ^ (D ^ C)): New simplifications if
23951 (~D ^ C) or (D ^ C) can be simplified.
23953 2021-01-13 Richard Biener <rguenther@suse.de>
23955 PR tree-optimization/92645
23956 * match.pd (BIT_FIELD_REF to conversion): Delay canonicalization
23957 until after vector lowering.
23959 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
23961 * config/aarch64/aarch64-sve.md (fnma<mode>4): Extend from SVE_FULL_I
23963 (@aarch64_pred_fnma<mode>, cond_fnma<mode>, *cond_fnma<mode>_2)
23964 (*cond_fnma<mode>_4, *cond_fnma<mode>_any): Likewise.
23966 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
23968 * config/aarch64/aarch64-sve.md (fma<mode>4): Extend from SVE_FULL_I
23970 (@aarch64_pred_fma<mode>, cond_fma<mode>, *cond_fma<mode>_2)
23971 (*cond_fma<mode>_4, *cond_fma<mode>_any): Likewise.
23973 2021-01-13 Richard Biener <rguenther@suse.de>
23975 PR tree-optimization/92645
23976 * tree-vect-slp.c (vect_build_slp_tree_1): Relax supported
23977 BIT_FIELD_REF argument.
23978 (vect_build_slp_tree_2): Record the desired vector type
23979 on the external vector def.
23980 (vectorizable_slp_permutation): Handle required punning
23981 of existing vector defs.
23983 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
23985 * rtl-ssa/accesses.h (def_lookup): Fix order of comparison results.
23987 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
23989 * config/sh/sh.md (movsf_ie): Remove operands[2] test.
23991 2021-01-13 Samuel Thibault <samuel.thibault@ens-lyon.org>
23993 * config.gcc [$target == *-*-gnu*]: Enable
23994 'default_gnu_indirect_function'.
23996 2021-01-13 Jakub Jelinek <jakub@redhat.com>
23999 * optabs.c (expand_vec_perm_const): Don't force v0 and v1 into
24000 registers before calling targetm.vectorize.vec_perm_const, only after
24002 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const): Handle
24003 two argument permutation when one operand is zero vector and only
24004 after that force operands into registers.
24005 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_1): New
24006 define_insn_and_split pattern.
24007 (*avx512bw_zero_extendv32qiv32hi2_1): Likewise.
24008 (*avx512f_zero_extendv16hiv16si2_1): Likewise.
24009 (*avx2_zero_extendv8hiv8si2_1): Likewise.
24010 (*avx512f_zero_extendv8siv8di2_1): Likewise.
24011 (*avx2_zero_extendv4siv4di2_1): Likewise.
24012 * config/mips/mips.c (mips_vectorize_vec_perm_const): Force operands
24014 * config/arm/arm.c (arm_vectorize_vec_perm_const): Likewise.
24015 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Likewise.
24016 * config/ia64/ia64.c (ia64_vectorize_vec_perm_const): Likewise.
24017 * config/aarch64/aarch64.c (aarch64_vectorize_vec_perm_const): Likewise.
24018 * config/rs6000/rs6000.c (rs6000_vectorize_vec_perm_const): Likewise.
24019 * config/gcn/gcn.c (gcn_vectorize_vec_perm_const): Likewise. Use std::swap.
24021 2021-01-13 Martin Liska <mliska@suse.cz>
24023 PR tree-optimization/98455
24024 * gimple-if-to-switch.cc (condition_info::record_phi_mapping):
24025 Record also virtual PHIs.
24026 (pass_if_to_switch::execute): Return TODO_cleanup_cfg only
24029 2021-01-13 Jonathan Wakely <jwakely@redhat.com>
24031 * doc/invoke.texi (C++ Modules): Fix typos.
24033 2021-01-13 Richard Biener <rguenther@suse.de>
24035 PR tree-optimization/98640
24036 * tree-ssa-sccvn.c (visit_nary_op): Do not try to
24037 handle plus or minus from a truncated operand to be
24040 2021-01-13 Jakub Jelinek <jakub@redhat.com>
24043 * config/i386/i386.md (*btr<mode>_1, *btr<mode>_2): New
24044 define_insn_and_split patterns.
24045 (splitter after *btr<mode>_2): New splitter.
24047 2021-01-13 Martin Liska <mliska@suse.cz>
24050 * cgraphunit.c (analyze_functions): Remove dead code.
24052 2021-01-13 Qian Jianhua <qianjh@cn.fujitsu.com>
24054 * config/aarch64/aarch64-cost-tables.h (a64fx_extra_costs): New.
24055 * config/aarch64/aarch64.c (a64fx_addrcost_table): New.
24056 (a64fx_regmove_cost, a64fx_vector_cost): New.
24057 (a64fx_tunings): Use the new added cost tables.
24059 2021-01-13 Jakub Jelinek <jakub@redhat.com>
24062 * config/i386/predicates.md (pmovzx_parallel): New predicate.
24063 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): New
24064 define_insn_and_split pattern.
24065 (*sse4_1_zero_extendv4hiv4si2_3): Likewise.
24066 (*sse4_1_zero_extendv2siv2di2_3): Likewise.
24068 2021-01-13 Julian Brown <julian@codesourcery.com>
24070 * config/gcn/gcn.c (gcn_conditional_register_usage): Remove dead code
24071 to fix v0 register.
24073 2021-01-13 Julian Brown <julian@codesourcery.com>
24075 * config/gcn/gcn.c (gcn_md_reorg): Fix case where EXEC reg is live
24078 2021-01-13 Julian Brown <julian@codesourcery.com>
24080 * config/gcn/gcn-valu.md (recip<mode>2<exec>, recip<mode>2): Use unspec
24081 for reciprocal-approximation instructions.
24082 (div<mode>3): Use fused multiply-accumulate operations for reciprocal
24083 refinement and division result.
24084 * config/gcn/gcn.md (UNSPEC_RCP): New unspec constant.
24086 2021-01-13 Julian Brown <julian@codesourcery.com>
24088 * config/gcn/gcn-valu.md (subdf): Rename to...
24091 2021-01-12 Martin Liska <mliska@suse.cz>
24093 * gcov.c (source_info::debug): Fix printf format for 32-bit hosts.
24095 2021-01-12 Andrea Corallo <andrea.corallo@arm.com>
24097 * function-abi.h: Fix typo.
24099 2021-01-12 Christophe Lyon <christophe.lyon@linaro.org>
24103 * config/arm/arm.h (ARM_HAVE_NEON_V8QI_LDST): New macro.
24104 (ARM_HAVE_NEON_V16QI_LDST, ARM_HAVE_NEON_V4HI_LDST): Likewise.
24105 (ARM_HAVE_NEON_V8HI_LDST, ARM_HAVE_NEON_V2SI_LDST): Likewise.
24106 (ARM_HAVE_NEON_V4SI_LDST, ARM_HAVE_NEON_V4HF_LDST): Likewise.
24107 (ARM_HAVE_NEON_V8HF_LDST, ARM_HAVE_NEON_V4BF_LDST): Likewise.
24108 (ARM_HAVE_NEON_V8BF_LDST, ARM_HAVE_NEON_V2SF_LDST): Likewise.
24109 (ARM_HAVE_NEON_V4SF_LDST, ARM_HAVE_NEON_DI_LDST): Likewise.
24110 (ARM_HAVE_NEON_V2DI_LDST): Likewise.
24111 (ARM_HAVE_V8QI_LDST, ARM_HAVE_V16QI_LDST): Likewise.
24112 (ARM_HAVE_V4HI_LDST, ARM_HAVE_V8HI_LDST): Likewise.
24113 (ARM_HAVE_V2SI_LDST, ARM_HAVE_V4SI_LDST, ARM_HAVE_V4HF_LDST): Likewise.
24114 (ARM_HAVE_V8HF_LDST, ARM_HAVE_V4BF_LDST, ARM_HAVE_V8BF_LDST): Likewise.
24115 (ARM_HAVE_V2SF_LDST, ARM_HAVE_V4SF_LDST, ARM_HAVE_DI_LDST): Likewise.
24116 (ARM_HAVE_V2DI_LDST): Likewise.
24117 * config/arm/mve.md (*movmisalign<mode>_mve_store): New pattern.
24118 (*movmisalign<mode>_mve_load): New pattern.
24119 * config/arm/neon.md (movmisalign<mode>): Move to ...
24120 * config/arm/vec-common.md: ... here.
24122 2021-01-12 Vladimir N. Makarov <vmakarov@redhat.com>
24125 * lra-eliminations.c (eliminate_regs_in_insn): Add transformation
24126 of pattern 'plus (plus (hard reg, const), pseudo)'.
24128 2021-01-12 Richard Biener <rguenther@suse.de>
24130 PR tree-optimization/98550
24131 * tree-vect-slp.c (vect_record_max_nunits): Check whether
24132 the group size is a multiple of the vector element count.
24133 (vect_build_slp_tree_1): When we need to fail because
24134 the vector type choosen causes unrolling do so lazily
24135 without affecting matches only at the end to guide group splitting.
24137 2021-01-12 Martin Liska <mliska@suse.cz>
24140 * optc-save-gen.awk: Compare also n_target_save vars with
24143 2021-01-12 Martin Liska <mliska@suse.cz>
24145 * gcov.c (source_info::debug): New.
24146 (print_usage): Add --debug (-D) option.
24147 (process_args): Likewise.
24148 (generate_results): Call src->debug after
24149 accumulate_line_counts.
24150 (read_graph_file): Properly assign id for EXIT_BLOCK.
24151 * profile.c (branch_prob): Dump function body before it is
24154 2021-01-12 Jakub Jelinek <jakub@redhat.com>
24156 PR tree-optimization/98629
24157 * tree-ssa-math-opts.c (arith_overflow_check_p): Don't update use_stmt
24158 unless returning non-zero.
24160 2021-01-12 Jakub Jelinek <jakub@redhat.com>
24162 PR tree-optimization/95731
24163 * tree-ssa-reassoc.c (optimize_range_tests_cmp_bitwise): Also optimize
24164 x < 0 && y < 0 && z < 0 into (x | y | z) < 0 for signed x, y, z.
24165 (optimize_range_tests): Call optimize_range_tests_cmp_bitwise
24166 only after optimize_range_tests_var_bound.
24168 2021-01-12 Jakub Jelinek <jakub@redhat.com>
24170 * configure.ac: Ensure c/Make-lang.in comes first in @all_lang_makefrags@.
24171 * configure: Regenerated.
24173 2021-01-12 liuhongt <hongtao.liu@intel.com>
24176 * config/i386/i386-builtins.h (BUILTIN_DESC_SWAP_OPERANDS):
24178 * config/i386/i386-expand.c (ix86_expand_sse_comi): Delete
24181 2021-01-12 Alexandre Oliva <oliva@adacore.com>
24183 * ssa-iterators.h (end_imm_use_stmt_traverse): Forward
24185 (auto_end_imm_use_stmt_traverse): New struct.
24186 (FOR_EACH_IMM_USE_STMT): Use it.
24187 (BREAK_FROM_IMM_USE_STMT, RETURN_FROM_IMM_USE_STMT): Remove,
24189 * gimple-ssa-strength-reduction.c: ... here, ...
24190 * graphite-scop-detection.c: ... here, ...
24191 * ipa-modref.c, ipa-pure-const.c, ipa-sra.c: ... here, ...
24192 * tree-predcom.c, tree-ssa-ccp.c: ... here, ...
24193 * tree-ssa-dce.c, tree-ssa-dse.c: ... here, ...
24194 * tree-ssa-loop-ivopts.c, tree-ssa-math-opts.c: ... here, ...
24195 * tree-ssa-phiprop.c, tree-ssa.c: ... here, ...
24196 * tree-vect-slp.c: ... and here, ...
24197 * doc/tree-ssa.texi: ... and the example here.
24199 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24201 * config/aarch64/aarch64-sve.md (sdiv_pow2<mode>3): Extend from
24202 SVE_FULL_I to SVE_I. Generate an UNSPEC_PRED_X.
24203 (*sdiv_pow2<mode>3): New pattern.
24204 (@cond_<sve_int_op><mode>): Extend from SVE_FULL_I to SVE_I.
24205 Wrap the ASRD in an UNSPEC_PRED_X.
24206 (*cond_<sve_int_op><mode>_2): Likewise. Replace the UNSPEC_PRED_X
24207 predicate with a constant PTRUE, if it isn't already.
24208 (*cond_<sve_int_op><mode>_z): Replace with...
24209 (*cond_<sve_int_op><mode>_any): ...this new pattern.
24211 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24213 * config/aarch64/aarch64-sve.md (*cond_bic<mode>_2): Extend from
24214 SVE_FULL_I to SVE_I.
24215 (*cond_bic<mode>_any): Likewise.
24217 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24219 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart)
24220 (@aarch64_pred_<MUL_HIGHPART:optab><mode>): Extend from SVE_FULL_I
24223 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24225 * config/aarch64/aarch64-sve.md (<su>abd<mode>_3): Extend from
24226 SVE_FULL_I to SVE_I.
24227 (*aarch64_cond_<su>abd<mode>_2): Likewise.
24228 (*aarch64_cond_<su>abd<mode>_any): Likewise.
24229 (@aarch64_pred_<su>abd<mode>): Likewise. Use UNSPEC_PRED_X
24230 for the max and min but not for the minus.
24231 (*aarch64_cond_<su>abd<mode>_3): New pattern.
24233 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24235 * config/aarch64/iterators.md (SVE_24I): New iterator.
24236 * config/aarch64/aarch64-sve.md (*aarch64_adr<mode>_shift): Extend from
24237 SVE_FULL_SDI to SVE_24I. Use containers rather than elements.
24239 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24241 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_BINARY:optab><mode>)
24242 (*cond_<SVE_INT_BINARY:optab><mode>_2): Extend from SVE_FULL_I
24244 (*cond_<SVE_INT_BINARY:optab><mode>_3): Likewise.
24245 (*cond_<SVE_INT_BINARY:optab><mode>_any): Likewise.
24246 (*cond_<SVE_INT_BINARY:optab><mode>_2_const): Likewise.
24247 (*cond_<SVE_INT_BINARY:optab><mode>_any_const): Likewise.
24249 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24251 * config/aarch64/aarch64-sve.md (<SVE_INT_BINARY_IMM:optab><mode>3)
24252 (@aarch64_pred_<SVE_INT_BINARY_IMM:optab><mode>)
24253 (*post_ra_<SVE_INT_BINARY_IMM:optab><mode>3): Extend from SVE_FULL_I
24256 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24258 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3)
24259 (v<ASHIFT:optab><mode>3, @aarch64_pred_<optab><mode>)
24260 (*post_ra_v<ASHIFT:optab><mode>3): Extend from SVE_FULL_I to SVE_I.
24262 2021-01-11 Martin Liska <mliska@suse.cz>
24265 * symtab-clones.h (clone_info::release): Release
24266 symtab::m_clones with ggc_delete as it's a GGC memory.
24268 2021-01-11 Matthias Klose <doko@ubuntu.com>
24270 * Makefile.in (LINK_PROGRESS): Show the link target.
24272 2021-01-11 Richard Biener <rguenther@suse.de>
24274 PR tree-optimization/91403
24275 * tree-vect-data-refs.c (vect_analyze_group_access_1): Cap
24276 single-element interleaving group size at 4096 elements.
24278 2021-01-11 Richard Biener <rguenther@suse.de>
24280 PR tree-optimization/98526
24281 * tree-vect-loop.c (vect_model_reduction_cost): Remove costing
24282 of the actual reduction op for the regular case.
24283 (vectorizable_reduction): Cost the stmts
24284 vect_transform_reduction produces here.
24286 2021-01-11 Andreas Krebbel <krebbel@linux.ibm.com>
24288 * tree-ssa-forwprop.c (simplify_vector_constructor): For
24289 big-endian, use UNPACK[_FLOAT]_HI.
24291 2021-01-11 Tamar Christina <tamar.christina@arm.com>
24293 * tree-vect-slp-patterns.c (class complex_pattern,
24294 class complex_add_pattern): Add parameters to matches.
24295 (complex_add_pattern::build): Free memory.
24296 (complex_add_pattern::matches): Move validation end of match.
24297 (complex_add_pattern::recognize): Likewise.
24299 2021-01-11 Tamar Christina <tamar.christina@arm.com>
24301 * tree-vect-slp-patterns.c (linear_loads_p): Fix externals.
24303 2021-01-11 Tamar Christina <tamar.christina@arm.com>
24305 * tree-vect-slp-patterns.c (is_linear_load_p): Fix ambiguity.
24307 2021-01-11 Jakub Jelinek <jakub@redhat.com>
24309 PR tree-optimization/95867
24310 * tree-ssa-math-opts.h: New header.
24311 * tree-ssa-math-opts.c: Include tree-ssa-math-opts.h.
24312 (powi_as_mults): No longer static. Use build_one_cst instead of
24313 build_real. Formatting fix.
24314 * tree-ssa-reassoc.c: Include tree-ssa-math-opts.h.
24315 (attempt_builtin_powi): Handle multiplication reassociation without
24316 powi_fndecl using powi_as_mults.
24317 (reassociate_bb): For integral types don't require
24318 -funsafe-math-optimizations to call attempt_builtin_powi.
24320 2021-01-11 Jakub Jelinek <jakub@redhat.com>
24322 PR tree-optimization/95852
24323 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): Change
24324 mul_stmts parameter type to vec<gimple *> &. Before cond_stmt
24325 allow in the bb any of the stmts in that vector, div_stmt and
24326 up to 3 cast stmts.
24327 (arith_cast_equal_p): New function.
24328 (arith_overflow_check_p): Add cast_stmt argument, handle signed
24329 multiply overflow checks.
24330 (match_arith_overflow): Adjust caller. Handle signed multiply
24333 2021-01-11 Jakub Jelinek <jakub@redhat.com>
24335 PR tree-optimization/95852
24336 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): New function.
24337 (uaddsub_overflow_check_p): Renamed to ...
24338 (arith_overflow_check_p): ... this. Handle also multiplication
24339 with overflow check.
24340 (match_uaddsub_overflow): Renamed to ...
24341 (match_arith_overflow): ... this. Add cfg_changed argument. Handle
24342 also multiplication with overflow check. Adjust function comment.
24343 (math_opts_dom_walker::after_dom_children): Adjust callers. Call
24344 match_arith_overflow also for MULT_EXPR.
24346 2021-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24348 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
24349 __builtin_convertvector.
24350 (vmovl_s16): Likewise.
24351 (vmovl_s32): Likewise.
24352 (vmovl_u8): Likewise.
24353 (vmovl_u16): Likewise.
24354 (vmovl_u32): Likewise.
24355 (vmovn_s16): Likewise.
24356 (vmovn_s32): Likewise.
24357 (vmovn_s64): Likewise.
24358 (vmovn_u16): Likewise.
24359 (vmovn_u32): Likewise.
24360 (vmovn_u64): Likewise.
24362 2021-01-11 Martin Liska <mliska@suse.cz>
24364 * gimple-if-to-switch.cc (struct condition_info): Use auto_var.
24365 (if_chain::is_beneficial): Delete clusters
24366 (find_conditions): Make second argument of conditions_in_bbs a
24367 pointer so that we control over it's lifetime.
24368 (pass_if_to_switch::execute): Delete them.
24370 2021-01-11 Kewen Lin <linkw@linux.ibm.com>
24372 * ira.c (move_unallocated_pseudos): Check other_reg and skip if
24375 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
24377 * config/vax/vax.md (cc): Remove mode attribute.
24378 (subst_<cc>, subst_f<cc>): Rename to...
24379 (subst_<mode>, subst_f<VAXccnz:mode>): ... these respectively.
24380 (*cbranch<VAXint:mode>4_<VAXcc:mode>): Update for `cc' removal.
24381 (*cbranch<VAXfp:mode>4_<VAXccnz:mode>): Likewise.
24382 (*branch_<mode>, *branch_<mode>_reversed): Likewise.
24384 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
24386 * config/vax/vax.md (subst_f<cc>): Add mode to operands and
24387 `const_double_zero'.
24389 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
24391 * config/pdp11/pdp11.md (PDPfp): New mode iterator.
24392 (fcc_cc, fcc_ccnz): Use it. Add mode to `const_double_zero' and
24395 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
24397 * genemit.c (gen_exp) <CONST_DOUBLE>: Handle `const_double_zero'
24399 * read-rtl.c (rtx_reader::read_rtx_code): Handle machine mode
24400 with `const_double_zero'.
24401 * doc/rtl.texi (Constant Expression Types): Document it.
24403 2021-01-09 Jakub Jelinek <jakub@redhat.com>
24406 * tree-cfg.c (verify_gimple_assign_binary): Allow lhs of
24407 POINTER_DIFF_EXPR to be any integral type.
24409 2021-01-09 Jakub Jelinek <jakub@redhat.com>
24411 PR rtl-optimization/98603
24412 * function.c (instantiate_virtual_regs_in_insn): For asm goto
24413 with impossible constraints, drop all SETs, CLOBBERs, drop PARALLEL
24414 if any, set ASM_OPERANDS mode to VOIDmode and change
24415 ASM_OPERANDS_OUTPUT_CONSTRAINT and ASM_OPERANDS_OUTPUT_IDX.
24417 2021-01-09 Alexandre Oliva <oliva@gnu.org>
24420 * final.c (notice_source_line): Narrow down the condition to
24421 skip a line-0 marker.
24423 2021-01-08 Sergei Trofimovich <siarheit@google.com>
24425 * ipa-modref.c (merge_call_side_effects): Fix
24426 linebreak split by reordering two print calls.
24428 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
24430 * config/s390/vector.md (*tf_to_fprx2_0): Rename from
24431 "*mov_tf_to_fprx2_0" for consistency, fix constraint.
24432 (*tf_to_fprx2_1): Rename from "*mov_tf_to_fprx2_1" for
24433 consistency, fix constraint.
24435 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
24437 * config/s390/s390-c.c (s390_def_or_undef_macro): Accept
24438 callables instead of mask values.
24439 (struct target_flag_set_p): New predicate.
24440 (s390_cpu_cpp_builtins_internal): Define or undefine
24441 __LONG_DOUBLE_VX__ macro.
24443 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
24446 * config/i386/i386.c (x86_function_profiler): Use R10 and R11
24447 to call mcount in large model with PIC for NO_PROFILE_COUNTERS
24450 2021-01-08 Richard Biener <rguenther@suse.de>
24452 * tree-ssa-sccvn.c (pass_fre::execute): Reset the SCEV hash table.
24454 2021-01-08 Richard Biener <rguenther@suse.de>
24456 * tree-vect-slp.c (scalar_stmts_to_slp_tree_map_t): Fix.
24457 (vect_build_slp_tree): On cache hit release the matched
24458 scalar stmts vector.
24459 * tree-vect-stmts.c (vectorizable_store): Properly free
24460 vec_oprnds before possibly gathering them again.
24462 2021-01-08 Richard Biener <rguenther@suse.de>
24464 PR tree-optimization/98544
24465 * tree-vect-slp.c (vect_optimize_slp): Always materialize
24466 permutes at a permute node.
24468 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
24471 * config/i386/i386.c (x86_function_profiler): Use R10 to call
24472 mcount in large model. Sorry for large model with PIC.
24474 2021-01-08 Jakub Jelinek <jakub@redhat.com>
24477 * config/i386/i386.opt (ix86_cmodel, ix86_incoming_stack_boundary_arg,
24478 ix86_pmode, ix86_preferred_stack_boundary_arg, ix86_regparm,
24479 ix86_veclibabi_type): Remove x_ prefix, use TargetVariable instead of
24480 TargetSave and initialize for variables with enum types.
24481 (mfentry, mstack-protector-guard-reg=, mstack-protector-guard-offset=,
24482 mstack-protector-guard-symbol=): Add Save.
24483 * config/i386/i386-options.c (ix86_function_specific_save,
24484 ix86_function_specific_restore): Don't save or restore x_ix86_cmodel,
24485 x_ix86_incoming_stack_boundary_arg, x_ix86_pmode,
24486 x_ix86_preferred_stack_boundary_arg, x_ix86_regparm,
24487 x_ix86_veclibabi_type.
24489 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
24491 * config/aarch64/aarch64-sve.md (*cnot<mode>): Extend from
24492 SVE_FULL_I to SVE_I.
24493 (*cond_cnot<mode>_2, *cond_cnot<mode>_any): Likewise.
24495 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
24497 * config/aarch64/aarch64-sve.md (*cond_uxt<mode>_2): Extend from
24498 SVE_FULL_I to SVE_I.
24499 (*cond_uxt<mode>_any): Likewise.
24501 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24503 * config/aarch64/iterators.md (Vwhalf): New iterator.
24504 * config/aarch64/aarch64-simd.md (aarch64_<sur>adalp<mode>_3):
24506 (aarch64_<sur>adalp<mode>): ... This. Make more
24508 (<sur>sadv16qi): Adjust callsite of the above.
24509 * config/aarch64/aarch64-simd-builtins.def (sadalp, uadalp): New
24511 * config/aarch64/arm_neon.h (vpadal_s8): Reimplement using
24513 (vpadal_s16): Likewise.
24514 (vpadal_u8): Likewise.
24515 (vpadal_u16): Likewise.
24516 (vpadalq_s8): Likewise.
24517 (vpadalq_s16): Likewise.
24518 (vpadalq_s32): Likewise.
24519 (vpadalq_u8): Likewise.
24520 (vpadalq_u16): Likewise.
24521 (vpadalq_u32): Likewise.
24523 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24525 * config/aarch64/aarch64-simd.md (aarch64_<su>abd<mode>_3):
24527 (aarch64_<su>abd<mode>): ... This.
24528 (<sur>sadv16qi): Adjust callsite of the above.
24529 * config/aarch64/aarch64-simd-builtins.def (sabd, uabd): Define
24531 * config/aarch64/arm_neon.h (vabd_s8): Reimplement using
24533 (vabd_s16): Likewise.
24534 (vabd_s32): Likewise.
24535 (vabd_u8): Likewise.
24536 (vabd_u16): Likewise.
24537 (vabd_u32): Likewise.
24538 (vabdq_s8): Likewise.
24539 (vabdq_s16): Likewise.
24540 (vabdq_s32): Likewise.
24541 (vabdq_u8): Likewise.
24542 (vabdq_u16): Likewise.
24543 (vabdq_u32): Likewise.
24545 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24547 * config/aarch64/aarch64-simd-builtins.def (saba, uaba): Define
24549 * config/aarch64/arm_neon.h (vaba_s8): Implement using builtin.
24550 (vaba_s16): Likewise.
24551 (vaba_s32): Likewise.
24552 (vaba_u8): Likewise.
24553 (vaba_u16): Likewise.
24554 (vaba_u32): Likewise.
24555 (vabaq_s8): Likewise.
24556 (vabaq_s16): Likewise.
24557 (vabaq_s32): Likewise.
24558 (vabaq_u8): Likewise.
24559 (vabaq_u16): Likewise.
24560 (vabaq_u32): Likewise.
24562 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24564 * config/aarch64/aarch64-simd.md (aba<mode>_3): Rename to...
24565 (aarch64_<su>aba<mode>): ... This. Handle uaba as well.
24566 Change RTL pattern to match.
24568 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
24570 * common/config/riscv/riscv-common.c (riscv_current_subset_list): New.
24571 * config/riscv/riscv-c.c (riscv-subset.h): New.
24572 (INCLUDE_STRING): Define.
24573 (riscv_cpu_cpp_builtins): Add new style architecture extension
24575 * config/riscv/riscv-subset.h (riscv_subset_list::begin): New.
24576 (riscv_subset_list::end): New.
24577 (riscv_current_subset_list): New.
24579 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
24581 * common/config/riscv/riscv-common.c (RISCV_DONT_CARE_VERSION):
24582 Move to riscv-subset.h.
24583 (struct riscv_subset_t): Ditto.
24584 (class riscv_subset_list): Ditto.
24585 * config/riscv/riscv-subset.h (RISCV_DONT_CARE_VERSION): Move
24586 from riscv-common.c.
24587 (struct riscv_subset_t): Ditto.
24588 (class riscv_subset_list): Ditto.
24589 * config/riscv/t-riscv ($(common_out_file)): Add file
24592 2021-01-07 Jakub Jelinek <jakub@redhat.com>
24595 * config/i386/i386.md (*bmi_blsi_<mode>_cmp, *bmi_blsi_<mode>_ccno):
24596 New define_insn patterns.
24598 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
24600 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_UNARY:optab><mode>)
24601 (*cond_<SVE_INT_UNARY:optab><mode>_2): Extend from SVE_FULL_I to SVE_I.
24602 (*cond_<SVE_INT_UNARY:optab><mode>_any): Likewise.
24604 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
24606 PR tree-optimization/98560
24607 * internal-fn.def (IFN_VCONDU, IFN_VCONDEQ): Use type vec_cond.
24608 * internal-fn.c (vec_cond_mask_direct): Get the data mode from
24610 (vec_cond_direct): Likewise argument 2.
24611 (vec_condu_direct, vec_condeq_direct): Delete.
24612 (expand_vect_cond_optab_fn): Rename to...
24613 (expand_vec_cond_optab_fn): ...this, replacing old macro.
24614 (expand_vec_condu_optab_fn, expand_vec_condeq_optab_fn): Delete.
24615 (expand_vect_cond_mask_optab_fn): Rename to...
24616 (expand_vec_cond_mask_optab_fn): ...this, replacing old macro.
24617 (direct_vec_cond_mask_optab_supported_p): Treat the optab as a
24619 (direct_vec_cond_optab_supported_p): Likewise.
24620 (direct_vec_condu_optab_supported_p): Delete.
24621 (direct_vec_condeq_optab_supported_p): Delete.
24622 * gimple-isel.cc: Include internal-fn.h.
24623 (gimple_expand_vec_cond_expr): Check that IFN_VCONDEQ is supported
24626 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
24628 PR tree-optimization/98560
24629 * gimple-isel.cc (gimple_expand_vec_cond_expr): If we fail to use
24630 IFN_VCOND{,U,EQ}, fall back on IFN_VCOND_MASK.
24632 2021-01-07 Uroš Bizjak <ubizjak@gmail.com>
24634 * config/i386/i386.md (insn): Merge from plusminus_insn, shift_insn,
24635 rotate_insn and optab code attributes.
24636 Update all uses to merged code attribute.
24637 * config/i386/sse.md: Update all uses to merged code attribute.
24638 * config/i386/mmx.md: Update all uses to merged code attribute.
24640 2021-01-07 Jakub Jelinek <jakub@redhat.com>
24642 PR tree-optimization/98568
24643 * gimple-ssa-store-merging.c (bswap_view_convert): New function.
24644 (bswap_replace): Use it.
24646 2021-01-06 Vladimir N. Makarov <vmakarov@redhat.com>
24648 PR rtl-optimization/97978
24649 * lra-int.h (lra_hard_reg_split_p): New external.
24650 * lra.c (lra_hard_reg_split_p): New global.
24651 (lra): Set up lra_hard_reg_split_p after splitting a hard reg.
24652 * lra-assigns.c (lra_assign): Don't check allocation correctness
24653 after hard reg splitting.
24655 2021-01-06 Martin Sebor <msebor@redhat.com>
24658 * builtins.c (new_delete_mismatch_p): New overload.
24659 (new_delete_mismatch_p (tree, tree)): Call it.
24661 2021-01-06 Alexandre Oliva <oliva@adacore.com>
24663 * Makefile.in (T_GLIMITS_H): New.
24664 (stmp-int-hdrs): Depend on it, use it.
24665 * config/t-vxworks (T_GLIMITS_H): Override it.
24666 (vxw-glimits.h): New.
24668 2021-01-06 Richard Biener <rguenther@suse.de>
24670 PR tree-optimization/98513
24671 * value-range.cc (intersect_ranges): Compare the upper bounds
24672 for the expected relation.
24674 2021-01-06 Gerald Pfeifer <gerald@pfeifer.com>
24677 2020-12-28 Gerald Pfeifer <gerald@pfeifer.com>
24679 * doc/standards.texi (HSAIL): Remove section.
24681 2021-01-05 Samuel Thibault <samuel.thibault@ens-lyon.org>
24683 * configure: Re-generate.
24685 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24687 * doc/invoke.texi (-std=c++20): Adjust for the publication of
24688 ISO 14882:2020 standard.
24689 * doc/standards.texi: Likewise.
24691 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24693 PR tree-optimization/94802
24694 * expr.h (maybe_optimize_sub_cmp_0): Declare.
24695 * expr.c: Include tree-pretty-print.h and flags.h.
24696 (maybe_optimize_sub_cmp_0): New function.
24697 (do_store_flag): Use it.
24698 * cfgexpand.c (expand_gimple_cond): Likewise.
24700 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
24702 * mux-utils.h (pointer_mux::m_ptr): Tweak description of contents.
24703 * rtlanal.c (simple_regno_set): Tweak description to clarify the
24706 2021-01-05 Richard Biener <rguenther@suse.de>
24708 PR tree-optimization/98516
24709 * tree-vect-slp.c (vect_optimize_slp): Permute the incoming
24710 lanes when materializing on a VEC_PERM node.
24711 (vectorizable_slp_permutation): Dump the permute properly.
24713 2021-01-05 Richard Biener <rguenther@suse.de>
24715 * tree-vect-slp.c (vect_slp_region): Move debug counter
24716 to cover individual subgraphs.
24718 2021-01-05 Richard Biener <rguenther@suse.de>
24720 PR tree-optimization/98428
24721 * tree-vect-slp.c (vect_build_slp_tree_1): Properly reject
24722 vector lane extracts for loop vectorization.
24724 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24726 PR tree-optimization/98514
24727 * tree-ssa-reassoc.c (bb_rank): Change type from long * to
24729 (operand_rank): Change type from hash_map<tree, long> to
24730 hash_map<tree, int64_t>.
24731 (phi_rank): Change return type from long to int64_t.
24732 (loop_carried_phi): Change block_rank variable type from long to
24734 (propagate_rank): Change return type, rank parameter type and
24735 op_rank variable type from long to int64_t.
24736 (find_operand_rank): Change return type from long to int64_t
24737 and change slot variable type from long * to int64_t *.
24738 (insert_operand_rank): Change rank parameter type from long to
24740 (get_rank): Change return type and rank variable type from long to
24741 int64_t. Use PRId64 instead of ld to print the rank.
24742 (init_reassoc): Change rank variable type from long to int64_t
24743 and adjust correspondingly bb_rank and operand_rank initialization.
24745 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24747 PR tree-optimization/96928
24748 * tree-ssa-phiopt.c (xor_replacement): New function.
24749 (tree_ssa_phiopt_worker): Call it.
24751 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24753 PR tree-optimization/96930
24754 * match.pd ((A / (1 << B)) -> (A >> B)): If A is extended
24755 from narrower value which has the same type as 1 << B, perform
24756 the right shift on the narrower value followed by extension.
24758 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24760 PR tree-optimization/96239
24761 * gimple-ssa-store-merging.c (maybe_optimize_vector_constructor): New
24763 (get_status_for_store_merging): Don't return BB_INVALID for blocks
24764 with potential bswap optimizable CONSTRUCTORs.
24765 (pass_store_merging::execute): Optimize vector CONSTRUCTORs with bswap
24768 2021-01-05 Richard Biener <rguenther@suse.de>
24770 PR tree-optimization/98381
24771 * tree.c (vector_element_bits): Properly compute bool vector
24773 * tree-vect-loop.c (vectorizable_live_operation): Properly
24774 compute the last lane bit offset.
24776 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
24779 * config/i386/sse.md (sse_cvtps2pi): Redefine as define_insn_and_split.
24780 Clear the top 64 bytes of the input XMM register.
24781 (sse_cvttps2pi): Ditto.
24783 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
24786 * config/i386/xopintrin.h (_mm256_cmov_si256): New.
24788 2021-01-05 H.J. Lu <hjl.tools@gmail.com>
24791 * config/i386/xmmintrin.h (_mm_extract_pi16): Cast to unsigned
24794 2021-01-05 Claudiu Zissulescu <claziss@synopsys.com>
24796 * config/arc/arc.md (maddsidi4_split): Use ACC_REG_FIRST.
24797 (umaddsidi4_split): Likewise.
24799 2021-01-05 liuhongt <hongtao.liu@intel.com>
24802 * config/i386/sse.md (*sse2_pmovskb_zexthisi): New
24803 define_insn_and_split for zero_extend of subreg HI of pmovskb
24805 (*sse2_pmovskb_zexthisi): Add new combine splitters for
24806 zero_extend of not of subreg HI of pmovskb result.
24808 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
24811 * explow.c (convert_memory_address_addr_space_1): Handle UNSPECs
24813 * config/aarch64/aarch64.c (aarch64_expand_mov_immediate): Use
24814 convert_memory_address to convert symbolic immediates to ptr_mode
24815 before forcing them to memory.
24817 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
24819 PR rtl-optimization/97144
24820 * recog.c (constrain_operands): Initialize matching_operand
24821 for each alternative, rather than only doing it once.
24823 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
24825 PR rtl-optimization/98403
24826 * rtl-ssa/changes.cc (function_info::finalize_new_accesses): Explain
24827 why we don't remove call clobbers.
24828 (function_info::apply_changes_to_insn): Don't attempt to add
24829 call clobbers here.
24831 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
24833 PR tree-optimization/98371
24834 * tree-vect-loop.c (vect_reanalyze_as_main_loop): New function.
24835 (vect_analyze_loop): If an epilogue loop appears to be cheaper
24836 than the main loop, re-analyze it as a main loop before adopting
24839 2021-01-05 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
24842 * configure.ac (NETLIBS): Determine using AX_LIB_SOCKET_NSL.
24843 * aclocal.m4, configure: Regenerate.
24844 * Makefile.in (NETLIBS): Define.
24845 (BACKEND): Remove $(CODYLIB).
24847 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24849 PR rtl-optimization/98334
24850 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
24851 Optimize (X - 1) * Y + Y to X * Y or (X + 1) * Y - Y to X * Y.
24853 2021-01-05 Bernd Edlinger <bernd.edlinger@hotmail.de>
24855 * tree-inline.c (expand_call_inline): Restore input_location.
24856 Return result from recursive call.
24858 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
24860 PR tree-optimization/95401
24861 * config/aarch64/aarch64-sve-builtins.cc
24862 (gimple_folder::load_store_cookie): Use bits rather than bytes
24863 for the alignment argument to IFN_MASK_LOAD and IFN_MASK_STORE.
24864 * gimple-fold.c (gimple_fold_mask_load_store_mem_ref): Likewise.
24865 * tree-vect-stmts.c (vectorizable_store): Likewise.
24866 (vectorizable_load): Likewise.
24868 2021-01-04 Richard Biener <rguenther@suse.de>
24870 PR tree-optimization/98308
24871 * tree-vect-stmts.c (vectorizable_load): Set invariant mask
24874 2021-01-04 Jakub Jelinek <jakub@redhat.com>
24876 PR tree-optimization/95771
24877 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Handle types
24878 with precision smaller than int's precision and types with precision
24879 twice as large as long long. Formatting fixes.
24881 2021-01-04 Richard Biener <rguenther@suse.de>
24883 PR tree-optimization/98464
24884 * tree-ssa-sccvn.c (vn_valueize_for_srt): Rename from ...
24885 (vn_valueize_wrapper): ... this. Temporarily adjust vn_context_bb.
24886 (process_bb): Adjust.
24888 2021-01-04 Matthew Malcomson <matthew.malcomson@arm.com>
24891 * doc/invoke.texi (-fsanitize=address): Fix wording describing
24892 clash with -fsanitize=hwaddress.
24894 2021-01-04 Richard Biener <rguenther@suse.de>
24896 PR tree-optimization/98282
24897 * tree-ssa-sccvn.c (vn_get_stmt_kind): Classify tcc_reference on
24898 invariants as VN_NARY.
24900 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
24903 * config/aarch64/aarch64-simd.md (aarch64_combine<mode>): Accept
24904 aarch64_simd_reg_or_zero for operand 2. Use the combinez patterns
24905 to handle zero operands.
24907 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
24909 * config/aarch64/aarch64.c (offset_6bit_signed_scaled_p): New function.
24910 (offset_6bit_unsigned_scaled_p): Fix typo in comment.
24911 (aarch64_sve_prefetch_operand_p): Accept MUL VLs in the range
24914 2021-01-04 Richard Biener <rguenther@suse.de>
24916 PR tree-optimization/98393
24917 * tree-vect-slp.c (vect_build_slp_tree): Properly zero matches
24918 when hitting the limit.
24920 2021-01-04 Richard Biener <rguenther@suse.de>
24922 PR tree-optimization/98291
24923 * tree-vect-loop.c (vectorizable_reduction): Bypass
24924 associativity check for SLP reductions with VF 1.
24926 2021-01-04 Jakub Jelinek <jakub@redhat.com>
24928 PR tree-optimization/96782
24929 * match.pd (x == ~x -> false, x != ~x -> true): New simplifications.
24931 2021-01-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
24933 * collect-utils.c (collect_execute): Check dumppfx.
24934 * collect2.c (maybe_run_lto_and_relink, do_link): Pass atsuffix
24935 to collect_execute.
24936 (do_link): Add new parameter atsuffix.
24937 (main): Handle -dumpdir option. Skip one argument for
24938 -o, -isystem and -B options.
24939 * gcc.c (make_at_file): New helper function.
24940 (close_at_file): Use it.
24942 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
24944 * config/darwin.h (MIN_LD64_NO_COAL_SECTS): Adjust.
24945 Amend handling for LD64_VERSION fallback defaults.
24947 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
24949 * config.gcc: Compute default version information
24950 from the configured target. Likewise defaults for
24952 * config/darwin10.h: Removed.
24953 * config/darwin12.h: Removed.
24954 * config/darwin9.h: Removed.
24955 * config/rs6000/darwin8.h: Removed.
24957 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
24959 * config/darwin9.h (ASM_OUTPUT_ALIGNED_COMMON): Delete.
24961 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
24963 * config/darwin9.h (STACK_CHECK_STATIC_BUILTIN): Move from here..
24964 * config/darwin.h (STACK_CHECK_STATIC_BUILTIN): .. to here.
24966 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
24968 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move from
24970 * config/darwin.h (LINK_GCC_C_SEQUENCE_SPEC): ... to here.
24972 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
24974 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move the spec
24975 for the Darwin10 unwinder stub from here ...
24976 * config/darwin.h (LINK_COMMAND_SPEC_A): ... to here.
24978 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
24980 * config/darwin.h (DSYMUTIL_SPEC): Default to DWARF
24981 (ASM_DEBUG_SPEC):Only define if the assembler supports
24983 (PREFERRED_DEBUGGING_TYPE): Default to DWARF.
24984 (DARWIN_PREFER_DWARF): Define.
24985 * config/darwin9.h (PREFERRED_DEBUGGING_TYPE): Remove.
24986 (DARWIN_PREFER_DWARF): Likewise
24987 (DSYMUTIL_SPEC): Likewise.
24988 (COLLECT_RUN_DSYMUTIL): Likewise.
24989 (ASM_DEBUG_SPEC): Likewise.
24990 (ASM_DEBUG_OPTION_SPEC): Likewise.
24992 2021-01-02 Jan Hubicka <jh@suse.cz>
24994 * cfg.c (free_block): ggc_free bb.
24996 2021-01-01 Jakub Jelinek <jakub@redhat.com>
24998 * gcc.c (process_command): Update copyright notice dates.
24999 * gcov-dump.c (print_version): Ditto.
25000 * gcov.c (print_version): Ditto.
25001 * gcov-tool.c (print_version): Ditto.
25002 * gengtype.c (create_file): Ditto.
25003 * doc/cpp.texi: Bump @copying's copyright year.
25004 * doc/cppinternals.texi: Ditto.
25005 * doc/gcc.texi: Ditto.
25006 * doc/gccint.texi: Ditto.
25007 * doc/gcov.texi: Ditto.
25008 * doc/install.texi: Ditto.
25009 * doc/invoke.texi: Ditto.
25011 2021-01-01 Jakub Jelinek <jakub@redhat.com>
25013 * ChangeLog-2020: Rotate ChangeLog. New file.
25016 Copyright (C) 2021 Free Software Foundation, Inc.
25018 Copying and distribution of this file, with or without modification,
25019 are permitted in any medium without royalty provided the copyright
25020 notice and this notice are preserved.