1 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
3 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
4 without a DECL in .data.rel.ro.local.
6 2020-01-30 Jakub Jelinek <jakub@redhat.com>
9 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
14 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
15 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
16 any_extend code iterator instead of always zero_extend.
17 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
18 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
19 Use any_extend code iterator instead of always zero_extend.
20 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
21 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
22 Use any_extend code iterator instead of always zero_extend.
23 (*sse2_pmovmskb_ext): New define_insn.
24 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
27 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
28 (*popcountsi2_zext_falsedep): New define_insn.
30 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
32 * config.in: Regenerated.
33 * configure: Regenerated.
35 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
38 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
39 LLVM's assembler changed the default in version 9.
41 2020-01-24 Jeff Law <law@redhat.com>
43 PR tree-optimization/89689
44 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
46 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
50 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
52 PR rtl-optimization/87763
53 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
54 simplification to handle subregs as well as bare regs.
55 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
57 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
60 * ira.c (ira): Revert use of simplified LRA algorithm.
62 2020-01-29 Martin Jambor <mjambor@suse.cz>
64 PR tree-optimization/92706
65 * tree-sra.c (struct access): Fields first_link, last_link,
66 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
67 next_rhs_queued and grp_rhs_queued respectively, new fields
68 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
69 (struct assign_link): Field next renamed to next_rhs, new field
70 next_lhs. Updated comment.
71 (work_queue_head): Renamed to rhs_work_queue_head.
72 (lhs_work_queue_head): New variable.
73 (add_link_to_lhs): New function.
74 (relink_to_new_repr): Also relink LHS lists.
75 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
76 (add_access_to_lhs_work_queue): New function.
77 (pop_access_from_work_queue): Renamed to
78 pop_access_from_rhs_work_queue.
79 (pop_access_from_lhs_work_queue): New function.
80 (build_accesses_from_assign): Also add links to LHS lists and to LHS
82 (child_would_conflict_in_lacc): Renamed to
83 child_would_conflict_in_acc. Adjusted parameter names.
84 (create_artificial_child_access): New parameter set_grp_read, use it.
85 (subtree_mark_written_and_enqueue): Renamed to
86 subtree_mark_written_and_rhs_enqueue.
87 (propagate_subaccesses_across_link): Renamed to
88 propagate_subaccesses_from_rhs.
89 (propagate_subaccesses_from_lhs): New function.
90 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
93 2020-01-29 Martin Jambor <mjambor@suse.cz>
95 PR tree-optimization/92706
96 * tree-sra.c (struct access): Adjust comment of
97 grp_total_scalarization.
98 (find_access_in_subtree): Look for single children spanning an entire
100 (scalarizable_type_p): Allow register accesses, adjust callers.
101 (completely_scalarize): Remove function.
102 (scalarize_elem): Likewise.
103 (create_total_scalarization_access): Likewise.
104 (sort_and_splice_var_accesses): Do not track total scalarization
106 (analyze_access_subtree): New parameter totally, adjust to new meaning
107 of grp_total_scalarization.
108 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
109 (can_totally_scalarize_forest_p): New function.
110 (create_total_scalarization_access): Likewise.
111 (create_total_access_and_reshape): Likewise.
112 (total_should_skip_creating_access): Likewise.
113 (totally_scalarize_subtree): Likewise.
114 (analyze_all_variable_accesses): Perform total scalarization after
115 subaccess propagation using the new functions above.
116 (initialize_constant_pool_replacements): Output initializers by
117 traversing the access tree.
119 2020-01-29 Martin Jambor <mjambor@suse.cz>
121 * tree-sra.c (verify_sra_access_forest): New function.
122 (verify_all_sra_access_forests): Likewise.
123 (create_artificial_child_access): Set parent.
124 (analyze_all_variable_accesses): Call the verifier.
126 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
128 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
129 if called on indirect edge.
130 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
131 speculative call if needed.
133 2020-01-29 Richard Biener <rguenther@suse.de>
135 PR tree-optimization/93428
136 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
137 permutation when the load node is created.
138 (vect_analyze_slp_instance): Re-use it here.
140 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
142 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
144 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
146 PR rtl-optimization/93272
147 * ira-lives.c (process_out_of_region_eh_regs): New function.
148 (process_bb_node_lives): Call it.
150 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
152 * coverage.c (read_counts_file): Make error message lowercase.
154 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
156 * profile-count.c (profile_quality_display_names): Fix ordering.
158 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
161 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
162 hash only when edge is first within the sequence.
163 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
164 (symbol_table::create_edge): Do not set target_prob.
165 (cgraph_edge::remove_caller): Watch for speculative calls when updating
167 (cgraph_edge::make_speculative): Drop target_prob parameter.
168 (cgraph_edge::speculative_call_info): Remove.
169 (cgraph_edge::first_speculative_call_target): New member function.
170 (update_call_stmt_hash_for_removing_direct_edge): New function.
171 (cgraph_edge::resolve_speculation): Rewrite to new API.
172 (cgraph_edge::speculative_call_for_target): New member function.
173 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
174 multiple speculation targets.
175 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
177 (verify_speculative_call): Verify that targets form an interval.
178 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
179 (cgraph_edge::first_speculative_call_target): New member function.
180 (cgraph_edge::next_speculative_call_target): New member function.
181 (cgraph_edge::speculative_call_target_ref): New member function.
182 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
183 (cgraph_edge): Remove target_prob.
184 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
185 Fix handling of speculative calls.
186 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
187 * ipa-fnsummary.c (analyze_function_body): Likewise.
188 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
189 * ipa-profile.c (dump_histogram): Fix formating.
190 (ipa_profile_generate_summary): Watch for overflows.
191 (ipa_profile): Do not require probablity to be 1/2; update to new API.
192 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
193 (update_indirect_edges_after_inlining): Update to new API.
194 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
196 * profile-count.h: (profile_probability::adjusted): New.
197 * tree-inline.c (copy_bb): Update to new speculative call API; fix
199 * value-prof.c (gimple_ic_transform): Rename to ...
200 (dump_ic_profile): ... this one; update dumping.
201 (stream_in_histogram_value): Fix formating.
202 (gimple_value_profile_transformations): Update.
204 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
207 * config/i386/i386.md (*movoi_internal_avx): Remove
208 TARGET_SSE_TYPELESS_STORES check.
209 (*movti_internal): Prefer TARGET_AVX over
210 TARGET_SSE_TYPELESS_STORES.
211 (*movtf_internal): Likewise.
212 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
213 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
214 from TARGET_SSE_TYPELESS_STORES.
216 2020-01-28 David Malcolm <dmalcolm@redhat.com>
218 * diagnostic-core.h (warning_at): Rename overload to...
219 (warning_meta): ...this.
220 (emit_diagnostic_valist): Delete decl of overload taking
222 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
223 (warning_at): Rename overload taking diagnostic_metadata to...
224 (warning_meta): ...this.
226 2020-01-28 Richard Biener <rguenther@suse.de>
228 PR tree-optimization/93439
229 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
230 * tree-cfg.c (move_sese_region_to_fn): ... here.
231 (verify_types_in_gimple_reference): Verify used cliques are
234 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
237 * config/i386/i386-options.c (set_ix86_tune_features): Add an
238 argument of a pointer to struct gcc_options and pass it to
239 parse_mtune_ctrl_str.
240 (ix86_function_specific_restore): Pass opts to
241 set_ix86_tune_features.
242 (ix86_option_override_internal): Likewise.
243 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
244 gcc_options and use it for x_ix86_tune_ctrl_string.
246 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
248 PR rtl-optimization/87763
249 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
250 simplification to handle subregs as well as bare regs.
251 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
253 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
255 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
256 for reduction chains that (now) include a call.
258 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
260 PR tree-optimization/92822
261 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
262 out the don't-care elements of a vector whose significant elements
263 are duplicates, make the don't-care elements duplicates too.
265 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
267 PR tree-optimization/93434
268 * tree-predcom.c (split_data_refs_to_components): Record which
269 components have had aliasing loads removed. Prevent store-store
270 commoning for all such components.
272 2020-01-28 Jakub Jelinek <jakub@redhat.com>
275 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
276 -1 or is_vshift is true, use new_vector with number of elts npatterns
277 rather than new_unary_operation.
279 PR tree-optimization/93454
280 * gimple-fold.c (fold_array_ctor_reference): Perform
281 elt_size.to_uhwi () just once, instead of calling it in every
282 iteration. Punt if that value is above size of the temporary
283 buffer. Decrease third native_encode_expr argument when
284 bufoff + elt_sz is above size of buf.
286 2020-01-27 Joseph Myers <joseph@codesourcery.com>
288 * config/mips/mips.c (mips_declare_object_name)
289 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
291 2020-01-27 Martin Liska <mliska@suse.cz>
293 PR gcov-profile/93403
294 * tree-profile.c (gimple_init_gcov_profiler): Generate
295 both __gcov_indirect_call_profiler_v4 and
296 __gcov_indirect_call_profiler_v4_atomic.
298 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
301 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
303 (@aarch64_split_simd_mov<mode>): Use it.
304 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
305 Leave the vec_extract patterns to handle 2-element vectors.
306 (aarch64_simd_mov_from_<mode>high): Likewise.
307 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
308 (vec_extractv2dfv1df): Likewise.
310 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
312 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
313 jump conditions for *compare_condjump<GPI:mode>.
315 2020-01-27 David Malcolm <dmalcolm@redhat.com>
318 * digraph.cc (test_edge::test_edge): Specify template for base
321 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
323 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
325 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
327 * config/arc/arc-protos.h (gen_mlo): Remove.
329 * config/arc/arc.c (AUX_MULHI): Define.
330 (arc_must_save_reister): Special handling for r58/59.
331 (arc_compute_frame_size): Consider mlo/mhi registers.
332 (arc_save_callee_saves): Emit fp/sp move only when emit_move
334 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
335 mlo/mhi name selection.
336 (arc_restore_callee_saves): Don't early restore blink when ISR.
337 (arc_expand_prologue): Add mlo/mhi saving.
338 (arc_expand_epilogue): Add mlo/mhi restoring.
341 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
342 numbering when MUL64 option is used.
343 (DWARF2_FRAME_REG_OUT): Define.
344 * config/arc/arc.md (arc600_stall): New pattern.
345 (VUNSPEC_ARC_ARC600_STALL): Define.
346 (mulsi64): Use correct mlo/mhi registers.
347 (mulsi_600): Clean it up.
348 * config/arc/predicates.md (mlo_operand): Remove any dependency on
350 (mhi_operand): Likewise.
352 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
353 Petro Karashchenko <petro.karashchenko@ring.com>
355 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
356 attributes if needed.
357 (prepare_move_operands): Generate special unspec instruction for
359 (arc_isuncached_mem_p): Propagate uncached attribute to each
361 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
362 (VUNSPEC_ARC_STDI): Likewise.
363 (ALLI): New mode iterator.
364 (mALLI): New mode attribute.
365 (lddi): New instruction pattern.
367 (stdidi_split): Split instruction for architectures which are not
368 supporting ll64 option.
369 (lddidi_split): Likewise.
371 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
373 PR rtl-optimization/92989
374 * lra-lives.c (process_bb_lives): Update the live-in set before
375 processing additional clobbers.
377 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
379 PR rtl-optimization/93170
380 * cselib.c (cselib_invalidate_regno_val): New function, split out
382 (cselib_invalidate_regno): ...here.
383 (cselib_invalidated_by_call_p): New function.
384 (cselib_process_insn): Iterate over all the hard-register entries in
385 REG_VALUES and invalidate any that cross call-clobbered registers.
387 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
389 * dojump.c (split_comparison): Use HONOR_NANS rather than
390 HONOR_SNANS when splitting LTGT.
392 2020-01-27 Martin Liska <mliska@suse.cz>
395 * opts.c (print_filtered_help): Exclude language-specific
396 options from --help=common unless enabled in all FEs.
398 2020-01-27 Martin Liska <mliska@suse.cz>
400 * opts.c (print_help): Exclude params from
401 all except --help=param.
403 2020-01-27 Martin Liska <mliska@suse.cz>
406 * config/i386/i386-features.c (make_resolver_func):
407 Align the code with ppc64 target implementation.
408 Do not generate a unique name for resolver function.
410 2020-01-27 Richard Biener <rguenther@suse.de>
412 PR tree-optimization/93397
413 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
414 converted reduction chain SLP graph adjustment.
416 2020-01-26 Marek Polacek <polacek@redhat.com>
419 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
422 2020-01-26 Jason Merrill <jason@redhat.com>
425 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
428 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
430 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
431 (rx_setmem): Likewise.
433 2020-01-26 Jakub Jelinek <jakub@redhat.com>
436 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
437 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
438 drop <di> from constraint of last operand.
441 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
442 TARGET_AVX2 and V4DFmode not in the split condition, but in the
443 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
445 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
448 * ipa-cp.c (get_info_about_necessary_edges): Remove value
451 2020-01-24 Jeff Law <law@redhat.com>
453 PR tree-optimization/92788
454 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
457 2020-01-24 Jakub Jelinek <jakub@redhat.com>
460 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
461 *avx_vperm_broadcast_<mode>,
462 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
463 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
464 Move before avx2_perm<mode>/avx512f_perm<mode>.
467 * simplify-rtx.c (simplify_const_unary_operation,
468 simplify_const_binary_operation): Punt for mode precision above
469 MAX_BITSIZE_MODE_ANY_INT.
471 2020-01-24 Andrew Pinski <apinski@marvell.com>
473 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
476 2020-01-24 Jeff Law <law@redhat.com>
479 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
480 for REGs. Call output_operand_lossage to get more reasonable
483 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
485 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
486 gcn_fp_compare_operator.
487 (vec_cmpu<mode>di): Use gcn_compare_operator.
488 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
489 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
490 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
491 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
492 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
493 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
494 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
495 gcn_fp_compare_operator.
496 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
497 gcn_fp_compare_operator.
498 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
499 gcn_fp_compare_operator.
500 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
501 gcn_fp_compare_operator.
503 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
505 * doc/install.texi (Cross-Compiler-Specific Options): Document
506 `--with-toolexeclibdir' option.
508 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
510 * target.def (flags_regnum): Also mention effect on delay slot filling.
511 * doc/tm.texi: Regenerate.
513 2020-01-23 Jeff Law <law@redhat.com>
516 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
518 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
521 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
524 2020-01-23 Jakub Jelinek <jakub@redhat.com>
526 PR rtl-optimization/93402
527 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
530 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
532 * config.in: Regenerated.
533 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
534 for TARGET_LIBC_GNUSTACK.
535 * configure: Regenerated.
536 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
537 found to be 2.31 or greater.
539 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
541 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
543 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
544 (mips_asm_file_end): New function. Delegate to
545 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
546 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
548 2020-01-23 Jakub Jelinek <jakub@redhat.com>
551 * config/i386/i386-modes.def (POImode): New mode.
552 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
553 * config/i386/i386.md (DPWI): New mode attribute.
554 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
556 (QPWI): ... this. Use POI instead of OI for TImode.
557 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
558 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
561 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
564 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
566 (speculation_tracker_rev): New pattern.
567 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
568 Use speculation_tracker_rev to track the inverse condition.
570 2020-01-23 Richard Biener <rguenther@suse.de>
572 PR tree-optimization/93381
573 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
574 alias-set of the def as argument and record the first one.
575 (vn_walk_cb_data::first_set): New member.
576 (vn_reference_lookup_3): Pass the alias-set of the current def
577 to push_partial_def. Fix alias-set used in the aggregate copy
579 (vn_reference_lookup): Consistently set *last_vuse_ptr.
580 * real.c (clear_significand_below): Fix out-of-bound access.
582 2020-01-23 Jakub Jelinek <jakub@redhat.com>
585 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
586 New define_insn patterns.
588 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
590 * doc/sourcebuild.texi (check-function-bodies): Add an
591 optional target/xfail selector.
593 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
595 PR rtl-optimization/93124
596 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
597 bare USE and CLOBBER insns.
599 2020-01-22 Andrew Pinski <apinski@marvell.com>
601 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
603 2020-01-22 David Malcolm <dmalcolm@redhat.com>
606 * gdbinit.in (break-on-saved-diagnostic): Update for move of
607 diagnostic_manager into "ana" namespace.
608 * selftest-run-tests.c (selftest::run_tests): Update for move of
609 selftest::run_analyzer_selftests to
610 ana::selftest::run_analyzer_selftests.
612 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
614 * cfgexpand.c (union_stack_vars): Update the size.
616 2020-01-22 Richard Biener <rguenther@suse.de>
618 PR tree-optimization/93381
619 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
620 throughout, handle all conversions the same.
622 2020-01-22 Jakub Jelinek <jakub@redhat.com>
625 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
626 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
627 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
628 Call force_reg on high_in2 unconditionally.
630 2020-01-22 Martin Liska <mliska@suse.cz>
632 PR tree-optimization/92924
633 * profile.c (compute_value_histograms): Divide
636 2020-01-22 Jakub Jelinek <jakub@redhat.com>
639 * output.h (assemble_name_resolve): Declare.
640 * varasm.c (assemble_name_resolve): New function.
641 (assemble_name): Use it.
642 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
644 2020-01-22 Joseph Myers <joseph@codesourcery.com>
646 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
647 update_web_docs_git instead of update_web_docs_svn.
649 2020-01-21 Andrew Pinski <apinski@marvell.com>
652 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
653 as PTR mode. Have operand 1 as being modeless, it can be P mode.
654 (*tlsgd_small_<mode>): Likewise.
655 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
656 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
657 register. Convert that register back to dest using convert_mode.
659 2020-01-21 Jim Wilson <jimw@sifive.com>
661 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
664 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
665 Uros Bizjak <ubizjak@gmail.com>
668 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
670 (legitimize_tls_address): Do GNU2 TLS address computation in
671 ptr_mode and zero-extend result to Pmode.
672 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
673 :P with :PTR and Pmode with ptr_mode.
674 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
675 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
676 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
678 2020-01-21 Jakub Jelinek <jakub@redhat.com>
681 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
682 the last two operands are CONST_INT_P before using them as such.
684 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
686 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
687 to get the integer element types.
689 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
691 * config/aarch64/aarch64-sve-builtins.h
692 (function_expander::convert_to_pmode): Declare.
693 * config/aarch64/aarch64-sve-builtins.cc
694 (function_expander::convert_to_pmode): New function.
695 (function_expander::get_contiguous_base): Use it.
696 (function_expander::prepare_gather_address_operands): Likewise.
697 * config/aarch64/aarch64-sve-builtins-sve2.cc
698 (svwhilerw_svwhilewr_impl::expand): Likewise.
700 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
703 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
704 cfun->machine->label_is_assembled.
705 (aarch64_print_patchable_function_entry): New.
706 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
707 * config/aarch64/aarch64.h (struct machine_function): New field,
710 2020-01-21 David Malcolm <dmalcolm@redhat.com>
713 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
716 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
719 * cgraph.c (cgraph_edge::resolve_speculation,
720 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
723 2020-01-21 Martin Liska <mliska@suse.cz>
725 * config/rs6000/rs6000.c (common_mode_defined): Remove
728 2020-01-21 Richard Biener <rguenther@suse.de>
730 PR tree-optimization/92328
731 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
732 type when value-numbering same-sized store by inserting a
734 (eliminate_dom_walker::eliminate_stmt): When eliminating
735 a redundant store handle bit-reinterpretation of the same value.
737 2020-01-21 Andrew Pinski <apinski@marvel.com>
740 * tree-into-ssa.c (prepare_block_for_update_1): Split out
742 (prepare_block_for_update): This. Use a worklist instead of
745 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
747 * gcc/config/arm/arm.c (clear_operation_p):
748 Initialise last_regno, skip first iteration
749 based on the first_set value and use ints instead
750 of the unnecessary HOST_WIDE_INTs.
752 2020-01-21 Jakub Jelinek <jakub@redhat.com>
755 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
756 compare_mode other than SFmode or DFmode.
758 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
761 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
762 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
763 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
765 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
767 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
769 2020-01-20 Andrew Pinski <apinski@marvell.com>
772 * targhooks.c (default_print_patchable_function_entry): Use
773 output_asm_insn to emit the nop instruction.
775 2020-01-20 Fangrui Song <maskray@google.com>
778 * targhooks.c (default_print_patchable_function_entry): Align to
781 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
784 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
785 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
786 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
787 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
788 (*tls_dynamic_gnu2_lea_64): Renamed to ...
789 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
790 Remove the {q} suffix from lea.
791 (*tls_dynamic_gnu2_call_64): Renamed to ...
792 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
793 (*tls_dynamic_gnu2_combine_64): Renamed to ...
794 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
795 Pass Pmode to gen_tls_dynamic_gnu2_64.
797 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
799 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
801 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
803 * config/aarch64/aarch64-sve-builtins-base.cc
804 (svld1ro_impl::memory_vector_mode): Remove parameter name.
806 2020-01-20 Richard Biener <rguenther@suse.de>
809 * dwarf2out.c (prune_unused_types): Unconditionally mark
810 called function DIEs.
812 2020-01-20 Martin Liska <mliska@suse.cz>
814 PR tree-optimization/93199
815 * tree-eh.c (struct leh_state): Add
816 new field outer_non_cleanup.
817 (cleanup_is_dead_in): Pass leh_state instead
818 of eh_region. Add a checking that state->outer_non_cleanup
819 points to outer non-clean up region.
820 (lower_try_finally): Record outer_non_cleanup
822 (lower_catch): Likewise.
823 (lower_eh_filter): Likewise.
824 (lower_eh_must_not_throw): Likewise.
825 (lower_cleanup): Likewise.
827 2020-01-20 Richard Biener <rguenther@suse.de>
829 PR tree-optimization/93094
830 * tree-vectorizer.h (vect_loop_versioning): Adjust.
831 (vect_transform_loop): Likewise.
832 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
833 loop_vectorized_call to vect_transform_loop.
834 * tree-vect-loop.c (vect_transform_loop): Pass down
835 loop_vectorized_call to vect_loop_versioning.
836 * tree-vect-loop-manip.c (vect_loop_versioning): Use
837 the earlier discovered loop_vectorized_call.
839 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
841 * doc/contribute.texi: Update for SVN -> Git transition.
842 * doc/install.texi: Likewise.
844 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
847 * cgraph.c (cgraph_edge::make_speculative): Increase number of
849 (verify_speculative_call): New function
850 (cgraph_node::verify_node): Use it.
851 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
854 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
857 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
858 (cgraph_edge::make_direct): Remove all indirect targets.
859 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
860 (cgraph_node::verify_node): Verify that only one call_stmt or
862 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
864 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
865 (lto_output_ref): Simplify streaming of stmt.
866 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
868 2020-01-18 Tamar Christina <tamar.christina@arm.com>
870 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
871 Mark parameter unused.
873 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
875 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
877 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
879 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
881 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
883 * Makefile.in: Add coroutine-passes.o.
884 * builtin-types.def (BT_CONST_SIZE): New.
885 (BT_FN_BOOL_PTR): New.
886 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
887 * builtins.def (DEF_COROUTINE_BUILTIN): New.
888 * coroutine-builtins.def: New file.
889 * coroutine-passes.cc: New file.
890 * function.h (struct GTY function): Add a bit to indicate that the
891 function is a coroutine component.
892 * internal-fn.c (expand_CO_FRAME): New.
893 (expand_CO_YIELD): New.
894 (expand_CO_SUSPN): New.
895 (expand_CO_ACTOR): New.
896 * internal-fn.def (CO_ACTOR): New.
900 * passes.def: Add pass_coroutine_lower_builtins,
901 pass_coroutine_early_expand_ifns.
902 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
903 (make_pass_coroutine_early_expand_ifns): New.
904 * doc/invoke.texi: Document the fcoroutines command line
907 2020-01-18 Jakub Jelinek <jakub@redhat.com>
909 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
912 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
913 after checking the argument is a REG. Don't use REGNO (reg)
914 again to set last_regno, reuse regno variable instead.
916 2020-01-17 David Malcolm <dmalcolm@redhat.com>
918 * doc/analyzer.texi (Limitations): Add note about NaN.
920 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
921 Sudakshina Das <sudi.das@arm.com>
923 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
925 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
926 (lshrdi3): Generate thumb2_lsrl for valid immediates.
927 * config/arm/constraints.md (Pg): New.
928 * config/arm/predicates.md (long_shift_imm): New.
929 (arm_reg_or_long_shift_imm): Likewise.
930 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
931 (thumb2_lsll): Likewise.
934 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
935 Sudakshina Das <sudi.das@arm.com>
937 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
938 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
939 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
940 register pairs for doubleword quantities for ARMv8.1M-Mainline.
941 * config/arm/thumb2.md (thumb2_asrl): New.
942 (thumb2_lsll): Likewise.
944 2020-01-17 Jakub Jelinek <jakub@redhat.com>
946 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
949 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
951 * gdbinit.in (help-gcc-hooks): New command.
952 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
953 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
956 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
958 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
959 correct target macro.
961 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
963 * config/aarch64/aarch64-protos.h
964 (aarch64_sve_ld1ro_operand_p): New.
965 * config/aarch64/aarch64-sve-builtins-base.cc
966 (class load_replicate): New.
967 (class svld1ro_impl): New.
968 (class svld1rq_impl): Change to inherit from load_replicate.
969 (svld1ro): New sve intrinsic function base.
970 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
971 New DEF_SVE_FUNCTION.
972 * config/aarch64/aarch64-sve-builtins-base.h
974 * config/aarch64/aarch64-sve-builtins.cc
975 (function_expander::add_mem_operand): Modify assert to allow
977 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
979 * config/aarch64/aarch64.c
980 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
981 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
982 (aarch64_sve_ld1ro_operand_p): New.
983 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
984 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
985 * config/aarch64/predicates.md
986 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
988 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
990 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
991 Introduce this ACLE specified predefined macro.
992 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
993 (fp): Disabling this disables f64mm.
994 (simd): Disabling this disables f64mm.
995 (fp16): Disabling this disables f64mm.
996 (sve): Disabling this disables f64mm.
997 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
998 (AARCH64_ISA_F64MM): New.
1000 * doc/invoke.texi (f64mm): Document new option.
1002 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
1004 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
1005 (neoversen1_tunings): Likewise.
1007 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
1010 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
1011 Add assert to ensure prolog has been emitted.
1012 (aarch64_split_atomic_op): Likewise.
1013 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
1014 Use epilogue_completed rather than reload_completed.
1015 (aarch64_atomic_exchange<mode>): Likewise.
1016 (aarch64_atomic_<atomic_optab><mode>): Likewise.
1017 (atomic_nand<mode>): Likewise.
1018 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
1019 (atomic_fetch_nand<mode>): Likewise.
1020 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
1021 (atomic_nand_fetch<mode>): Likewise.
1023 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
1026 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
1028 (REVERSE_CONDITION): Delete.
1029 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
1030 (CCFP_CCFPE): Likewise.
1031 (e): New mode attribute.
1032 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
1033 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
1034 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
1035 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
1036 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
1037 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
1038 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
1039 name of generator from gen_ccmpdi to gen_ccmpccdi.
1040 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
1041 the previous comparison but aren't able to, use the new ccmp_rev
1044 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
1046 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
1047 than testing directly for INTEGER_CST.
1048 (gimplify_target_expr, gimplify_omp_depend): Likewise.
1050 2020-01-17 Jakub Jelinek <jakub@redhat.com>
1052 PR tree-optimization/93292
1053 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
1054 get_vectype_for_scalar_type returns NULL.
1056 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
1058 * params.opt (-param=max-predicted-iterations): Increase range from 0.
1059 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
1061 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
1063 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
1065 * params.opt: (max-predicted-iterations): Set bounds.
1066 * predict.c (real_almost_one, real_br_prob_base,
1067 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
1068 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
1069 probabilities; do not truncate to reg_br_prob_bases.
1070 (estimate_loops_at_level): Pass max_cyclic_prob.
1071 (estimate_loops): Compute max_cyclic_prob.
1072 (estimate_bb_frequencies): Do not initialize real_*; update calculation
1074 * profile-count.c (profile_probability::to_sreal): New.
1075 * profile-count.h (class sreal): Move up in file.
1076 (profile_probability::to_sreal): Declare.
1078 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1081 (arm_invalid_conversion): New function for target hook.
1082 (arm_invalid_unary_op): New function for target hook.
1083 (arm_invalid_binary_op): New function for target hook.
1085 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1087 * config.gcc: Add arm_bf16.h.
1088 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
1089 (arm_simd_builtin_std_type): Add BFmode.
1090 (arm_init_simd_builtin_types): Define element types for vector types.
1091 (arm_init_bf16_types): New function.
1092 (arm_init_builtins): Add arm_init_bf16_types function call.
1093 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
1094 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
1095 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
1096 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
1097 (arm_vector_mode_supported_p): Add V4BF, V8BF.
1098 (arm_mangle_type): Add __bf16.
1099 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
1100 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
1101 arm_bf16_ptr_type_node.
1102 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
1103 define_split between ARM registers.
1104 * config/arm/arm_bf16.h: New file.
1105 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
1106 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
1108 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
1109 * config/arm/vfp.md: Add BFmode to movhf patterns.
1111 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
1112 Andre Vieira <andre.simoesdiasvieira@arm.com>
1114 * config/arm/arm-cpus.in (mve, mve_float): New features.
1115 (dsp, mve, mve.fp): New options.
1116 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
1117 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
1118 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
1120 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1121 Thomas Preud'homme <thomas.preudhomme@arm.com>
1123 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
1125 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
1126 error for using -mcmse when targeting Armv8.1-M Mainline.
1128 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1129 Thomas Preud'homme <thomas.preudhomme@arm.com>
1131 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
1132 address in r4 when targeting Armv8.1-M Mainline.
1133 (nonsecure_call_value_internal): Likewise.
1134 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
1135 a register match_operand again. Emit BLXNS when targeting
1137 (nonsecure_call_value_reg_thumb2): Likewise.
1139 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1140 Thomas Preud'homme <thomas.preudhomme@arm.com>
1142 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
1143 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
1144 variable as true when floating-point ABI is not hard. Replace
1145 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
1146 Generate VLSTM and VLLDM instruction respectively before and
1147 after a function call to cmse_nonsecure_call function.
1148 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
1149 (VUNSPEC_VLLDM): Likewise.
1150 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
1151 (lazy_load_multiple_insn): Likewise.
1153 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1154 Thomas Preud'homme <thomas.preudhomme@arm.com>
1156 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
1157 (arm_emit_vfp_multi_reg_pop): Likewise.
1158 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
1159 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
1160 restore callee-saved VFP registers.
1162 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1163 Thomas Preud'homme <thomas.preudhomme@arm.com>
1165 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
1166 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
1167 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
1168 callee-saved GPRs as well as clear ip register before doing a nonsecure
1169 call then restore callee-saved GPRs after it when targeting
1171 (arm_reorg): Adapt to function rename.
1173 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1174 Thomas Preud'homme <thomas.preudhomme@arm.com>
1176 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
1177 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
1178 clear_vfp_multiple pattern based on a new vfp parameter.
1179 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
1180 targeting Armv8.1-M Mainline.
1181 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
1182 unconditionally when targeting Armv8.1-M Mainline architecture. Check
1183 whether VFP registers are available before looking call_used_regs for a
1185 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
1186 of prototype of clear_operation_p.
1187 (clear_vfp_multiple_operation): New predicate.
1188 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
1189 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
1191 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1192 Thomas Preud'homme <thomas.preudhomme@arm.com>
1194 * config/arm/arm-protos.h (clear_operation_p): Declare.
1195 * config/arm/arm.c (clear_operation_p): New function.
1196 (cmse_clear_registers): Generate clear_multiple instruction pattern if
1197 targeting Armv8.1-M Mainline or successor.
1198 (output_return_instruction): Only output APSR register clearing if
1199 Armv8.1-M Mainline instructions not available.
1200 (thumb_exit): Likewise.
1201 * config/arm/predicates.md (clear_multiple_operation): New predicate.
1202 * config/arm/thumb2.md (clear_apsr): New define_insn.
1203 (clear_multiple): Likewise.
1204 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
1206 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1207 Thomas Preud'homme <thomas.preudhomme@arm.com>
1209 * config/arm/arm.c (fp_sysreg_names): Declare and define.
1210 (use_return_insn): Also return false for Armv8.1-M Mainline.
1211 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
1212 Mainline instructions are available.
1213 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
1214 when targeting Armv8.1-M Mainline Security Extensions.
1215 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
1216 Mainline entry function.
1217 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
1218 targeting Armv8.1-M Mainline or successor.
1219 (arm_expand_epilogue): Fix indentation of caller-saved register
1220 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
1222 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
1223 (FP_SYSREGS): Likewise.
1224 (enum vfp_sysregs_encoding): Define enum.
1225 (fp_sysreg_names): Declare.
1226 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
1227 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
1228 (pop_fpsysreg_insn): Likewise.
1230 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1231 Thomas Preud'homme <thomas.preudhomme@arm.com>
1233 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
1234 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
1235 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
1236 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
1237 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
1238 (ARMv8_1m_main): New feature group.
1239 (armv8.1-m.main): New architecture.
1240 * config/arm/arm-tables.opt: Regenerate.
1241 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
1242 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
1243 (arm_options_perform_arch_sanity_checks): Error out when targeting
1244 Armv8.1-M Mainline Security Extensions.
1245 * config/arm/arm.h (arm_arch8_1m_main): Declare.
1247 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1249 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
1250 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
1251 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
1252 aarch64_bfdot_laneq): New.
1253 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
1254 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
1255 vbfdotq_laneq_f32): New.
1256 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
1257 VBFMLA_W, VBF): New.
1258 (isquadop): Add V4BF, V8BF.
1260 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1262 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
1263 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
1264 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
1265 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
1266 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
1267 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
1268 usdot_laneq, sudot_lane,sudot_laneq): New.
1269 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
1270 (aarch64_<sur>dot_lane): New.
1271 * config/aarch64/arm_neon.h (vusdot_s32): New.
1273 (vusdot_lane_s32): New.
1274 (vsudot_lane_s32): New.
1275 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
1276 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
1278 2020-01-16 Martin Liska <mliska@suse.cz>
1280 * value-prof.c (dump_histogram_value): Fix
1281 obvious spacing issue.
1283 2020-01-16 Andrew Pinski <apinski@marvell.com>
1285 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
1286 !storage_order_barrier_p.
1288 2020-01-16 Andrew Pinski <apinski@marvell.com>
1290 * sched-int.h (_dep): Add unused bit-field field for the padding.
1291 * sched-deps.c (init_dep_1): Init unused field.
1293 2020-01-16 Andrew Pinski <apinski@marvell.com>
1295 * optabs.h (create_expand_operand): Initialize target field also.
1297 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
1299 PR tree-optimization/92429
1300 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
1301 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
1303 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
1306 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
1308 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
1309 aarch64_sve_int_mode to each mode.
1311 2020-01-15 David Malcolm <dmalcolm@redhat.com>
1313 * doc/analyzer.texi (Overview): Add note about
1314 -fdump-ipa-analyzer.
1316 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
1318 PR tree-optimization/93231
1319 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
1320 input_type is unsigned. Use tree_to_shwi for shift constant.
1321 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
1322 (simplify_count_trailing_zeroes): Add test to handle known non-zero
1323 inputs more efficiently.
1325 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
1327 * config/i386/i386.md (*movsf_internal): Do not require
1328 SSE2 ISA for alternatives 14 and 15.
1330 2020-01-15 Richard Biener <rguenther@suse.de>
1333 * tree-eh.c (sink_clobbers): If we already visited the destination
1334 block do not defer insertion.
1335 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
1336 the purpose of defered insertion.
1338 2020-01-15 Jakub Jelinek <jakub@redhat.com>
1340 * BASE-VER: Bump to 10.0.1.
1342 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
1344 PR tree-optimization/93247
1345 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
1346 type of the stmt that we're going to vectorize.
1348 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
1350 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
1351 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
1354 2020-01-15 Martin Liska <mliska@suse.cz>
1356 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
1357 2 calls of streamer_read_hwi in a function call.
1359 2020-01-15 Richard Biener <rguenther@suse.de>
1361 * alias.c (record_alias_subset): Avoid redundant work when
1362 subset is already recorded.
1364 2020-01-14 David Malcolm <dmalcolm@redhat.com>
1366 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
1367 the analyzer options provide CWE identifiers.
1369 2020-01-14 David Malcolm <dmalcolm@redhat.com>
1371 * tree-diagnostic-path.cc (path_summary::event_range::print):
1372 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
1373 using get_pure_location.
1375 2020-01-15 Jakub Jelinek <jakub@redhat.com>
1377 PR tree-optimization/93262
1378 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
1379 perform head trimming only if the last argument is constant,
1380 either all ones, or larger or equal to head trim, in the latter
1381 case decrease the last argument by head_trim.
1383 PR tree-optimization/93249
1384 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
1385 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
1386 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
1387 perform head trim unless we can prove there are no '\0' chars
1388 from the source among the first head_trim chars.
1390 2020-01-14 David Malcolm <dmalcolm@redhat.com>
1392 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
1394 2020-01-15 Jakub Jelinek <jakub@redhat.com>
1397 * config/i386/sse.md
1398 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
1399 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
1400 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
1401 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
1402 just a single alternative instead of two, make operands 1 and 2
1405 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
1408 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
1411 2020-01-14 David Malcolm <dmalcolm@redhat.com>
1413 * Makefile.in (lang_opt_files): Add analyzer.opt.
1414 (ANALYZER_OBJS): New.
1415 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
1416 tristate.o and ANALYZER_OBJS.
1417 (TEXI_GCCINT_FILES): Add analyzer.texi.
1418 * common.opt (-fanalyzer): New driver option.
1419 * config.in: Regenerate.
1420 * configure: Regenerate.
1421 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
1422 (gccdepdir): Also create depdir for "analyzer" subdir.
1423 * digraph.cc: New file.
1424 * digraph.h: New file.
1425 * doc/analyzer.texi: New file.
1426 * doc/gccint.texi ("Static Analyzer") New menu item.
1427 (analyzer.texi): Include it.
1428 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
1429 ("Warning Options"): Add static analysis warnings to the list.
1430 (-Wno-analyzer-double-fclose): New option.
1431 (-Wno-analyzer-double-free): New option.
1432 (-Wno-analyzer-exposure-through-output-file): New option.
1433 (-Wno-analyzer-file-leak): New option.
1434 (-Wno-analyzer-free-of-non-heap): New option.
1435 (-Wno-analyzer-malloc-leak): New option.
1436 (-Wno-analyzer-possible-null-argument): New option.
1437 (-Wno-analyzer-possible-null-dereference): New option.
1438 (-Wno-analyzer-null-argument): New option.
1439 (-Wno-analyzer-null-dereference): New option.
1440 (-Wno-analyzer-stale-setjmp-buffer): New option.
1441 (-Wno-analyzer-tainted-array-index): New option.
1442 (-Wno-analyzer-use-after-free): New option.
1443 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
1444 (-Wno-analyzer-use-of-uninitialized-value): New option.
1445 (-Wanalyzer-too-complex): New option.
1446 (-fanalyzer-call-summaries): New warning.
1447 (-fanalyzer-checker=): New warning.
1448 (-fanalyzer-fine-grained): New warning.
1449 (-fno-analyzer-state-merge): New warning.
1450 (-fno-analyzer-state-purge): New warning.
1451 (-fanalyzer-transitivity): New warning.
1452 (-fanalyzer-verbose-edges): New warning.
1453 (-fanalyzer-verbose-state-changes): New warning.
1454 (-fanalyzer-verbosity=): New warning.
1455 (-fdump-analyzer): New warning.
1456 (-fdump-analyzer-callgraph): New warning.
1457 (-fdump-analyzer-exploded-graph): New warning.
1458 (-fdump-analyzer-exploded-nodes): New warning.
1459 (-fdump-analyzer-exploded-nodes-2): New warning.
1460 (-fdump-analyzer-exploded-nodes-3): New warning.
1461 (-fdump-analyzer-supergraph): New warning.
1462 * doc/sourcebuild.texi (dg-require-dot): New.
1463 (dg-check-dot): New.
1464 * gdbinit.in (break-on-saved-diagnostic): New command.
1465 * graphviz.cc: New file.
1466 * graphviz.h: New file.
1467 * ordered-hash-map-tests.cc: New file.
1468 * ordered-hash-map.h: New file.
1469 * passes.def (pass_analyzer): Add before
1470 pass_ipa_whole_program_visibility.
1471 * selftest-run-tests.c (selftest::run_tests): Call
1472 selftest::ordered_hash_map_tests_cc_tests.
1473 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
1475 * shortest-paths.h: New file.
1476 * timevar.def (TV_ANALYZER): New timevar.
1477 (TV_ANALYZER_SUPERGRAPH): Likewise.
1478 (TV_ANALYZER_STATE_PURGE): Likewise.
1479 (TV_ANALYZER_PLAN): Likewise.
1480 (TV_ANALYZER_SCC): Likewise.
1481 (TV_ANALYZER_WORKLIST): Likewise.
1482 (TV_ANALYZER_DUMP): Likewise.
1483 (TV_ANALYZER_DIAGNOSTICS): Likewise.
1484 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
1485 * tree-pass.h (make_pass_analyzer): New decl.
1486 * tristate.cc: New file.
1487 * tristate.h: New file.
1489 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
1492 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
1493 alternatives 9 and 10.
1495 2020-01-14 David Malcolm <dmalcolm@redhat.com>
1497 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
1498 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
1499 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
1500 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
1501 (selftest::hash_map_tests_c_tests): Call it.
1502 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
1503 New static constant, using the value of = H::empty_zero_p.
1504 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
1505 from default_hash_traits <Value>.
1506 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
1508 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
1509 * hash-table.h (hash_table::alloc_entries): Guard the loop of
1510 calls to mark_empty with !Descriptor::empty_zero_p.
1511 (hash_table::empty_slow): Conditionalize the memset call with a
1512 check that Descriptor::empty_zero_p; otherwise, loop through the
1513 entries calling mark_empty on them.
1514 * hash-traits.h (int_hash::empty_zero_p): New static constant.
1515 (pointer_hash::empty_zero_p): Likewise.
1516 (pair_hash::empty_zero_p): Likewise.
1517 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
1519 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
1520 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
1521 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
1522 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
1523 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
1524 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
1525 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
1526 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
1528 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
1531 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
1533 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
1534 fix typo on return value.
1536 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
1539 * cgraph.c (symbol_table::create_edge): Init speculative_id and
1541 (cgraph_edge::make_speculative): Add param for setting speculative_id
1543 (cgraph_edge::speculative_call_info): Update comments and find reference
1544 by speculative_id for multiple indirect targets.
1545 (cgraph_edge::resolve_speculation): Decrease the speculations
1546 for indirect edge, drop it's speculative if not direct target
1547 left. Update comments.
1548 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
1549 (cgraph_node::dump): Print num_speculative_call_targets.
1550 (cgraph_node::verify_node): Don't report error if speculative
1551 edge not include statement.
1552 (cgraph_edge::num_speculative_call_targets_p): New function.
1553 * cgraph.h (int common_target_id): Remove.
1554 (int common_target_probability): Remove.
1555 (num_speculative_call_targets): New variable.
1556 (make_speculative): Add param for setting speculative_id.
1557 (cgraph_edge::num_speculative_call_targets_p): New declare.
1558 (target_prob): New variable.
1559 (speculative_id): New variable.
1560 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
1561 call summaries for multiple speculative call targets.
1562 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
1563 * ipa-profile.c (struct speculative_call_target): New struct.
1564 (class speculative_call_summary): New class.
1565 (class speculative_call_summaries): New class.
1566 (call_sums): New variable.
1567 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
1568 (ipa_profile_write_edge_summary): New function.
1569 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
1570 (ipa_profile_dump_all_summaries): New function.
1571 (ipa_profile_read_edge_summary): New function.
1572 (ipa_profile_read_summary_section): New function.
1573 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
1574 (ipa_profile): Generate num_speculative_call_targets from
1576 * ipa-ref.h (speculative_id): New variable.
1577 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
1578 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
1579 common_target_probability. Stream out speculative_id and
1580 num_speculative_call_targets.
1581 (input_edge): Likewise.
1582 * predict.c (dump_prediction): Remove edges count assert to be
1584 * symtab.c (symtab_node::create_reference): Init speculative_id.
1585 (symtab_node::clone_references): Clone speculative_id.
1586 (symtab_node::clone_referring): Clone speculative_id.
1587 (symtab_node::clone_reference): Clone speculative_id.
1588 (symtab_node::clear_stmts_in_references): Clear speculative_id.
1589 * tree-inline.c (copy_bb): Duplicate all the speculative edges
1590 if indirect call contains multiple speculative targets.
1591 * value-prof.h (check_ic_target): Remove.
1592 * value-prof.c (gimple_value_profile_transformations):
1593 Use void function gimple_ic_transform.
1594 * value-prof.c (gimple_ic_transform): Handle topn case.
1595 Fix comment typos. Change it to a void function.
1597 2020-01-13 Andrew Pinski <apinski@marvell.com>
1599 * config/aarch64/aarch64-cores.def (octeontx2): New define.
1600 (octeontx2t98): New define.
1601 (octeontx2t96): New define.
1602 (octeontx2t93): New define.
1603 (octeontx2f95): New define.
1604 (octeontx2f95n): New define.
1605 (octeontx2f95mm): New define.
1606 * config/aarch64/aarch64-tune.md: Regenerate.
1607 * doc/invoke.texi (-mcpu=): Document the new cpu types.
1609 2020-01-13 Jason Merrill <jason@redhat.com>
1611 PR c++/33799 - destroy return value if local cleanup throws.
1612 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
1614 2020-01-13 Martin Liska <mliska@suse.cz>
1616 * ipa-cp.c (get_max_overall_size): Use newly
1617 renamed param param_ipa_cp_unit_growth.
1618 * params.opt: Remove legacy param name.
1620 2020-01-13 Martin Sebor <msebor@redhat.com>
1622 PR tree-optimization/93213
1623 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
1624 stores to be eliminated.
1626 2020-01-13 Martin Liska <mliska@suse.cz>
1628 * opts.c (print_help): Do not print CL_PARAM
1629 and CL_WARNING for CL_OPTIMIZATION.
1631 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
1634 * doc/invoke.texi (Warning Options): Add caveat about some warnings
1635 depending on optimization settings.
1637 2020-01-13 Jakub Jelinek <jakub@redhat.com>
1639 PR tree-optimization/90838
1640 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
1641 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
1642 argument rather than to initialize temporary for targets that
1643 don't use the mode argument at all. Initialize ctzval to avoid
1646 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
1648 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
1649 * tree-core.h: Document it.
1650 * gimplify.c (gimplify_omp_workshare): Set it.
1651 * omp-low.c (lower_omp_target): Use it.
1652 * tree-pretty-print.c (dump_omp_clause): Print it.
1654 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
1655 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
1657 2020-01-10 David Malcolm <dmalcolm@redhat.com>
1659 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
1660 * common.opt (fdiagnostics-path-format=): New option.
1661 (diagnostic_path_format): New enum.
1662 (fdiagnostics-show-path-depths): New option.
1663 * coretypes.h (diagnostic_event_id_t): New forward decl.
1664 * diagnostic-color.c (color_dict): Add "path".
1665 * diagnostic-event-id.h: New file.
1666 * diagnostic-format-json.cc (json_from_expanded_location): Make
1668 (json_end_diagnostic): Call context->make_json_for_path if it
1669 exists and the diagnostic has a path.
1670 (diagnostic_output_format_init): Clear context->print_path.
1671 * diagnostic-path.h: New file.
1672 * diagnostic-show-locus.c (colorizer::set_range): Special-case
1673 when printing a run of events in a diagnostic_path so that they
1674 all get the same color.
1675 (layout::m_diagnostic_path_p): New field.
1676 (layout::layout): Initialize it.
1677 (layout::print_any_labels): Don't colorize the label text for an
1678 event in a diagnostic_path.
1679 (gcc_rich_location::add_location_if_nearby): Add
1680 "restrict_to_current_line_spans" and "label" params. Pass the
1681 former to layout.maybe_add_location_range; pass the latter
1682 when calling add_range.
1683 * diagnostic.c: Include "diagnostic-path.h".
1684 (diagnostic_initialize): Initialize context->path_format and
1685 context->show_path_depths.
1686 (diagnostic_show_any_path): New function.
1687 (diagnostic_path::interprocedural_p): New function.
1688 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
1689 (simple_diagnostic_path::num_events): New function.
1690 (simple_diagnostic_path::get_event): New function.
1691 (simple_diagnostic_path::add_event): New function.
1692 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
1693 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
1694 (debug): New overload taking a diagnostic_path *.
1695 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
1696 * diagnostic.h (enum diagnostic_path_format): New enum.
1697 (json::value): New forward decl.
1698 (diagnostic_context::path_format): New field.
1699 (diagnostic_context::show_path_depths): New field.
1700 (diagnostic_context::print_path): New callback field.
1701 (diagnostic_context::make_json_for_path): New callback field.
1702 (diagnostic_show_any_path): New decl.
1703 (json_from_expanded_location): New decl.
1704 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
1705 (-fdiagnostics-show-path-depths): New option.
1706 (-fdiagnostics-color): Add "path" to description of default
1707 GCC_COLORS; describe it.
1708 (-fdiagnostics-format=json): Document how diagnostic paths are
1709 represented in the JSON output format.
1710 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
1711 Add optional params "restrict_to_current_line_spans" and "label".
1712 * opts.c (common_handle_option): Handle
1713 OPT_fdiagnostics_path_format_ and
1714 OPT_fdiagnostics_show_path_depths.
1715 * pretty-print.c: Include "diagnostic-event-id.h".
1716 (pp_format): Implement "%@" format code for printing
1717 diagnostic_event_id_t *.
1718 (selftest::test_pp_format): Add tests for "%@".
1719 * selftest-run-tests.c (selftest::run_tests): Call
1720 selftest::tree_diagnostic_path_cc_tests.
1721 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
1722 * toplev.c (general_init): Initialize global_dc->path_format and
1723 global_dc->show_path_depths.
1724 * tree-diagnostic-path.cc: New file.
1725 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
1726 non-static. Drop "diagnostic" param in favor of storing the
1727 original value of "where" and re-using it.
1728 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
1729 maybe_unwind_expanded_macro_loc.
1730 (tree_diagnostics_defaults): Initialize context->print_path and
1731 context->make_json_for_path.
1732 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
1734 (default_tree_make_json_for_path): New decl.
1735 (maybe_unwind_expanded_macro_loc): New decl.
1737 2020-01-10 Jakub Jelinek <jakub@redhat.com>
1739 PR tree-optimization/93210
1740 * fold-const.h (native_encode_initializer,
1741 can_native_interpret_type_p): Declare.
1742 * fold-const.c (native_encode_string): Fix up handling with off != -1,
1744 (native_encode_initializer): New function, moved from dwarf2out.c.
1745 Adjust to native_encode_expr compatible arguments, including dry-run
1746 and partial extraction modes. Don't handle STRING_CST.
1747 (can_native_interpret_type_p): No longer static.
1748 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
1749 offset / BITS_PER_UNIT fits into int and don't call it if
1750 can_native_interpret_type_p fails. If suboff is NULL and for
1751 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
1752 native_encode_initializer.
1753 (fold_const_aggregate_ref_1): Formatting fix.
1754 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
1755 (tree_add_const_value_attribute): Adjust caller.
1757 PR tree-optimization/90838
1758 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
1759 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
1760 CTZ_DEFINED_VALUE_AT_ZERO.
1762 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
1765 * lra-constraints.c (match_reload): Permit input operands have the
1766 same mode as output while other input operands have a different
1769 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
1771 PR tree-optimization/90838
1772 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
1773 (check_ctz_string): Likewise.
1774 (optimize_count_trailing_zeroes): Likewise.
1775 (simplify_count_trailing_zeroes): Likewise.
1776 (pass_forwprop::execute): Try ctz simplification.
1777 * match.pd: Add matching for ctz idioms.
1779 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1781 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
1783 (aarch64_invalid_unary_op): New function for target hook.
1784 (aarch64_invalid_binary_op): New function for target hook.
1786 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1788 * config.gcc: Add arm_bf16.h.
1789 * config/aarch64/aarch64-builtins.c
1790 (aarch64_simd_builtin_std_type): Add BFmode.
1791 (aarch64_init_simd_builtin_types): Define element types for vector
1793 (aarch64_init_bf16_types): New function.
1794 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
1795 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
1797 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
1798 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
1800 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
1801 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
1802 * config/aarch64/aarch64.c
1803 (aarch64_classify_vector_mode): Add support for BF types.
1804 (aarch64_gimplify_va_arg_expr): Add support for BF types.
1805 (aarch64_vq_mode): Add support for BF types.
1806 (aarch64_simd_container_mode): Add support for BF types.
1807 (aarch64_mangle_type): Add support for BF scalar type.
1808 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
1809 * config/aarch64/arm_bf16.h: New file.
1810 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
1811 * config/aarch64/iterators.md: Add BF types to mode attributes.
1812 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
1814 2020-01-10 Jason Merrill <jason@redhat.com>
1816 PR c++/93173 - incorrect tree sharing.
1817 * gimplify.c (copy_if_shared): No longer static.
1818 * gimplify.h: Declare it.
1820 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
1822 * doc/invoke.texi (-msve-vector-bits=): Document that
1823 -msve-vector-bits=128 now generates VL-specific code for
1824 little-endian targets.
1825 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
1826 build_vector_type_for_mode to construct the data vector types.
1827 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
1828 VL-specific code for -msve-vector-bits=128 on little-endian targets.
1829 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
1830 for 128-bit vectors.
1832 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
1834 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
1837 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
1839 * config/aarch64/aarch64-builtins.c
1840 (aarch64_builtin_vectorized_function): Check for specific vector modes,
1841 rather than checking the number of elements and the element mode.
1843 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
1845 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
1846 get_related_vectype_for_scalar_type rather than build_vector_type
1847 to create the index type for a conditional reduction.
1849 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
1851 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
1852 for any type of gather or scatter, including strided accesses.
1854 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
1856 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
1859 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
1861 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
1863 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
1864 parameter and its use to reset DR_OFFSET's.
1865 (vect_transform_loop): Remove orig_drs_init argument.
1866 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
1867 member of dr_vec_info rather than the offset of the associated
1868 data_reference's innermost_loop_behavior.
1869 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
1870 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
1871 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
1872 get_dr_vinfo_offset.
1873 (vectorizable_store): Likewise.
1874 (vectorizable_load): Likewise.
1876 2020-01-10 Richard Biener <rguenther@suse.de>
1878 * gimple-ssa-store-merging
1879 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
1881 2020-01-10 Martin Liska <mliska@suse.cz>
1884 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
1885 encapsulation that was there before r280040.
1887 2020-01-10 Richard Biener <rguenther@suse.de>
1890 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
1891 sequences to avoid walking them again for secondary opportunities.
1892 (pass_lower_eh_dispatch::execute): Instead actually insert
1895 2020-01-10 Richard Biener <rguenther@suse.de>
1898 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
1899 (cleanup_all_empty_eh): Walk landing pads in reverse order to
1900 avoid quadraticness.
1902 2020-01-10 Martin Jambor <mjambor@suse.cz>
1904 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
1905 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
1906 to get param_ipa_sra_max_replacements.
1907 (param_splitting_across_edge): Pass the caller to
1908 pull_accesses_from_callee.
1910 2020-01-10 Martin Jambor <mjambor@suse.cz>
1912 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
1913 * ipa-cp.c (max_new_size): Removed.
1914 (orig_overall_size): New variable.
1915 (get_max_overall_size): New function.
1916 (estimate_local_effects): Use it. Adjust dump.
1917 (decide_about_value): Likewise.
1918 (ipcp_propagate_stage): Do not calculate max_new_size, just store
1919 orig_overall_size. Adjust dump.
1920 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
1922 2020-01-10 Martin Jambor <mjambor@suse.cz>
1924 * params.opt (param_ipa_max_agg_items): Mark as Optimization
1925 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
1926 instead of param_ipa_max_agg_items.
1927 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
1928 optimization info for the callee.
1930 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
1932 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
1933 markers if debug_inline_points is false.
1935 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
1937 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
1939 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
1940 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
1941 aarch64-sve-builtins-sve2.h.
1942 (aarch64-sve-builtins-sve2.o): New rule.
1943 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
1944 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
1945 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
1946 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
1947 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
1948 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
1950 * config/aarch64/aarch64-sve.md: Update comments with SVE2
1951 instructions that are handled here.
1952 (@cond_asrd<mode>): Generalize to...
1953 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
1954 (*cond_asrd<mode>_2): Generalize to...
1955 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
1956 (*cond_asrd<mode>_z): Generalize to...
1957 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
1958 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
1959 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
1960 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
1961 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
1963 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
1964 (@aarch64_scatter_stnt<mode>): Likewise.
1965 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
1966 (@aarch64_mul_lane_<mode>): Likewise.
1967 (@aarch64_sve_suqadd<mode>_const): Likewise.
1968 (*<sur>h<addsub><mode>): Generalize to...
1969 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
1971 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
1972 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
1973 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
1974 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
1975 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
1976 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
1977 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
1978 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
1979 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
1980 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
1981 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
1982 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
1983 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
1984 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
1985 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
1986 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
1987 (@aarch64_sve2_xar<mode>): Likewise.
1988 (@aarch64_sve2_bcax<mode>): Likewise.
1989 (*aarch64_sve2_eor3<mode>): Rename to...
1990 (@aarch64_sve2_eor3<mode>): ...this.
1991 (@aarch64_sve2_bsl<mode>): New expander.
1992 (@aarch64_sve2_nbsl<mode>): Likewise.
1993 (@aarch64_sve2_bsl1n<mode>): Likewise.
1994 (@aarch64_sve2_bsl2n<mode>): Likewise.
1995 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
1996 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
1997 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
1998 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
1999 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
2000 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
2001 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
2002 (<su>mull<bt><Vwide>): Generalize to...
2003 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
2005 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
2006 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
2007 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
2008 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2009 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
2010 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2011 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
2012 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2013 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
2014 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2015 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
2016 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
2017 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
2018 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
2019 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
2020 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
2021 (<SHRNB:r>shrnb<mode>): Generalize to...
2022 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
2024 (<SHRNT:r>shrnt<mode>): Generalize to...
2025 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
2027 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
2028 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
2029 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
2030 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
2031 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
2032 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
2033 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
2034 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
2035 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
2036 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
2037 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
2038 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
2039 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
2040 (@aarch64_sve2_cvtnt<mode>): Likewise.
2041 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
2042 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
2043 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
2044 (@aarch64_sve2_cvtxnt<mode>): Likewise.
2045 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
2046 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
2047 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
2048 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
2049 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
2050 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
2051 (@aarch64_sve2_pmul<mode>): Likewise.
2052 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
2053 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
2054 (@aarch64_sve2_tbl2<mode>): Likewise.
2055 (@aarch64_sve2_tbx<mode>): Likewise.
2056 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
2057 (@aarch64_sve2_histcnt<mode>): Likewise.
2058 (@aarch64_sve2_histseg<mode>): Likewise.
2059 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
2060 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
2061 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
2062 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
2063 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
2064 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
2065 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
2066 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
2067 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
2068 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
2069 (SVE2_PMULL_PAIR_I): New mode iterators.
2070 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
2071 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
2072 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
2073 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
2074 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
2075 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
2076 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
2077 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
2078 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
2079 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
2080 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
2081 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
2082 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
2083 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
2084 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
2085 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
2086 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
2087 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
2088 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
2089 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
2090 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
2091 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
2092 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
2093 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
2094 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
2095 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
2096 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
2097 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
2098 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
2099 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
2100 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
2102 (VNARROW, Ventype): New mode attributes.
2103 (Vewtype): Handle VNx2DI. Fix typo in comment.
2104 (VDOUBLE): New mode attribute.
2105 (sve_lane_con): Handle VNx8HI.
2106 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
2107 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
2108 (sve_int_op, sve_int_op_rev): Handle the above codes.
2109 (sve_pred_int_rhs2_operand): Likewise.
2110 (MULLBT, SHRNB, SHRNT): Delete.
2111 (SVE_INT_SHIFT_IMM): New int iterator.
2112 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
2113 and UNSPEC_WHILEHS for TARGET_SVE2.
2114 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
2115 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
2116 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
2117 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
2118 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
2119 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
2120 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
2121 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
2122 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
2123 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
2124 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
2125 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
2126 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
2127 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
2128 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
2129 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
2130 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
2131 (optab): Handle the new unspecs.
2132 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
2134 (lr): Handle the new unspecs.
2136 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
2137 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
2138 (sve_int_qsub_op): New int attributes.
2139 (sve_fp_op, rot): Handle the new unspecs.
2140 * config/aarch64/aarch64-sve-builtins.h
2141 (function_resolver::require_matching_pointer_type): Declare.
2142 (function_resolver::resolve_unary): Add an optional boolean argument.
2143 (function_resolver::finish_opt_n_resolution): Add an optional
2144 type_suffix_index argument.
2145 (gimple_folder::redirect_call): Declare.
2146 (gimple_expander::prepare_gather_address_operands): Add an optional
2148 * config/aarch64/aarch64-sve-builtins.cc: Include
2149 aarch64-sve-builtins-sve2.h.
2150 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
2151 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
2152 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
2153 (TYPES_hsd_integer): Use TYPES_hsd_signed.
2154 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
2155 (TYPES_s_unsigned): Likewise.
2156 (TYPES_s_integer): Use TYPES_s_unsigned.
2157 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
2158 (TYPES_sd_integer): Use them.
2159 (TYPES_d_unsigned): New macro.
2160 (TYPES_d_integer): Use it.
2161 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
2162 (TYPES_cvt_narrow): Likewise.
2163 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
2164 (preds_mx): New variable.
2165 (function_builder::add_overloaded_function): Allow the new feature
2166 set to be more restrictive than the original one.
2167 (function_resolver::infer_pointer_type): Remove qualifiers from
2168 the pointer type before printing it.
2169 (function_resolver::require_matching_pointer_type): New function.
2170 (function_resolver::resolve_sv_displacement): Handle functions
2171 that don't support 32-bit vector indices or svint32_t vector offsets.
2172 (function_resolver::finish_opt_n_resolution): Take the inferred type
2173 as a separate argument.
2174 (function_resolver::resolve_unary): Optionally treat all forms in
2175 the same way as normal merging functions.
2176 (gimple_folder::redirect_call): New function.
2177 (function_expander::prepare_gather_address_operands): Add an argument
2178 that says whether scaled forms are available. If they aren't,
2179 handle scaling of vector indices and don't add the extension and
2181 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
2182 fall back to using cond_* instead.
2183 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
2184 Split out the member variables into...
2185 (rtx_code_function_base): ...this new base class.
2186 (rtx_code_function_rotated): Inherit rtx_code_function_base.
2187 (unspec_based_function): Split out the member variables into...
2188 (unspec_based_function_base): ...this new base class.
2189 (unspec_based_function_rotated): Inherit unspec_based_function_base.
2190 (unspec_based_function_exact_insn): New class.
2191 (unspec_based_add_function, unspec_based_add_lane_function)
2192 (unspec_based_lane_function, unspec_based_pred_function)
2193 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
2194 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
2195 (unspec_based_sub_function, unspec_based_sub_lane_function): New
2197 (unspec_based_fused_function): New class.
2198 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
2199 (unspec_based_fused_lane_function): New class.
2200 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
2202 (CODE_FOR_MODE1): New macro.
2203 (fixed_insn_function): New class.
2204 (while_comparison): Likewise.
2205 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
2206 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
2207 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
2208 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
2209 (load_gather_sv_restricted, shift_left_imm_long): Declare.
2210 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
2211 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
2212 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
2213 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
2214 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
2215 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
2216 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
2217 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
2218 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
2219 Also add an initial argument for unary_convert_narrowt, regardless
2220 of the predication type.
2221 (build_32_64): Allow loads and stores to specify MODE_none.
2222 (build_sv_index64, build_sv_uint_offset): New functions.
2223 (long_type_suffix): New function.
2224 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
2225 (binary_imm_long_base, load_gather_sv_base): Likewise.
2226 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
2227 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
2228 (unary_narrowb_base, unary_narrowt_base): Likewise.
2229 (binary_long_lane_def, binary_long_lane): New shape.
2230 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
2231 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
2232 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
2233 (binary_to_uint_def, binary_to_uint): Likewise.
2234 (binary_wide_def, binary_wide): Likewise.
2235 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
2236 (compare_def, compare): Likewise.
2237 (compare_ptr_def, compare_ptr): Likewise.
2238 (load_ext_gather_index_restricted_def,
2239 load_ext_gather_index_restricted): Likewise.
2240 (load_ext_gather_offset_restricted_def,
2241 load_ext_gather_offset_restricted): Likewise.
2242 (load_gather_sv_def): Inherit from load_gather_sv_base.
2243 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
2244 (shift_left_imm_def, shift_left_imm): Likewise.
2245 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
2246 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
2247 (store_scatter_index_restricted_def,
2248 store_scatter_index_restricted): Likewise.
2249 (store_scatter_offset_restricted_def,
2250 store_scatter_offset_restricted): Likewise.
2251 (tbl_tuple_def, tbl_tuple): Likewise.
2252 (ternary_long_lane_def, ternary_long_lane): Likewise.
2253 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
2254 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
2255 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
2256 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
2257 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
2258 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
2259 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
2260 (ternary_uint_def, ternary_uint): Likewise.
2261 (unary_convert): Fix typo in comment.
2262 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
2263 (unary_long_def, unary_long): Likewise.
2264 (unary_narrowb_def, unary_narrowb): Likewise.
2265 (unary_narrowt_def, unary_narrowt): Likewise.
2266 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
2267 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
2268 (unary_to_int_def, unary_to_int): Likewise.
2269 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
2270 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
2271 (svasrd_impl): Delete.
2272 (svcadd_impl::expand): Handle integer operations too.
2273 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
2274 new functions to derive the unspec numbers.
2275 (svmla_svmls_lane_impl): Replace with...
2276 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
2277 integer operations too.
2278 (svwhile_impl): Rename to...
2279 (svwhilelx_impl): ...this and inherit from while_comparison.
2280 (svasrd): Use unspec_based_function.
2281 (svmla_lane): Use svmla_lane_impl.
2282 (svmls_lane): Use svmls_lane_impl.
2283 (svrecpe, svrsqrte): Handle unsigned integer operations too.
2284 (svwhilele, svwhilelt): Use svwhilelx_impl.
2285 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
2286 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
2287 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
2288 * config/aarch64/aarch64-sve-builtins.def: Include
2289 aarch64-sve-builtins-sve2.def.
2291 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2293 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
2294 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
2295 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
2296 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
2297 immediates as well as vector ones.
2298 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
2299 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
2300 (aarch64_sve_qsub_immediate): Update calls accordingly.
2302 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2304 * config/aarch64/aarch64-sve2.md: Add banner comments.
2305 (<su>mulh<r>s<mode>3): Move further up file.
2306 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
2307 (*aarch64_sve2_sra<mode>): Move further down file.
2308 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
2310 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2312 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
2314 (while_optab_cmp): Handle them.
2315 * config/aarch64/aarch64-sve.md
2316 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
2317 and add a "@" marker.
2318 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
2319 instead of gen_aarch64_sve2_while_ptest.
2320 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
2322 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2324 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
2325 (UNSPEC_WHILELE): ...this.
2326 (UNSPEC_WHILE_LO): Rename to...
2327 (UNSPEC_WHILELO): ...this.
2328 (UNSPEC_WHILE_LS): Rename to...
2329 (UNSPEC_WHILELS): ...this.
2330 (UNSPEC_WHILE_LT): Rename to...
2331 (UNSPEC_WHILELT): ...this.
2332 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
2333 (cmp_op, while_optab_cmp): Likewise.
2334 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
2335 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
2336 (svwhilelt): Likewise.
2338 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2340 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
2341 (unary_to_uint): Define.
2342 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
2343 (unary_count): Rename to...
2344 (unary_to_uint_def, unary_to_uint): ...this.
2345 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
2347 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2349 * config/aarch64/aarch64-sve-builtins-functions.h
2350 (code_for_mode_function): New class.
2351 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
2352 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
2353 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
2354 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
2355 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
2357 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2359 * config/aarch64/iterators.md (addsub): New code attribute.
2360 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
2362 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
2363 in the asm string and attributes. Fix indentation.
2364 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
2366 (@aarch64_sve_<optab><mode>): ...this.
2367 * config/aarch64/aarch64-sve-builtins.h
2368 (function_expander::expand_signed_unpred_op): Delete.
2369 * config/aarch64/aarch64-sve-builtins.cc
2370 (function_expander::expand_signed_unpred_op): Likewise.
2371 (function_expander::map_to_rtx_codes): If the optab isn't defined,
2372 try using code_for_aarch64_sve instead.
2373 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
2374 (svqsub_impl): Likewise.
2375 (svqadd, svqsub): Use rtx_code_function instead.
2377 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2379 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
2380 (HADDSUB, sur, addsub): Remove them.
2382 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2384 * tree-nrv.c (pass_return_slot::execute): Handle all internal
2385 functions the same way, rather than singling out those that
2386 aren't mapped directly to optabs.
2388 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2390 * target.def (compatible_vector_types_p): New target hook.
2391 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
2392 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
2393 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
2394 * doc/tm.texi: Regenerate.
2395 * gimple-expr.c: Include target.h.
2396 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
2397 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
2399 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
2400 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
2401 Use the original predicate if it already has a suitable type.
2403 2020-01-09 Martin Jambor <mjambor@suse.cz>
2405 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
2406 resolve_speculation and redirect_call_stmt_to_callee static. Change
2407 return type of set_call_stmt to cgraph_edge *.
2408 * auto-profile.c (afdo_indirect_call): Adjust call to
2409 redirect_call_stmt_to_callee.
2410 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
2411 make the this pointer explicit, adjust self-recursive calls and the
2412 call top make_direct. Return the resulting edge.
2413 (cgraph_edge::remove): Make this pointer explicit.
2414 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
2415 (cgraph_edge::make_direct): Likewise, adjust call to
2416 resolve_speculation.
2417 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
2418 call to set_call_stmt.
2419 (cgraph_update_edges_for_call_stmt_node): Update call to
2420 set_call_stmt and remove.
2421 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
2422 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
2423 (cgraph_node::create_edge_including_clones): Moved "first" definition
2424 of edge to the block where it was used. Adjusted calls to
2426 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
2427 cgraph_edge::remove.
2428 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
2429 make_direct and redirect_call_stmt_to_callee.
2430 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
2431 resolve_speculation and make_direct.
2432 * ipa-inline-transform.c (inline_transform): Adjust call to
2433 redirect_call_stmt_to_callee.
2434 (check_speculations_1):: Adjust call to resolve_speculation.
2435 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
2436 resolve-speculation.
2437 (inline_small_functions): Adjust call to resolve_speculation.
2438 (ipa_inline): Likewise.
2439 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
2441 * ipa-visibility.c (function_and_variable_visibility): Make iteration
2442 safe with regards to edge removal, adjust calls to
2443 redirect_call_stmt_to_callee.
2444 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
2445 and redirect_call_stmt_to_callee.
2446 * multiple_target.c (create_dispatcher_calls): Adjust call to
2447 redirect_call_stmt_to_callee
2448 (redirect_to_specific_clone): Likewise.
2449 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
2450 Adjust calls to cgraph_edge::remove.
2451 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
2452 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
2453 (expand_call_inline): Adjust call to cgraph_edge::remove.
2455 2020-01-09 Martin Liska <mliska@suse.cz>
2457 * params.opt: Set Optimization for
2458 param_max_speculative_devirt_maydefs.
2460 2020-01-09 Martin Sebor <msebor@redhat.com>
2464 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
2466 2020-01-09 Martin Liska <mliska@suse.cz>
2468 * auto-profile.c (auto_profile): Use opt_for_fn
2470 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
2471 (propagate_vals_across_arith_jfunc): Likewise.
2472 (hint_time_bonus): Likewise.
2473 (incorporate_penalties): Likewise.
2474 (good_cloning_opportunity_p): Likewise.
2475 (perform_estimation_of_a_value): Likewise.
2476 (estimate_local_effects): Likewise.
2477 (ipcp_propagate_stage): Likewise.
2478 * ipa-fnsummary.c (decompose_param_expr): Likewise.
2479 (set_switch_stmt_execution_predicate): Likewise.
2480 (analyze_function_body): Likewise.
2481 * ipa-inline-analysis.c (offline_size): Likewise.
2482 * ipa-inline.c (early_inliner): Likewise.
2483 * ipa-prop.c (ipa_analyze_node): Likewise.
2484 (ipcp_transform_function): Likewise.
2485 * ipa-sra.c (process_scan_results): Likewise.
2486 (ipa_sra_summarize_function): Likewise.
2487 * params.opt: Rename ipcp-unit-growth to
2488 ipa-cp-unit-growth. Add Optimization for various
2489 IPA-related parameters.
2491 2020-01-09 Richard Biener <rguenther@suse.de>
2494 * gimplify.c (gimplify_expr): Deal with NOP definitions.
2496 2020-01-09 Richard Biener <rguenther@suse.de>
2498 PR tree-optimization/93040
2499 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
2501 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
2503 * common/config/avr/avr-common.c (avr_option_optimization_table)
2504 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
2506 2020-01-09 Martin Liska <mliska@suse.cz>
2508 * cgraphclones.c (symbol_table::materialize_all_clones):
2509 Use cgraph_node::dump_name.
2511 2020-01-09 Jakub Jelinek <jakub@redhat.com>
2514 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
2515 output_operand_lossage instead of gcc_unreachable.
2516 * doc/md.texi (riscv f constraint): Fix typo.
2519 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
2520 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
2521 CONST_SCALAR_INT_P instead of CONST_INT_P.
2522 (*subv<mode>4_1): Rename to ...
2523 (subv<mode>4_1): ... this.
2524 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
2525 define_insn_and_split patterns.
2526 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
2529 2020-01-08 David Malcolm <dmalcolm@redhat.com>
2531 * vec.c (class selftest::count_dtor): New class.
2532 (selftest::test_auto_delete_vec): New test.
2533 (selftest::vec_c_tests): Call it.
2534 * vec.h (class auto_delete_vec): New class template.
2535 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
2537 2020-01-08 David Malcolm <dmalcolm@redhat.com>
2539 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
2541 2020-01-08 Jim Wilson <jimw@sifive.com>
2543 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
2544 use of TLS_MODEL_LOCAL_EXEC when not pic.
2546 2020-01-08 David Malcolm <dmalcolm@redhat.com>
2548 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
2551 2020-01-08 Jakub Jelinek <jakub@redhat.com>
2554 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
2555 *stack_protect_set_3 peephole2): Also check that the second
2556 insns source is general_operand.
2559 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
2560 predicate for output operand instead of register_operand.
2561 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
2562 memory destination and non-memory operands[2].
2564 2020-01-08 Martin Liska <mliska@suse.cz>
2566 * cgraph.c (cgraph_node::dump): Use ::dump_name or
2567 ::dump_asm_name instead of (::name or ::asm_name).
2568 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
2569 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
2570 (analyze_functions): Likewise.
2571 (expand_all_functions): Likewise.
2572 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
2573 (propagate_bits_across_jump_function): Likewise.
2574 (dump_profile_updates): Likewise.
2575 (ipcp_store_bits_results): Likewise.
2576 (ipcp_store_vr_results): Likewise.
2577 * ipa-devirt.c (dump_targets): Likewise.
2578 * ipa-fnsummary.c (analyze_function_body): Likewise.
2579 * ipa-hsa.c (check_warn_node_versionable): Likewise.
2580 (process_hsa_functions): Likewise.
2581 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
2582 (set_alias_uids): Likewise.
2583 * ipa-inline-transform.c (save_inline_function_body): Likewise.
2584 * ipa-inline.c (recursive_inlining): Likewise.
2585 (inline_to_all_callers_1): Likewise.
2586 (ipa_inline): Likewise.
2587 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
2588 (ipa_propagate_frequency): Likewise.
2589 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2590 (remove_described_reference): Likewise.
2591 * ipa-pure-const.c (worse_state): Likewise.
2592 (check_retval_uses): Likewise.
2593 (analyze_function): Likewise.
2594 (propagate_pure_const): Likewise.
2595 (propagate_nothrow): Likewise.
2596 (dump_malloc_lattice): Likewise.
2597 (propagate_malloc): Likewise.
2598 (pass_local_pure_const::execute): Likewise.
2599 * ipa-visibility.c (optimize_weakref): Likewise.
2600 (function_and_variable_visibility): Likewise.
2601 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
2602 (ipa_discover_variable_flags): Likewise.
2603 * lto-streamer-out.c (output_function): Likewise.
2604 (output_constructor): Likewise.
2605 * tree-inline.c (copy_bb): Likewise.
2606 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
2607 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
2609 2020-01-08 Richard Biener <rguenther@suse.de>
2612 * tree-eh.c (sink_clobbers): Update virtual operands for
2613 the first and last stmt only. Add a dry-run capability.
2614 (pass_lower_eh_dispatch::execute): Perform clobber sinking
2615 after CFG manipulations and in RPO order to catch all
2616 secondary opportunities reliably.
2618 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
2621 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
2623 2019-01-08 Richard Biener <rguenther@suse.de>
2626 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
2627 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
2628 virtual operand, also updating SSA use.
2629 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
2630 Update stmt after resetting virtual operand.
2631 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
2632 * gimple-iterator.c (gsi_remove): When not removing the stmt
2633 permanently do not delink immediate uses or mark the stmt modified.
2635 2020-01-08 Martin Liska <mliska@suse.cz>
2637 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
2638 (ipa_call_context::estimate_size_and_time): Likewise.
2639 (inline_analyze_function): Likewise.
2641 2020-01-08 Martin Liska <mliska@suse.cz>
2643 * cgraph.c (cgraph_node::dump): Use systematically
2646 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
2648 Add -nodevicespecs option for avr.
2651 * config/avr/avr.opt (-nodevicespecs): New driver option.
2652 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
2653 "-specs=device-specs/..." if that option is not set.
2654 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
2656 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
2658 Implement 64-bit double functions for avr.
2661 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
2662 --with-double-comparison.
2663 * doc/install.texi: Document them.
2664 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
2665 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
2666 <WITH_DOUBLE_COMPARISON>: New built-in defines.
2667 * doc/invoke.texi (AVR Built-in Macros): Document them.
2668 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
2669 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
2670 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
2672 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
2675 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
2676 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
2677 when only building rm-profile multilibs.
2679 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
2682 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
2683 lattice for a value to check.
2684 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
2685 finite propagation in self-recursive scc.
2687 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
2689 * ipa-inline.c (caller_growth_limits): Restore the AND.
2691 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
2693 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
2694 (VEC_ALLREG_ALT): New iterator.
2695 (VEC_ALLREG_INT_MODE): New iterator.
2696 (VCMP_MODE): New iterator.
2697 (VCMP_MODE_INT): New iterator.
2698 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
2699 (vec_cmp<u>v64qidi): New define_expand.
2700 (vec_cmp<mode>di_exec): Use VCMP_MODE.
2701 (vec_cmpu<mode>di_exec): New define_expand.
2702 (vec_cmp<u>v64qidi_exec): New define_expand.
2703 (vec_cmp<mode>di_dup): Use VCMP_MODE.
2704 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
2705 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
2706 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
2707 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
2708 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
2709 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
2710 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
2711 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
2712 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
2714 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
2715 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
2717 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
2719 * config/gcn/constraints.md (DA): Update description and match.
2721 (Db): New constraint.
2722 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
2724 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
2725 Implement 'Db' mixed immediate type.
2726 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
2727 (addcv64si3_dup<exec_vcc>): Delete.
2728 (subcv64si3<exec_vcc>): Rework constraints.
2729 (addv64di3): Rework constraints.
2730 (addv64di3_exec): Rework constraints.
2731 (subv64di3): Rework constraints.
2732 (addv64di3_dup): Delete.
2733 (addv64di3_dup_exec): Delete.
2734 (addv64di3_zext): Rework constraints.
2735 (addv64di3_zext_exec): Rework constraints.
2736 (addv64di3_zext_dup): Rework constraints.
2737 (addv64di3_zext_dup_exec): Rework constraints.
2738 (addv64di3_zext_dup2): Rework constraints.
2739 (addv64di3_zext_dup2_exec): Rework constraints.
2740 (addv64di3_sext_dup2): Rework constraints.
2741 (addv64di3_sext_dup2_exec): Rework constraints.
2743 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
2745 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
2746 existing target checks.
2748 2020-01-07 Richard Biener <rguenther@suse.de>
2750 * doc/install.texi: Bump minimal supported MPC version.
2752 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
2754 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
2755 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
2756 * langhooks.c: Include stor-layout.h.
2757 (lhd_simulate_enum_decl): New function.
2758 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
2759 handle_arm_sve_h for the LTO frontend.
2760 (register_vector_type): Cope with null returns from pushdecl.
2762 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
2764 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
2765 (aarch64_sve::nvectors_if_data_type): Replace with...
2766 (aarch64_sve::builtin_type_p): ...this.
2767 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
2768 (find_vector_type): Delete.
2769 (add_sve_type_attribute): New function.
2770 (lookup_sve_type_attribute): Likewise.
2771 (register_builtin_types): Add an "SVE type" attribute to each type.
2772 (register_tuple_type): Likewise.
2773 (svbool_type_p, nvectors_if_data_type): Delete.
2774 (mangle_builtin_type): Use lookup_sve_type_attribute.
2775 (builtin_type_p): Likewise. Add an overload that returns the
2776 number of constituent vector and predicate registers.
2777 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
2778 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
2779 instead of aarch64_sve_argument_p.
2780 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
2781 (aarch64_pass_by_reference): Likewise.
2782 (aarch64_function_value_1): Likewise.
2783 (aarch64_return_in_memory): Likewise.
2784 (aarch64_layout_arg): Likewise.
2786 2020-01-07 Jakub Jelinek <jakub@redhat.com>
2788 PR tree-optimization/93156
2789 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
2790 least significant bit is always clear.
2792 PR tree-optimization/93118
2793 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
2794 simplifier with two intermediate conversions.
2796 2020-01-07 Martin Liska <mliska@suse.cz>
2798 * params.opt: Add Optimization for various parameters.
2800 2020-01-07 Martin Liska <mliska@suse.cz>
2803 * doc/extend.texi: Explain cloning for target_clone
2806 2020-01-07 Martin Liska <mliska@suse.cz>
2808 PR tree-optimization/92860
2809 * common.opt: Make in Optimization option
2810 as it is affected by -O0, which is an Optimization
2812 * tree-inline.c (tree_inlinable_function_p):
2813 Use opt_for_fn for warn_inline.
2814 (expand_call_inline): Likewise.
2816 2020-01-07 Martin Liska <mliska@suse.cz>
2818 PR tree-optimization/92860
2819 * common.opt: Make flag_ree as optimization
2822 2020-01-07 Martin Liska <mliska@suse.cz>
2824 PR optimization/92860
2825 * params.opt: Mark param_min_crossjump_insns with Optimization
2828 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
2830 * ipa-inline-analysis.c (estimate_growth): Fix typo.
2831 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
2833 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
2835 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
2836 helper function to return the valid addressing formats for a given
2837 hard register and mode.
2838 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
2840 * config/rs6000/constraints.md (Q constraint): Update
2842 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
2845 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
2846 Use 'Q' for doing vector extract from memory.
2847 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
2849 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
2850 doing vector extract from memory.
2851 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
2852 extract from memory.
2854 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
2855 for the offset being 34-bits when -mcpu=future is used.
2857 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
2859 * config/pa/pa.md: Revert change to use ordered_comparison_operator
2860 instead of cmpib_comparison_operator in cmpib patterns.
2861 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
2862 of cmpib_comparison_operator. Revise comment.
2864 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
2866 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
2867 in an IFN_DIV_POW2 node to be equal.
2869 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
2871 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
2872 (vect_check_scalar_mask): ...this.
2873 (vectorizable_store, vectorizable_load): Update call accordingly.
2874 (vectorizable_call): Use vect_check_scalar_mask to check the mask
2875 argument in calls to conditional internal functions.
2877 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
2879 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
2880 '0' matching inputs.
2881 (subv64di3_exec): Likewise.
2883 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
2885 * config/mips/mips.c (vr4130_align_insns): Fix typo.
2886 * doc/md.texi (movstr): Likewise.
2888 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
2890 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
2893 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
2895 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
2897 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
2898 to a temporary file and use move-if-change to update the real
2899 file where necessary.
2901 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
2903 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
2904 rather than Upa for CPY /M.
2906 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
2908 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
2911 2020-01-06 Martin Liska <mliska@suse.cz>
2913 PR tree-optimization/92860
2914 * params.opt: Mark param_max_combine_insns with Optimization
2917 2020-01-05 Jakub Jelinek <jakub@redhat.com>
2920 * config/i386/i386.md (SWIDWI): New mode iterator.
2921 (DWI, dwi): Add TImode variants.
2922 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
2923 <general_hilo_operand> instead of <general_operand>. Use
2924 CONST_SCALAR_INT_P instead of CONST_INT_P.
2925 (*addv<mode>4_1): Rename to ...
2926 (addv<mode>4_1): ... this.
2927 (QWI): New mode attribute.
2928 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
2929 define_insn_and_split patterns.
2930 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
2932 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
2933 <general_hilo_operand> instead of <general_operand>.
2934 (*addcarry<mode>_1): New define_insn.
2935 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
2937 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
2939 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
2940 Use "call" instead of "set".
2942 2020-01-03 Martin Jambor <mjambor@suse.cz>
2945 * ipa-cp.c (print_all_lattices): Skip functions without info.
2947 2020-01-03 Jakub Jelinek <jakub@redhat.com>
2950 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
2951 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
2952 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
2953 for 'e' simd clones.
2956 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
2958 (mprefer-vector-width=): Add Save.
2959 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
2960 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
2961 (ix86_debug_options, ix86_function_specific_print): Adjust
2962 ix86_target_string callers.
2963 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
2964 (ix86_valid_target_attribute_tree): Likewise.
2965 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
2966 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
2967 ix86_target_string caller.
2970 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
2971 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
2972 instead of gen_int_shift_amount + convert_modes.
2974 PR rtl-optimization/93088
2975 * loop-iv.c (find_single_def_src): Punt after looking through
2976 128 reg copies for regs with single definitions. Move definitions
2979 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
2981 * config/arm/arm-c.c (arm_cpu_builtins): Define
2982 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
2983 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
2984 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
2985 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
2986 * config/arm/arm-tables.opt: Regenerated.
2987 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
2988 arm_arch_i8mm and arm_arch_bf16 when enabled.
2989 * config/arm/arm.h (TARGET_I8MM): New macro.
2990 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
2991 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
2992 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
2993 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
2994 (v8_6_a_simd_variants): New.
2995 (v8_*_a_simd_variants): Add i8mm and bf16.
2996 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
2998 2020-01-02 Jakub Jelinek <jakub@redhat.com>
3001 * predict.c (compute_function_frequency): Don't call
3002 warn_function_cold on functions that already have cold attribute.
3004 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
3007 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
3008 COMDAT group function labels in .data.rel.ro.local section.
3009 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
3012 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
3013 comparison_operator in B and S integer comparisons. Likewise, use
3014 ordered_comparison_operator instead of cmpib_comparison_operator in
3016 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
3018 2020-01-01 Jakub Jelinek <jakub@redhat.com>
3020 Update copyright years.
3022 * gcc.c (process_command): Update copyright notice dates.
3023 * gcov-dump.c (print_version): Ditto.
3024 * gcov.c (print_version): Ditto.
3025 * gcov-tool.c (print_version): Ditto.
3026 * gengtype.c (create_file): Ditto.
3027 * doc/cpp.texi: Bump @copying's copyright year.
3028 * doc/cppinternals.texi: Ditto.
3029 * doc/gcc.texi: Ditto.
3030 * doc/gccint.texi: Ditto.
3031 * doc/gcov.texi: Ditto.
3032 * doc/install.texi: Ditto.
3033 * doc/invoke.texi: Ditto.
3035 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
3037 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
3040 2020-01-01 Jakub Jelinek <jakub@redhat.com>
3042 PR tree-optimization/93098
3043 * match.pd (popcount): For shift amounts, use integer_onep
3044 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
3045 tests. Make sure that precision is power of two larger than or equal
3046 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
3047 instead of ULL suffixed constants. Formatting fixes.
3049 Copyright (C) 2020 Free Software Foundation, Inc.
3051 Copying and distribution of this file, with or without modification,
3052 are permitted in any medium without royalty provided the copyright
3053 notice and this notice are preserved.