1 2018-11-06 Alexandre Oliva <aoliva@redhat.com>
3 * auto-inc-dec.c: Include valtrack.h. Improve comments.
4 (reg_next_debug_use): New.
5 (attempt_change): Propagate adjusted expression into affected
7 (merge_in_block): Track uses in debug insns.
8 (pass_inc_dec::execute): Allocate and release
11 2018-11-06 Wei Xiao <wei3.xiao@intel.com>
13 * config/i386/avx512fintrin.h: Update VFIXUPIMM* intrinsics.
14 (_mm512_fixupimm_round_pd): Update parameters and builtin.
15 (_mm512_maskz_fixupimm_round_pd): Ditto.
16 (_mm512_fixupimm_round_ps): Ditto.
17 (_mm512_maskz_fixupimm_round_ps): Ditto.
18 (_mm_fixupimm_round_sd): Ditto.
19 (_mm_maskz_fixupimm_round_sd): Ditto.
20 (_mm_fixupimm_round_ss): Ditto.
21 (_mm_maskz_fixupimm_round_ss): Ditto.
22 (_mm512_fixupimm_pd): Ditto.
23 (_mm512_maskz_fixupimm_pd): Ditto.
24 (_mm512_fixupimm_ps): Ditto.
25 (_mm512_maskz_fixupimm_ps): Ditto.
26 (_mm_fixupimm_sd): Ditto.
27 (_mm_maskz_fixupimm_sd): Ditto.
28 (_mm_fixupimm_ss): Ditto.
29 (_mm_maskz_fixupimm_ss): Ditto.
30 (_mm512_mask_fixupimm_round_pd): Update builtin.
31 (_mm512_mask_fixupimm_round_ps): Ditto.
32 (_mm_mask_fixupimm_round_sd): Ditto.
33 (_mm_mask_fixupimm_round_ss): Ditto.
34 (_mm512_mask_fixupimm_pd): Ditto.
35 (_mm512_mask_fixupimm_ps): Ditto.
36 (_mm_mask_fixupimm_sd): Ditto.
37 (_mm_mask_fixupimm_ss): Ditto.
38 * config/i386/avx512vlintrin.h:
39 (_mm256_fixupimm_pd): Update parameters and builtin.
40 (_mm256_maskz_fixupimm_pd): Ditto.
41 (_mm256_fixupimm_ps): Ditto.
42 (_mm256_maskz_fixupimm_ps): Ditto.
43 (_mm_fixupimm_pd): Ditto.
44 (_mm_maskz_fixupimm_pd): Ditto.
45 (_mm_fixupimm_ps): Ditto.
46 (_mm_maskz_fixupimm_ps): Ditto.
47 (_mm256_mask_fixupimm_pd): Update builtin.
48 (_mm256_mask_fixupimm_ps): Ditto.
49 (_mm_mask_fixupimm_pd): Ditto.
50 (_mm_mask_fixupimm_ps): Ditto.
51 * config/i386/i386-builtin-types.def: Add new types and remove useless ones.
52 * config/i386/i386-builtin.def: Update builtin definitions.
53 * config/i386/i386.c: Handle new builtin types and remove useless ones.
54 * config/i386/sse.md: Update VFIXUPIMM* patterns.
55 (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
56 (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
57 (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Update.
58 (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
59 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
60 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Update.
61 * config/i386/subst.md:
62 (round_saeonly_sd_mask_operand4): Add new subst_attr.
63 (round_saeonly_sd_mask_op4): Ditto.
64 (round_saeonly_expand_operand5): Ditto.
65 (round_saeonly_expand): Update.
67 2018-11-05 Max Filippov <jcmvbkbc@gmail.com>
69 * config/xtensa/uclinux.h (XTENSA_ALWAYS_PIC): Change to 0.
71 2018-11-05 Segher Boessenkool <segher@kernel.crashing.org>
73 PR rtl-optimization/87871
74 * combine.c (make_more_copies): Skip if dest is frame_pointer_rtx.
76 2018-11-05 Paul Koning <ni1d@arrl.net>
78 * doc/sourcebuild.texi (target attributes): Document new "inf"
79 effective target keyword.
81 2018-11-05 Robin Dapp <rdapp@linux.ibm.com>
83 * config/s390/s390.c (s390_register_move_cost): Increase costs for
84 moves involving the CC reg.
86 2018-11-05 Richard Biener <rguenther@suse.de>
88 PR tree-optimization/87873
89 * tree-ssa-loop-manip.h (split_loop_exit_edge): Add copy_constants_p
91 * tree-ssa-loop-manip.c (split_loop_exit_edge): Likewise.
92 * tree-vect-loop.c (vect_transform_loop): When splitting the
93 loop exit also create forwarder PHIs for constants.
94 * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
95 Handle constant to_arg, add extra checking we match up the correct
98 2018-11-05 Robin Dapp <rdapp@linux.ibm.com>
100 * config/s390/s390.md: QImode and HImode for load on condition.
102 2018-11-05 Robin Dapp <rdapp@linux.ibm.com>
104 * config/s390/predicates.md: Fix typo.
105 * config/s390/s390.md: Allow immediates for load on condition.
107 2018-11-05 Martin Liska <mliska@suse.cz>
109 * alloc-pool.h (struct pool_usage): Use SIZE_AMOUNT.
110 * bitmap.h (struct bitmap_usage): Likewise.
111 * ggc-common.c (SCALE): Remove.
113 (struct ggc_usage): Use SIZE_AMOUNT. And update
115 * ggc-page.c (SCALE): Remove.
116 (STAT_LABEL): Remove.
117 (ggc_print_statistics): Use SIZE_AMOUNT.
118 * gimple.h (SCALE): Remove.
120 * input.c (ONE_K): Remove.
123 (STAT_LABEL): Likewise.
124 (FORMAT_AMOUNT): Likewise.
125 (dump_line_table_statistics): Use SIZE_AMOUNT.
126 * mem-stats.h (struct mem_usage): Likewise.
127 * rtl.c (dump_rtx_statistics): Likewise.
128 (rtx_alloc_counts): Change type to size_t.
129 (rtx_alloc_sizes): Likewise.
130 (rtx_count_cmp): New.
131 (dump_rtx_statistics): Sort first based on counts.
132 * tree.c (tree_nodes_cmp): New.
133 (tree_codes_cmp): New.
134 (dump_tree_statistics): Sort first based on counts.
135 * system.h (ONE_K): New.
137 (SIZE_SCALE): Likewise.
138 (SIZE_LABEL): Likewise.
139 (SIZE_AMOUNT): Likewise.
140 * tree-cfg.c (dump_cfg_stats): Use SIZE_AMOUNT.
141 * tree-dfa.c (dump_dfa_stats): Likewise.
142 * tree-phinodes.c (phinodes_print_statistics): Likewise.
143 * tree-ssanames.c (ssanames_print_statistics): Likewise.
144 * tree.c (dump_tree_statistics): Likewise.
145 * vec.c (struct vec_usage): Likewise.
146 * trans-mem.c (tm_mangle): Enlarge buffer in order to not
147 trigger a -Werror=format-overflow with
148 --enable-gather-detailed-stats.
150 2018-11-05 Martin Liska <mliska@suse.cz>
152 * mem-stats.h (mem_alloc_description::release_instance_overhead):
154 * vec.c (struct vec_usage): Register m_element_size.
155 (vec_prefix::register_overhead): New arguments: elements and
157 (vec_prefix::release_overhead): Subtract elements.
158 * vec.h (struct vec_prefix): Change signature.
159 (va_heap::reserve): Pass proper arguments.
160 (va_heap::release): Likewise.
162 2018-11-05 Martin Liska <mliska@suse.cz>
164 * mem-stats.h (mem_alloc_description::get_list): Fix GNU coding
168 2018-11-05 Richard Biener <rguenther@suse.de>
170 * tree-scalar-evolution.h (final_value_replacement_loop): Update
172 * tree-scalar-evolution.c (final_value_replacement_loop): Return
173 whether anything was done.
174 (scev_const_prop): Remove constant propagation part, fold
176 * tree-ssa-loop.c (pass_scev_cprop::execute): ... here.
177 (pass_data_scev_cprop): TODO_cleanup_cfg is now done
180 2018-11-05 Jakub Jelinek <jakub@redhat.com>
182 PR tree-optimization/87859
183 * gimple-ssa-store-merging.c (struct merged_store_group): Add
184 only_constants and first_nonmergeable_order members.
185 (merged_store_group::merged_store_group): Initialize them.
186 (merged_store_group::do_merge): Clear only_constants member if
187 adding something other than INTEGER_CST store.
188 (imm_store_chain_info::coalesce_immediate_stores): Don't merge
189 stores with order >= first_nonmergeable_order. Use
190 merged_store->only_constants instead of always recomputing it.
191 Set merged_store->first_nonmergeable_order if we've skipped any
192 stores. Attempt to merge overlapping INTEGER_CST stores that
193 we would otherwise skip.
196 * match.pd (X + Y < X): Don't optimize if TYPE_OVERFLOW_SANITIZED.
198 2018-11-05 Xuepeng Guo <xuepeng.guo@intel.com>
201 * config/i386/emmintrin.h (__v16qs): New to cope with option
203 (_mm_cmpeq_epi8): Replace __v16qi with __v16qs.
204 (_mm_cmplt_epi8): Likewise.
205 (_mm_cmpgt_epi8): Likewise.
207 2018-11-05 Richard Biener <rguenther@suse.de>
209 PR rtl-optimization/87852
210 * fwprop.c (use_killed_between): Only consider single-defs of the
211 use whose definition statement dominates the use.
213 2018-11-05 Martin Liska <mliska@suse.cz>
216 * doc/invoke.texi: Remove options that are
217 not disabled with -Os.
219 2018-11-05 Martin Liska <mliska@suse.cz>
222 * doc/extend.texi: Update constrain about the last argument
223 of __builtin_expect_with_probability.
225 2018-11-05 Martin Liska <mliska@suse.cz>
228 * predict.c (expr_expected_value_1): Verify
229 that last argument is a real constants and emit
232 2018-11-05 Martin Liska <mliska@suse.cz>
234 PR gcov-profile/77698
235 * ipa-profile.c (ipa_profile): Adjust hotness threshold
238 2018-11-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
240 PR tree-optimization/86572
241 * builtins.c (c_strlen): Handle negative offsets in a safe way.
243 2018-11-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
245 PR tree-optimization/87672
246 * gimple-fold.c (gimple_fold_builtin_stxcpy_chk): Gimplify.
247 * tree-ssa-strlen.c (handle_builtin_strcat): Adjust object size.
249 2018-11-04 Uros Bizjak <ubizjak@gmail.com>
252 * cfgexpand.c (pass_expand::execute): Move the call to
253 finish_eh_generation in front of the call to expand_stack_alignment.
255 2018-11-04 Venkataramanan Kumar <venkataramanan.kumar@amd.com>
257 * common/config/i386/i386-common.c (processor_alias_table): Add
259 * config.gcc (i[34567]86-*-linux* | ...): Add znver2.
260 (case ${target}): Add znver2.
261 * config/i386/driver-i386.c: (host_detect_local_cpu): Let
262 -march=native recognize znver2 processors.
263 * config/i386/i386-c.c (ix86_target_macros_internal): Add znver2.
264 * config/i386/i386.c (m_znver2): New definition.
265 (m_ZNVER): New definition.
266 (m_AMD_MULTIPLE): Includes m_znver2.
267 (processor_cost_table): Add znver2 entry.
268 (processor_target_table): Add znver2 entry.
269 (get_builtin_code_for_version): Set priority for
271 (processor_model): Add M_AMDFAM17H_ZNVER2.
272 (arch_names_table): Ditto.
273 (ix86_reassociation_width): Include znver2.
274 * config/i386/i386.h (TARGET_znver2): New definition.
275 (struct ix86_size_cost): Add TARGET_ZNVER2.
276 (enum processor_type): Add PROCESSOR_ZNVER2.
277 * config/i386/i386.md (define_attr "cpu"): Add znver2.
278 * config/i386/x86-tune-costs.h: (processor_costs) Add znver2 costs.
279 * config/i386/x86-tune-sched.c: (ix86_issue_rate): Add znver2.
280 (ix86_adjust_cost): Add znver2.
281 * config/i386/x86-tune.def: Replace m_ZNVER1 by m_ZNVER.
282 * gcc/doc/extend.texi: Add details about znver2.
283 * gcc/doc/invoke.texi: Add details about znver2.
285 2018-11-03 Sandra Loosemore <sandra@codesourcery.com>
289 * config/nios2/nios2.c (nios2_rtx_costs): Recognize <mul>sidi3
292 2018-11-02 Aaron Sawdey <acsawdey@linux.ibm.com>
294 * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Pay
295 attention to TARGET_AVOID_XFORM and BYTES_BIG_ENDIAN.
297 2018-11-02 Richard Earnshaw <rearnsha@arm.com>
299 * config/aarch64/aarch64.c ((aarch64_override_options): Disable
300 shrink-wrapping when -mtrack-speculation.
302 2018-11-02 Richard Biener <rguenther@suse.de>
304 * tree-ssa-coalesce.c (struct coalesce_list): Add obstack member.
305 (pop_cost_one_pair): Do not free pair.
306 (pop_best_coalesce): Likewise.
307 (create_coalesce_list): Initialize obstack.
308 (delete_coalesce_list): Free obstack.
309 (find_coalesce_pair): Obstack-allocate coalesce pairs.
310 (add_cost_one_coalesce): Likewise.
311 (struct live_track): Remove bitmap pointer indirections.
312 (new_live_track): Adjust.
313 (delete_live_track): Likewise.
314 (live_track_remove_partition): Likewise.
315 (live_track_add_partition): Likewise.
316 (live_track_live_p): Likewise.
317 (live_track_process_def): Likewise.
318 (live_track_clear_base_vars): Likewise.
320 2018-11-02 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
322 * configure.ac (gcc_cv_as_sparc_register_op): Remove.
323 * configure: Regenerate.
324 * config.in: Regenerate.
325 * config/sparc/sparc.c (sparc_output_scratch_registers): Remove
326 HAVE_AS_REGISTER_PSEUDO_OP guard.
327 * config/sparc/sparc.h (ASM_ARCH64_SPEC): Remove
328 !HAVE_AS_REGISTER_PSEUDO_OP support.
329 (ASM_DECLARE_REGISTER_GLOBAL): Remove HAVE_AS_REGISTER_PSEUDO_OP
332 2018-11-02 Richard Biener <rguenther@suse.de>
334 * dwarf2out.c (build_abbrev_table): Guard lookup_external_ref call
337 2018-11-02 Richard Biener <rguenther@suse.de>
339 PR tree-optimization/87776
340 * tree-ssa-sccvn.c (do_rpo_vn): Do not mark backedges
341 executable when iterating but running into rpo-vn-max-loop-depth
342 and not eliding the iteration.
344 2018-11-30 Jan Hubicka <jh@suse.cz>
346 * tree.c (free_lang_data_in_decl): Free TREE_TYPE and alignment of
349 2018-11-01 Aaron Sawdey <acsawdey@linux.ibm.com>
351 * config/rs6000/rs6000-protos.h (rs6000_address_for_fpconvert): Remove
353 * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem):
354 Combine with rs6000_address_for_fpconvert.
355 (rs6000_address_for_fpconvert): Combine with
356 rs6000_force_indexed_or_indirect_mem.
357 (rs6000_expand_vector_init): Change function call from
358 rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
359 * config/rs6000/rs6000.md (floatsi<mode>2_lfiwax): Change call from
360 rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
361 (floatsi<mode>2_lfiwax_mem): Ditto.
362 (floatunssi<mode>2_lfiwzx): Ditto.
363 (floatunssi<mode>2_lfiwzx_mem): Ditto.
364 (float<QHI:mode><FP_ISA3:mode>2): Ditto.
365 (floatuns<QHI:mode><FP_ISA3:mode>2): Ditto.
366 (fix_trunc<mode>si2_stfiwx): Ditto.
367 (fixuns_trunc<mode>si2_stfiwx): Ditto.
368 (float_<mode>si2_hw): Ditto.
369 (floatuns_<mode>si2_hw): Ditto.
370 * config/rs6000/vsx.md (*vsx_extract_si): Ditto.
371 (vsx_splat_<mode>): Ditto.
373 2018-11-01 Joseph Myers <joseph@codesourcery.com>
375 * configure.ac (gcc_cv_as_ia64_ltoffx_ldxmov_relocs): Add newline
376 at end of assembler input text.
377 * configure: Regenerate.
379 2018-11-01 Jakub Jelinek <jakub@redhat.com>
381 PR tree-optimization/87826
382 * gimple-ssa-store-merging.c (do_shift_rotate): Punt if count is
383 negative or larger or equal to type's precision.
385 2018-10-31 Alexandre Oliva <aoliva@redhat.com>
387 * opts.c (default_options_table): Do not enable
388 OPT_fdelayed_branch at -Og.
389 * doc/invoke.texi (-fdelayed-branch): Document it.
391 2018-10-31 Richard Henderson <richard.henderson@linaro.org>
393 * optabs-libfuncs.c (build_libfunc_function_visibility):
394 New, split out from...
395 (build_libfunc_function): ... here.
396 (init_one_libfunc_visibility): New, split out from ...
397 (init_one_libfunc): ... here.
399 * config/aarch64/atomics.md (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse):
400 scratch register need not be early-clobber. Document the reason
401 why we cannot use ST<OP>.
403 2018-10-31 Joseph Myers <joseph@codesourcery.com>
406 * configure.ac: Remove AC_PREREQ. Use AC_LANG_SOURCE. Use single
407 line for second argument of AC_DEFINE_UNQUOTED.
408 * doc/install.texi (Tools/packages necessary for modifying GCC):
409 Update to autoconf 2.69 and automake 1.15.1.
410 * aclocal.m4, config.in, configure: Regenerate.
412 2018-10-31 Pat Haugen <pthaugen@us.ibm.com>
414 * tree-ssa-loop-ivcanon.c (try_peel_loop): Fix typo and remove dead
417 2018-10-31 Martin Liska <mliska@suse.cz>
420 * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
421 Add new argument invalid_extension.
422 (aarch64_get_all_extension_candidates): New function.
423 (aarch64_rewrite_selected_cpu): Add NULL to function call.
424 * config/aarch64/aarch64-protos.h (aarch64_parse_extension): Add
426 (aarch64_get_all_extension_candidates): New function.
427 * config/aarch64/aarch64.c (aarch64_parse_arch): Add new
428 argument invalid_extension.
429 (aarch64_parse_cpu): Likewise.
430 (aarch64_print_hint_for_extensions): New function.
431 (aarch64_validate_mcpu): Provide hint about invalid extension.
432 (aarch64_validate_march): Likewise.
433 (aarch64_handle_attr_arch): Pass new argument.
434 (aarch64_handle_attr_cpu): Provide hint about invalid extension.
435 (aarch64_handle_attr_isa_flags): Likewise.
437 2018-10-31 Richard Biener <rguenther@suse.de>
441 * tree-outof-ssa.c (insert_backedge_copies): Restrict
442 copy generation to useful cases. Place the copy before
443 the definition of the backedge value when possible.
445 2018-10-31 Claudiu Zissulescu <claziss@synopsys.com>
447 * config/arc/arc-arch.h (ARC_TUNE_ARC7XX): New tune value.
448 * config/arc/arc.c (arc_active_insn): New function.
449 (check_store_cacheline_hazard): Likewise.
450 (workaround_arc_anomaly): Use check_store_cacheline_hazard.
451 (arc_override_options): Disable delay slot scheduler for older
453 (arc_store_addr_hazard_p): New implementation, old one renamed to
455 (arc_store_addr_hazard_internal_p): Renamed.
456 (arc_reorg): Don't combine into brcc instructions which are part
457 of hardware hazard solution.
458 * config/arc/arc.md (attr tune): Consider new arc7xx tune value.
459 (tune_arc700): Likewise.
460 * config/arc/arc.opt (arc7xx): New tune value.
461 * config/arc/arc700.md: Improve A7 scheduler.
463 2018-10-31 Claudiu Zissulescu <claziss@synopsys.com>
465 * config/arc/arc.c (arc_override_options): Remove
466 TARGET_COMPACT_CASESI.
467 * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Update.
468 (CASE_VECTOR_MODE): Likewise.
469 (CASE_VECTOR_PC_RELATIVE): Likewise.
470 (CASE_VECTOR_SHORTEN_MODE): Likewise.
471 (CASE_VECTOR_SHORTEN_MODE1): Delete.
472 (ADDR_VEC_ALIGN): Update.
473 (ASM_OUTPUT_CASE_LABEL): Undefine.
474 (ASM_OUTPUT_BEFORE_CASE_LABEL): Undefine.
475 (TARGET_BI_BIH): Define.
476 (DEFAULT_BRANCH_INDEX): Likewise.
477 * config/arc/arc.md (casesi): Rework to accept BI/BIH
478 instructions, remove compact_casesi use case.
479 (casesi_compact_jump): Remove.
480 (casesi_dispatch): New pattern.
481 * config/arc/arc.opt: Add mbranch-index option. Deprecate
482 compact_casesi option.
483 * doc/invoke.texi: Document mbranch-index option.
485 2018-10-31 Claudiu Zissulescu <claziss@synopsys.com>
487 * config/arc/arc.c (arc_get_tp): Remove function.
488 (arc_emit_call_tls_get_addr): Likewise.
489 (arc_call_tls_get_addr): New function.
490 (arc_legitimize_tls_address): Make use of arc_call_tls_get_addr.
491 * config/arc/arc.md (tls_load_tp_soft): Remove.
492 (tls_gd_get_addr): Likewise.
494 2018-10-31 Claudiu Zissulescu <claziss@synopsys.com>
496 * config/arc/arc.md (mulsi3): Remove call to mulsi_600_lib.
497 (mulsi3_600_lib): Remove pattern.
498 (umulsi3_highpart_600_lib_le): Likewise.
499 (umulsi3_highpart): Remove call to umulsi3_highpart_600_lib_le.
500 (umulsidi3): Remove call to umulsidi3_600_lib.
501 (umulsidi3_600_lib): Remove pattern.
502 (peephole2): Remove peephole using the above deprecated patterns.
504 2018-10-31 Thomas Preud'homme <thomas.preudhomme@linaro.org>
507 * config/arm/arm.c (arm_option_check_internal): Disable the combined
508 use of -mslow-flash-data and -mword-relocations.
509 (arm_option_override): Enable -mword-relocations if -fpic or -fPIC.
510 * config/arm/arm.md (SYMBOL_REF MOVT splitter): Stop checking for
512 * doc/invoke.texi (-mword-relocations): Mention conflict with
514 (-mslow-flash-data): Reciprocally.
516 2018-10-31 Richard Henderson <richard.henderson@linaro.org>
518 * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Force
519 16-byte modes held in GP registers to use an even regno.
521 * config/aarch64/aarch64.c (aarch64_emit_bic): Remove.
522 (aarch64_atomic_ldop_supported_p): Remove.
523 (aarch64_gen_atomic_ldop): Remove.
524 * config/aarch64/atomic.md (atomic_<atomic_optab><ALLI>):
525 Fully expand LSE operations here.
526 (atomic_fetch_<atomic_optab><ALLI>): Likewise.
527 (atomic_<atomic_optab>_fetch<ALLI>): Likewise.
528 (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse): Drop atomic_op iterator
529 and use ATOMIC_LDOP instead; use register_operand for the input;
530 drop the split and emit insns directly.
531 (aarch64_atomic_fetch_<ATOMIC_LDOP><ALLI>_lse): Likewise.
532 (aarch64_atomic_<atomic_op>_fetch<ALLI>_lse): Remove.
533 (@aarch64_atomic_load<ATOMIC_LDOP><ALLI>): Remove.
535 * config/aarch64/aarch64.c (aarch64_emit_atomic_swap): Remove.
536 (aarch64_gen_atomic_ldop): Don't call it.
537 * config/aarch64/atomics.md (atomic_exchange<ALLI>):
538 Use aarch64_reg_or_zero.
539 (aarch64_atomic_exchange<ALLI>): Likewise.
540 (aarch64_atomic_exchange<ALLI>_lse): Remove split; remove & from
541 operand 0; use aarch64_reg_or_zero for input; merge ...
542 (@aarch64_atomic_swp<ALLI>): ... this and remove.
544 * config/aarch64/aarch64.c (aarch64_gen_compare_reg_maybe_ze): New.
545 (aarch64_split_compare_and_swap): Use it.
546 (aarch64_expand_compare_and_swap): Likewise. Remove convert_modes;
547 test oldval against the proper predicate.
548 * config/aarch64/atomics.md (@atomic_compare_and_swap<ALLI>):
549 Use nonmemory_operand for expected.
550 (cas_short_expected_pred): New.
551 (@aarch64_compare_and_swap<SHORT>): Use it; use "rn" not "rI" to match.
552 (@aarch64_compare_and_swap<GPI>): Use "rn" not "rI" for expected.
553 * config/aarch64/predicates.md (aarch64_plushi_immediate): New.
554 (aarch64_plushi_operand): New.
556 * config/aarch64/aarch64.c (aarch64_expand_compare_and_swap):
557 Force oldval into the rval register for TARGET_LSE; emit the compare
558 during initial expansion so that it may be deleted if unused.
559 (aarch64_gen_atomic_cas): Remove.
560 * config/aarch64/atomics.md (@aarch64_compare_and_swap<SHORT>_lse):
561 Change =&r to +r for operand 0; use match_dup for operand 2;
562 remove is_weak and mod_f operands as unused. Drop the split
564 (@aarch64_atomic_cas<SHORT>): ... this pattern's output; remove.
565 (@aarch64_compare_and_swap<GPI>_lse): Similarly.
566 (@aarch64_atomic_cas<GPI>): Similarly.
568 2018-10-31 Richard Biener <rguenther@suse.de>
570 * tree-eh.c (replace_trapping_overflow): Simplify ABS_EXPR case
573 2018-10-31 Sameera Deshpande <sameera.deshpande@linaro.org>
575 * config/aarch64/aarch64-cores.def (saphira): Use saphira pipeline.
576 * config/aarch64/aarch64.md: Include saphira.md
577 * config/aarch64/saphira.md: New file for pipeline description.
579 2018-10-30 Martin Sebor <msebor@redhat.com>
582 * gimple-ssa-sprintf.c (format_directive): Use %G to include
584 (sprintf_dom_walker::compute_format_length):
585 Avoid setting POSUNDER4K here.
586 (get_destination_size): Handle null argument values.
587 (get_user_idx_format): New function.
588 (sprintf_dom_walker::handle_gimple_call): Handle all printf-like
589 functions, including user-defined with attribute format printf.
590 Use %G to include inlining context.
593 2018-10-30 Jan Hubicka <jh@suse.cz>
595 * params.def (lto-partitions): Bump from 32 to 128.
597 2018-10-30 Jan Hubicka <jh@suse.cz>
600 (free_lang_data_d, add_tree_to_fld_list, fld_worklist_push): Move
602 (free_lang_data_in_type): Forward declare.
603 (fld_type_variant_equal_p): New function.
604 (fld_type_variant): New function
605 (fld_incomplete_types): New hash.
606 (fld_incomplete_type_of): New function
607 (fld_simplfied-type): New function.
608 (free_lang_data_in_decl): Add fld parameter; simplify type of FIELD_DECL
609 (free_lang_data): Allocate and free fld_incomplete_type; update call
610 of free_lang_data_in_decl.
612 2018-10-30 Eric Botcazou <ebotcazou@adacore.com>
614 * gcov.c (output_lines): Remove duplicate line.
616 2018-10-30 Aaron Sawdey <acsawdey@linux.ibm.com>
618 * config/rs6000/rs6000.md (bswapdi2): Force address into register
619 if not in indexed or indirect form.
620 (bswapdi2_load): Change predicate to indexed_or_indirect_operand.
621 (bswapdi2_store): Ditto.
622 * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem): New
624 * config/rs6000/rs6000-protos.h (rs6000_force_indexed_or_indirect_mem):
625 Prototype for helper function.
627 2018-10-30 Martin Sebor <msebor@redhat.com>
629 * doc/extend.texi (optimize): Clarify/expand attribute documentation.
630 (target, pragma GCC optimize, pragma GCC target): Ditto.
632 2018-10-30 Jonathan Wakely <jwakely@redhat.com>
634 * doc/extend.texi: Fix prototype and description of
635 __builtin_expect_with_probability.
637 2018-10-30 Michael Ploujnikov <michael.ploujnikov@oracle.com>
639 * cgraph.h (clone_function_name_1): Replaced by new
640 clone_function_name_numbered that takes name as string; for
641 privatize_symbol_name_1 use only.
642 (clone_function_name): Renamed to
643 clone_function_name_numbered to be explicit about numbering.
644 (clone_function_name): New two-argument function that does
645 not number its output.
646 (clone_function_name): New three-argument function that
647 takes a number to append to its output.
648 * cgraphclones.c (duplicate_thunk_for_node):
649 (clone_function_name_1): Renamed.
650 (clone_function_name_numbered): Two new functions.
651 (clone_function_name): Improved documentation.
652 (cgraph_node::create_virtual_clone): Use clone_function_name_numbered.
653 * config/rs6000/rs6000.c (make_resolver_func): Ditto.
654 * final.c (final_scan_insn_1): Use the new clone_function_name
656 * multiple_target.c (create_dispatcher_calls): Ditto.
657 (create_target_clone): Ditto.
658 * omp-expand.c (grid_expand_target_grid_body): Ditto.
659 * omp-low.c (create_omp_child_function_name): Ditto.
660 * omp-simd-clone.c (simd_clone_create): Ditto.
661 * symtab.c (simd_symtab_node::noninterposable_alias): Use the
662 new clone_function_name without numbering.
664 2018-10-30 Richard Earnshaw <rearnsha@arm.com>
666 * alloc-pool.h (base_pool_allocator <TBlockAllocator>::initialize):
667 Assert that the allocation size is not zero.
669 2018-10-30 Richard Biener <rguenther@suse.de>
671 PR tree-optimization/87800
672 * tree-vect-slp.c (vect_build_slp_tree_2): Reject any
673 non-induction or reduction PHIs.
675 2018-10-30 Sameera Deshpande <sameera.deshpande@linaro.org
677 * config/aarch64/falkor-tag-collision-avoidance.c
678 (execute_tag_collision_avoidance): Call df_note_add_problem.
680 2018-10-30 Martin Liska <mliska@suse.cz>
682 * doc/extend.texi: Fix typo in documentation
683 of __builtin_expect_with_probability.
685 2018-10-29 David Malcolm <dmalcolm@redhat.com>
688 * input.c (get_substring_ranges_for_loc): Detect if
689 linemap_resolve_location gives us a NULL map, and reject
692 2018-10-29 Iain Buclaw <ibuclaw@gdcproject.org>
694 * config.gcc (xstormy16-*-elf): Set tm_d_file.
696 2018-10-29 David Malcolm <dmalcolm@redhat.com>
697 Martin Sebor <msebor@redhat.com>
698 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
700 * doc/ux.texi (Quoting): New subsection, adapted from material at
701 https://gcc.gnu.org/wiki/DiagnosticsGuidelines written by
702 MartinSebor and ManuelLopezIbanez.
703 (Fix-it hints): Note that fix-it hints shouldn't be marked for
706 2018-10-29 Kugan Vivekanandarajah <kuganv@linaro.org>
709 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Fix niter
712 2018-10-29 Paul A. Clarke <pc@us.ibm.com>
714 * gcc.target/powerpc/ssse3-check.h: Remove inaccurate comment.
716 2018-10-29 Paul Koning <ni1d@arrl.net>
718 * doc/rtl.texi (CONST_WIDE_INT_ELT): Give correct macro name.
720 2018-10-29 Paul A. Clarke <pc@us.ibm.com>
722 * gcc/config/rs6000/mmintrin.h (_mm_packs_pi16, _mm_packs_pi32,
723 _mm_packs_pu16, _mm_unpackhi_pi8, _mm_unpacklo_pi8, _mm_add_pi8,
724 _mm_add_pi16, _mm_add_pi32, _mm_sub_pi8, _mm_sub_pi16, _mm_sub_pi32,
725 _mm_cmpgt_pi8, _mm_cmpeq_pi16, _mm_cmpgt_pi16, _mm_cmpeq_pi32,
726 _mm_cmpgt_pi32, _mm_adds_pi8, _mm_adds_pi16, _mm_adds_pu8,
727 _mm_adds_pu16, _mm_subs_pi8, _mm_subs_pi16, _mm_subs_pu8,
728 _mm_subs_pu16, _mm_madd_pi16, _mm_mulhi_pi16, _mm_mullo_pi16,
729 _mm_sll_pi16, _mm_sra_pi16, _mm_srl_pi16, _mm_set1_pi16, _mm_set1_pi8):
730 Change 'vector' to '__vector'.
731 * gcc/config/rs6000/xmmintrin.h (_mm_cvtps_pi32, _mm_cvttps_pi32,
732 _mm_cvtps_pi16, _mm_cvtps_pi8, _mm_max_pi16, _mm_max_pu8, _mm_min_pi16,
733 _mm_min_pu8, _mm_mulhi_pu16, _mm_shuffle_pi16, _mm_avg_pu8,
734 _mm_avg_pu16): Likewise. And, whitespace corrections.
736 2018-10-29 Richard Biener <rguenther@suse.de>
738 PR tree-optimization/87785
739 * tree-vect-slp.c (vect_gather_slp_loads): Only gather
742 2018-10-29 Olivier Hainque <hainque@adacore.com>
744 * VXWORKS_NET_LIBS_RTP: New macro, network part of VXWORKS_LIBS_RTP.
746 2018-10-29 Olivier Hainque <hainque@adacore.com>
748 * config/vxworks.h (VXWORKS_PERSONALITY): New VxWorks
749 ports configuration macro, defaults to "gnu".
750 (VXWORKS_OS_CPP_BUILTINS): Use it. Feed TOOL and
751 TOOL_FAMILY instead of _VX_TOOL/_VXTOOL_FAMILY.
753 2018-10-29 Olivier Hainque <hainque@adacore.com>
755 * config/vxworks.h (CLEAR_INSN_CACHE): #undef before #define.
757 2018-10-29 Richard Biener <rguenther@suse.de>
759 PR tree-optimization/87790
760 * tree-vect-slp.c (vect_mark_slp_stmts): Simplify.
761 (vect_make_slp_decision): Adjust.
762 (vect_slp_analyze_bb_1): Likewise.
763 (vect_detect_hybrid_slp_stmts): Properly union SLP type over
766 2018-10-29 Richard Biener <rguenther@suse.de>
768 PR tree-optimization/87785
769 * tree-vect-slp.c (vect_build_slp_tree_2): Remove loads argument
771 (vect_build_slp_tree): Likewise.
772 (vect_gather_slp_loads): New function.
773 (vect_analyze_slp_instance): Gather loads separately from the
776 2018-10-29 Martin Liska <mliska@suse.cz>
778 * Makefile.in: Make dependency to json.o.
779 * doc/gcov.texi: Document new JSON format, remove
780 old intermediate format documentation.
781 * gcov.c (struct function_info): Come up with m_name and
783 (function_info::function_info): Initialize it.
784 (function_info::~function_info): Release it.
785 (main): Rename flag_intermediate_format to flag_json_format.
786 (print_usage): Describe --json-format.
787 (process_args): Set flag_json_format.
788 (output_intermediate_line): Remove.
789 (output_intermediate_json_line): Likewise.
790 (get_gcov_intermediate_filename): Return new extension
792 (output_intermediate_file): Implement JSON emission.
793 (output_json_intermediate_file): Implement JSON emission.
794 (generate_results): Use ::get_name for function name.
795 Handle JSON output file.
796 (read_graph_file): Use ::get_name instead of cplus_demangle.
797 (read_count_file): Likewise.
798 (solve_flow_graph): Likewise.
799 (add_line_counts): Likewise.
800 (accumulate_line_counts): Use new flag_json_format.
801 (output_function_details): Use ::get_name instead of cplus_demangle.
802 (output_lines): Likewise.
803 * json.cc (test_writing_literals): Add new tests.
804 * json.h (class literal): Add new boolean constructor.
806 2018-10-29 Segher Boessenkool <segher@kernel.crashing.org>
808 PR rtl-optimization/87701
809 PR rtl-optimization/87780
810 * combine.c (make_more_copies): Rewrite.
812 2018-10-28 Kugan Vivekanandarajah <kuganv@linaro.org>
814 * doc/generic.texi (ABSU_EXPR): Document.
815 * match.pd (absu(x)*absu(x) -> x*x): Handle.
816 (absu(absu(X)) -> absu(X)): Likewise.
817 (absu(-X) -> absu(X)): Likewise.
818 (absu(X) where X is nonnegative -> X): Likewise.
820 2018-10-28 Iain Buclaw <ibuclaw@gdcproject.org>
822 * Makefile.in (tm_d_file_list, tm_d_include_list): New variables.
823 (TM_D_H, D_TARGET_DEF, D_TARGET_H, D_TARGET_OBJS): New variables.
824 (tm_d.h, cs-tm_d.h, default-d.o): New rules.
825 (d/d-target-hooks-def.h, s-d-target-hooks-def-h): New rules.
826 (s-tm-texi): Also check timestamp on d-target.def.
827 (generated_files): Add TM_D_H and d-target-hooks-def.h.
828 (build/genhooks.o): Also depend on D_TARGET_DEF.
829 * config.gcc (tm_d_file, d_target_objs, target_has_targetdm): New
831 * config/aarch64/aarch64-d.c: New file.
832 * config/aarch64/aarch64-linux.h (GNU_USER_TARGET_D_CRITSEC_SIZE):
834 * config/aarch64/aarch64-protos.h (aarch64_d_target_versions): New
836 * config/aarch64/aarch64.h (TARGET_D_CPU_VERSIONS): Define.
837 * config/aarch64/t-aarch64 (aarch64-d.o): New rule.
838 * config/arm/arm-d.c: New file.
839 * config/arm/arm-protos.h (arm_d_target_versions): New prototype.
840 * config/arm/arm.h (TARGET_D_CPU_VERSIONS): Define.
841 * config/arm/linux-eabi.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
842 * config/arm/t-arm (arm-d.o): New rule.
843 * config/default-d.c: New file.
844 * config/glibc-d.c: New file.
845 * config/gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
846 * config/i386/i386-d.c: New file.
847 * config/i386/i386-protos.h (ix86_d_target_versions): New prototype.
848 * config/i386/i386.h (TARGET_D_CPU_VERSIONS): Define.
849 * config/i386/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
850 (GNU_USER_TARGET_D_CRITSEC_SIZE): Define.
851 * config/i386/t-i386 (i386-d.o): New rule.
852 * config/kfreebsd-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
853 * config/kopensolaris-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
854 * config/linux-android.h (ANDROID_TARGET_D_OS_VERSIONS): Define.
855 * config/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
856 * config/mips/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
857 * config/mips/mips-d.c: New file.
858 * config/mips/mips-protos.h (mips_d_target_versions): New prototype.
859 * config/mips/mips.h (TARGET_D_CPU_VERSIONS): Define.
860 * config/mips/t-mips (mips-d.o): New rule.
861 * config/powerpcspe/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
862 * config/powerpcspe/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
863 * config/powerpcspe/powerpcspe-d.c: New file.
864 * config/powerpcspe/powerpcspe-protos.h (rs6000_d_target_versions):
866 * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
867 Support GNU D by using 0 as the language type.
868 * config/powerpcspe/powerpcspe.h (TARGET_D_CPU_VERSIONS): Define.
869 * config/powerpcspe/t-powerpcspe (powerpcspe-d.o): New rule.
870 * config/riscv/riscv-d.c: New file.
871 * config/riscv/riscv-protos.h (riscv_d_target_versions): New
873 * config/riscv/riscv.h (TARGET_D_CPU_VERSIONS): Define.
874 * config/riscv/t-riscv (riscv-d.o): New rule.
875 * config/rs6000/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
876 * config/rs6000/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
877 * config/rs6000/rs6000-d.c: New file.
878 * config/rs6000/rs6000-protos.h (rs6000_d_target_versions): New
880 * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
881 Support GNU D by using 0 as the language type.
882 * config/rs6000/rs6000.h (TARGET_D_CPU_VERSIONS): Define.
883 * config/rs6000/t-rs6000 (rs6000-d.o): New rule.
884 * config/s390/s390-d.c: New file.
885 * config/s390/s390-protos.h (s390_d_target_versions): New prototype.
886 * config/s390/s390.h (TARGET_D_CPU_VERSIONS): Define.
887 * config/s390/t-s390 (s390-d.o): New rule.
888 * config/sparc/sparc-d.c: New file.
889 * config/sparc/sparc-protos.h (sparc_d_target_versions): New
891 * config/sparc/sparc.h (TARGET_D_CPU_VERSIONS): Define.
892 * config/sparc/t-sparc (sparc-d.o): New rule.
893 * config/t-glibc (glibc-d.o): New rule.
894 * configure: Regenerated.
895 * configure.ac (tm_d_file): New variable.
896 (tm_d_file_list, tm_d_include_list, d_target_objs): Add substitutes.
897 * doc/contrib.texi (Contributors): Add self for the D frontend.
898 * doc/frontends.texi (G++ and GCC): Mention D as a supported language.
899 * doc/install.texi (Configuration): Mention libphobos as an option for
900 --enable-shared. Mention d as an option for --enable-languages.
901 (Testing): Mention check-d as a target.
902 * doc/invoke.texi (Overall Options): Mention .d, .dd, and .di as file
903 name suffixes. Mention d as a -x option.
904 * doc/sourcebuild.texi (Top Level): Mention libphobos.
905 * doc/standards.texi (Standards): Add section on D language.
906 * doc/tm.texi: Regenerated.
907 * doc/tm.texi.in: Add @node for D language and ABI, and @hook for
908 TARGET_CPU_VERSIONS, TARGET_D_OS_VERSIONS, and TARGET_D_CRITSEC_SIZE.
909 * dwarf2out.c (is_dlang): New function.
910 (gen_compile_unit_die): Use DW_LANG_D for D.
911 (declare_in_namespace): Return module die for D, instead of adding
912 extra declarations into the namespace.
913 (gen_namespace_die): Generate DW_TAG_module for D.
914 (gen_decl_die): Handle CONST_DECLSs for D.
915 (dwarf2out_decl): Likewise.
916 (prune_unused_types_walk_local_classes): Handle DW_tag_interface_type.
917 (prune_unused_types_walk): Handle DW_tag_interface_type same as other
919 * gcc.c (default_compilers): Add entries for .d, .dd and .di.
920 * genhooks.c: Include d/d-target.def.
922 2018-10-28 Iain Sandoe <iain@sandoe.co.uk>
925 * config/rs6000/darwin.h (STACK_BOUNDARY): New.
926 (RS6000_STARTING_FRAME_OFFSET): Adjust to preserve 16byte alignment.
927 (STACK_DYNAMIC_OFFSET): Likewise.
929 2018-10-27 Sandra Loosemore <sandra@codesourcery.com>
932 * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Fix
935 2018-10-26 Tulio Magno Quites Machado Filho <tuliom@linux.ibm.com>
937 * doc/extend.texi (PowerPC builtins): Fix __builtin_unpack_ibm128
938 return type and other typos.
940 2018-10-26 Aaron Sawdey <acsawdey@linux.ibm.com>
942 * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Change to
943 a shorter sequence with fewer branches.
944 (emit_final_str_compare_gpr): Ditto.
946 2018-10-26 Paul A. Clarke <pc@us.ibm.com>
948 * config/rs6000/tmmintrin.h: New file.
949 * config.gcc (powerpc*-*-*): Add tmmintrin.h to extra_headers.
951 2018-10-26 Paul A. Clarke <pc@us.ibm.com>
953 * config/rs6000/mmintrin.h: Enable 32bit compilation.
954 * config/rs6000/xmmintrin.h: Likewise.
956 2018-10-26 Paul A. Clarke <pc@us.ibm.com>
958 * config/rs6000/xmmintrin.h (_mm_extract_pi16): Fix for big-endian.
960 2018-10-26 Richard Biener <rguenther@suse.de>
962 * tree-vect-slp.c (vect_mark_slp_stmts): Add visited hash_set
964 (vect_mark_slp_stmts_relevant): Likewise.
965 (vect_detect_hybrid_slp_stmts): Likewise.
966 (vect_bb_slp_scalar_cost): Likewise.
967 (vect_remove_slp_scalar_calls): Likewise.
969 2018-10-26 Jan Hubicka <jh@suse.cz>
971 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix recursion.
972 (warn_types_mismatch): Fix walk of DECL_NAME.
973 (odr_types_equivalent_p): Fix overactive assert.
975 2018-10-26 Richard Biener <rguenther@suse.de>
977 PR tree-optimization/87105
978 * tree-vectorizer.h (_slp_tree::refcnt): New member.
979 * tree-vect-slp.c (vect_free_slp_tree): Decrement and honor
981 (vect_create_new_slp_node): Initialize refcnt to one.
983 (scalar_stmts_set_t, bst_fail): Remove.
984 (vect_build_slp_tree_2): Add bst_map argument and adjust calls.
985 (vect_build_slp_tree): Add bst_map argument and lookup
986 already created SLP nodes.
987 (vect_print_slp_tree): Handle a SLP graph, print SLP node
989 (vect_slp_rearrange_stmts): Handle a SLP graph.
990 (vect_analyze_slp_instance): Adjust and free SLP nodes from
991 the CSE map. Fix indenting.
992 (vect_schedule_slp_instance): Add short-cut.
994 2018-10-26 Martin Liska <mliska@suse.cz>
997 * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Hash with
998 addr_expr and not with pointers.
1000 2018-10-26 Jan Hubicka <jh@suse.cz>
1002 * tree.c (free_lang_data_in_type): Only check main variants.
1003 * ipa-devirt.c (warn_odr): Make static.
1004 (types_same_for_odr): Drop strict variant.
1005 (types_odr_comparable): Likewise.
1006 (odr_or_derived_type_p): Look for main variants.
1007 (odr_name_hasher::equal): Cleanup comment.
1008 (odr_subtypes_equivalent): Add warn and warned arguments; check main
1010 (type_variants_equivalent_p): break out from ...
1011 (odr_types_equivalent): ... here; go for main variants where needed.
1012 (warn_odr): ... here; turn static.
1013 (warn_types_mismatch): Compare mangled names of main variants.
1014 * ipa-utils.h (types_odr_comparable): Drop strict parameter.
1015 (type_with_linkage_p): Sanity check that we look at main variant.
1016 * lto.c (lto_read_decls): Only consider main variant to be ODR type.
1017 * tree.h (types_same_for_odr): Drop strict argument.
1019 2018-10-26 Richard Biener <rguenther@suse.de>
1021 PR tree-optimization/87746
1022 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
1023 Simplify and fix WRT strided store groups with size not
1024 equal to step in element count.
1025 (vect_analyze_group_access_1): Dump the whole group.
1027 2018-10-25 Carl Love <cel@us.ibm.com>
1029 * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_VSCEDPGT,
1030 P9V_BUILTIN_VEC_VSCEDPLT, P9V_BUILTIN_VEC_VSCEDPEQ,
1031 P9V_BUILTIN_VEC_VSCEDPUO): Rename base overloaded name. Add quad
1032 precicion entry for each overloaded builtin.
1033 * config/rs6000/rs6000-builtin.def (VSCEDPGT, VSCEDPLT, VSCEDPEQ,
1034 VSCEDPUO): Rename overloaded name.
1035 (VSCEDPGT, VSCEQPGT, VSCEDPLT, VSCEQPLT, VSCEDPEQ, VSCEQPEQ,
1036 VSCEDPUO, VSCEQPUO): Add defitions for overloaded builtins.
1037 * config/rs6000/vsx.md (xscmpexpqp_<code>_<mode>): Add
1038 define_expand for xscmpexqp instruction.
1039 (*xscmpexpqp): Add define_insn for the xscmpexqp instruction.
1041 2018-10-25 Bill Schmidt <wschmidt@linux.ibm.com>
1042 Jinsong Ji <jji@us.ibm.com>
1044 * config/rs6000/emmintrin.h (_mm_slli_epi16): Replace deprecated
1045 function with vec_sl.
1046 (_mm_slli_epi32): Likewise.
1047 (_mm_slli_epi64): Likewise.
1048 (_mm_srai_epi16): Replace deprecated function with vec_sra.
1049 (_mm_srai_epi32): Likewise.
1050 (_mm_srli_epi16): Replace deprecated function with vec_sr.
1051 (_mm_srli_epi32): Likewise.
1052 (_mm_srli_epi64): Likewise.
1053 (_mm_sll_epi16): Replace deprecated function with vec_sl.
1054 (_mm_sll_epi32): Likewise.
1055 (_mm_sll_epi64): Likewise.
1056 (_mm_sra_epi16): Replace deprecated function with vec_sra.
1057 (_mm_sra_epi32): Likewise.
1058 (_mm_srl_epi16): Replace deprecated function with vec_sr.
1059 (_mm_srl_epi32): Likewise.
1060 (_mm_srl_epi64): Likewise.
1062 2018-10-25 Bill Schmidt <wschmidt@linux.ibm.com>
1063 Jinsong Ji <jji@us.ibm.com>
1065 * gcc/config/rs6000/emmintrin.h (_mm_sll_epi16): Replace
1066 comparison operators with vec_cmp* for compatibility due to
1067 unfortunate history; clean up formatting and use types more
1069 (_mm_sll_epi32): Likewise.
1070 (_mm_sll_epi64): Likewise.
1071 (_mm_srl_epi16): Likewise.
1072 (_mm_srl_epi32): Likewise.
1073 (_mm_srl_epi64): Likewise.
1075 2018-10-25 Bill Schmidt <wschmidt@linux.ibm.com>
1076 Jinsong Ji <jji@us.ibm.com>
1078 * config/rs6000/emmintrin.h (_mm_sll_epi64): Remove wrong cast.
1079 * config/rs6000/xmmintrin.h (_mm_min_ps): Change m's type to
1080 __vector __bool int. Use vec_cmpgt in preference to deprecated
1081 function vec_vcmpgtfp.
1082 (_mm_max_ps): Likewise.
1084 2018-10-25 Jeff Law <law@redhat.com>
1086 * config/rl78/rl78.c (insn_ok_now): Always re-recognize the insn
1089 2018-10-25 Martin Sebor <msebor@redhat.com>
1091 * doc/extend.texi (aligned): Expand attribute description.
1092 (Alignment): Rename section. Discuss function arguments.
1094 2018-10-25 Jan Hubicka <jh@suse.cz>
1096 * ipa-devirt.c (main_odr_variant): Remove.
1097 (hash_odr_name, types_same_for_odr, types_odr_comparable,
1098 odr_name_hasher::equal, odr_subtypes_equivalent_p):
1099 Drop use of main_odr_variant.
1100 (add_type_duplicate): Silence confused warnings on integer types.
1101 (get_odr_type): Always look for main variant.
1102 (register_odr_type): Simplify.
1104 2018-10-25 Richard Biener <rguenther@suse.de>
1106 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses):
1107 Initialize ng to silence error with release checking bootstrap.
1109 2018-10-25 Richard Biener <rguenther@suse.de>
1111 * tree-if-conv.c: Include tree-ssa-sccvn.h.
1112 (tree_if_conversion): Run CSE on the if-converted loop body.
1114 2018-10-25 Ilya Leoshkevich <iii@linux.ibm.com>
1116 * config/s390/constraints.md (ZL): New constraint.
1117 * config/s390/s390.c (legitimate_pic_operand_p): Accept LARL
1119 * config/s390/s390.md (movdi_larl): Remove.
1120 (movdi_64): Add the LARL alternative.
1122 2018-10-25 Ilya Leoshkevich <iii@linux.ibm.com>
1125 * rtl.c (RTX_CODE_HWINT_P_1): New helper macro.
1126 (RTX_CODE_HWINT_P): New macro.
1127 (rtx_code_size): Use RTX_CODE_HWINT_P ().
1129 2018-10-25 Jan Hubicka <jh@suse.cz>
1131 * ipa-devirt.c (odr_types_equivalent_p): Do not ICE if one of types
1134 2018-10-25 Richard Biener <rguenther@suse.de>
1136 PR tree-optimization/87665
1137 PR tree-optimization/87745
1138 * tree-vectorizer.h (get_earlier_stmt): Remove.
1139 (get_later_stmt): Pick up UID from the original non-pattern stmt.
1141 2018-10-25 Sam Tebbs <sam.tebbs@arm.com>
1143 * options.texi (Deprecated): Move list to Var section.
1145 2018-10-24 Bill Schmidt <wschmidt@linux.ibm.com>
1146 Jinsong Ji <jji@us.ibm.com>
1148 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Change deprecated
1149 __vector long to __vector long long.
1150 (_mm_cvtpd_ps): Likewise.
1151 (_mm_cvttpd_epi32): Likewise.
1152 (_mm_cvtpi32_pd): Likewise.
1153 (_mm_unpackhi_epi64): Likewise.
1154 (_mm_unpacklo_epi64): Likewise.
1156 2018-10-24 Segher Boessenkool <segher@kernel.crashing.org>
1158 PR rtl-optimization/87720
1159 * combine.c (make_more_copies): Skip if the dest is pc_rtx.
1161 2018-10-24 Alexandre Oliva <aoliva@redhat.com>
1163 * gimple-ssa-isolate-paths.c
1164 (find_implicit_erroneous_behavior): Do not change code if the
1165 pass is running for warnings only.
1166 (find_explicit_erroneous_behavior): Likewise.
1168 2018-10-24 Michael Meissner <meissner@linux.ibm.com>
1170 * config/rs6000/rs6000.c (TARGET_MANGLE_DECL_ASSEMBLER_NAME):
1171 Define as rs6000_mangle_decl_assembler_name.
1172 (rs6000_mangle_decl_assembler_name): If the user switched from IBM
1173 long double to IEEE long double, switch the names of the long
1174 double built-in functions to be <func>f128 instead of <func>l.
1176 2018-10-24 Martin Sebor <msebor@redhat.com>
1178 * doc/extend.texi (nonnull): List no-argument form. Reference
1179 -fno-delete-null-pointer-checks and -fisolate-erroneous-paths-attribute.
1181 2018-10-24 Richard Biener <rguenther@suse.de>
1183 * tree-ssa-sccvn.c (do_rpo_vn): Free rpo_state.
1185 2018-10-24 Martin Liska <mliska@suse.cz>
1187 PR tree-optimization/84436
1188 * tree-switch-conversion.c (switch_conversion::contains_same_values_p):
1190 (switch_conversion::contains_linear_function_p): New.
1191 (switch_conversion::build_one_array): Support linear
1192 transformation on input.
1193 * tree-switch-conversion.h (struct switch_conversion): Add
1194 contains_linear_function_p declaration.
1196 2018-10-24 Richard Biener <rguenther@suse.de>
1198 * varasm.c (const_hash_1): Return hash of ADDR_EXPR
1199 if its argument is CONSTANT_CLASS_P.
1201 2018-10-11 Jan Hubicka <hubicka@ucw.cz>
1203 * ipa-utils.h (type_with_linkage_p): No longer check for TYPE_STUB_DECL;
1204 it is wrong for forward declarations.
1206 2018-10-24 Ilya Leoshkevich <iii@linux.ibm.com>
1208 * config/s390/s390.c (s390_check_qrst_address): Add the missing
1209 SYMBOL_REF_P () check.
1211 2018-10-24 Richard Biener <rguenther@suse.de>
1213 PR tree-optimization/87105
1214 * tree-vect-data-refs.c (vect_analyze_group_access_1): Adjust
1215 dump classification.
1216 (vect_analyze_data_ref_accesses): Handle duplicate loads and
1217 stores by splitting the affected group after the fact.
1218 * tree-vect-slp.c (vect_build_slp_tree_2): Dump when we
1219 fail the SLP build because of size constraints.
1221 2018-10-24 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1223 * configure.ac (gcc_cv_ld_aligned_shf_merge): New test.
1224 * configure: Regenerate.
1225 * config.in: Regenerate.
1226 * varasm.c (mergeable_string_section): Use readonly_data_section
1227 if linker doesn't support SHF_MERGE with alignment > 8.
1228 (mergeable_constant_section): Likewise.
1230 2018-10-24 Richard Biener <rguenther@suse.de>
1232 PR tree-optimization/84013
1233 * tree-ssa-structalias.c (struct msdi_data): New struct for
1234 marshalling data to walk_stmt_load_store_ops.
1235 (maybe_set_dependence_info): Refactor as callback for
1236 walk_stmt_load_store_ops.
1237 (compute_dependence_clique): Set restrict info on all stmt kinds.
1239 2018-10-24 Martin Liska <mliska@suse.cz>
1241 * cgraph.c (cgraph_node::dump):
1242 Remove reduntant dumps and make tp_first_run dump more compact.
1244 2018-10-24 Richard Biener <rguenther@suse.de>
1246 PR tree-optimization/87665
1247 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Adjust
1250 2018-10-12 Jeff Law <law@redhat.com>
1252 * config/h8300/h8300.c (h8300_expand_prologue): Fix stm generation
1255 2018-10-23 Richard Biener <rguenther@suse.de>
1257 * tree-vrp.c (add_assert_info): Guard dump_printf with
1259 * gimple-ssa-evrp-analyze.c
1260 (evrp_range_analyzer::record_ranges_from_incoming_edge):
1261 Use value_range::ignore_equivs_equal_p.
1263 2018-10-23 Richard Biener <rguenther@suse.de>
1265 PR tree-optimization/87105
1266 PR tree-optimization/87608
1267 * passes.def (pass_all_early_optimizations): Add early phi-opt
1269 * tree-ssa-phiopt.c (value_replacement): Ignore NOPs and predicts in
1270 addition to debug stmts.
1271 (tree_ssa_phiopt_worker): Add early_p argument, do only min/max
1272 and abs replacement early.
1273 * tree-cfg.c (gimple_empty_block_p): Likewise.
1275 2018-10-23 Richard Earnshaw <rearnsha@arm.com>
1278 * config.gcc (arm*-*-netbsdelf*): Default to StrongARM if no CPU
1279 specified to configure.
1280 (arm*-*-*): Use ARM7TDMI as the target CPU if no default provided.
1282 2018-10-23 Richard Biener <rguenther@suse.de>
1284 PR tree-optimization/87700
1285 * tree-ssa-copy.c (set_copy_of_val): Fix change detection logic.
1287 2018-10-23 Jakub Jelinek <jakub@redhat.com>
1290 * config/i386/avx512vlintrin.h (_mm_mask_mullo_epi32): Change type of
1291 second argument from __mmask16 to __mmask8.
1292 * config/i386/avx512vlbwintrin.h (_mm_mask_packus_epi32,
1293 _mm_mask_packs_epi32): Likewise.
1294 * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i64scatter_ps):
1296 (_mm512_mask_prefetch_i64scatter_pd): Likewise. Formatting fix.
1298 2018-10-23 Richard Biener <rguenther@suse.de>
1300 * tree-vect-stmts.c (vect_analyze_stmt): Fix typo in comment.
1302 2018-10-23 Richard Biener <rguenther@suse.de>
1304 PR tree-optimization/86144
1305 * tree-vect-stmts.c (vect_analyze_stmt): Prefer -mveclibabi
1306 over simd attribute.
1308 2018-10-23 Richard Biener <rguenther@suse.de>
1310 PR tree-optimization/87693
1311 * tree-ssa-threadedge.c (thread_around_empty_blocks): Handle
1312 the case we do not find the taken edge.
1314 2018-10-22 Bill Schmidt <wschmidt@linux.ibm.com>
1315 Jinsong Ji <jji@us.ibm.com>
1317 * config/rs6000/emmintrin.h (_MM_SHUFFLE2): Comment cleanup.
1318 (_mm_store_pd): Use unaligned vector type for pointer cast.
1319 (_mm_maskmoveu_si128): Likewise.
1320 * config/rs6000/xmmintrin.h (__m128_u): New typedef.
1321 (_mm_store_ps): Use unaligned vector type for pointer cast.
1323 2018-10-22 Paul Koning <ni1d@arrl.net>
1325 * symtab.c (symtab_node::increase_alignment): Correct max
1328 2018-10-22 Yury Gribov <tetra2005@gmail.com>
1330 PR tree-optimization/87633
1331 * match.pd: Do not generate unordered integer comparisons.
1333 2018-10-22 Segher Boessenkool <segher@kernel.crashing.org>
1335 PR rtl-optimization/87600
1336 * combine.c: Add include of expr.h.
1337 (cant_combine_insn_p): Do not combine moves from any hard non-fixed
1338 register to a pseudo.
1339 (make_more_copies): New function, add a copy to a new pseudo after
1340 the moves from hard registers into pseudos.
1341 (rest_of_handle_combine): Declare rebuild_jump_labels_after_combine
1342 later. Call make_more_copies.
1344 2018-10-22 Andrew Stubbs <ams@codesourcery.com>
1346 * lra-constraints.c (process_alt_operands): New local array,
1347 matching_early_clobber. Check matching_early_clobber before
1348 decrementing reject, and set matching_early_clobber after.
1350 2018-10-22 Segher Boessenkool <segher@kernel.crashing.org>
1353 * config/rs6000/rs6000.c (print_operand_address): For unexpected RTL
1354 call output_addr_const and hope for the best.
1356 2018-10-22 Richard Biener <rguenther@suse.de>
1358 * gimple-ssa-evrp-analyze.c
1359 (evrp_range_analyzer::record_ranges_from_incoming_edge): Be
1360 smarter about what ranges to use.
1361 * tree-vrp.c (add_assert_info): Dump here.
1362 (register_edge_assert_for_2): Instead of here at multiple but
1365 * gcc.dg/tree-ssa/evrp12.c: New testcase.
1366 * gcc.dg/predict-6.c: Adjust.
1367 * gcc.dg/tree-ssa/vrp33.c: Disable EVRP.
1368 * gcc.dg/tree-ssa/vrp02.c: Likewise.
1369 * gcc.dg/tree-ssa/cunroll-9.c: Likewise.
1371 2018-10-22 Steven Bosscher <steven@gcc.gnu.org>
1372 Richard Biener <rguenther@suse.de>
1374 * bitmap.h: Update data structure documentation, including a
1375 description of bitmap views as either linked-lists or splay trees.
1376 (struct bitmap_element_def): Update comments for splay tree bitmaps.
1377 (struct bitmap_head_def): Likewise.
1378 (bitmap_list_view, bitmap_tree_view): New prototypes.
1379 (bitmap_initialize_stat): Initialize a bitmap_head's indx and
1381 (bmp_iter_set_init): Assert the iterated bitmaps are in list form.
1382 (bmp_iter_and_init, bmp_iter_and_compl_init): Likewise.
1383 * bitmap.c (bitmap_elem_to_freelist): Unregister overhead of a
1384 released bitmap element here.
1385 (bitmap_element_free): Remove.
1386 (bitmap_elt_clear_from): Work on splay tree bitmaps.
1387 (bitmap_list_link_element): Renamed from bitmap_element_link. Move
1388 this function similar ones such that linked-list bitmap implementation
1389 functions are grouped.
1390 (bitmap_list_unlink_element): Renamed from bitmap_element_unlink,
1391 and moved for grouping.
1392 (bitmap_list_insert_element_after): Renamed from
1393 bitmap_elt_insert_after, and moved for grouping.
1394 (bitmap_list_find_element): New function spliced from bitmap_find_bit.
1395 (bitmap_tree_link_left, bitmap_tree_link_right,
1396 bitmap_tree_rotate_left, bitmap_tree_rotate_right, bitmap_tree_splay,
1397 bitmap_tree_link_element, bitmap_tree_unlink_element,
1398 bitmap_tree_find_element): New functions for splay-tree bitmap
1400 (bitmap_element_link, bitmap_element_unlink, bitmap_elt_insert_after):
1401 Renamed and moved, see above entries.
1402 (bitmap_tree_listify_from): New function to convert part of a splay
1403 tree bitmap to a linked-list bitmap.
1404 (bitmap_list_view): Convert a splay tree bitmap to linked-list form.
1405 (bitmap_tree_view): Convert a linked-list bitmap to splay tree form.
1406 (bitmap_find_bit): Remove.
1407 (bitmap_clear, bitmap_clear_bit, bitmap_set_bit,
1408 bitmap_single_bit_set_p, bitmap_first_set_bit, bitmap_last_set_bit):
1409 Handle splay tree bitmaps.
1410 (bitmap_copy, bitmap_count_bits, bitmap_and, bitmap_and_into,
1411 bitmap_elt_copy, bitmap_and_compl, bitmap_and_compl_into,
1412 bitmap_compl_and_into, bitmap_elt_ior, bitmap_ior, bitmap_ior_into,
1413 bitmap_xor, bitmap_xor_into, bitmap_equal_p, bitmap_intersect_p,
1414 bitmap_intersect_compl_p, bitmap_ior_and_compl,
1415 bitmap_ior_and_compl_into, bitmap_set_range, bitmap_clear_range,
1416 bitmap_hash): Reject trying to act on splay tree bitmaps. Make
1417 corresponding changes to use linked-list specific bitmap_element
1418 manipulation functions as applicable for efficiency.
1419 (bitmap_tree_to_vec): New function.
1420 (debug_bitmap_elt_file): New function split out from ...
1421 (debug_bitmap_file): ... here. Handle splay tree bitmaps.
1422 (bitmap_print): Likewise.
1424 PR tree-optimization/63155
1425 * tree-ssa-propagate.c (ssa_prop_init): Use tree-view for the
1427 * tree-ssa-coalesce.c (coalesce_ssa_name): Populate used_in_copies
1430 2018-10-22 Martin Liska <mliska@suse.cz>
1432 PR tree-optimization/87686
1434 2018-08-29 Martin Liska <mliska@suse.cz>
1436 * tree-switch-conversion.c (switch_conversion::expand):
1437 Strenghten assumption about gswitch statements.
1439 2018-10-22 Martin Liska <mliska@suse.cz>
1441 * ipa-icf.c (sem_item::compare_attributes): Remove.
1442 (sem_item::compare_referenced_symbol_properties): Use
1443 attribute_list_equal instead.
1444 (sem_function::equals_wpa): Likewise.
1445 * ipa-icf.h: Remove compare_attributes.
1447 2018-10-22 Richard Biener <rguenther@suse.de>
1450 * mem-stats.h (mem_usage::operator==): Fix pasto.
1452 2018-10-22 Richard Biener <rguenther@suse.de>
1454 PR tree-optimization/87640
1455 * tree-vrp.c (set_value_range_with_overflow): Decompose
1457 (extract_range_from_binary_expr_1): Adjust.
1459 2018-10-22 Martin Jambor <mjambor@suse.cz>
1461 * tree-eh.h (stmt_could_throw_p): Add function parameter.
1462 (stmt_can_throw_external): Likewise.
1463 (stmt_can_throw_internal): Likewise.
1464 * tree-eh.c (lower_eh_constructs_2): Pass cfun to stmt_could_throw_p.
1465 (lower_eh_constructs_2): Likewise.
1466 (stmt_could_throw_p): Add fun parameter, use it instead of cfun.
1467 (stmt_can_throw_external): Likewise.
1468 (stmt_can_throw_internal): Likewise.
1469 (maybe_clean_eh_stmt_fn): Pass cfun to stmt_could_throw_p.
1470 (maybe_clean_or_replace_eh_stmt): Pass cfun to stmt_could_throw_p.
1471 (maybe_duplicate_eh_stmt_fn): Pass new_fun to stmt_could_throw_p.
1472 (maybe_duplicate_eh_stmt): Pass cfun to stmt_could_throw_p.
1473 (pass_lower_eh_dispatch::execute): Pass cfun to
1474 stmt_can_throw_external.
1475 (cleanup_empty_eh): Likewise.
1476 (verify_eh_edges): Pass cfun to stmt_could_throw_p.
1477 * cgraph.c (cgraph_edge::set_call_stmt): Pass a function to
1478 stmt_can_throw_external instead of pushing it to cfun.
1479 (symbol_table::create_edge): Likewise.
1480 * gimple-fold.c (fold_builtin_atomic_compare_exchange): Pass cfun to
1481 stmt_can_throw_internal.
1482 * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Pass cfun
1483 to stmt_could_throw_p.
1484 * gimple-ssa-store-merging.c (handled_load): Pass cfun to
1485 stmt_can_throw_internal.
1486 (pass_store_merging::execute): Likewise.
1487 * gimple-ssa-strength-reduction.c
1488 (find_candidates_dom_walker::before_dom_children): Pass cfun to
1490 * gimplify-me.c (gimple_regimplify_operands): Pass cfun to
1491 stmt_can_throw_internal.
1492 * ipa-pure-const.c (check_call): Pass cfun to stmt_could_throw_p and
1493 to stmt_can_throw_external.
1494 (check_stmt): Pass cfun to stmt_could_throw_p.
1495 (check_stmt): Pass cfun to stmt_can_throw_external.
1496 (pass_nothrow::execute): Likewise.
1497 * trans-mem.c (expand_call_tm): Pass cfun to stmt_can_throw_internal.
1498 * tree-cfg.c (is_ctrl_altering_stmt): Pass cfun to
1499 stmt_can_throw_internal.
1500 (verify_gimple_in_cfg): Pass cfun to stmt_could_throw_p.
1501 (stmt_can_terminate_bb_p): Pass cfun to stmt_can_throw_external.
1502 (gimple_purge_dead_eh_edges): Pass cfun to stmt_can_throw_internal.
1503 * tree-complex.c (expand_complex_libcall): Pass cfun to
1504 stmt_could_throw_p and to stmt_can_throw_internal.
1505 (expand_complex_multiplication): Pass cfun to stmt_can_throw_internal.
1506 * tree-inline.c (copy_edges_for_bb): Likewise.
1507 (maybe_move_debug_stmts_to_successors): Likewise.
1508 * tree-outof-ssa.c (ssa_is_replaceable_p): Pass cfun to
1510 * tree-parloops.c (oacc_entry_exit_ok_1): Likewise.
1511 * tree-sra.c (scan_function): Pass cfun to stmt_can_throw_external.
1512 * tree-ssa-alias.c (stmt_kills_ref_p): Pass cfun to
1513 stmt_can_throw_internal.
1514 * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
1515 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Pass cfun to
1517 (mark_aliased_reaching_defs_necessary_1): Pass cfun to
1518 stmt_can_throw_internal.
1519 * tree-ssa-forwprop.c (pass_forwprop::execute): Likewise.
1520 * tree-ssa-loop-im.c (movement_possibility): Pass cfun to
1522 * tree-ssa-loop-ivopts.c (find_givs_in_stmt_scev): Likewise.
1523 (add_autoinc_candidates): Pass cfun to stmt_can_throw_internal.
1524 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
1525 (convert_mult_to_fma_1): Likewise.
1526 (convert_to_divmod): Likewise.
1527 * tree-ssa-phiprop.c (propagate_with_phi): Likewise.
1528 * tree-ssa-pre.c (compute_avail): Pass cfun to stmt_could_throw_p.
1529 * tree-ssa-propagate.c
1530 (substitute_and_fold_dom_walker::before_dom_children): Likewise.
1531 * tree-ssa-reassoc.c (suitable_cond_bb): Likewise.
1532 (maybe_optimize_range_tests): Likewise.
1533 (linearize_expr_tree): Likewise.
1534 (reassociate_bb): Likewise.
1535 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
1536 * tree-ssa-scopedtables.c (hashable_expr_equal_p): Likewise.
1537 * tree-ssa-strlen.c (adjust_last_stmt): Likewise.
1538 (handle_char_store): Likewise.
1539 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Pass cfun to
1540 stmt_can_throw_internal.
1541 * tree-vect-patterns.c (check_bool_pattern): Pass cfun to
1543 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
1544 (vectorizable_call): Pass cfun to stmt_can_throw_internal.
1545 (vectorizable_simd_clone_call): Likewise.
1546 * value-prof.c (gimple_ic): Pass cfun to stmt_could_throw_p.
1547 (gimple_stringop_fixed_value): Likewise.
1549 2018-10-22 Ilya Leoshkevich <iii@linux.ibm.com>
1551 * config/s390/s390.c (s390_loadrelative_operand_p): Accept
1552 literal pool references.
1553 (s390_check_qrst_address): Adapt to the new behavior of
1554 s390_loadrelative_operand_p ().
1556 2018-10-22 H.J. Lu <hongjiu.lu@intel.com>
1559 * config/i386/sse.md (*andnot<mode>3_bcst): New.
1561 2018-10-22 H.J. Lu <hongjiu.lu@intel.com>
1564 * config/i386/sse.md (*<code><mode>3_bcst): New.
1566 2018-10-22 H.J. Lu <hongjiu.lu@intel.com>
1569 * config/i386/sse.md (avx512bcst): Updated for V4SI, V2DI, V8SI,
1570 V4DI, V16SI and V8DI.
1571 (*sub<mode>3<mask_name>_bcst): New.
1572 (*add<mode>3<mask_name>_bcst): Likewise.
1574 2018-10-21 Bill Schmidt <wschmidt@linux.ibm.com>
1575 Jinsong Ji <jji@us.ibm.com>
1577 * config/rs6000/emmintrin.h (_mm_movemask_pd): Replace __vector
1578 __m64 with __vector unsigned long long for compatibility.
1579 (_mm_movemask_epi8): Likewise.
1580 * config/rs6000/xmmintrin.h (_mm_cvtps_pi32): Likewise.
1581 (_mm_cvttps_pi32): Likewise.
1582 (_mm_cvtpi32_ps): Likewise.
1583 (_mm_cvtps_pi16): Likewise.
1584 (_mm_loadh_pi): Likewise.
1585 (_mm_storeh_pi): Likewise.
1586 (_mm_movehl_ps): Likewise.
1587 (_mm_movelh_ps): Likewise.
1588 (_mm_loadl_pi): Likewise.
1589 (_mm_storel_pi): Likewise.
1590 (_mm_movemask_ps): Likewise.
1591 (_mm_shuffle_pi16): Likewise.
1593 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
1596 * config/i386/avx512fintrin.h (_mm512_fnmsub_round_pd): Use
1597 __builtin_ia32_vfnmsubpd512_mask.
1598 (_mm512_mask_fnmsub_round_pd): Likewise.
1599 (_mm512_fnmsub_pd): Likewise.
1600 (_mm512_mask_fnmsub_pd): Likewise.
1601 (_mm512_maskz_fnmsub_round_pd): Use
1602 __builtin_ia32_vfnmsubpd512_maskz.
1603 (_mm512_maskz_fnmsub_pd): Likewise.
1604 (_mm512_fnmsub_round_ps): Use __builtin_ia32_vfnmsubps512_mask.
1605 (_mm512_mask_fnmsub_round_ps): Likewise.
1606 (_mm512_fnmsub_ps): Likewise.
1607 (_mm512_mask_fnmsub_ps): Likewise.
1608 (_mm512_maskz_fnmsub_round_ps): Use
1609 __builtin_ia32_vfnmsubps512_maskz.
1610 (_mm512_maskz_fnmsub_ps): Likewise.
1611 * config/i386/avx512vlintrin.h (_mm256_mask_fnmsub_pd): Use
1612 __builtin_ia32_vfnmsubpd256_mask.
1613 (_mm256_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256_maskz.
1614 (_mm_mask_fnmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
1615 (_mm_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd128_maskz.
1616 (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
1617 (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
1618 (_mm256_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_maskz.
1619 (_mm_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_mask.
1620 (_mm_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_maskz.
1621 * config/i386/fmaintrin.h (_mm_fnmsub_pd): Use
1622 __builtin_ia32_vfnmsubpd.
1623 (_mm256_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256.
1624 (_mm_fnmsub_ps): Use __builtin_ia32_vfnmsubps.
1625 (_mm256_fnmsub_ps): Use __builtin_ia32_vfnmsubps256.
1626 (_mm_fnmsub_sd): Use __builtin_ia32_vfnmsubsd3.
1627 (_mm_fnmsub_ss): Use __builtin_ia32_vfnmsubss3.
1628 * config/i386/i386-builtin.def: Add
1629 __builtin_ia32_vfnmsubpd256_mask,
1630 __builtin_ia32_vfnmsubpd256_maskz,
1631 __builtin_ia32_vfnmsubpd128_mask,
1632 __builtin_ia32_vfnmsubpd128_maskz,
1633 __builtin_ia32_vfnmsubps256_mask,
1634 __builtin_ia32_vfnmsubps256_maskz,
1635 __builtin_ia32_vfnmsubps128_mask,
1636 __builtin_ia32_vfnmsubps128_maskz,
1637 __builtin_ia32_vfnmsubpd512_mask,
1638 __builtin_ia32_vfnmsubpd512_maskz,
1639 __builtin_ia32_vfnmsubps512_mask,
1640 __builtin_ia32_vfnmsubps512_maskz, __builtin_ia32_vfnmsubss3,
1641 __builtin_ia32_vfnmsubsd3, __builtin_ia32_vfnmsubps,
1642 __builtin_ia32_vfnmsubpd, __builtin_ia32_vfnmsubps256 and.
1643 __builtin_ia32_vfnmsubpd256.
1644 * config/i386/sse.md (fma4i_fnmsub_<mode>): New.
1645 (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Likewise.
1646 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1):
1648 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2):
1650 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3):
1652 (fmai_vmfnmsub_<mode><round_name>): Likewise.
1654 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
1657 * config/i386/avx512fintrin.h (_mm512_fnmadd_round_pd): Use
1658 __builtin_ia32_vfnmaddpd512_mask.
1659 (_mm512_mask_fnmadd_round_pd): Likewise.
1660 (_mm512_fnmadd_pd): Likewise.
1661 (_mm512_mask_fnmadd_pd): Likewise.
1662 (_mm512_maskz_fnmadd_round_pd): Use
1663 __builtin_ia32_vfnmaddpd512_maskz.
1664 (_mm512_maskz_fnmadd_pd): Likewise.
1665 (_mm512_fnmadd_round_ps): Use __builtin_ia32_vfnmaddps512_mask.
1666 (_mm512_mask_fnmadd_round_ps): Likewise.
1667 (_mm512_fnmadd_ps): Likewise.
1668 (_mm512_mask_fnmadd_ps): Likewise.
1669 (_mm512_maskz_fnmadd_round_ps): Use
1670 __builtin_ia32_vfnmaddps512_maskz.
1671 (_mm512_maskz_fnmadd_ps): Likewise.
1672 * config/i386/avx512vlintrin.h (_mm256_mask_fnmadd_pd): Use
1673 __builtin_ia32_vfnmaddpd256_mask.
1674 (_mm256_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256_maskz.
1675 (_mm_mask_fnmadd_pd): Use __builtin_ia32_vfmaddpd128_mask
1676 (_mm_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd128_maskz.
1677 (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
1678 (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
1679 (_mm256_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_maskz.
1680 (_mm_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_mask.
1681 (_mm_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_maskz.
1682 * config/i386/fmaintrin.h (_mm_fnmadd_pd): Use
1683 __builtin_ia32_vfnmaddpd.
1684 (_mm256_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256.
1685 (_mm_fnmadd_ps): Use __builtin_ia32_vfnmaddps.
1686 (_mm256_fnmadd_ps): Use __builtin_ia32_vfnmaddps256.
1687 (_mm_fnmadd_sd): Use __builtin_ia32_vfnmaddsd3.
1688 (_mm_fnmadd_ss): Use __builtin_ia32_vfnmaddss3.
1689 * config/i386/i386-builtin.def: Add
1690 __builtin_ia32_vfnmaddpd256_mask,
1691 __builtin_ia32_vfnmaddpd256_maskz,
1692 __builtin_ia32_vfnmaddpd128_mask,
1693 __builtin_ia32_vfnmaddpd128_maskz,
1694 __builtin_ia32_vfnmaddps256_mask,
1695 __builtin_ia32_vfnmaddps256_maskz,
1696 __builtin_ia32_vfnmaddps128_mask,
1697 __builtin_ia32_vfnmaddps128_maskz,
1698 __builtin_ia32_vfnmaddpd512_mask,
1699 __builtin_ia32_vfnmaddpd512_maskz,
1700 __builtin_ia32_vfnmaddps512_mask,
1701 __builtin_ia32_vfnmaddps512_maskz, __builtin_ia32_vfnmaddss3,
1702 __builtin_ia32_vfnmaddsd3, __builtin_ia32_vfnmaddps,
1703 __builtin_ia32_vfnmaddpd, __builtin_ia32_vfnmaddps256 and.
1704 __builtin_ia32_vfnmaddpd256.
1705 * config/i386/sse.md (fma4i_fnmadd_<mode>): New.
1706 (<avx512>_fnmadd_<mode>_maskz<round_expand_name>): Likewise.
1707 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1):
1709 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2):
1711 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3):
1713 (fmai_vmfnmadd_<mode><round_name>): Likewise.
1715 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
1718 * config/i386/avx512fintrin.h (_mm512_fmsub_round_pd): Use
1719 __builtin_ia32_vfmsubpd512_mask.
1720 (_mm512_mask_fmsub_round_pd): Likewise.
1721 (_mm512_fmsub_pd): Likewise.
1722 (_mm512_mask_fmsub_pd): Likewise.
1723 (_mm512_maskz_fmsub_round_pd): Use
1724 __builtin_ia32_vfmsubpd512_maskz.
1725 (_mm512_maskz_fmsub_pd): Likewise.
1726 (_mm512_fmsub_round_ps): Use __builtin_ia32_vfmsubps512_mask.
1727 (_mm512_mask_fmsub_round_ps): Likewise.
1728 (_mm512_fmsub_ps): Likewise.
1729 (_mm512_mask_fmsub_ps): Likewise.
1730 (_mm512_maskz_fmsub_round_ps): Use
1731 __builtin_ia32_vfmsubps512_maskz.
1732 (_mm512_maskz_fmsub_ps): Likewise.
1733 * config/i386/avx512vlintrin.h (_mm256_mask_fmsub_pd): Use
1734 __builtin_ia32_vfmsubpd256_mask.
1735 (_mm256_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd256_maskz.
1736 (_mm_mask_fmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
1737 (_mm_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd128_maskz.
1738 (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
1739 (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
1740 (_mm256_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps256_maskz.
1741 (_mm_mask_fmsub_ps): Use __builtin_ia32_vfmsubps128_mask.
1742 (_mm_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps128_maskz.
1743 * config/i386/fmaintrin.h (_mm_fmsub_pd): Use
1744 __builtin_ia32_vfmsubpd.
1745 (_mm256_fmsub_pd): Use __builtin_ia32_vfmsubpd256.
1746 (_mm_fmsub_ps): Use __builtin_ia32_vfmsubps.
1747 (_mm256_fmsub_ps): Use __builtin_ia32_vfmsubps256.
1748 (_mm_fmsub_sd): Use __builtin_ia32_vfmsubsd3.
1749 (_mm_fmsub_ss): Use __builtin_ia32_vfmsubss3.
1750 * config/i386/i386-builtin.def: Add
1751 __builtin_ia32_vfmsubpd256_mask,
1752 __builtin_ia32_vfmsubpd256_maskz,
1753 __builtin_ia32_vfmsubpd128_mask,
1754 __builtin_ia32_vfmsubpd128_maskz,
1755 __builtin_ia32_vfmsubps256_mask,
1756 __builtin_ia32_vfmsubps256_maskz,
1757 __builtin_ia32_vfmsubps128_mask,
1758 __builtin_ia32_vfmsubps128_maskz,
1759 __builtin_ia32_vfmsubpd512_mask,
1760 __builtin_ia32_vfmsubpd512_maskz,
1761 __builtin_ia32_vfmsubps512_mask,
1762 __builtin_ia32_vfmsubps512_maskz, __builtin_ia32_vfmsubss3,
1763 __builtin_ia32_vfmsubsd3, __builtin_ia32_vfmsubps,
1764 __builtin_ia32_vfmsubpd, __builtin_ia32_vfmsubps256 and.
1765 __builtin_ia32_vfmsubpd256.
1766 * config/i386/sse.md (fma4i_fmsub_<mode>): New.
1767 (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Likewise.
1768 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1):
1770 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2):
1772 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3):
1774 (fmai_vmfmsub_<mode><round_name>): Likewise.
1776 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
1778 * config/i386/sse.md (*<plusminus_insn><mode>3<mask_name>_bcst_1):
1779 Remove plus. Renamed to ...
1780 (*sub<mode>3<mask_name>_bcst): This.
1781 (*add<mode>3<mask_name>_bcst_2): Renamede to ...
1782 (*add<mode>3<mask_name>_bcst): This.
1784 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
1787 * config/i386/sse.md (*mul<mode>3<mask_name>_bcst): New.
1789 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
1792 * i386/avx512vlintrin.h (_mm256_or_epi32): New.
1793 (_mm_or_epi32): Likewise.
1794 (_mm256_xor_epi32): Likewise.
1795 (_mm_xor_epi32): Likewise.
1796 (_mm256_or_epi64): Likewise.
1797 (_mm_or_epi64): Likewise.
1798 (_mm256_xor_epi64): Likewise.
1799 (_mm_xor_epi64): Likewise.
1801 2018-10-20 H.J. Lu <hongjiu.lu@intel.com>
1804 * config/i386/sse.md (*<avx512>_div<mode>3<mask_name>_bcst): New.
1806 2018-10-20 Jakub Jelinek <jakub@redhat.com>
1809 * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR.
1811 2018-10-20 Andreas Schwab <schwab@linux-m68k.org>
1813 * doc/ux.texi: Move @section directly after @node.
1815 2018-10-19 Jakub Jelinek <jakub@redhat.com>
1819 * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
1820 depend closely nested inside of loop with ordered clause with
1823 2018-10-19 David Malcolm <dmalcolm@redhat.com>
1825 * Makefile.in (TEXI_GCCINT_FILES): Add ux.texi.
1826 * doc/gccint.texi: Include ux.texi and use it in top-level menu.
1827 * doc/ux.texi: New file.
1829 2018-10-19 Segher Boessenkool <segher@kernel.crashing.org>
1831 * config/rs6000/rs6000.h (REG_ALLOC_ORDER): Move 68 (that is, CR0) to
1832 be the first CR field allocated.
1834 2018-10-19 Richard Biener <rguenther@suse.de>
1837 * config/i386/i386.c (ix86_builtin_vectorization_cost): Use
1838 TYPE_VECTOR_SUBPARTS and avoid relying on vector mode.
1840 2018-10-19 H.J. Lu <hongjiu.lu@intel.com>
1843 * config/i386/sse.md
1844 (*<plusminus_insn><mode>3<mask_name>_bcst_1): New.
1845 (*add<mode>3<mask_name>_bcst_2): Likewise.
1847 2018-10-19 H.J. Lu <hongjiu.lu@intel.com>
1849 * config/i386/sse.md
1850 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
1851 Replace nonimmediate_operand with register_operand.
1852 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
1854 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
1857 2018-10-19 Ilya Leoshkevich <iii@linux.ibm.com>
1859 PR rtl-optimization/87596
1860 * lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () +
1861 lra_get_insn_recog_data () instead of lra_insn_recog_data[]
1862 for instructions in FROM..TO range.
1864 2018-10-19 Eric Botcazou <ebotcazou@adacore.com>
1866 * cfgexpand.c (expand_one_var): Use specific wording in error message
1867 for non-local frame variables.
1868 * stor-layout.c (layout_decl): Do not issue a warning for them.
1870 2018-10-19 Robin Dapp <rdapp@linux.ibm.com>
1872 * haifa-sched.c (priority): Add force_recompute parameter.
1873 (apply_replacement): Call priority () with force_recompute = true.
1874 (restore_pattern): Likewise.
1876 2018-10-18 H.J. Lu <hongjiu.lu@intel.com>
1878 * simplify-rtx.c (simplify_subreg): Limit mask of vec_merge to
1879 HOST_BITS_PER_WIDE_INT.
1880 (test_vector_ops_duplicate): Likewise.
1882 2018-10-18 H.J. Lu <hongjiu.lu@intel.com>
1885 * config/i386/sse.md (VF_AVX512): New.
1886 (avx512bcst): Likewise.
1887 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
1889 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
1891 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
1894 2018-10-18 Jonathan Wakely <jwakely@redhat.com>
1896 * doc/invoke.texi (-dumpversion): Improve grammar.
1897 (-dumpfullversion): Make more consistent with -dumpversion.
1899 2018-10-18 Uros Bizjak <ubizjak@gmail.com>
1901 * config/i386/i386.c (ix86_emit_fp_unordered_jump):
1902 Set JUMP_LABEL to the jump insn.
1903 (ix86_emit_i387_log1p): Use ix86_expand_branch to expand branch.
1904 Predict emitted jump and add label to jump insn.
1906 2018-10-18 David Malcolm <dmalcolm@redhat.com>
1908 PR tree-optimization/87562
1909 * input.c (get_substring_ranges_for_loc): Use
1910 LRK_SPELLING_LOCATION rather than LRK_MACRO_EXPANSION_POINT when
1911 getting the linemap for the endpoint. Verify that it's either
1912 in the same linemap as the start point's spelling location, or
1913 at least in the same file.
1915 2018-10-18 Richard Biener <rguenther@suse.de>
1917 * config/i386/i386.c (ix86_builtin_vectorization_cost): Do not
1918 feed width-specific load/store costs through ix86_vec_cost.
1919 * config/i386/x86-tune-costs.h (athlon_cost): Adjust.
1920 (k8_cost): Likewise.
1921 (bdver_cost): Likewise.
1922 (znver1_cost): Likewise.
1923 (btver1_cost): Likewise.
1924 (btver2_cost): Likewise.
1926 2018-10-18 H.J. Lu <hongjiu.lu@intel.com>
1928 * simplify-rtx.c (simplify_subreg): Call simplify_gen_subreg
1929 to simplify subreg of vec_merge.
1931 2018-10-18 Richard Biener <rguenther@suse.de>
1933 * config/i386/i386.c: Fix costing of vector FMA.
1935 2018-10-18 Richard Biener <rguenther@suse.de>
1937 * config/i386/i386.c (ix86_vec_cost): Remove !parallel path
1939 (ix86_builtin_vectorization_cost): For vec_construct properly
1940 cost insertion into SSE regs.
1941 (...): Adjust calls to ix86_vec_cost.
1943 2018-10-18 Richard Biener <rguenther@suse.de>
1947 2018-02-07 Richard Biener <rguenther@suse.de>
1949 PR tree-optimization/84204
1950 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
1953 2018-10-18 H.J. Lu <hongjiu.lu@intel.com>
1956 * simplify-rtx.c (simplify_subreg): Simplify subreg of vec_merge
1958 (test_vector_ops_duplicate): Add test for a scalar subreg of a
1959 VEC_MERGE of a VEC_DUPLICATE.
1961 2018-10-17 Joseph Myers <joseph@codesourcery.com>
1963 * doc/cpp.texi (__STDC_VERSION__): Document C2X handling.
1964 * doc/invoke.texi (-std=c2x, -std=gnu2x): Document new options.
1965 * doc/standards.texi (C Language): Document C2X.
1966 * dwarf2out.c (highest_c_language), config/rl78/rl78.c
1967 (rl78_option_override): Handle "GNU C2X" language name.
1969 2018-10-17 Joseph Myers <joseph@codesourcery.com>
1971 * doc/invoke.texi (-std=c17), doc/standards.texi (C Language):
1972 Document C17 as published in 2018.
1974 2018-10-17 Eric Botcazou <ebotcazou@adacore.com>
1977 * fold-const.c (fold_truth_andor_1): If the right side is not constant,
1978 bail out if both sides do not have the same storage order.
1980 2018-10-17 Aldy Hernandez <aldyh@redhat.com>
1982 * bitmap.c (bitmap_head::dump): New.
1983 * bitmap.h (bitmap_head): Add dump().
1984 * gimple-ssa-evrp-analyze.c
1985 (evrp_range_analyzer::try_find_new_range): Adjust for value_range API.
1986 (evrp_range_analyzer::set_ssa_range_info): Same.
1987 (evrp_range_analyzer::record_ranges_from_phis): Same.
1988 (evrp_range_analyzer::record_ranges_from_stmt): Same.
1989 * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Same.
1990 * gimple-ssa-sprintf.c (get_int_range): Same.
1991 (format_integer): Same.
1992 (sprintf_dom_walker::handle_gimple_call): Same.
1993 * ipa-cp.c (ipcp_vr_lattice::meet_with_1): Same.
1994 (ipcp_vr_lattice::top_p): Same.
1995 (ipcp_vr_lattice::bottom_p): Same.
1996 (ipcp_vr_lattice::set_to_bottom): Same.
1997 (ipa_vr_operation_and_type_effects): Same.
1998 (propagate_vr_across_jump_function): Same.
1999 (ipcp_store_vr_results): Same.
2000 * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Same.
2001 (ipa_print_node_jump_functions_for_edge): Same.
2002 (ipa_get_value_range): Same.
2003 (ipa_compute_jump_functions_for_edge): Same.
2004 (ipa_write_jump_function): Same.
2005 * tree-ssa-dom.c (simplify_stmt_for_jump_threading): Same.
2006 * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
2008 * vr-values.c (set_value_range_to_nonnegative): Same.
2009 (set_value_range_to_truthvalue): Same.
2010 (vr_values::get_value_range): Same.
2011 (vr_values::set_defs_to_varying): Same.
2012 (vr_values::update_value_range): Same.
2013 (symbolic_range_based_on_p): Same.
2014 (vr_values::op_with_boolean_value_range_p): Same.
2015 (vr_values::extract_range_for_var_from_comparison_expr): Same.
2016 (vr_values::extract_range_from_ssa_name): Same.
2017 (vr_values::extract_range_from_binary_expr): Same.
2018 (vr_values::extract_range_from_unary_expr): Same.
2019 (vr_values::extract_range_from_cond_expr): Same.
2020 (vr_values::extract_range_from_comparison): Same.
2021 (vr_values::check_for_binary_op_overflow): Same.
2022 (vr_values::extract_range_basic): Same.
2023 (vr_values::extract_range_from_assignment): Same.
2024 (compare_ranges): Same.
2025 (compare_range_with_value): Same.
2026 (vr_values::adjust_range_with_scev): Same.
2027 (vrp_valueize): Same.
2028 (vrp_valueize_1): Same.
2029 (vr_values::get_vr_for_comparison): Same.
2030 (vr_values::compare_name_with_value): Same.
2031 (vr_values::compare_names): Same.
2032 (vr_values::vrp_evaluate_conditional): Same.
2033 (find_case_label_ranges): Same.
2034 (vr_values::vrp_visit_switch_stmt): Same.
2035 (vr_values::extract_range_from_phi_node): Same.
2036 (vr_values::simplify_div_or_mod_using_ranges): Same.
2037 (vr_values::simplify_bit_ops_using_ranges): Same.
2038 (test_for_singularity): Same.
2039 (range_fits_type_p): Same.
2040 (vr_values::simplify_cond_using_ranges_1): Same.
2041 (vr_values::simplify_switch_using_ranges): Same.
2042 (vr_values::simplify_float_conversion_using_ranges): Same.
2043 (vr_values::two_valued_val_range_p): Same.
2044 (vr_values::add_equivalence): Move to value_range::equiv_add.
2045 * vr-values.h (vr_values::add_equivalence): Remove.
2046 (VR_INITIALIZER): Remove.
2047 * tree-vrp.c (value_range::set): New.
2048 (value_range::equiv_add): New.
2049 (value_range::value_range): New.
2050 (value_range::deep_copy): New.
2051 (value_range::check): New.
2052 (value_range::equal_p): New.
2053 (value_range::ignore_equivs_equal_p): New.
2054 (value_range::operator==): New.
2055 (value_range::operator!=): New.
2056 (value_range::symbolic_p): New.
2057 (value_range::numeric_p): New.
2058 (value_range::set_undefined): New.
2059 (value_range::set_varying): New.
2060 (value_range::may_contain_p): New.
2061 (value_range::equiv_clear): New.
2062 (value_range::singleton_p): New.
2063 (value_range::intersect): New.
2064 (value_range::dump): New.
2065 (value_range::set_and_canonicalize): New.
2066 (set_value_range): Adjust for value_range API.
2067 (set_value_range_to_undefined): Same.
2068 (set_value_range_to_varying): Same.
2069 (set_and_canonicalize_value_range): Same.
2070 (set_value_range_to_nonnull): Same.
2071 (set_value_range_to_null): Same.
2072 (range_is_null): Same.
2073 (range_is_nonnull): Same.
2074 (range_int_cst_p): Same.
2075 (range_int_cst_singleton_p): Same.
2076 (symbolic_range_p): Same.
2077 (range_includes_zero_p): Same.
2078 (value_range_constant_singleton): Same.
2079 (vrp_set_zero_nonzero_bits): Same.
2080 (ranges_from_anti_range): Same.
2081 (extract_range_into_wide_ints): Same.
2082 (extract_range_from_multiplicative_op): Same.
2083 (set_value_range_with_overflow): Same.
2084 (extract_range_from_binary_expr_1): Same.
2085 (extract_range_from_unary_expr): Same.
2086 (dump_value_range): Same.
2087 (debug_value_range): Same.
2088 (vrp_prop::check_array_ref): Same.
2089 (vrp_prop::check_mem_ref): Same.
2090 (vrp_prop::vrp_initialize): Same.
2091 (vrp_prop::visit_stmt): Same.
2092 (intersect_ranges): Same.
2093 (vrp_prop::visit_phi): Same.
2094 (vrp_prop::vrp_finalize): Same.
2095 (determine_value_range_1): Same.
2096 (determine_value_range): Same.
2097 (vrp_intersect_ranges_1): Rename to...
2098 (vrp_intersect_1): this.
2099 (vrp_intersect_ranges): Rename to...
2100 (value_range::intersect_helper): ...this.
2101 (vrp_meet_1): Rename to...
2102 (value_range::union_helper): ...this.
2103 (vrp_meet): Rename to...
2104 (value_range::union_): ...this.
2105 (copy_value_range): Remove.
2106 * tree-vrp.h (struct value_range): Rewrite into a proper class.
2107 (value_range::vrtype): New.
2108 (value_range::type): New.
2109 (value_range::equiv): New.
2110 (value_range::min): New.
2111 (value_range::max): New.
2112 (value_range::varying_p): New.
2113 (value_range::undefined_p): New.
2114 (value_range::null_p): New.
2115 (value_range::equiv_add): New.
2116 (copy_value_range): Remove.
2118 2018-10-17 David Malcolm <dmalcolm@redhat.com>
2120 * Makefile.in (SELFTEST_TARGETS): New.
2121 (selftest) Change from s-selftest-c to $(SELFTEST_TARGETS).
2122 (C_SELFTEST_FLAGS, C_SELFTEST_DEPS, s-selftest-c, selftest-c-gdb)
2123 (selftest-gdb, selftest-c-valgrind, selftest-valgrind): Move to
2125 (CPP_SELFTEST_FLAGS, CPP_SELFTEST_DEPS, s-selftest-c++)
2126 (selftest-c++-gdb, selftest-c++-valgrind): Move to
2128 * configure: Regenerate.
2129 * configure.ac (selftest_languages): New.
2131 2018-10-17 Aldy Hernandez <aldyh@redhat.com>
2133 * tree-vrp.c (extract_range_from_multiplicative_op): Remove
2134 overflow wraps argument.
2135 (extract_range_from_binary_expr_1): Do not pass overflow wraps to
2136 wide_int_range_multiplicative_op.
2137 * wide-int-range.cc (wide_int_range_mult_wrapping): Remove
2138 overflow wraps argument.
2139 (wide_int_range_multiplicative_op): Same.
2140 (wide_int_range_lshift): Same.
2141 (wide_int_range_div): Same.
2142 * wide-int-range.h (wide_int_range_multiplicative_op): Same.
2143 (wide_int_range_lshift): Same.
2144 (wide_int_range_div): Same.
2146 2018-10-17 Aldy Hernandez <aldyh@redhat.com>
2148 * wide-int-range.h (wide_int_range_shift_undefined_p): Adjust to
2149 use sign as argument.
2150 * tree-vrp.c (extract_range_from_binary_expr_1): Pass sign to
2151 wide_int_range_shift_undefined_p.
2153 2018-10-16 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2155 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>):
2157 (@despeculate_copy<ALLI_TI:mode>): ... This.
2158 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): Remove
2161 2018-10-16 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2163 * config.gcc: Obsolete *-*-solaris2.10*.
2164 * doc/install.texi (Specific, *-*-solaris2*): Document it.
2166 2018-10-12 Jeff Law <law@redhat.com>
2168 * config/ft32/ft32.md (ft32_general_movsrc_operand): Disable
2169 reg + sym +- const_int addressing modes.
2171 2018-10-15 David Malcolm <dmalcolm@redhat.com>
2173 * common.opt (fdiagnostics-minimum-margin-width=): New option.
2174 * diagnostic-show-locus.c (layout::layout): Apply the minimum
2176 (layout::start_annotation_line): Only print up to 3 of the
2177 margin character, to avoid touching the left-hand side.
2178 (selftest::test_diagnostic_show_locus_fixit_lines): Update for
2179 minimum margin width, as set by test_diagnostic_context's ctor.
2180 (selftest::test_fixit_insert_containing_newline): Likewise.
2181 (selftest::test_fixit_insert_containing_newline_2): Likewise.
2182 (selftest::test_line_numbers_multiline_range): Clear
2183 dc.min_margin_width.
2184 * diagnostic.c (diagnostic_initialize): Initialize
2186 * diagnostic.h (struct diagnostic_context): Add field
2188 * doc/invoke.texi: Add -fdiagnostics-minimum-margin-width=.
2189 * opts.c (common_handle_option): Handle
2190 OPT_fdiagnostics_minimum_margin_width_.
2191 * selftest-diagnostic.c
2192 (selftest::test_diagnostic_context::test_diagnostic_context):
2193 Initialize min_margin_width to 6.
2194 * toplev.c (general_init): Initialize global_dc->min_margin_width.
2196 2018-10-15 David Malcolm <dmalcolm@redhat.com>
2198 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
2199 Fix usage of "error_at_rich_loc" in the comment.
2201 2018-10-15 Renlin Li <renlin.li@arm.com>
2204 * tree-vectorizer.c (try_vectorize_loop_1): Don't use
2205 if-conversioned loop when it contains ifn with types not
2206 supported by backend.
2207 * internal-fn.c (expand_direct_optab_fn): Add an assert.
2208 (direct_internal_fn_supported_p): New helper function.
2209 * internal-fn.h (direct_internal_fn_supported_p): Declare.
2211 2018-10-15 Jakub Jelinek <jakub@redhat.com>
2214 * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
2215 Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
2216 OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
2218 2018-10-15 Bin Cheng <bin.cheng@linux.alibaba.com>
2220 PR tree-optimization/87022
2221 * tree-loop-distribution.c (pg_add_dependence_edges): Check all
2222 bits in dist vector rather than the first one.
2224 2018-10-15 Richard Biener <rguenther@suse.de>
2227 * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
2228 (visit_loadstore): When a used restrict tag escaped verify that
2229 the points-to solution of "other" pointers do not include
2231 (compute_dependence_clique): If a used restrict tag escaped
2232 communicated that down to visit_loadstore.
2234 2018-10-15 Andreas Krebbel <krebbel@linux.ibm.com>
2236 * config/s390/s390.c (s390_expand_vec_init): Force vector element
2237 into reg if it isn't a general operand.
2239 2018-10-14 H.J. Lu <hongjiu.lu@intel.com>
2242 * config/i386/sse.md (*vec_dupv2di): Add register source to
2245 2018-10-14 H.J. Lu <hongjiu.lu@intel.com>
2248 * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
2249 Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
2250 OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
2251 OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
2253 2018-10-13 Eric Botcazou <ebotcazou@adacore.com>
2255 * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
2256 (notice_args_size): Set it in the current trace if no insn that can
2257 throw internally has been seen yet.
2258 (connect_traces): When connecting args_size between traces, allow the
2259 incoming values not to match if there is an insn setting it before the
2260 first insn that can throw internally; in that case, force the creation
2261 of a CFI note on this latter insn.
2263 2018-10-13 Jonathan Wakely <jwakely@redhat.com>
2265 * opt-problem.h (opt_wrapper): Use template-argument-list when naming
2266 the base class, because using the injected-class-name was not clearly
2267 specified until DR 176.
2269 2018-10-12 Paul Koning <ni1d@arrl.net>
2271 * config/pdp11/pdp11.md (doloop_end): New expander.
2272 (doloop_end_insn): renamed from "doloop_end".
2273 (addqi3): New pattern.
2274 (subqi3): New pattern.
2275 * config/pdp11/predicates.md (incdec_operand): New predicate.
2277 2018-10-12 Yury Gribov <tetra2005@gmail.com>
2280 * real.c (format_helper::can_represent_integral_type_p): New function
2281 * real.h (format_helper::can_represent_integral_type_p): Ditto.
2282 * match.pd: New pattern.
2284 2018-10-12 Alexandre Oliva <oliva@adacore.com>
2286 * configure.ac: Introduce --enable-large-address-aware
2287 to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
2288 * doc/install.texi: Document it.
2289 * configure, config.in: Rebuilt.
2290 * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
2291 based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
2292 (LINK_SPEC): Insert it.
2293 * config/i386/mingw-w64.h: Likewise.
2295 * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
2297 2018-10-12 Peter Bergner <bergner@linux.ibm.com>
2299 PR rtl-optimization/87600
2300 * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
2302 2018-10-12 Paul Koning <ni1d@arrl.net>
2304 * doc/md.texi (doloop_end): Document that the pattern code may
2305 need to check operand mode.
2307 2018-10-12 Wilco Dijkstra <wdijkstr@arm.com>
2309 * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
2310 to zero-extend between int and floating-point registers.
2311 (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
2312 ldp into floating-point registers. Add type and arch attributes.
2313 (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
2314 Use f_loads for type attribute.
2316 2018-10-11 Martin Sebor <msebor@redhat.com>
2318 * doc/extend.texi (attribute packed): Correct typos.
2320 2018-10-11 Martin Sebor <msebor@redhat.com>
2322 * doc/extend.texi (attribute flatten): Mention interaction with
2325 2018-10-11 Jan Hubicka <hubicka@ucw.cz>
2328 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
2329 Set new_decl virtual flag to zero.
2331 2018-10-11 Martin Sebor <msebor@redhat.com>
2334 * doc/extend.texi (attribute format_arg): Discuss using multiple
2335 attributes on a single function.
2337 2018-10-11 Giuliano Belinassi <giuliano.belinassi@usp.br>
2339 PR tree-optimization/86829
2340 * match.pd (sin (atan (x))): New simplification rules.
2341 (cos (atan (x))): Likewise.
2342 * real.c (build_sinatan_real): New function.
2343 * real.h (build_sinatan_real): Prototype.
2345 2018-10-11 Will Schmidt <will_schmidt@vnet.ibm.com>
2347 * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
2349 (fold_mergeeo_helper): New helper function.
2350 (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
2351 intrinsics. Correct some whitespace indentation issues.
2353 2018-10-11 Wilco Dijkstra <wdijkstr@arm.com>
2356 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
2357 Use HOST_WIDE_INT_1U for shift.
2359 2018-10-11 Doug Rupp <rupp@adacore.com>
2360 Olivier Hainque <hainque@adacore.com>
2362 * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
2363 Pass --relax to the linker for RTPs.
2364 (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
2366 2018-10-11 Andrew Stubbs <ams@codesourcery.com>
2367 Jan Hubicka <jh@suse.cz>
2368 Martin Jambor <mjambor@suse.cz>
2370 * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
2371 the same elements are repeated rather than printing all of them.
2372 * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
2373 "repeated" elements.
2374 * read-rtl-function.c (test_loading_repeat): New function.
2375 (read_rtl_function_c_tests): Call test_loading_repeat.
2376 * rtl-tests.c (test_dumping_repeat): New function.
2377 (rtl_tests_c_tests): Call test_dumping_repeat.
2379 2018-10-11 Richard Biener <rguenther@suse.de>
2381 * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
2382 bdver?_cost): Unify to ...
2383 (bdver_memcpy, bdver_memset, bdver_cost): ... this.
2384 * config/i386/i386.c (processor_cost_table): Adjust.
2386 2018-10-10 Eric Botcazou <ebotcazou@adacore.com>
2389 * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
2390 the thunk when expanding to GIMPLE.
2392 2018-10-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
2394 * varasm.c (mergeable_string_section): Don't try to move zero-length
2395 strings to the merge section.
2397 2018-10-10 Uros Bizjak <ubizjak@gmail.com>
2400 * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
2402 2018-10-10 Jakub Jelinek <jakub@redhat.com>
2405 * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
2406 to special_args set.
2408 2018-10-10 Richard Biener <rguenther@suse.de>
2410 * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
2411 reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
2412 reduc_plus_scal_v4sf): Merge into pattern reducing to half width
2413 and recursing and pattern terminating the recursion on SSE
2414 vector width using ix86_expand_reduc.
2415 (reduc_sminmax_scal_<mode>): Split into part reducing to half
2416 width and recursing and SSE2 vector variant doing the final
2417 reduction with ix86_expand_reduc.
2418 (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
2419 with terminating the recursion at AVX level, splitting that
2422 2018-10-09 David Malcolm <dmalcolm@redhat.com>
2424 * genmatch.c (error_cb): Rename to...
2425 (diagnostic_cb): ...this, converting int params to enums.
2426 (fatal_at): Update for renaming.
2427 (warning_at): Likewise.
2429 * input.c (selftest::ebcdic_execution_charset::apply):
2430 Update for renaming of...
2431 (selftest::ebcdic_execution_charset::on_error): ...this, renaming
2433 (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
2434 converting level and reason to enums.
2435 (class selftest::lexer_error_sink): Rename to...
2436 (class selftest::lexer_test_options): ...this, renaming field
2437 "m_errors" to "m_diagnostics".
2438 (selftest::lexer_test_options::apply): Update for renaming of...
2439 (selftest::lexer_test_options::on_error): ...this, renaming to...
2440 (selftest::lexer_test_options::on_diagnostic): ...this
2441 converting level and reason to enums.
2442 (selftest::test_lexer_string_locations_raw_string_unterminated):
2443 Update for renamings.
2444 * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
2447 2018-10-09 Paul A. Clarke <pc@us.ibm.com>
2449 * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
2450 * config/rs6000/pmmintrin.h: New file.
2452 2018-10-09 Eric Botcazou <ebotcazou@adacore.com>
2454 PR tree-optimization/86659
2455 * gimple-match.h (gimple_match_op constructors): Initialize reverse.
2457 2018-10-09 Richard Biener <rguenther@suse.de>
2459 PR tree-optimization/63155
2460 * tree-ssa-structalias.c: Include tree-ssa.h.
2461 (get_constraint_for_ssa_var): For undefs return nothing_id.
2462 (find_func_aliases): Cleanup PHI handling.
2464 2018-10-09 Richard Biener <rguenther@suse.de>
2466 * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
2469 2018-10-09 Martin Liska <mliska@suse.cz>
2471 * asan.c (asan_emit_stack_protection): If a stack variable
2472 is located in a same file as current function, then emit
2473 line info into variable definition string.
2475 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
2477 * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
2480 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
2482 * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
2485 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
2487 PR tree-optimization/86659
2488 * gimple-match.h (struct gimple_match_op): Add reverse field.
2489 (gimple_match_op::set_op): New overloaded method.
2490 * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
2491 the REF_REVERSE_STORAGE_ORDER flag on the value.
2492 (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
2493 REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
2495 2018-10-08 Richard Sandiford <richard.sandiford@arm.com>
2498 * gimple-ssa-backprop.c (backprop::intersect_uses): Use
2499 FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
2501 2018-10-08 H.J. Lu <hongjiu.lu@intel.com>
2504 * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
2505 Defined with __builtin_ia32_vfmaddsubpd512_mask.
2507 2018-10-08 Richard Biener <rguenther@suse.de>
2509 * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
2510 cost the same as AVX128 ones.
2512 2018-10-08 Paul Koning <ni1d@arrl.net>
2514 * config/pdp11/pdp11-protos.h (output_block_move): Remove.
2515 (expand_block_move): New function.
2516 * config/pdp11/pdp11.c (output_block_move): Remove.
2517 (expand_block_move): New function.
2518 * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
2519 * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
2520 (*movmemhi1): Remove.
2522 2018-10-08 Robin Dapp <rdapp@linux.ibm.com>
2524 * config/s390/2827.md: Increase latencies for some FP instructions.
2526 2018-10-08 Richard Biener <rguenther@suse.de>
2528 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
2530 * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
2531 * tree-vectorizer.h (dump_stmt_cost): Adjust.
2532 (add_stmt_cost): Dump return value of the hook.
2534 2018-10-08 Richard Biener <rguenther@suse.de>
2536 PR tree-optimization/63155
2537 * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
2538 (ssa_propagation_engine::ssa_propagate): Remove redundant
2539 bitmap bit clearing.
2541 2018-10-05 Peter Bergner <bergner@linux.ibm.com>
2543 PR rtl-optimization/86939
2544 PR rtl-optimization/87479
2545 * ira.h (non_conflicting_reg_copy_p): New prototype.
2546 * ira-lives.c (ignore_reg_for_conflicts): New static variable.
2547 (make_hard_regno_dead): Don't add conflicts for register
2548 ignore_reg_for_conflicts.
2549 (make_object_dead): Likewise.
2550 (non_conflicting_reg_copy_p): New function.
2551 (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
2552 Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
2553 * lra-lives.c (ignore_reg_for_conflicts): New static variable.
2554 (make_hard_regno_dead): Don't add conflicts for register
2555 ignore_reg_for_conflicts. Remove special conflict handling of
2556 REAL_PIC_OFFSET_TABLE_REGNUM. Remove now unused argument
2557 check_pic_pseudo_p and update callers.
2558 (mark_pseudo_dead): Don't add conflicts for register
2559 ignore_reg_for_conflicts.
2560 (process_bb_lives): Set ignore_reg_for_conflicts for copies.
2562 2018-10-05 Andrew Waterman <andrew@sifive.com>
2563 Jim Wilson <jimw@sifive.com>
2565 * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
2566 Add define_expand. Add ! HONOR_SNANS check to current pattern. Add
2567 new pattern using HONOR_SNANS that emits one extra instruction.
2569 2018-10-05 Segher Boessenkool <segher@kernel.crashing.org>
2571 * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
2572 patterns): Merge SI and DI patterns to a GPR pattern.
2573 (unnamed define_insn and define_split for record form of that): Merge
2574 to a single define_insn_and_split pattern.
2576 2018-10-05 David Malcolm <dmalcolm@redhat.com>
2579 * input.c (expand_location_to_spelling_point): Add param "aspect"
2580 and use rather than hardcoding LOCATION_ASPECT_CARET.
2581 (get_substring_ranges_for_loc): Handle the case of a single token
2582 within a macro expansion.
2583 * input.h (expand_location_to_spelling_point): Add "aspect" param,
2584 defaulting to LOCATION_ASPECT_CARET.
2586 2018-10-05 Paul Koning <ni1d@arrl.net>
2588 * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
2589 (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
2590 (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
2591 (pdp11_guard_type): New function.
2593 2018-10-05 Paul Koning <ni1d@arrl.net>
2595 * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
2596 * config/pdp11/pdp11.opt (mfloat32): Remove.
2598 * doc/invoke.texi (pdp11 -mfloat32): Remove:
2599 (pdp11 -mfloat64): Remove.
2601 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
2603 * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
2604 (*cmp<mode>_cc_i387): Ditto.
2605 (*cmpu<mode>_cc_i387): Ditto.
2606 (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
2607 * config/i386/i386.c (ix86_expand_fp_compare): Remove
2609 <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
2610 Emit x86_sahf_1 pattern.
2611 (ix86_expand_compare): Update call to ix86_expand_fp_compare.
2612 (ix86_expand_carry_flag_compare): Ditto.
2614 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
2616 * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
2617 to reg_or_0_operand. Add "C" constraint.
2618 (*cmpxf_cc_i387): Ditto.
2619 (*cmp<mode>_i387): Change operand 2 predicate
2620 to nonimm_or_0_operand. Add "C" constraint.
2621 (*cmp<mode>_cc_i387): Ditto.
2622 (*cmp<mode>_0_i387): Remove insn pattern.
2623 (*cmp<mode>_0_cc_i387): Ditto.
2625 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
2627 * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
2628 * config/i386/predicates.md (nonimm_or_0_operand): Rename
2629 from vector_move_operand. Update all uses.
2631 2018-10-05 Martin Sebor <msebor@redhat.com>
2633 PR tree-optimization/87490
2634 * builtins.c (expand_builtin_strnlen): Handle a null data.decl
2637 2018-10-05 Richard Biener <rguenther@suse.de>
2639 PR tree-optimization/63155
2640 * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
2641 vertical space in dumpfiles.
2642 * tree-ssa-propagate.h
2643 (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
2644 * tree-ssa-propagate.c (cfg_blocks_back): New global.
2645 (ssa_edge_worklist_back): Likewise.
2646 (curr_order): Likewise.
2647 (cfg_blocks_get): Remove abstraction.
2648 (cfg_blocks_add): Likewise.
2649 (cfg_blocks_empty_p): Likewise.
2650 (add_ssa_edge): Add to current or next worklist based on
2652 (add_control_edge): Likewise.
2653 (ssa_propagation_engine::process_ssa_edge_worklist): Fold
2655 (ssa_propagation_engine::ssa_propagate): ... here. Unify
2656 iteration from CFG and SSA edge worklist so we process
2657 everything in RPO order, prioritizing forward progress
2659 (ssa_prop_init): Allocate new worklists, do not dump
2661 (ssa_prop_fini): Free new worklists.
2663 2018-10-05 Richard Biener <rguenther@suse.de>
2665 * tree-core.h (tree_block::abstract_flag): Remove.
2666 (tree_block::block_num): Make full 32bits.
2667 * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
2668 * tree.h (BLOCK_ABSTRACT): Remove.
2669 * dwarf2out.c (gen_lexical_block_die): Remove dead code
2670 resulting from BLOCK_ABSTRACT being always false.
2671 (gen_inlined_subroutine_die): Likewise.
2672 (gen_block_die): Likewise.
2673 * tree.c (block_ultimate_origin): Likewise.
2674 * tree-pretty-print.c (dump_block_node): Remove code dealing
2675 with BLOCK_ABSTRACT.
2676 * tree-ssa-live.c (dump_scope_block): Likewise.
2677 * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
2678 * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
2680 2018-10-05 Richard Biener <rguenther@suse.de>
2682 * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
2683 is asked for initialize mode to the component mode of the
2686 2018-10-05 H.J. Lu <hongjiu.lu@intel.com>
2689 * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
2690 assembler for -mavx.
2691 * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
2693 2018-10-05 Segher Boessenkool <segher@kernel.crashing.org>
2696 * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
2698 * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
2699 to be DImode. When using mffscrn, force the operand to a register.
2701 2018-10-04 Uros Bizjak <ubizjak@gmail.com>
2703 * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
2704 from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
2705 X87MODEF mode iterator.
2706 (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
2707 *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
2708 X87MODEF mode iterator.
2710 2018-10-04 Vinay Kumar <vinay.kumar@blackfigtech.com>
2712 * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
2713 -Wno-prio-ctor-dtor.
2715 2018-10-04 David Malcolm <dmalcolm@redhat.com>
2717 * Makefile.in (OBJS): Add opt-problem.o.
2718 * dump-context.h: Include "selftest.h.
2719 (selftest::temp_dump_context): New forward decl.
2720 (class dump_context): Make friend of class
2721 selftest::temp_dump_context.
2722 (dump_context::dump_loc_immediate): New decl.
2723 (class dump_pretty_printer): Move here from dumpfile.c.
2724 (class temp_dump_context): Move to namespace selftest.
2725 (temp_dump_context::temp_dump_context): Add param
2726 "forcibly_enable_dumping".
2727 (selftest::verify_dumped_text):
2728 (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
2729 (selftest::verify_item):
2730 (ASSERT_IS_TEXT): Move here from dumpfile.c.
2731 (ASSERT_IS_TREE): Likewise.
2732 (ASSERT_IS_GIMPLE): Likewise.
2733 * dumpfile.c (dump_context::dump_loc): Move immediate dumping
2735 (dump_context::dump_loc_immediate): ...this new function.
2736 (class dump_pretty_printer): Move to dump-context.h.
2737 (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
2738 (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
2739 (temp_dump_context::temp_dump_context): Move to "selftest"
2740 namespace. Add param "forcibly_enable_dumping", and use it to
2741 conditionalize the use of m_pp;
2742 (selftest::verify_dumped_text): Make non-static.
2743 (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
2744 (selftest::verify_item): Make non-static.
2745 (ASSERT_IS_TEXT): Move to dump-context.h.
2746 (ASSERT_IS_TREE): Likewise.
2747 (ASSERT_IS_GIMPLE): Likewise.
2748 (selftest::test_capture_of_dump_calls): Pass "true" for new
2749 param of temp_dump_context.
2750 * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
2751 it to MSG_ALL_PRIORITIES. Update values of TDF_COMPARE_DEBUG and
2753 * opt-problem.cc: New file.
2754 * opt-problem.h: New file.
2755 * optinfo-emit-json.cc
2756 (selftest::test_building_json_from_dump_calls): Pass "true" for
2757 new param of temp_dump_context.
2758 * optinfo.cc (optinfo_kind_to_dump_flag): New function.
2759 (optinfo::emit_for_opt_problem): New function.
2760 (optinfo::emit): Clarity which emit_item is used.
2761 * optinfo.h (optinfo::get_dump_location): New accessor.
2762 (optinfo::emit_for_opt_problem): New decl.
2763 (optinfo::emit): Make const.
2764 * selftest-run-tests.c (selftest::run_tests): Call
2765 selftest::opt_problem_cc_tests.
2766 * selftest.h (selftest::opt_problem_cc_tests): New decl.
2767 * tree-data-ref.c (dr_analyze_innermost): Convert return type from
2768 bool to opt_result, converting fprintf messages to
2769 opt_result::failure_at calls. Add "stmt" param for use by the
2771 (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
2772 (runtime_alias_check_p): Convert return type from bool to
2773 opt_result, converting dump_printf calls to
2774 opt_result::failure_at, using the statement DDR_A for their
2776 (find_data_references_in_stmt): Convert return type from bool to
2777 opt_result, converting "return false" to opt_result::failure_at
2779 * tree-data-ref.h: Include "opt-problem.h".
2780 (dr_analyze_innermost): Convert return type from bool to opt_result,
2781 and add a const gimple * param.
2782 (find_data_references_in_stmt): Convert return type from bool to
2784 (runtime_alias_check_p): Likewise.
2785 * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
2786 dr_analyze_innermost.
2787 * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
2788 Convert return type from bool to opt_result, adding a message for
2789 the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
2790 (vect_analyze_data_ref_dependence): Convert return type from bool
2791 to opt_result. Change sense of return type from "false"
2792 effectively meaning "no problems" to "false" meaning a problem,
2793 so that "return false" becomes "return opt_result::success".
2794 Convert "return true" calls to opt_result::failure_at, using
2795 the location of statement A rather than vect_location.
2796 (vect_analyze_data_ref_dependences): Convert return type from bool
2798 (verify_data_ref_alignment): Likewise, converting dump_printf_loc
2799 calls to opt_result::failure_at, using the stmt location rather
2801 (vect_verify_datarefs_alignment): Convert return type from bool
2803 (vect_enhance_data_refs_alignment): Likewise. Split local "stat"
2804 into multiple more-tightly-scoped copies.
2805 (vect_analyze_data_refs_alignment): Convert return type from bool
2807 (vect_analyze_data_ref_accesses): Likewise, converting a
2808 "return false" to a "return opt_result::failure_at", adding a
2810 (vect_prune_runtime_alias_test_list): Convert return type from
2811 bool to opt_result, converting dump_printf_loc to
2812 opt_result::failure_at. Add a %G to show the pertinent statement,
2813 and use the stmt's location rather than vect_location.
2814 (vect_find_stmt_data_reference): Convert return type from
2815 bool to opt_result, converting dump_printf_loc to
2816 opt_result::failure_at, using stmt's location.
2817 (vect_analyze_data_refs): Convert return type from bool to
2818 opt_result. Convert "return false" to "return
2819 opt_result::failure_at", adding messages as needed.
2820 * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
2821 type from bool to opt_result.
2822 (vect_determine_vf_for_stmt): Likewise.
2823 (vect_determine_vectorization_factor): Likewise, converting
2824 dump_printf_loc to opt_result::failure_at, using location of phi
2825 rather than vect_location.
2826 (vect_analyze_loop_form_1): Convert return type from bool to
2827 opt_result, converting dump_printf_loc calls, retaining the use of
2829 (vect_analyze_loop_form): Convert return type from loop_vec_info
2830 to opt_loop_vec_info.
2831 (vect_analyze_loop_operations): Convert return type from bool to
2832 opt_result, converting dump_printf_loc calls, using the location
2833 of phi/stmt rather than vect_location where available. Convert
2834 various "return false" to "return opt_result::failure_at" with
2835 "unsupported phi" messages.
2836 (vect_get_datarefs_in_loop): Convert return type from bool to
2837 opt_result. Add a message for the
2838 PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
2839 (vect_analyze_loop_2): Convert return type from bool to
2840 opt_result. Ensure "ok" is set to a opt_result::failure_at before
2841 each "goto again;", adding new messages where needed.
2842 Add "unsupported grouped {store|load}" messages.
2843 (vect_analyze_loop): Convert return type from loop_vec_info to
2845 * tree-vect-slp.c (vect_analyze_slp): Convert return type from
2847 * tree-vect-stmts.c (process_use): Likewise, converting
2848 dump_printf_loc call and using stmt location, rather than
2850 (vect_mark_stmts_to_be_vectorized): Likeise.
2851 (vect_analyze_stmt): Likewise, adding a %G.
2852 (vect_get_vector_types_for_stmt): Convert return type from bool to
2853 opt_result, converting dump_printf_loc calls and using stmt
2854 location, rather than vect_location.
2855 (vect_get_mask_type_for_stmt): Convert return type from tree to
2856 opt_tree, converting dump_printf_loc calls and using stmt location.
2857 * tree-vectorizer.c: Include "opt-problem.h.
2858 (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
2859 MSG_PRIORITY_INTERNALS. Convert local "loop_vinfo" from
2860 loop_vec_info to opt_loop_vec_info. If if fails, and dumping is
2861 enabled, use it to report at the top level "couldn't vectorize
2862 loop" followed by the problem.
2863 * tree-vectorizer.h (opt_loop_vec_info): New typedef.
2864 (vect_mark_stmts_to_be_vectorized): Convert return type from bool
2866 (vect_analyze_stmt): Likewise.
2867 (vect_get_vector_types_for_stmt): Likewise.
2868 (tree vect_get_mask_type_for_stmt): Likewise.
2869 (vect_analyze_data_ref_dependences): Likewise.
2870 (vect_enhance_data_refs_alignment): Likewise.
2871 (vect_analyze_data_refs_alignment): Likewise.
2872 (vect_verify_datarefs_alignment): Likewise.
2873 (vect_analyze_data_ref_accesses): Likewise.
2874 (vect_prune_runtime_alias_test_list): Likewise.
2875 (vect_find_stmt_data_reference): Likewise.
2876 (vect_analyze_data_refs): Likewise.
2877 (vect_analyze_loop): Convert return type from loop_vec_info to
2879 (vect_analyze_loop_form): Likewise.
2880 (vect_analyze_slp): Convert return type from bool to opt_result.
2882 2018-10-04 David Malcolm <dmalcolm@redhat.com>
2884 * doc/invoke.texi (-fopt-info): Document new "internals"
2886 * dump-context.h (dump_context::apply_dump_filter_p): New decl.
2887 * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
2889 (optinfo_verbosity_options): Add "internals".
2890 (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
2891 (dump_context::apply_dump_filter_p): New member function.
2892 (dump_context::dump_loc): Use apply_dump_filter_p rather than
2893 explicitly masking the dump_kind.
2894 (dump_context::begin_scope): Increment the scope depth first. Use
2895 apply_dump_filter_p rather than explicitly masking the dump_kind.
2896 (dump_context::emit_item): Use apply_dump_filter_p rather than
2897 explicitly masking the dump_kind.
2898 (dump_dec): Likewise.
2899 (dump_hex): Likewise.
2900 (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
2901 (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
2902 (opt_info_switch_p): Update handling of default
2903 MSG_OPTIMIZED_LOCATIONS to cope with default of
2904 MSG_PRIORITY_USER_FACING.
2905 (dump_basic_block): Use apply_dump_filter_p rather than explicitly
2906 masking the dump_kind.
2907 (selftest::test_capture_of_dump_calls): Update test_dump_context
2908 instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
2909 than MSG_ALL. Generalize scope test to be run at all four
2910 combinations of with/without MSG_PRIORITY_USER_FACING and
2911 MSG_PRIORITY_INTERNALS, adding examples of explicit priority
2912 for each of the two values.
2913 * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
2914 Rename MSG_ALL to MSG_ALL_KINDS. Add MSG_PRIORITY_USER_FACING,
2915 MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
2916 values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
2917 (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
2918 with MSG_PRIORITY_*.
2919 * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
2920 dump messages as MSG_PRIORITY_USER_FACING.
2921 * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
2922 about the interaction with MSG_PRIORITY_*.
2924 2018-10-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
2926 * varasm.c (output_constant): Add new parameter merge_strings.
2927 Make strings properly zero terminated in merge string sections.
2928 (mergeable_string_section): Don't fail if the last char is non-zero.
2929 (assemble_variable_contents): Handle merge string sections.
2930 (assemble_variable): Likewise.
2931 (assemble_constant_contents): Likewise.
2932 (output_constant_def_contents): Likewise.
2933 (output_constructor_array_range,
2934 output_constructor_regular_field): Adjust call to output_constant.
2935 (output_object_block): Adjust call to assemble_constant_contents
2936 and assemble_variable_contents.
2938 2018-10-04 Martin Liska <mliska@suse.cz>
2941 * cgraphunit.c (process_function_and_variable_attributes):
2942 Warn about a function with alias attribute and a body.
2944 2018-10-04 Martin Liska <mliska@suse.cz>
2947 * multiple_target.c (redirect_to_specific_clone): New function.
2948 (ipa_target_clone): Use it.
2949 * tree-inline.c: Fix comment.
2951 2018-10-04 David Malcolm <dmalcolm@redhat.com>
2953 * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
2955 (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
2956 (gcc::dump_manager::register_pass): New member function, adapted
2957 from loop body in gcc::pass_manager::register_pass, adding a
2958 call to update_dfi_for_opt_info.
2959 (gcc::dump_manager::opt_info_enable_passes): Store the
2960 -fopt-info options into the new fields. Move the loop
2962 (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
2964 * dumpfile.h (struct opt_pass): New forward decl.
2965 (gcc::dump_manager::register_pass): New decl.
2966 (gcc::dump_manager::update_dfi_for_opt_info): New decl.
2967 (class gcc::dump_manager): Add fields "m_optgroup_flags",
2968 "m_optinfo_flags", and "m_optinfo_filename".
2969 * passes.c (gcc::pass_manager::register_pass): Move all of the
2970 dump-handling code to gcc::dump_manager::register_pass.
2972 2018-10-04 Peter Bergner <bergner@linux.ibm.com>
2974 PR rtl-optimization/87466
2975 * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
2976 * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
2977 * doc/tm.texi: Regenerate.
2978 * ira-lives.c (process_bb_node_lives): Use the new target hook.
2979 * lra-lives.c (process_bb_lives): Likewise.
2980 * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
2983 2018-10-04 Tamar Christina <tamar.christina@arm.com>
2985 * params.c (add_params): Fix initialization.
2987 2018-10-04 Martin Liska <mliska@suse.cz>
2989 PR gcov-profile/84107
2990 * tree-profile.c (init_ic_make_global_vars):
2991 Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
2992 Come up with new ic_tuple* variables. Emit
2993 __gcov_indirect_call{,_topn} variables.
2994 (gimple_gen_ic_profiler): Access the variable
2996 (gimple_gen_ic_func_profiler): Access
2997 __gcov_indirect_call.callee field.
2998 (gimple_init_gcov_profiler): Use ptr_type_node.
2999 * value-prof.c (gimple_ic): Use ptr_type_node.
3001 2018-10-04 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3003 PR tree-optimization/85787
3004 * ipa-pure-const.c (malloc_candidate_p_1): Move most of malloc_candidate_p
3005 into this function and add support for detecting multiple phis.
3006 (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
3008 2018-10-04 Martin Liska <mliska@suse.cz>
3011 * ipa-inline.c (inline_to_all_callers_1):
3012 Call ultimate_alias_target for node being inlined.
3014 2018-10-03 Jeff Law <law@redhat.com>
3016 * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
3018 * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
3019 * tree.h (get_typenode_from_name): Prototype.
3021 2018-10-03 Uros Bizjak <ubizjak@gmail.com>
3023 * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
3024 Change operand 2 predicate to nonimmediate_operand.
3025 (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
3027 2018-10-03 Martin Sebor <msebor@redhat.com>
3028 Jeff Law <law@redhat.com>
3030 * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
3032 (get_string_length): Detect unterminated arrays.
3033 (format_string): Same.
3034 (format_directive): Warn about unterminated arrays.
3035 (handle_gimple_call): Mark statements with no_warning as needed.
3037 2018-10-03 Jim Wilson <jimw@sifive.com>
3039 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
3040 also define __riscv_abi_rve. Delete trailing white space.
3042 2018-10-03 Paul Koning <ni1d@arrl.net>
3044 Enable LRA register allocator for PDP11.
3045 * config/pdp11/constraints.md (Q): Use define_memory_constraint.
3048 * config/pdp11/pdp11.c (pdp11_lra_p): New function.
3049 * config/pdp11/pdp11.opt (-mlra): New option.
3050 * doc/invoke.texi (PDP-11 Options): Document -mlra.
3052 2018-10-03 Uros Bizjak <ubizjak@gmail.com>
3054 * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
3055 (*<absneg:code>extend<mode>xf2): Ditto.
3057 2018-10-03 Aldy Hernandez <aldyh@redhat.com>
3059 PR tree-optimization/87415
3060 * tree-vrp.c (set_value_range_with_overflow): Special case one bit
3063 2018-10-02 Jeff Law <law@redhat.com>
3065 * gimple-fold.c (get_range_strlen): Only set *nonstr when
3066 an unterminated string is discovered. Bubble up range
3067 even for unterminated strings.
3068 (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
3069 indicates the string was not terminated via NONSTR.
3071 2018-10-03 Aldy Hernandez <aldyh@redhat.com>
3073 * tree-vrp.c (extract_range_from_unary_expr): Special case all
3074 pointer conversions.
3075 Do not do anything special for anti-ranges.
3077 2018-10-03 Jérôme Lambourg <lambourg@adacore.com>
3079 * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
3080 DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
3082 2018-10-03 Martin Liska <mliska@suse.cz>
3084 PR gcov-profile/86109
3085 * coverage.c (coverage_begin_function): Do not
3086 mark lambdas as artificial.
3087 * tree-core.h (struct GTY): Remove tm_clone_flag
3088 and introduce new lambda_function.
3089 * tree.h (DECL_LAMBDA_FUNCTION): New macro.
3091 2018-10-02 Aaron Sawdey <acsawdey@linux.ibm.com>
3094 * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
3095 P8_VECTOR and VSX are enabled.
3097 2018-10-02 Andreas Krebbel <krebbel@linux.ibm.com>
3099 * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
3100 0x3907 as CPU model number.
3102 2018-10-02 Andreas Krebbel <krebbel@linux.ibm.com>
3104 * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
3105 * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
3106 PF_Z14. Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
3107 TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
3108 TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
3109 * config/s390/s390.md: Likewise. Rename also the cpu attribute
3110 value from arch12 to z14.
3112 2018-10-02 Uros Bizjak <ubizjak@gmail.com>
3114 * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
3116 (isinf<mode>2): Ditto.
3118 2018-10-02 Uros Bizjak <ubizjak@gmail.com>
3120 * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
3121 before emitting fxam. Perform calculations in XFmode.
3123 2018-10-02 Marc Glisse <marc.glisse@inria.fr>
3125 * match.pd (((X /[ex] A) +- B) * A): New transformation.
3127 2018-10-02 Marc Glisse <marc.glisse@inria.fr>
3130 * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
3131 * tree.c (signed_or_unsigned_type_for): Handle complex.
3133 2018-10-02 Jeff Law <law@redhat.com>
3135 * gimple-fold.c (get_range_strlen): Remove dead code.
3137 2018-10-02 Martin Sebor <msebor@redhat.com>
3138 Jeff Law <law@redhat.com>
3140 * builtins.c (unterminated_array): Add new arguments.
3141 If argument is not terminated, bubble up size and exact
3143 (expand_builtin_strnlen): Detect, avoid expanding
3144 and diagnose unterminated arrays.
3145 (c_strlen): Fill in offset of start of unterminated strings.
3146 * builtins.h (unterminated_array): Update prototype.
3148 2018-10-02 Richard Biener <rguenther@suse.de>
3150 * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
3151 of haddv4df, first reduce to SSE width and exploit the fact
3152 that we only need element zero with the reduction result.
3153 (reduc_plus_scal_v2df): Likewise.
3155 2018-10-02 Eric Botcazou <ebotcazou@adacore.com>
3157 * dojump.h (do_jump): Delete.
3158 (do_jump_1): Likewise.
3159 (split_comparison): Move around.
3160 * dojump.c (do_jump): Make static.
3161 (do_jump_1): Likewise.
3162 (jumpifnot): Move around.
3163 (jumpifnot_1): Likewise.
3165 (jumpif_1): Likewise.
3166 * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
3168 2018-10-02 Eric Botcazou <ebotcazou@adacore.com>
3170 * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
3171 insns in the delay slot and add_insn_after for the jump insn.
3173 2018-10-02 Richard Biener <rguenther@suse.de>
3175 * tree-inline.c (expand_call_inline): Use the location of
3176 the callee declaration for the inline-entry marker.
3177 * final.c (notice_source_line): Remove special-casing of
3178 NOTE_INSN_INLINE_ENTRY.
3180 2018-10-01 Carl Love <cel@us.ibm.com>
3183 * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
3184 (__builtin_mtfsb0): New.
3185 (__builtin_mtfsb1): New.
3186 ( __builtin_set_fpscr_rn): New.
3187 (__builtin_set_fpscr_drn): New.
3188 * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
3189 (rs6000_expand_set_fpscr_rn_builtin): Add.
3190 (rs6000_expand_set_fpscr_drn_builtin): Add.
3191 (rs6000_expand_builtin): Add case statement entries for
3192 RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
3193 RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
3194 RS6000_BUILTIN_MFFSL.
3195 (rs6000_init_builtins): Add ftype initialization and def_builtin
3196 calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
3197 __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
3198 * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
3199 rs6000_mffscdrn): Add define_insn.
3200 (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
3201 * doc/extend.texi: Add documentation for the builtins.
3203 2018-10-01 Richard Biener <rguenther@suse.de>
3205 PR tree-optimization/87465
3206 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
3207 causing branch miscounts.
3209 2018-10-01 Tamar Christina <tamar.christina@arm.com>
3211 * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
3212 aarch64_option_default_param): New.
3213 (params.h): Include.
3214 (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
3215 * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
3216 stack-clash protection validation code.
3218 2018-10-01 Tamar Christina <tamar.christina@arm.com>
3220 * params.c (validate_param): New.
3221 (add_params): Use it.
3222 (set_param_value): Refactor param validation into validate_param.
3223 (diagnostic.h): Include.
3224 * diagnostic.h (diagnostic_ready_p): New.
3226 2018-10-01 Tamar Christina <tamar.christina@arm.com>
3228 * params.c (set_param_value):
3229 Add index of parameter being validated.
3230 * common/common-target.def (option_validate_param): New.
3231 * common/common-targhooks.h (default_option_validate_param): New.
3232 * common/common-targhooks.c (default_option_validate_param): New.
3233 * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
3234 * doc/tm.texi: Regenerate.
3236 2018-10-01 Tamar Christina <tamar.christina@arm.com>
3239 * config/aarch64/aarch64.c (aarch64_override_options_internal):
3240 Add validation for stack-clash parameters and set defaults.
3242 2018-10-01 Tamar Christina <tamar.christina@arm.com>
3245 * configure.ac: Add stack-clash-protection-guard-size.
3246 * doc/install.texi: Document it.
3247 * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
3248 * params.def: Update comment for guard-size.
3249 (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
3250 PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
3251 * configure: Regenerate.
3253 2018-10-01 Tamar Christina <tamar.christina@arm.com>
3256 * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
3257 STACK_DYNAMIC_OFFSET): New.
3258 * config/aarch64/aarch64.c (aarch64_layout_frame):
3259 Update outgoing args size.
3260 (aarch64_stack_clash_protection_alloca_probe_range,
3261 TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
3263 2018-10-01 Tamar Christina <tamar.christina@arm.com>
3266 * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
3268 * target.def (stack_clash_protection_alloca_probe_range): New.
3269 (stack_clash_protection_final_dynamic_probe): Remove.
3270 * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
3271 (default_stack_clash_protection_final_dynamic_probe): Remove.
3272 * targhooks.c: Likewise.
3273 * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
3274 (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
3275 * doc/tm.texi: Regenerate.
3277 2018-10-01 Tamar Christina <tamar.christina@arm.com>
3280 * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
3281 * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
3282 aarch64_clamp_to_uimm12_shift): New.
3283 (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
3284 * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
3286 2018-10-01 Tamar Christina <tamar.christina@arm.com>
3289 * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
3291 2018-10-01 Jeff Law <law@redhat.com>
3292 Richard Sandiford <richard.sandiford@linaro.org>
3293 Tamar Christina <tamar.christina@arm.com>
3296 * config/aarch64/aarch64.md
3297 (probe_stack_range): Add k (SP) constraint.
3298 * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
3299 STACK_CLASH_MAX_UNROLL_PAGES): New.
3300 * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
3301 stack probes for stack clash.
3302 (aarch64_allocate_and_probe_stack_space): New.
3303 (aarch64_expand_prologue): Use it.
3304 (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
3305 (aarch64_sub_sp): Add emit_move_imm optional param.
3307 2018-10-01 MCC CS <deswurstes@users.noreply.github.com>
3309 PR tree-optimization/87261
3310 * match.pd: Remove trailing whitespace.
3311 Add (x & y) | ~(x | y) -> ~(x ^ y),
3312 (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
3314 2018-10-01 Claudiu Zissulescu <claziss@synopsys.com>
3316 * config/arc/arc.md (*add_n): Clean up pattern, update instruction
3318 (ashlsi3_insn): Update instruction constraints.
3319 (ashrsi3_insn): Likewise.
3320 (rotrsi3): Likewise.
3321 (add_shift): Likewise.
3322 * config/arc/constraints.md (Csz): New 32 bit constraint. It
3323 avoids placing in the limm field small constants which, otherwise,
3324 could end into a small instruction.
3326 2018-10-01 Claudiu Zissulescu <claziss@synopsys.com>
3328 * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
3329 destination register is not odd-even.
3330 (umaddsidi4_split): Likewise.
3332 2018-10-01 Richard Biener <rguenther@suse.de>
3334 * tree-inline.c (expand_call_inline): Store origin of fn
3335 in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
3336 * tree.c (block_ultimate_origin): Simplify and do some
3339 2018-09-30 Uros Bizjak <ubizjak@gmail.com>
3341 * config/i386/mmx.md (EMMS): New int iterator.
3342 (emms): New int attribute.
3343 (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
3344 EMMS int iterator. Explicitly declare clobbers.
3345 (mmx_emms): Remove expander.
3347 * config/i386/predicates.md (emms_operation): Remove predicate.
3348 (vzeroall_pattern): New predicate.
3349 (vzeroupper_pattern): Rename from vzeroupper_operation.
3350 * config/i386/i386.c (ix86_avx_u128_mode_after): Use
3351 vzeroupper_pattern and vzeroall_pattern predicates.
3353 2018-09-30 Peter Bergner <bergner@linux.ibm.com>
3355 PR rtl-optimization/86939
3356 * ira-lives.c (make_hard_regno_born): Rename from this...
3357 (make_hard_regno_live): ... to this. Remove update to conflict
3358 information. Update function comment.
3359 (make_hard_regno_dead): Add conflict information update. Update
3361 (make_object_born): Rename from this...
3362 (make_object_live): ... to this. Remove update to conflict information.
3363 Update function comment.
3364 (make_object_dead): Add conflict information update. Update function
3366 (mark_pseudo_regno_live): Call make_object_live.
3367 (mark_pseudo_regno_subword_live): Likewise.
3368 (mark_hard_reg_dead): Update function comment.
3369 (mark_hard_reg_live): Call make_hard_regno_live.
3370 (process_bb_node_lives): Likewise.
3371 * lra-lives.c (make_hard_regno_born): Rename from this...
3372 (make_hard_regno_live): ... to this. Remove update to conflict
3373 information. Remove now uneeded check_pic_pseudo_p argument.
3374 Update function comment.
3375 (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
3376 to conflict information. Update function comment.
3377 (mark_pseudo_live): Remove update to conflict information. Update
3379 (mark_pseudo_dead): Add conflict information update.
3380 (mark_regno_live): Call make_hard_regno_live.
3381 (mark_regno_dead): Call make_hard_regno_dead with new arguement.
3382 (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
3384 2018-09-29 H.J. Lu <hongjiu.lu@intel.com>
3387 * config/i386/i386.c (construct_container): Use TImode for
3388 BLKmode values in 2 integer registers.
3390 2018-09-29 Jeff Law <law@redhat.com>
3392 * builtins.c (unterminated_array): Pass in c_strlen_data * to
3393 c_strlen rather than just a tree *.
3394 (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
3395 Update recursive calls appropriately. If caller did not provide a
3396 suitable data pointer, create a local one. When a non-terminated
3397 string is discovered, bubble up information about the string via the
3398 c_strlen_data object.
3399 * builtins.h (c_strlen): Update prototype.
3400 (c_strlen_data): New structure.
3401 * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
3402 For a type 2 call, if c_strlen indicates a non-terminated string
3403 use the length of the non-terminated string.
3404 (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
3406 2018-09-29 Jakub Jelinek <jakub@redhat.com>
3409 * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
3410 __m512d type for __A argument rather than __m512.
3412 2018-09-28 John David Anglin <danglin@gcc.gnu.org>
3414 * match.pd (simple_comparison): Don't optimize if either operand is
3415 a function pointer when target needs function pointer canonicalization.
3417 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
3419 * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
3420 power5 .. power9 to remove indirection.
3421 * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
3422 ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
3423 ASM_CPU_476_SPEC): Delete.
3424 (ASM_CPU_SPEC): Adjust.
3425 (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
3426 asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
3428 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
3430 * config.in: Delete HAVE_AS_DCI.
3431 * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
3432 * config/rs6000/rs6000.h: Ditto.
3433 * configure.ac: Delete HAVE_AS_DCI.
3434 * configure: Regenerate.
3436 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
3438 * config.in (HAVE_AS_LWSYNC): Delete.
3439 * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
3440 * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
3442 * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
3443 * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
3445 * configure.ac: Delete HAVE_AS_LWSYNC.
3446 * configure: Regenerate.
3448 2018-09-28 Eric Botcazou <ebotcazou@adacore.com>
3449 Pierre-Marie de Rodat <derodat@adacore.com>
3451 * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
3452 * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
3453 (cgraph_node::create_thunk): Add indirect_offset parameter.
3454 (thunk_adjust): Likewise.
3455 * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
3456 and initialize the corresponding field with it.
3457 (cgraph_node::dump): Dump indirect_offset field.
3458 * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
3459 * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
3460 (thunk_adjust): Add indirect_offset parameter and deal with it.
3461 (cgraph_node::expand_thunk): Deal with the indirect_offset field and
3462 pass it to thunk_adjust. Do not call the target hook if it's non-zero
3463 or if the thunk is external or local. Fix formatting. Do not chain
3464 the RESULT_DECL to BLOCK_VARS. Pass the static chain to the target,
3465 if any, in the GIMPLE representation.
3466 * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
3467 * lto-cgraph.c (lto_output_node): Write indirect_offset field.
3468 (input_node): Read indirect_offset field.
3469 * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
3470 call to thunk_adjust.
3471 * tree-nested.c (struct nesting_info): Add thunk_p field.
3472 (create_nesting_tree): Set it.
3473 (convert_all_function_calls): Copy static chain from targets to thunks.
3474 (finalize_nesting_tree_1): Return early for thunks.
3475 (unnest_nesting_tree_1): Do not finalize thunks.
3476 (gimplify_all_functions): Do not gimplify thunks.
3478 2018-09-28 David Malcolm <dmalcolm@redhat.com>
3480 * opt-suggestions.c (option_proposer::build_option_suggestions):
3481 Release "option_values".
3483 2018-09-28 David Malcolm <dmalcolm@redhat.com>
3485 * coverage.c (get_coverage_counts): Convert problem-reporting dump
3486 messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
3487 * dumpfile.c (kind_as_string): New function.
3488 (dump_loc): Rather than a hardcoded prefix of "note: ", use
3489 kind_as_string to vary the prefix based on dump_kind.
3490 (selftest::test_capture_of_dump_calls): Update for above.
3492 2018-09-28 Uros Bizjak <ubizjak@gmail.com>
3494 * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
3495 (GET_SSE_REGNO): Rename from SSE_REGNO. Update all uses for rename.
3497 2018-09-28 Uros Bizjak <ubizjak@gmail.com>
3499 * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
3500 * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
3501 INVALID_REGNUM instead of FPSR_REG.
3502 (ix86_md_asm_adjust): Do not clobber FPSR_REG.
3503 * config/i386/i386.md: Update comment of FP compares.
3504 (fldenv): Do not clobber FPSR_REG.
3506 2018-09-28 Richard Biener <rguenther@suse.de>
3508 * tree.h (BLOCK_ORIGIN): New.
3509 * omp-expand.c (grid_expand_target_grid_body): Assign
3510 BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
3511 * tree-inline.c (remap_block): Likewise.
3512 * auto-profile.c (get_function_decl_from_block): Simplify
3513 by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
3514 * langhooks.c (lhd_print_error_function): Likewise.
3515 * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
3517 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
3518 * tree.c (block_nonartificial_location): Likewise.
3519 (block_ultimate_origin): Likewise.
3520 * tree-pretty-print.c (percent_K_format): Likewise. Remove
3521 no longer needed LTO case.
3523 2018-09-28 Andrew Stubbs <ams@codesourcery.com>
3524 Jan Hubicka <jh@suse.cz>
3525 Martin Jambor <mjambor@suse.cz>
3527 * simplify-rtx.c (simplify_merge_mask): New function.
3528 (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
3529 same masks are used in op1 or op2.
3530 (test_vec_merge): New function.
3531 (test_vector_ops): Call test_vec_merge.
3533 2018-09-28 Eric Botcazou <ebotcazou@adacore.com>
3535 * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
3536 * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
3537 * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
3538 (cypress_costs): Set it.
3539 (supersparc_costs): Likewise.
3540 (hypersparc_costs): Likewise.
3541 (leon_cost): Likewise.
3542 (leon3_costs): Likewise.
3543 (sparclet_costs): Likewise.
3544 (ultrasparc_costs): Likewise.
3545 (ultrasparc_costs): Likewise.
3546 (niagara_costs): Likewise.
3547 (niagara2_costs): Likewise.
3548 (niagara3_costs): Likewise.
3549 (niagara4_costs): Likewise.
3550 (niagara7_costs): Likewise.
3551 (m8_costs): Likewise.
3552 (TARGET_CAN_FOLLOW_JUMP): Define.
3553 (pass_work_around_errata::gate): Minor tweak.
3554 (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
3555 Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
3556 Automaitcally clear MASK_FSMULD mask for V7 processors.
3557 (sparc_can_follow_jump): New static function.
3558 (output_ubranch): Deal with CROSSING_JUMP_P.
3559 (sparc_use_sched_lookahead): Rewrite using switch statement.
3560 (sparc_issue_rate): Reorder.
3561 (sparc_branch_cost): New function.
3563 2018-09-27 Martin Sebor <msebor@redhat.com>
3565 * tree.h (tree_to_shwi): Add attribute nonnull and pure.
3566 (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
3567 (int_fits_type_p): Same.
3569 2018-09-27 Uros Bizjak <ubizjak@gmail.com>
3571 * config/i386/i386.md (FPCR_REG): Remove.
3572 (UNSPEC_FLDCW): Remove.
3573 (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
3574 (x86_fldcw_1): Remove insn pattern.
3575 (fnstenv): Do not clobber FPCR_REG.
3577 * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
3578 (CALL_USED_REGISTERS): Ditto.
3579 (REG_ALLOC_ORDER): Ditto.
3580 (REG_CLASS_CONTENTS): Ditto.
3581 (HI_REGISTER_NAMES): Ditto.
3582 (ADDITIONAL_REGISTER_NAMES): Use defines instead
3583 of numerical constants.
3584 * config/i386/i386.c (regclass_map): Remove fpsr register.
3585 (dbx_register_map): Ditto.
3586 (dbx64_register_map): Ditto.
3587 (svr4_dbx_register_map): Ditto.
3588 (print_reg): Do not handle FPCR_REG.
3590 2018-09-27 Segher Boessenkool <segher@kernel.crashing.org>
3593 * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
3594 HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
3595 Delete, always treat as true.
3596 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
3597 Ditto. Simplify remaining code.
3598 * config/powerpcspe/powerpcspe.h: Ditto.
3599 * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
3600 Simplify remaining code.
3601 (rs6000_expand_builtin): Ditto.
3602 * config/rs6000/rs6000.h: Ditto.
3603 * configure.ac: Ditto.
3604 * configure: Regenerate.
3606 2018-09-27 Martin Liska <mliska@suse.cz>
3608 * coverage.c (get_coverage_counts): Revert the formatting
3609 of missing profile opt info.
3611 2018-09-27 Richard Biener <rguenther@suse.de>
3615 * dwarf2out.c (set_block_origin_self): Do not mark outermost
3616 block as we do not output that.
3617 (gen_inlined_subroutine_die): Elide the originally outermost
3618 block, matching what we do for concrete instances.
3619 (decls_for_scope): Add parameter specifying whether to recurse
3622 2018-09-27 Andrew Stubbs <ams@codesourcery.com>
3623 Tom de Vries <tom@codesourcery.com>
3627 * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
3628 STORE_FLAG_VALUE == 1.
3630 2018-09-27 Andreas Krebbel <krebbel@linux.ibm.com>
3632 * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
3633 constant definitions.
3634 ("tx_assist"): Replace magic number with PPA_TX_ABORT.
3635 ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
3636 ("speculation_barrier"): New expander definition.
3638 2018-09-26 Indu Bhagat <indu.bhagat@oracle.com>
3640 PR gcov-profile/86957
3641 * common.opt: New warning option -Wmissing-profile.
3642 * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
3643 * doc/invoke.texi: Document -Wmissing-profile.
3645 2018-09-26 Jim Wilson <jimw@sifive.com>
3647 * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
3648 (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
3650 2018-09-26 Martin Sebor <msebor@redhat.com>
3652 * tree.c (zerop): Change return type to bool.
3653 (integer_zerop, integer_onep, integer_each_onep): Same.
3654 (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
3655 (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
3656 (real_onep, real_minus_onep, chain_index): Same.
3657 (print_type_hash_statistics, type_list_equal): Same.
3658 * tree.h (zerop): Same.
3659 (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
3660 (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
3661 (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
3662 (real_onep, real_minus_onep, chain_index): Same.
3663 (print_type_hash_statistics, type_list_equal): Same.
3665 2018-09-26 Jim Wilson <jimw@sifive.com>
3667 * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
3669 2018-09-26 Jakub Jelinek <jakub@redhat.com>
3672 * config/i386/i386.c: Include debug.h and dwarf2out.h.
3673 (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
3676 2018-09-25 Andrew Stubbs <ams@codesourcery.com>
3678 * builtins.c (get_builtin_sync_mem): Force address mode conversion.
3680 2018-09-26 Uros Bizjak <ubizjak@gmail.com>
3682 * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
3683 and FP_SECOND_SSE_REGS.
3684 (REG_CLASS_NAMES): Ditto.
3685 (REG_CLASS_CONTENTS): Ditto.
3686 * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
3687 FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
3688 (ix86_preferred_output_reload_class): Ditto.
3689 * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
3690 clobber constraint to "=&f".
3691 (fix_truncdi_i387): Ditto.
3692 (lrintxfdi2): Ditto.
3693 (fistdi2_<rounding>): Ditto.
3694 (fpremxf4_i387): Change "=u" constraint to "=f".
3695 (fprem1xf4_i387): Ditto.
3697 (fptanxf4_i387): Ditto.
3698 (fxtractxf3_i387): Ditto.
3699 (fscalexf4_i387): Ditto.
3700 (atan2xf3): Change "u" constraint to "f".
3701 (fyl2xxf3_i387): Ditto.
3702 (fyl2xp1xf3_i387): Ditto.
3704 2018-09-26 Uros Bizjak <ubizjak@gmail.com>
3707 * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
3708 for removed I387_MASK_PM entity.
3711 2018-09-26 Jeff Law <law@redhat.com>
3713 2018-09-26 Alexey Neyman <stilor@att.net>
3715 * graphite.h: Include <isl/id.h> and <isl/space.h>; these
3716 headers are no longer pulled in by <isl/val.h>.
3718 2018-09-26 Richard Biener <rguenther@suse.de>
3721 * dwarf2out.c (gen_lexical_block_die): Do not equate inline
3722 or concrete instance DIE to the tree. Create abstract origin
3723 attributes also for concrete instances.
3725 2018-09-26 Alexey Neyman <stilor@att.net>
3727 * graphite.h: Include <isl/id.h> and <isl/space.h>; these
3728 headers are no longer pulled in by <isl/val.h>.
3730 2018-09-26 Matthew Malcomson <matthew.malcomson@arm.com>
3732 * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
3733 Use new helper functions.
3734 * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
3735 Use new helper functions.
3736 * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
3737 aarch_mm_needs_release): New declarations.
3738 * config/arm/aarch-common.c (aarch_mm_needs_acquire,
3739 aarch_mm_needs_release): New.
3741 2018-09-26 Eric Botcazou <ebotcazou@adacore.com>
3743 * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
3744 (arm32_output_mi_thunk): Deal with long calls.
3746 2018-09-26 Richard Biener <rguenther@suse.de>
3750 * tree-inline.c (expand_call_inline): When the location
3751 of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
3752 or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
3753 the inserted BLOCK to make inlined_function_outer_scope_p
3755 * dwarf2out.c (add_call_src_coords_attributes): Do not add
3756 coords for reserved locations.
3758 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
3760 * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
3761 (*call_indirect_nonlocal_sysv<mode>): Ditto.
3762 (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
3763 (*sibcall_nonlocal_sysv<mode>): Ditto.
3764 (*sibcall_value_nonlocal_sysv<mode>): Ditto.
3765 (<bd>_<mode>): Ditto.
3766 (<bd>tf_<mode>): Ditto.
3768 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
3770 * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
3771 control string as a list of templates instead of as C code.
3772 (*altivec_movti): Ditto.
3773 * config/rs6000/darwin.md (movdf_low_di): Ditto.
3775 2018-09-25 Jim Wilson <jimw@sifive.com>
3777 * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
3778 when target symbol is weak.
3780 2018-09-25 Bernd Edlinger <bernd.edlinger@hotmail.de>
3783 * builtins.c (unterminated_array): Simplify.
3784 * expr.c (string_constant): Handle SSA_NAME. Add more exceptions
3785 where pointer arithmetic is safe.
3787 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
3790 * config/rs6000/altivec.md (altivec_vspltb): Use
3791 const_0_to_15_operand instead of u5bit_cint_operand.
3792 (*altivec_vspltb_internal): Ditto.
3793 (altivec_vspltb_direct): Ditto.
3794 (altivec_vsplth): Use const_0_to_7_operand instead of
3796 (*altivec_vsplth_internal): Ditto.
3797 (altivec_vsplth_direct): Ditto.
3798 (altivec_vspltw): Use const_0_to_3_operand instead of
3800 (*altivec_vspltw_internal): Ditto.
3801 (altivec_vspltw_direct): Ditto.
3802 (altivec_vspltsf): Ditto.
3803 (*altivec_vspltsf_internal): Ditto.
3804 * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
3805 various splats with the proper size immediate. Reorder the various
3806 cases by ascending size of immediate, and put all such together.
3808 2018-09-25 Richard Biener <rguenther@suse.de>
3811 * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
3813 (maybe_create_die_with_external_ref): Do not create
3814 DW_TAG_imported_unit here.
3815 (add_abstract_origin_attribute): Handle external BLOCK refs.
3816 (dwarf2out_abstract_function): Simplify LTO case.
3817 (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
3818 rather than using maybe_create_die_with_external_ref.
3820 2018-09-25 Uros Bizjak <ubizjak@gmail.com>
3823 * config/i386/i386.md (frndintxf2_mask_pm): Remove.
3824 (frndintxf2_mask_pm_i387): Ditto.
3825 (nearbyintxf2): Rewrite expander pattern to match rintxf2.
3826 Enable for !flag_trapping_math.
3827 (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
3828 Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
3829 Change operand 1 predicate to nonimmediate_operand.
3830 (attr "i387_cw"): Remove mask_pm.
3831 * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
3832 (enum ix86_entity): Remove I387_MASK_PM.
3833 * config/i386/i386.c (ix86_i387_mode_needed): Do not
3834 handle I387_MASK_PM.
3835 (ix86_mode_needed): Ditto.
3836 (ix86_mode_after): Ditto.
3837 (ix86_mode_entry): Ditto.
3838 (ix86_mode_exit): Ditto.
3839 (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
3841 2018-09-25 Jakub Jelinek <jakub@redhat.com>
3843 * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
3844 to_update_switch_stmts to vNULL instead of calling create on them
3847 2018-09-25 Richard Biener <rguenther@suse.de>
3849 PR tree-optimization/87402
3850 * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
3851 (visit_phi): Re-instantiate handling of supposed to be VARYING
3852 but non-VARYING backedge value.
3854 2018-09-25 Richard Biener <rguenther@suse.de>
3857 * dwarf2out.c (struct sym_off_pair): New.
3858 (external_die_map): New global.
3859 (lookup_decl_die): When in LTO create DIEs lazily from the
3861 (lookup_block_die): New function, create DIEs lazily in LTO.
3862 (equate_block_to_die): New function.
3863 (dwarf2out_die_ref_for_decl): During WPA get the association
3864 from the external DIE map.
3865 (dwarf2out_register_external_die): Record mapping into the
3867 (maybe_create_die_with_external_ref): New function split out from
3868 DIE generation part of old dwarf2out_register_external_die.
3869 (add_abstract_origin_attribute): Do not return the DIE. When
3870 in LTO reference externals directly.
3871 (dwarf2out_abstract_function): When in LTO ignore calls for
3872 decls with external DIEs (already present abstract instances).
3873 (gen_call_site_die): Adjust.
3874 (add_high_low_attributes): Likewise.
3875 (gen_lexical_block_die): Likewise.
3876 (gen_inlined_subroutine_die): Likewie.
3877 (gen_block_die): Likewise.
3878 (dwarf2out_inline_entry): Likewise.
3879 (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
3882 2018-09-25 Martin Liska <mliska@suse.cz>
3884 * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
3885 integers and not by a float value.
3887 2018-09-25 Martin Liska <mliska@suse.cz>
3890 * dbgcnt.c (dbg_cnt_process_single_pair): Return false
3892 * dumpfile.c (dump_enable_all): Remove extra parenthesis.
3893 * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
3894 * godump.c (go_format_type): Remove extra parenthesis.
3896 2018-09-25 Martin Liska <mliska@suse.cz>
3898 * alias.c (set_dest_equal_p): Remove unused function.
3899 * config/i386/i386.c (def_builtin_pure2): Likewise.
3900 * diagnostic-show-locus.c (class layout): Remove
3902 (layout::layout): Likewise here.
3903 * dump-context.h (class temp_dump_context): Likewise.
3904 * dwarf2out.c (add_AT_fde_ref): Remove unused function.
3905 (add_AT_loclistsptr): Likewise.
3906 (add_AT_offset): Likewise.
3907 (get_AT_hi_pc): Likewise.
3908 (is_comdat_die): Likewise.
3909 (type_is_enum): Likewise.
3910 (ceiling): Likewise.
3911 (add_AT_vms_delta): Likewise.
3912 (is_class_die): Likewise.
3913 * edit-context.c (class line_event): Remove unused field.
3914 * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
3916 * ipa-cp.c (ipa_get_vr_lat): Likewise.
3917 * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
3918 (ok_for_base_p_nonstrict): Likewise.
3919 * tree-chrec.c (is_not_constant_evolution): Likewise.
3920 (chrec_fold_poly_cst): Likewise.
3921 * tree-if-conv.c (has_pred_critical_p): Likewise.
3922 * tree-ssa-coalesce.c (print_exprs): Likewise.
3923 * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
3924 * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
3925 * tree-vrp.c (value_ranges_intersect_p): Likewise.
3926 (value_range_nonnegative_p): Likewise.
3928 2018-09-25 Martin Liska <mliska@suse.cz>
3930 * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
3931 Do not handle "GNU Pascal".
3932 * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
3934 * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
3935 from documentation. Likewise.
3936 * dbxout.c (dbxout_range_type): Likewise.
3937 * doc/cpp.texi: Likewise.
3938 * doc/extend.texi: Likewise.
3939 * doc/frontends.texi: Likewise.
3940 * doc/invoke.texi: Remove Pascal entry.
3941 * tree.def (CLEANUP_POINT_EXPR): Likewise.
3942 * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
3944 2018-09-25 Martin Liska <mliska@suse.cz>
3947 * doc/invoke.texi: Document all parameters and remove default
3950 2018-09-25 Ilya Leoshkevich <iii@linux.ibm.com>
3953 * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
3954 contains HOST_WIDE_INTs when computing its size.
3956 2018-09-24 Jim Wilson <jimw@sifive.com>
3959 * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
3961 (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
3963 2018-09-24 Andrew Pinski <apinski@marvell.com>
3965 * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
3966 access prev before checking it for NULLness in the
3967 AARCH64_FUSE_CMP_BRANCH case.
3969 2018-09-24 H.J. Lu <hongjiu.lu@intel.com>
3972 * config/i386/i386.c (rest_of_insert_endbranch): Set
3973 endbr_queued_at_entrance to true and don't insert ENDBR if
3974 x86_function_profiler will be called.
3975 (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
3977 * config/i386/i386.h (machine_function): Add
3978 endbr_queued_at_entrance.
3980 2018-09-24 Ilya Leoshkevich <iii@linux.ibm.com>
3982 * genattrtab.c (mk_attr_alt): Use alternative_mask.
3983 (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
3985 (check_attr_test): Use alternative_mask.
3986 (get_attr_value): Likewise.
3987 (compute_alternative_mask): Use alternative_mask and XWINT.
3988 (make_alternative_compare): Use alternative_mask.
3989 (attr_alt_subset_p): Use XWINT.
3990 (attr_alt_subset_of_compl_p): Likewise.
3991 (attr_alt_intersection): Use alternative_mask and XWINT.
3992 (attr_alt_union): Likewise.
3993 (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
3994 (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
3995 (simplify_test_exp): Use alternative_mask and XWINT.
3996 (write_test_expr): Use alternative_mask and XWINT, adjust bit
3997 number calculation to support 64 bits. Generate code that
3998 checks 64-bit masks.
3999 (main): Use alternative_mask.
4000 * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
4002 2018-09-24 Ilya Leoshkevich <iii@linux.ibm.com>
4005 * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
4006 RETURN+USE when returning via %r14.
4008 2018-09-24 Martin Liska <mliska@suse.cz>
4010 * gcov.c (output_lines): Print colorization legend
4011 for both flag_use_colors and flag_use_hotness_colors.
4014 2018-09-24 Martin Liska <mliska@suse.cz>
4016 * coverage.c (get_coverage_counts): Use warning_at
4017 with current_function_decl location. Use %qD in warning
4020 2018-09-24 Martin Liska <mliska@suse.cz>
4022 * memory-block.h (memory_block_pool::release): Annotate with
4023 valgrind that the memory is not accessible.
4025 2018-09-24 Martin Liska <mliska@suse.cz>
4028 * asan.c: Make asan_handled_variables extern.
4030 * cfgexpand.c (expand_stack_vars): Make sure
4031 a representative is unpoison if another
4032 variable in the partition is handled by
4033 use-after-scope sanitization.
4035 2018-09-24 Richard Biener <rguenther@suse.de>
4037 PR tree-optimization/63155
4038 * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
4039 the worklist when the edge of the respective argument isn't
4042 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
4044 * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
4045 ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
4046 (MASK_CLASS_P): Update for rename.
4047 (MAYBE_MASK_CLASS_P): Ditto.
4048 (REG_CLASS_NAMES): Update.
4049 (REG_CLASS_CONTENT): Update.
4050 * config/i386/i386.c (regclass_map): Update for MASK_REG
4051 and ALL_MASK_REGS rename.
4052 * config/i386/constraints.md (Yk): Update for rename.
4055 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
4057 * config/i386/i386.h (enum reg_class): Remove
4058 EVEX_SSE_REGS and MOD4_SSE_REGS.
4059 (REG_CLASS_NAMES): Update.
4060 (REG_CLASS_CONTENT): Update.
4061 * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
4062 registers as ALL_SSE_REGS.
4063 (ix86_additional_allocno_class_p): Remove.
4064 (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
4065 (ix86_register_priority): Lower priority of EVEX SSE registers.
4066 Use IN_RANGE macro where appropriate.
4067 (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
4068 AVX-5124VNNIW checks.
4069 (ix86_modes_tieable_p): Tie 512-bit SSE modes.
4070 * config/i386/sse.md (avx5124fmaddps_4fmaddps)
4071 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
4072 (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
4073 (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
4074 (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
4075 (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
4076 (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
4077 (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
4078 (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
4079 (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
4080 * config/i386/constraints.md (Yh): Remove.
4082 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
4084 * config/i386/i386.c (regclass_map): Declare integer REX registers
4087 2018-09-23 Gerald Pfeifer <gerald@pfeifer.com>
4089 * doc/service.texi (Service): Switch the fsf.org link to https.
4091 2018-09-22 Chung-Ju Wu <jasonwucj@gmail.com>
4094 * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4095 Define to speculation_safe_value_not_needed.
4097 2018-09-21 Florian Weimer <fweimer@redhat.com>
4100 * doc/extend.texi (Common Function Attributes): Mention that
4101 noreturn suppresses tail call optimization.
4103 2018-09-21 Jeff Law <law@redhat.com>
4105 * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
4106 vr_values::cleanup_edges_and_switches.
4107 * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
4109 (identify_jump_threads): Remove EDGE_IGNORE handling.
4110 (execute_vrp): Move handling of to_remove_edges and
4111 to_update_switch_stmts into vr_values class member functions.
4112 * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
4113 (to_update_switch_stmts): Likewise.
4114 * vr-values.c: Include cfghooks.h.
4115 (vr_values::vr_values): Initialize to_remove_edges and
4116 to_update_switch_stmts.
4117 (vr_values::~vr_values): Verify to_remove_edges and
4118 to_update_switch_stmts are empty.
4119 (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
4120 (vr_values::cleanup_edges_and_switches): New member function.
4121 * vr-values.h (vr_values): Add cleanup_edges_and_switches member
4122 function. Add new data members.
4124 2018-09-21 David Malcolm <dmalcolm@redhat.com>
4126 PR tree-optimization/87309
4127 * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
4128 calls with pflags and alt_flags.
4129 (selftest::test_capture_of_dump_calls): Add test of interaction of
4130 MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
4132 2018-09-21 Olivier Hainque <hainque@adacore.com>
4134 * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
4136 2018-09-21 Olivier Hainque <hainque@adacore.com>
4138 * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
4140 2018-09-21 Olivier Hainque <hainque@adacore.com>
4142 * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
4143 Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
4144 (VXWORKS_LIBS_RTP): Minor reordering.
4146 2018-09-21 Olivier Hainque <hainque@adacore.com>
4148 * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
4149 (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
4151 2018-09-21 Olivier Hainque <hainque@adacore.com>
4153 * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
4154 (PTRDIFF_TYPE): Likewise.
4156 2018-09-21 Olivier Hainque <hainque@adacore.com>
4158 * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
4159 triplet, similar to support for VxWorks7.
4160 * config/vxworks-dummy.h: Provide a default definition
4161 of TARGET_VXWORKS64 to 0.
4163 2018-09-21 Olivier Hainque <hainque@adacore.com>
4165 * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
4166 * config/vxworks-dummy.h: here.
4168 2018-09-21 Olivier Hainque <hainque@adacore.com>
4170 * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
4172 2018-09-21 Shaokun Zhang <zhangshaokun@hisilicon.com>
4173 Bo Zhou <zbo.zhou@hisilicon.com>
4175 * config/aarch64/aarch64-cores.def (tsv110): New CPU.
4176 * config/aarch64/aarch64-tune.md: Regenerated.
4177 * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
4178 * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
4179 * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
4181 2018-09-21 Andrew Stubbs <ams@codesourcery.com>
4182 Julian Brown <julian@codesourcery.com>
4184 * builtins.c (get_builtin_sync_mem): Handle address spaces.
4186 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
4188 * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
4189 if the call takes a static chain.
4191 2018-09-21 Martin Liska <mliska@suse.cz>
4193 * auto-profile.c (autofdo_source_profile::read): Do not
4195 (read_profile): Do not add working sets.
4196 (read_autofdo_file): Remove sum_all.
4197 (afdo_callsite_hot_enough_for_early_inline): Remove const
4199 * coverage.c (struct counts_entry): Remove gcov_summary.
4200 (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
4201 do not support GCOV_TAG_PROGRAM_SUMMARY.
4202 (get_coverage_counts): Remove summary and expected
4204 * coverage.h (get_coverage_counts): Likewise.
4205 * doc/gcov-dump.texi: Remove -w option.
4206 * gcov-dump.c (dump_working_sets): Remove.
4207 (main): Do not support '-w' option.
4208 (print_usage): Likewise.
4209 (tag_summary): Likewise.
4210 * gcov-io.c (gcov_write_summary): Do not dump
4212 (gcov_read_summary): Likewise.
4213 (gcov_histo_index): Remove.
4214 (gcov_histogram_merge): Likewise.
4215 (compute_working_sets): Likewise.
4216 * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
4218 (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
4219 (GCOV_TAG_SUMMARY_LENGTH): Adjust.
4220 (GCOV_HISTOGRAM_SIZE): Remove.
4221 (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
4222 (struct gcov_summary): Simplify rapidly just
4223 to runs and sum_max fields.
4224 (gcov_histo_index): Remove.
4225 (NUM_GCOV_WORKING_SETS): Likewise.
4226 (compute_working_sets): Likewise.
4227 * gcov-tool.c (print_overlap_usage_message): Remove
4228 trailing empty line.
4229 * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
4230 (output_lines): Remove program related line.
4231 * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
4232 * lto-cgraph.c (output_profile_summary): Do not stream GCOV
4234 (input_profile_summary): Do not read it.
4235 (merge_profile_summaries): And do not merge it.
4236 (input_symtab): Do not call removed function.
4237 * modulo-sched.c (sms_schedule): Do not print sum_max.
4238 * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
4239 removed when histogram method was invented.
4240 (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
4242 * postreload-gcse.c (eliminate_partially_redundant_load): Fix
4244 * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
4245 and dump selected value.
4246 * profile.c (add_working_set): Remove.
4247 (get_working_sets): Likewise.
4248 (find_working_set): Likewise.
4249 (get_exec_counts): Do not work with working sets.
4250 (read_profile_edge_counts): Do not inform as sum_max is removed.
4251 (compute_branch_probabilities): Likewise.
4252 (compute_value_histograms): Remove argument for call of
4253 get_coverage_counts.
4254 * profile.h: Do not make gcov_summary const.
4256 2018-09-21 Monk Chiang <sh.chiang04@gmail.com>
4258 * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
4260 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
4262 PR tree-optimization/86990
4263 * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
4264 Check that the entire merged store group is made of constants only for
4267 2018-09-20 Allan Sandfeld Jensen <allan.jensen@qt.io>
4269 * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
4270 (VTABLE_VERIFICATION_SPEC): Likewise.
4271 (SANITIZER_EARLY_SPEC): Likewise.
4272 (SANITIZER_SPEC): Likewise.
4273 * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
4274 * doc/invoke.texi (Link Options): Document -r.
4276 2018-09-20 Richard Biener <rguenther@suse.de>
4279 * gimplify.c (gimplify_expr): Retain alignment of
4280 addressable lvalue in dereference.
4282 2018-09-20 Alexandre Oliva <aoliva@redhat.com>
4285 * configure.ac: Check for .loc is_stmt support.
4286 * configure, config.in: Rebuilt.
4287 * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
4290 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
4292 * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
4293 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
4296 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
4298 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
4300 * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
4303 2018-09-20 Marek Polacek <polacek@redhat.com>
4305 * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
4307 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
4309 PR tree-optimization/87288
4310 * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
4311 into account when determining PEELING_FOR_NITERS.
4313 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
4315 PR tree-optimization/86877
4316 * tree-vect-loop.c (vect_analyze_loop_2): Call
4317 vect_verify_datarefs_alignment.
4319 2018-09-19 Marek Polacek <polacek@redhat.com>
4321 * doc/invoke.texi: Document -Wclass-conversion.
4323 2018-09-19 John David Anglin <danglin@gcc.gnu.org>
4325 * config/pa/pa.c (pa_adjust_priority): Delete.
4326 (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
4328 * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
4329 (atomic_storehi): Likewise.
4330 (atomic_storesi): Likewise.
4331 (atomic_loaddi): Restore compare and swap exchange loop code.
4333 2018-09-19 Segher Boessenkool <segher@kernel.crashing.org>
4335 PR rtl-optimization/86902
4336 * combine.c (try_combine): When changing the CC mode used, don't change
4337 an unrelated mode in other_insn to that new CC mode.
4339 2018-09-19 David Malcolm <dmalcolm@redhat.com>
4341 * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
4342 with %T in place of calls to dump_generic_expr.
4343 (prune_runtime_alias_test_list): Likewise.
4344 (create_runtime_alias_checks): Likewise.
4345 * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
4346 (vect_analyze_data_ref_dependence): Likewise.
4347 (vect_slp_analyze_data_ref_dependence): Likewise.
4348 (vect_record_base_alignment): Likewise. Use %G in place of call
4349 to dump_gimple_stmt.
4350 (vect_compute_data_ref_alignment): Likewise.
4351 (verify_data_ref_alignment): Likewise.
4352 (vect_find_same_alignment_drs): Likewise.
4353 (vect_analyze_group_access_1): Likewise.
4354 (vect_analyze_data_ref_accesses): Likewise.
4355 (dependence_distance_ge_vf): Likewise.
4356 (dump_lower_bound): Likewise.
4357 (vect_prune_runtime_alias_test_list): Likewise.
4358 (vect_find_stmt_data_reference): Likewise.
4359 (vect_analyze_data_refs): Likewise.
4360 (vect_create_addr_base_for_vector_ref): Likewise.
4361 (vect_create_data_ref_ptr): Likewise.
4362 * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
4363 (vect_can_advance_ivs_p): Likewise.
4364 (vect_update_ivs_after_vectorizer): Likewise.
4365 (vect_gen_prolog_loop_niters): Likewise.
4366 (vect_prepare_for_masked_peels): Likewise.
4367 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
4368 (vect_determine_vectorization_factor): Likewise.
4369 (vect_is_simple_iv_evolution): Likewise.
4370 (vect_analyze_scalar_cycles_1): Likewise.
4371 (vect_analyze_loop_operations): Likewise.
4372 (report_vect_op): Likewise.
4373 (vect_is_slp_reduction): Likewise.
4374 (check_reduction_path): Likewise.
4375 (vect_is_simple_reduction): Likewise.
4376 (vect_create_epilog_for_reduction): Likewise.
4377 (vect_finalize_reduction:): Likewise.
4378 (vectorizable_induction): Likewise.
4379 (vect_transform_loop_stmt): Likewise.
4380 (vect_transform_loop): Likewise.
4381 (optimize_mask_stores): Likewise.
4382 * tree-vect-patterns.c (vect_pattern_detected): Likewise.
4383 (vect_split_statement): Likewise.
4384 (vect_recog_over_widening_pattern): Likewise.
4385 (vect_recog_average_pattern): Likewise.
4386 (vect_determine_min_output_precision_1): Likewise.
4387 (vect_determine_precisions_from_range): Likewise.
4388 (vect_determine_precisions_from_users): Likewise.
4389 (vect_mark_pattern_stmts): Likewise.
4390 (vect_pattern_recog_1): Likewise.
4391 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
4392 (vect_record_max_nunits): Likewise.
4393 (vect_build_slp_tree_1): Likewise.
4394 (vect_build_slp_tree_2): Likewise.
4395 (vect_print_slp_tree): Likewise.
4396 (vect_analyze_slp_instance): Likewise.
4397 (vect_detect_hybrid_slp_stmts): Likewise.
4398 (vect_detect_hybrid_slp_1): Likewise.
4399 (vect_slp_analyze_operations): Likewise.
4400 (vect_slp_analyze_bb_1): Likewise.
4401 (vect_transform_slp_perm_load): Likewise.
4402 (vect_schedule_slp_instance): Likewise.
4403 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
4404 (vect_mark_stmts_to_be_vectorized): Likewise.
4405 (vect_init_vector_1): Likewise.
4406 (vect_get_vec_def_for_operand): Likewise.
4407 (vect_finish_stmt_generation_1): Likewise.
4408 (vect_check_load_store_mask): Likewise.
4409 (vectorizable_call): Likewise.
4410 (vectorizable_conversion): Likewise.
4411 (vectorizable_operation): Likewise.
4412 (vectorizable_load): Likewise.
4413 (vect_analyze_stmt): Likewise.
4414 (vect_is_simple_use): Likewise.
4415 (vect_get_vector_types_for_stmt): Likewise.
4416 (vect_get_mask_type_for_stmt): Likewise.
4417 * tree-vectorizer.c (increase_alignment): Likewise.
4419 2018-09-19 Andrew Stubbs <ams@codesourcery.com>
4421 * doc/rtl.texi: Adjust vec_select description.
4422 * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
4423 non-constant selectors.
4425 2018-09-19 Matthew Malcomson <matthew.malcomson@arm.com>
4427 * config/aarch64/aarch64-protos.h
4428 (aarch64_offset_9bit_signed_unscaled_p): New declaration.
4429 * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
4430 (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
4431 * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
4432 (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
4433 (AARCH64_FL_PROFILE): Move index so flags are ordered.
4434 (AARCH64_ISA_RCPC8_4): New flag.
4435 * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
4436 to aarch64_offset_9bit_signed_unscaled_p.
4437 * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
4439 * config/aarch64/constraints.md (Ust): New constraint.
4440 * config/aarch64/predicates.md.
4441 (aarch64_9bit_offset_memory_operand): New predicate.
4442 (aarch64_rcpc_memory_operand): New predicate.
4444 2018-09-19 Eric Botcazou <ebotcazou@adacore.com>
4446 PR rtl-optimization/87361
4447 * rtlanal.c (nonzero_bits1): Revert accidental change.
4449 2018-09-19 Richard Biener <rguenther@suse.de>
4451 PR tree-optimization/87349
4452 PR tree-optimization/87342
4453 * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
4455 2018-09-18 Marek Polacek <polacek@redhat.com>
4457 P1064R0 - Allowing Virtual Function Calls in Constant Expressions
4458 * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
4460 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
4462 * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
4464 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
4466 PR rtl-optimization/86882
4467 * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
4469 2018-09-18 Uros Bizjak <ubizjak@gmail.com>
4471 * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
4472 *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
4474 2018-09-18 Jonathan Wakely <jwakely@redhat.com>
4477 * doc/invoke.texi (Link Options): Fix formatting and grammar.
4479 2018-09-18 Richard Biener <rguenther@suse.de>
4482 * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
4483 (compute_samebase_partition_bases): Likewise.
4484 (coalesce_ssa_name): Always use compute_optimized_partition_bases.
4485 (gimple_can_coalesce_p): Simplify.
4487 2018-09-18 Hans-Peter Nilsson <hp@bitrange.com>
4489 Handle a library implementation of ffs calling __builtin_ffs.
4490 * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
4491 (mmix_init_libfuncs): New function: make __builtin_ffs expand
4494 2018-09-17 David Malcolm <dmalcolm@redhat.com>
4496 * diagnostic-show-locus.c (class layout_range): Add field
4498 (layout_range::layout_range): Add "original_idx" param and use it
4499 to initialize new field.
4500 (make_range): Use 0 for original_idx.
4501 (layout::layout): Pass in index to calls to
4502 maybe_add_location_range.
4503 (layout::maybe_add_location_range): Add param "original_idx" and
4504 pass it on to layout_range.
4505 (layout::print_any_labels): Pass on range->m_original_idx to
4507 (gcc_rich_location::add_location_if_nearby): Use 0 for
4509 * gcc-rich-location.h (text_range_label::get_text): Update for new
4511 (range_label_for_type_mismatch::get_text): Likewise.
4513 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
4515 * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
4517 2018-09-17 David Malcolm <dmalcolm@redhat.com>
4519 * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
4520 format_string_diagnostic_t.
4521 (fmtwarn_n): Likewise.
4522 * substring-locations.c
4523 (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
4524 (format_warning_n_va): Convert to...
4525 (format_string_diagnostic_t::emit_warning_n_va): ...this.
4526 (format_warning_va): Convert to...
4527 (format_string_diagnostic_t::emit_warning_va): ...this.
4528 (format_warning_at_substring): Convert to...
4529 (format_string_diagnostic_t::emit_warning): ...this.
4530 (format_warning_at_substring_n): Convert to...
4531 (format_string_diagnostic_t::emit_warning_n): ...this.
4532 * substring-locations.h (class format_string_diagnostic_t): New
4534 (format_warning_va): Convert to
4535 format_string_diagnostic_t::emit_warning_va.
4536 (format_warning_n_va): Convert to
4537 format_string_diagnostic_t::emit_warning_n_va.
4538 (format_warning_at_substring): Convert to
4539 format_string_diagnostic_t::emit_warning.
4540 (format_warning_at_substring_n): Convert to
4541 format_string_diagnostic_t::emit_warning_n.
4543 2018-09-17 Cesar Philippidis <cesar@codesourcery.com>
4544 Bernd Schmidt <bernds_cb1@t-online.de>
4546 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
4549 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
4551 * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
4552 operand 0 predicate to nonimmediate operand.
4553 (rint<mode>2_frndint): Remove insn pattern.
4554 (rint<mode>2): Change operand 1 predicate to general_operand.
4555 Extend operand 1 to XFmode and generate rintxf2 insn.
4556 (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
4557 Do not use X87MODEF mode macro.
4558 (frndintxf2_<rounding>_i387): Rename from
4559 frndint<mode>2_<rounding>_i387. Do not use X87MODEF mode macro.
4560 (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
4561 to XFmode and generate significandxf3 insn.
4563 2018-09-17 Richard Biener <rguenther@suse.de>
4565 PR tree-optimization/87328
4566 * tree-ssa-sccvn.c (process_bb): Remove assertion about not
4567 visiting unexecutable backedges when not iterating.
4568 (do_rpo_vn): Mark all edges not executable even when not
4571 2018-09-17 Martin Jambor <mjambor@suse.cz>
4574 * doc/invoke.texi (Warning Options): Likewise.
4576 2018-09-17 Richard Biener <rguenther@suse.de>
4578 PR tree-optimization/87301
4579 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
4580 clean EH info from leftover copy assignments.
4582 2018-09-17 Martin Liska <mliska@suse.cz>
4584 PR gcov-profile/85871
4585 * gcov.c (output_intermediate_file): Fix out of bounds
4588 2018-09-17 Vineet Gupta <vgupta@synopsys.com>
4590 * config/arc/arc.c: Object attributes for core4 not reflected
4592 * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
4595 2018-09-17 Alexey Brodkin <abrodkin@synopsys.com>
4597 * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
4599 2018-09-17 Martin Liska <mliska@suse.cz>
4601 * doc/gcov.texi: Document new option --use-hotness-colors.
4602 * gcov.c (struct source_info): Declare new field.
4603 (source_info::source_info): Set default for maximum_count.
4604 (print_usage): Add new -q option.
4605 (process_args): Process it.
4606 (accumulate_line_info): Save src->maximum_count.
4607 (output_line_beginning): Make color line number if
4608 flag_use_hotness_colors is set.
4609 (output_line_details): Pass default argument value.
4610 (output_lines): Pass src->maximum_count.
4612 2018-09-17 Martin Liska <mliska@suse.cz>
4614 * common/config/i386/i386-common.c (ix86_get_valid_option_values):
4615 Use processor_names table.
4616 * config/i386/i386.c (ix86_default_align): Use
4617 processor_cost_table for alignment values.
4618 (ix86_option_override_internal): Use processor_names.
4619 (ix86_function_specific_print): Likewise.
4620 * config/i386/i386.h (struct processor_costs):
4621 Add alignment values.
4622 (struct ptt): Remove and replace with const char *.
4623 * config/i386/x86-tune-costs.h (struct processor_costs):
4624 Declare default alignments for all costs.
4626 2018-09-17 Aldy Hernandez <aldyh@redhat.com>
4628 * tree-vrp.c (extract_range_from_unary_expr): Do not special case
4629 symbolics or VR_VARYING ranges for ABS_EXPR.
4630 * wide-int-range.cc (wide_int_range_abs): Return positive numbers
4631 when range will wrap.
4633 2018-09-15 Eric Botcazou <ebotcazou@adacore.com>
4636 * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
4637 before and after a JUMP_TABLE_DATA.
4639 2018-09-14 John David Anglin <danglin@gcc.gnu.org>
4642 * dojump.c (do_compare_and_jump): Canonicalize function pointers
4643 when one operand is a function pointer. Use POINTER_TYPE_P and
4644 FUNC_OR_METHOD_TYPE_P.
4645 * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
4646 * fold-const.c (build_range_check): Likewise.
4647 * match.pd (simple_comparison): Likewise.
4649 2018-09-14 David Malcolm <dmalcolm@redhat.com>
4652 * spellcheck.c (get_edit_distance_cutoff): New function.
4653 (selftest::test_edit_distance_unit_test_oneway): Rename to...
4654 (selftest::test_get_edit_distance_one_way): ...this.
4655 (selftest::test_get_edit_distance_unit): Rename to...
4656 (selftest::test_get_edit_distance_both_ways): ...this.
4657 (selftest::test_edit_distances): Move tests to this new function,
4658 and test some more pairs of strings. Update for above renaming.
4659 (selftest::get_old_cutoff): New function.
4660 (selftest::test_get_edit_distance_cutoff): New function.
4661 (selftest::assert_suggested_for): New function.
4662 (ASSERT_SUGGESTED_FOR): New macro.
4663 (selftest::assert_not_suggested_for): New function.
4664 (ASSERT_NOT_SUGGESTED_FOR): New macro.
4665 (selftest::test_suggestions): New function.
4666 (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
4667 tests to selftest::test_edit_distances and call it. Add calls to
4668 selftest::test_get_edit_distance_cutoff and
4669 selftest::test_suggestions.
4670 * spellcheck.h (get_edit_distance_cutoff): New function declaration.
4671 (best_match::consider): Replace hard-coded cutoff calculation with
4673 (best_match::get_cutoff): New declaration.
4674 (best_match::get_best_meaningful_candidate): Likewise.
4676 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
4678 * builtins.c (fold_builtin_strlen): Remove TODO comment.
4680 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
4683 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
4685 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
4686 terminated string literal.
4688 2018-09-14 Martin Sebor <msebor@redhat.com>
4690 * builtins.c (unterminated_array): Handle ARRAY_REF.
4691 (expand_builtin_stpcpy_1): Detect unterminated char arrays.
4692 * builtins.h (unterminated_array): Declare extern.
4693 * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
4695 (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
4698 2018-09-14 Martin Sebor <msebor@redhat.com>
4699 Jeff Law <law@redhat.com>
4701 * builtins.c (unterminated_array): New.
4702 (expand_builtin_strcpy): Adjust.
4703 (expand_builtin_strcpy_args): Detect unterminated arrays.
4704 * gimple-fold.c (get_maxval_strlen): Add argument. Detect
4705 unterminated arrays.
4706 * gimple-fold.h (get_maxval_strlen): Add argument.
4707 (gimple_fold_builtin_strcpy): Detec unterminated arrays.
4709 * gimple-fold.c (get_range_strlen): Add argument.
4710 (get_maxval_strlen): Adjust.
4711 * gimple-fold.h (get_range_strlen): Add argument.
4713 2018-09-14 Wei Xiao <wei3.xiao@intel.com>
4715 * config/i386/movdirintrin.h: Fix copyright year.
4717 2018-09-14 Uros Bizjak <ubizjak@gmail.com>
4719 * reg-stack.c: Include regs.h.
4720 (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
4721 (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
4722 FIRST_STACK_REG, not DFmode.
4723 (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
4724 FIRST_STACK_REG, not XFmode. Explicitly construct swap RTX.
4725 (change stack): Default register mode to the reg_raw_mode of
4726 FIRST_STACK_REG, not DFmode.
4727 * config/i386/i386.md (*swap<mode>): Remove insn pattern.
4728 (*swapxf): Rename from swapxf.
4730 2018-09-14 Carl Love <cel@us.ibm.com>
4732 * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
4733 * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
4735 2018-09-14 Segher Boessenkool <segher@kernel.crashing.org>
4738 * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
4741 2018-09-14 Sam Tebbs <sam.tebbs@arm.com>
4744 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
4746 2018-09-14 Jason Merrill <jason@redhat.com>
4748 Fix --enable-gather-detailed-mem-stats.
4749 * hash-table.c (hash_table_usage): Change from variable to function.
4750 * hash-table.h: Adjust.
4751 * Makefile.in: Add missing dependencies on hash-table.h.
4753 2018-09-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4755 PR tree-optimization/87259
4757 (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
4758 execute_cse_reciprocals_1 has tried transforming.
4760 2018-09-14 Aldy Hernandez <aldyh@redhat.com>
4762 * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
4763 VR_VARYING for PLUS/MINUS_EXPR.
4765 2018-09-14 Ilya Leoshkevich <iii@linux.ibm.com>
4767 * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
4770 2018-09-14 Richard Biener <rguenther@suse.de>
4773 * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
4774 bits for the merged partition.
4776 2018-09-13 Martin Sebor <msebor@redhat.com>
4777 Bernd Edlinger <bernd.edlinger@hotmail.de>
4779 * builtins.h (c_srlen): Add argument.
4780 * builtins.c (warn_string_no_nul): New function.
4781 (c_strlen): Add argument and use it. Update recursive calls.
4782 Pass DECL argument to string_constant to get info on non
4783 terminated strings. Update *NONSTR as needed.
4784 (fold_builtin_strlen): Add argument to calls to c_strlen.
4785 Warn for unterminated arrays.
4786 (warn_string_no_null): Add prototype.
4787 * expr.c (string_constant): Update arguments. Update recursive
4788 calls appropriately. Detect missing NUL terminator and outermost
4789 declaration its missing in.
4790 Improve checks for arrays with nonzero lower bound or elements
4791 that are not a single byte. Simplify offset computation.
4792 Simplify checks for non-NUL terminated strings.
4793 * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
4794 * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
4796 2018-09-13 Bernd Edlinger <bernd.edlinger@hotmail.de>
4798 * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
4800 * fold-const.c (c_getstr): Fix function comment. Remove unused third
4801 argument. Fix range checks.
4802 * fold-const.h (c_getstr): Adjust protoype.
4803 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
4804 string is constant but contains no NUL byte.
4806 * expr.c (string_constant): Adjust function comment.
4807 Remove bogus check for zero termination.
4809 * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
4811 * varasm.c (compare_constant): Compare type size of STRING_CSTs.
4812 (get_constant_size): Don't make STRING_CSTs larger than they are.
4813 (check_string_literal): New check function for STRING_CSTs.
4814 (output_constant): Use it.
4816 2018-09-13 Eric Botcazou <ebotcazou@adacore.com>
4819 * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
4821 2018-09-13 Richard Biener <rguenther@suse.de>
4823 PR tree-optimization/87263
4824 * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
4825 (struct unwind_state): Add max_rpo field.
4826 (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
4827 Compute max_rpo, the max RPO number a block can be backwards reached
4828 from. Re-write non-iterating mode to a RPO ordered worklist approach,
4829 separating it from the iterating mode.
4831 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
4833 * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
4834 (rfs_decision): New scheduling decision.
4836 2018-09-13 Richard Biener <rguenther@suse.de>
4839 * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
4840 (vn_nary_op_insert_pieces_predicated): Do not write useless
4841 valid_dominated_by_p entry outside of the allocated storage.
4843 2018-09-13 Omar Sandoval <osandov@osandov.com>
4844 Tom de Vries <tdevries@suse.de>
4847 * dwarf2out.c (is_c): New function.
4848 (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
4850 2018-09-13 Sam Tebbs <sam.tebbs@arm.com>
4853 * config/aarch64/aarch64.md (*aarch64_bfxil):
4855 * config/aarch64/constraints.md (Ulc): Define.
4856 * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
4858 * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
4861 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
4863 * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
4864 * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
4865 aarch64_layout_frame call.
4866 (aarch64_expand_epilogue): Likewise.
4867 (aarch64_initial_elimination_offset): Likewise.
4868 (aarch64_get_separate_components): Likewise.
4869 (aarch64_use_return_insn_p): Likewise.
4870 (aarch64_layout_frame): Remove unneeded check.
4872 2018-09-13 Jakub Jelinek <jakub@redhat.com>
4874 * configure.ac: Only append
4875 " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
4876 gcc_config_arguments if it was never reconfigured or last reconfigure
4877 was with different arguments.
4878 * configure: Regenerated.
4880 2018-09-13 Jakub Jelinek <jakub@redhat.com>
4881 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4884 * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
4885 (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
4887 2018-09-13 Jakub Jelinek <jakub@redhat.com>
4889 PR tree-optimization/87287
4890 * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
4891 X % C == 0 to X % (unsigned) C == 0 optimization to ...
4892 * match.pd (X % C == 0): ... here. New optimization.
4894 2018-09-12 Jakub Jelinek <jakub@redhat.com>
4897 * expr.h (maybe_optimize_mod_cmp): Declare.
4898 * expr.c (mod_inv): New function.
4899 (maybe_optimize_mod_cmp): New function.
4900 (do_store_flag): Use it.
4901 * cfgexpand.c (expand_gimple_cond): Likewise.
4903 2018-09-09 Cesar Philippidis <cesar@codesourcery.com>
4904 Julian Brown <julian@codesourcery.com>
4907 * gimplify.c (gimplify_scan_omp_clauses): Set
4908 target_firstprivatize_array_bases in OpenACC parallel and kernels
4909 region contexts. Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
4910 OpenACC data regions.
4912 2018-09-12 Uros Bizjak <ubizjak@gmail.com>
4914 * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
4915 (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
4917 2018-09-12 Richard Biener <rguenther@suse.de>
4919 PR tree-optimization/87280
4920 * tree-ssa-sccvn.c (process_bb): Handle the case of executable
4921 edge but unreachable target.
4922 (do_rpo_vn): For conservatively handling a PHI only mark
4923 the backedge executable but not the block reachable.
4925 2018-09-12 Richard Biener <rguenther@suse.de>
4927 PR tree-optimization/87266
4928 * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
4931 2018-09-12 Andreas Krebbel <krebbel@linux.ibm.com>
4933 * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
4935 ("trunc<BFP:mode><DFP_ALL:mode>2")
4936 ("trunc<DFP_ALL:mode><BFP:mode>2")
4937 ("extend<BFP:mode><DFP_ALL:mode>2")
4938 ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
4939 according to the target operand type.
4941 2018-09-12 Jakub Jelinek <jakub@redhat.com>
4942 Andreas Krebbel <krebbel@linux.ibm.com>
4944 PR tree-optimization/86844
4945 * gimple-ssa-store-merging.c
4946 (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
4947 there are any overlapping stores in between them, make sure they are
4948 also coalesced or we give up completely.
4950 2018-09-12 Jakub Jelinek <jakub@redhat.com>
4953 * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
4954 BIT_AND_EXPR's second operand is a power of two. Formatting fix.
4956 2018-09-12 Tom de Vries <tdevries@suse.de>
4958 * common.opt (gdescribe-dies): Add option.
4959 * dwarf2out.c (add_name_and_src_coords_attributes): Add description
4960 attribute for artifical and nameless decls.
4961 (dwarf2out_register_external_die): Add description attribute to
4962 external reference die.
4963 (add_desc_attribute): New functions.
4964 (gen_subprogram_die): Add description attribute to
4965 DW_TAG_call_site_parameter.
4966 * tree-pretty-print.c (print_generic_expr_to_str): New function.
4967 * tree-pretty-print.h (print_generic_expr_to_str): Declare.
4968 * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
4970 (@item -gdescribe-dies): Add.
4972 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
4974 * tree-vrp.c (vrp_shift_undefined_p): Remove.
4975 (extract_range_from_binary_expr_1: Call
4976 wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
4977 * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
4980 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
4982 * gimple-ssa-warn-alloca.c
4983 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
4984 field for ALLOCA_BOUND_*_LARGE.
4986 2018-09-11 Nathan Sidwell <nathan@acm.org>
4988 * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
4990 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
4992 * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
4993 for clobbers. Remove obsolete comment.
4995 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
4997 * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
4998 mpxchk, mpxld and mpxst types.
4999 (define_attr length_immediate): Remove all processing of mpx types.
5000 (define_attr prefix_0f): Ditto.
5001 (define_attr memory): Ditto.
5003 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
5005 * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
5006 (log<mode>2): Change operand 1 predicate to general_operand.
5007 Extend operand 1 to XFmode and generate logxf3 insn.
5008 (log10<mode>2): Change operand 1 predicate to general_operand.
5009 Extend operand 1 to XFmode and generate log10xf3 insn.
5010 (log2<mode>2): Change operand 1 predicate to general_operand.
5011 Extend operand 1 to XFmode and generate log2xf3 insn.
5012 (fyl2xp1_extend<mode>xf3_i387): Remove.
5013 (log1p<mode>2): Change operand 1 predicate to general_operand.
5014 Extend operand 1 to XFmode and generate log1pxf3 insn.
5015 (fxtract_extend<mode>xf3_i387): Remove.
5016 (logb<mode>2): Change operand 1 predicate to general_operand.
5017 Extend operand 1 to XFmode and generate logbxf3 insn.
5018 (ilogb<mode>2): Change operand 1 predicate to general_operand.
5019 Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
5020 (significand<mode>2): Change operand 1 predicate to general_operand.
5021 Extend operand 1 to XFmode and generate significandxf3 insn.
5023 2018-09-11 Nathan Sidwell <nathan@acm.org>
5025 * gcc.c (perror_with_name, pfatal_with_name): Delete.
5026 (load_specs): Use fatal_error.
5027 (DELETE_IF_ORDINARY, process_command): Use error.
5028 (execute, run_attempt): Use fatal_error.
5030 2018-09-11 Andrew Stubbs <ams@codesourcery.com>
5032 * diagnostic-core.h (sorry_at): New prototype.
5033 * diagnostic.c (sorry_at): New function.
5035 2018-09-11 Aldy Hernandez <aldyh@redhat.com>
5037 * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
5038 by zero as VR_UNDEFINED.
5040 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
5042 * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
5043 (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
5044 (<sincos>mode2): New expander.
5045 (sincos_extend<mode>xf3_i387): Remove insn pattern.
5046 (sincos -> sin, cos splitters): Remove splitter patterns.
5047 (sincos<mode>3): Change operand 2 predicate to general_operand.
5048 Extend operand 2 to XFmode and generate sincosxf3 insn.
5049 (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
5050 Change operand 3 predicate to const1_operand.
5051 (fptan_extend<mode>xf4_i387): Remove insn pattern.
5052 (tanxf2): Update operands in the call to fptanxf4_i387.
5053 (tan<mode>2): Change operand 1 predicate to general_operand.
5054 Extend operand 1 to XFmode and generate tanxf3 insn.
5055 (atan2xf3): Rename from *fpatanxf3_i387.
5056 (fpatan_extend<mode>xf3_i387): Remove insn pattern.
5057 (atan2xf3): Remove expander.
5058 (atan2<mode<3): Change operand 1 and 2 predicates to general_operand.
5059 Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
5060 (atan<mode>2): Change operand 1 predicate to general_operand.
5061 Extend operand 1 to XFmode and generate atanxf3 insn.
5063 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
5065 * config/i386/i386.md (x87/SSE constant load splitter): Use
5066 memory_operand instead of nonimmediate_operand for input operand
5069 2018-09-09 Uros Bizjak <ubizjak@gmail.com>
5071 * config/i386/i386.md (float partial SSE register stall splitter): Move
5072 splitter near its instruction pattern.
5073 (float_extend partial SSE register stall splitter): Ditto.
5074 (float_truncate partial SSE register stall splitter): Ditto.
5076 2018-09-09 Hans-Peter Nilsson <hp@bitrange.com>
5079 * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
5080 to speculation_safe_value_not_needed.
5083 * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
5084 non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
5085 (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
5086 leaf_function_p, instead use has_hard_reg_initial_val.
5088 2018-09-09 Nathan Sidwell <nathan@acm.org>
5090 * gcc.h (pfatal_with_name): Don't declare here.
5091 * gcc.c (pfatal_with_name): Make static.
5093 2018-09-09 Xianmiao Qu <xianmiao_qu@c-sky.com>
5095 * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
5098 2018-09-08 John David Anglin <danglin@gcc.gnu.org>
5100 PR rtl-optimization/85458
5101 * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
5102 priority hook to reduce the priority of EXPR.
5104 2018-09-07 Uros Bizjak <ubizjak@gmail.com>
5106 * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
5107 DImode for x87 on 32bit targets. Conditionally disable x87 modes
5108 with X87_ENABLE_FLOAT. Remove preparation code.
5109 (*float<SWI48:mode><MODEF:mode>2): Rename from
5110 *float<SWI48:mode><MODEF:mode>2_mixed. Handle x87, SSE and mixed
5111 math using "enabled" attribute.
5112 (*floatdi<MODEF:mode>2_i387): Rename from
5113 *float<SWI48x:mode><MODEF:mode>2_i387. Handle only DImode and
5114 enable for 32bit targets only.
5115 (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
5117 (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
5118 as operand 1 predicate. Rewrite as define_insn_and_split.
5119 (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
5121 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
5123 * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
5124 to fallthru to FLOAT case.
5126 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
5129 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
5130 around folding of vec_sl to handle out of range shift values.
5132 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
5134 * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
5135 Update callers to gen_fix_trunc<mode>_i387_fisttp
5136 (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
5137 nonimmediate_operand.
5138 (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
5139 and corresponding splitters.
5140 (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
5141 (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
5142 (fix_truncdi_i387_with_temp): Remove insn pattern
5143 and corresponding splitters.
5144 (fix_trunc<mode>_i387): Change operand 0 predicate to
5145 nonimmediate_operand.
5146 (fix_trunc<mode>_i387_with_temp): Remove insn pattern
5147 and corresponding splitters.
5148 (*fistdi2_1): Remove.
5150 (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
5151 (lrintxfdi2): Remove expander. Reimplement as define_insn.
5152 (*fist<mode>2_1): Remove.
5153 (fist<mode>2): Ditto.
5154 (fist<mode>2_with_temp): Remove insn pattern and corresponding
5156 (lrintxf<mode>2): Remove expander. Reimplement as define_insn.
5157 (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
5158 (fistdi2_<rounding>): Change operand 0 predicate to
5159 nonimmediate_operand.
5160 (fistdi2_<rounding>_with_temp): Remove insn pattern
5161 and corresponding splitters.
5162 (fist<mode>2_<rounding>): Change operand 0 predicate to
5163 nonimmediate_operand.
5164 (fist<mode>2_<rounding>_with_temp): Remove insn pattern
5165 and corresponding splitters.
5167 (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
5169 2018-09-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
5171 * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
5174 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
5176 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
5177 early gimple folding of vec_splat().
5178 * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
5179 * gimple-fold.h: Add an extern define for tree_vec_extract().
5181 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
5183 * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
5184 wrappers around TREE_TYPE comparisons.
5186 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
5189 * config/s390/predicates.md: Add nonsym_memory_operand.
5190 * config/s390/s390.c (s390_legitimize_cs_operand): If operand
5191 contains a SYMBOL_REF, load it into an intermediate pseudo.
5192 (s390_emit_compare_and_swap): Legitimize operand.
5193 * config/s390/s390.md: Use the new nonsym_memory_operand
5194 with UNSPECV_CAS patterns.
5196 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
5199 * config/s390/s390-passes.def: New file.
5200 * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
5202 (make_pass_s390_early_mach): Add declaration.
5203 * config/s390/s390.c (make_pass_s390_early_mach):
5204 (s390_option_override): Remove dynamic registration.
5205 * config/s390/t-s390: Add s390-passes.def.
5207 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
5209 * config/s390/s390.c (s390_decompose_constant_pool_ref):
5210 Remove UNSPEC_LTREL_BASE check.
5211 (annotate_constant_pool_refs): Likewise.
5212 (find_constant_pool_ref): Likewise.
5213 (find_ltrel_base): Removed.
5214 (replace_ltrel_base): Removed.
5215 (s390_mainpool_finish): Remove replace_ltrel_base call.
5216 (s390_chunkify_start): Remove pending LTREL_BASE logic.
5217 (s390_chunkify_finish): Remove replace_ltrel_base call.
5218 * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
5220 2018-09-06 Hans-Peter Nilsson <hp@axis.com>
5223 * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
5224 to speculation_safe_value_not_needed.
5226 2018-09-05 Cesar Philippidis <cesar@codesourcery.com>
5227 Bernd Schmidt <bernds_cb1@t-online.de>
5229 * config/nvptx/nvptx-opts.h: New file.
5230 * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
5231 * config/nvptx/nvptx.h: Include "nvptx-opts.h".
5233 (TARGET_SM35): New macro.
5234 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
5236 * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
5238 (misa=): New option.
5239 * doc/invoke.texi (Nvidia PTX Options): Document -misa.
5241 2018-09-05 Uros Bizjak <ubizjak@gmail.com>
5243 * config/i386/i386.md (truncdfsf2): Remove expander.
5244 (truncdfsf2_with_temp): Ditto.
5245 (truncxf<mode>2): Ditto.
5246 (*truncdfsf_fast_mixed): Remove insn pattern.
5247 (*truncdfsf_fast_i387): Ditto.
5248 (*truncdfsf_mixed): Ditto.
5249 (*truncdfsf_i387): Ditto.
5250 (*truncdfsf2_i387_1): Ditto.
5251 (*truncxfsf2_mixed): Ditto.
5252 (*truncxfdf2_mixed): Ditto.
5253 (*truncxf<mode>2_i387_noop): Ditto. Update callers
5254 to call gen_truncxf<mode>2 instead.
5255 (*truncxf<mode>2_i387): Remove.
5256 (reg->reg splitters): Remove splitter pattern.
5257 (reg->mem splitters): Ditto.
5259 (truncdfsf2): New insn pattern.
5260 (truncxf<mode>2): Ditto.
5262 2018-09-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5264 * tree-ssa-math-opts.c (is_mult_by): New function.
5265 (is_square_of): Use the above.
5266 (optimize_recip_sqrt): New function.
5267 (pass_cse_reciprocals::execute): Use the above.
5269 2018-09-05 Richard Biener <rguenther@suse.de>
5272 * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
5273 to zero-init the emplaced vec.
5275 2018-09-05 Martin Liska <mliska@suse.cz>
5277 PR tree-optimization/87205
5278 * tree-switch-conversion.c (pass_lower_switch::execute):
5279 Group cases for switch statements.
5281 2018-09-05 Richard Biener <rguenther@suse.de>
5283 PR tree-optimization/87217
5284 * tree-ssa-sccvn.c (vuse_valueize): New.
5285 (vn_reference_lookup_pieces): Use it.
5286 (vn_reference_lookup): Likewise.
5288 2018-09-05 Nathan Sidwell <nathan@acm.org>
5291 * stor-layout.c (place_field): Scan forwards to check last
5292 bitfield when ms_bitfield_placement is in effect.
5294 2018-09-05 Richard Biener <rguenther@suse.de>
5297 * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
5300 2018-09-05 Siddhesh Poyarekar <siddhesh@sourceware.org>
5301 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
5303 * config/aarch64/falkor-tag-collision-avoidance.c: New file.
5304 * config.gcc (extra_objs): Build it.
5305 * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
5307 * config/aarch64/aarch64-passes.def
5308 (pass_tag_collision_avoidance): New pass.
5309 * config/aarch64/aarch64.c (qdf24xx_tunings): Add
5310 AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
5311 (aarch64_classify_address): Remove static qualifier.
5312 (aarch64_address_info, aarch64_address_type): Move to...
5313 * config/aarch64/aarch64-protos.h: ... here.
5314 (make_pass_tag_collision_avoidance): New function.
5315 * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
5318 2018-09-05 Martin Liska <mliska@suse.cz>
5320 * doc/gcov.texi: Update documentation of humar
5322 * gcov.c (format_count): Print one decimal place, it provides
5323 more fine number of situations like '1G' vs. '1.4G'.
5325 2018-09-05 Martin Liska <mliska@suse.cz>
5328 * config/rs6000/rs6000.opt: Mark the option as Deprecated.
5329 * optc-gen.awk: Allow 'Var' for Deprecated options in order
5330 to generate a MASK value.
5332 2018-09-04 H.J. Lu <hongjiu.lu@intel.com>
5335 * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
5336 if frame pointer isn't used.
5337 (compute_frame_pointer_to_fb_displacement): Likewise.
5339 2018-09-04 Jakub Jelinek <jakub@redhat.com>
5342 * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
5343 OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
5344 OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
5345 (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
5346 and OPTION_MASK_ISA_XSAVEC_UNSET.
5348 2018-09-04 Max Filippov <jcmvbkbc@gmail.com>
5350 * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
5351 XOR operations in NAND case.
5353 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
5355 * wide-int-range.cc (wide_int_range_convert): New.
5356 * wide-int-range.h (wide_int_range_convert): New.
5357 * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
5358 code into wide_int_range_convert.
5359 (extract_range_into_wide_ints): Do not munge anti range constants
5360 into the entire domain. Just return the range back.
5362 2018-09-04 Martin Liska <mliska@suse.cz>
5364 * genmatch.c (output_line_directive): Add new argument
5366 (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
5368 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
5370 * doc/invoke.texi (Option Summary): Add whitespace.
5372 * doc/invoke.texi (Option Summary): Add -Waligned-new.
5374 2018-09-04 Richard Biener <rguenther@suse.de>
5376 PR tree-optimization/87211
5377 * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
5378 backedge value we're supposed to treat as VARYING also number
5379 the PHI to VARYING in case it got a different value-number already.
5381 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
5383 * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
5384 (extract_range_from_binary_expr_1): Do not call
5385 vrp_can_optimize_bit_op.
5386 * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
5388 (wide_int_range_get_mask_and_bounds): New.
5389 (wide_int_range_optimize_bit_op): New.
5390 (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
5391 (wide_int_range_bit_and): Same.
5392 * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
5393 (wide_int_range_optimize_bit_op): New.
5394 (wide_int_range_get_mask_and_bounds): New.
5396 2018-09-04 Richard Biener <rguenther@suse.de>
5398 PR tree-optimization/87176
5399 * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
5400 variable. When value-numbering a virtual PHI node make sure
5401 to not value-number to the backedge value.
5403 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
5405 * doc/extend.texi (Long Long, Hex Floats): Document support for
5406 long long and hex floats in more recent versions of ISO C++.
5408 2018-09-03 Richard Biener <rguenther@suse.de>
5410 PR tree-optimization/87177
5411 * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
5414 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
5416 * bb-reorder.c (edge_order): Convert to C-qsort-style
5417 tri-state comparator.
5418 (reorder_basic_blocks_simple): Change std::stable_sort to
5421 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
5423 * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
5424 tri-state comparator.
5425 (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
5427 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
5429 * sort.cc (struct sort_ctx): New field 'nlim'. Use it...
5430 (mergesort): ... here as maximum count for using netsort.
5431 (gcc_qsort): Set nlim to 3 if stable sort is requested.
5432 (gcc_stablesort): New.
5433 * system.h (gcc_stablesort): Declare.
5435 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
5437 * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
5438 * system.h (qsort): Always redirect to gcc_qsort. Update comment.
5439 * vec.c (qsort_chk): Do not call gcc_qsort. Update comment.
5441 2018-09-03 Segher Boessenkool <segher@kernel.crashing.org>
5443 * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
5444 lxsdx and stxsdx alternatives.
5445 (*mov<mode>_hardfloat64): Ditto.
5446 * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
5448 2018-09-03 Richard Biener <rguenther@suse.de>
5450 PR tree-optimization/87200
5451 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
5454 2018-09-03 Martin Liska <mliska@suse.cz>
5456 PR tree-optimization/87201
5457 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
5458 Fix parenthesis in an expression.
5460 2018-09-03 Richard Biener <rguenther@suse.de>
5462 PR tree-optimization/87197
5463 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
5464 visited. CSE the VN_INFO hashtable lookup.
5466 PR tree-optimization/87169
5467 * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
5468 iterating make sure there's no extra backedges from irreducible
5469 regions feeding the header. Mark the destination block
5472 2018-09-03 Martin Liska <mliska@suse.cz>
5475 * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
5476 * common/common-targhooks.c (default_get_valid_option_values):
5478 * common/common-targhooks.h (default_get_valid_option_values):
5480 * common/config/i386/i386-common.c: Move processor_target_table
5482 (ix86_get_valid_option_values): New function.
5483 (TARGET_GET_VALID_OPTION_VALUES): New macro.
5484 * config/i386/i386.c (struct ptt): Move to i386-common.c.
5485 (PTA_*): Move all defined masks into i386-common.c.
5486 (ix86_function_specific_restore): Use new processor_cost_table.
5487 * config/i386/i386.h (struct ptt): Moved from i386.c.
5488 (struct pta): Likewise.
5489 * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
5490 * doc/tm.texi.in: Likewise.
5491 * opt-suggestions.c (option_proposer::suggest_option):
5492 Pass prefix to build_option_suggestions.
5493 (option_proposer::get_completions): Likewise.
5494 (option_proposer::build_option_suggestions): Use the new target
5496 * opts.c (struct option_help_tuple): New struct.
5497 (print_filtered_help): Use the new target hook.
5499 2018-09-03 Martin Liska <mliska@suse.cz>
5502 * predict.c (set_even_probabilities): Add likely_edges
5503 argument and handle cases where we have precisely one
5505 (combine_predictions_for_bb): Catch also likely_edges.
5506 (tree_predict_by_opcode): Handle gswitch statements.
5507 * tree-cfg.h (find_case_label_for_value): New declaration.
5508 (find_taken_edge_switch_expr): Likewise.
5509 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
5510 Find pivot in decision tree based on probabily, not by number of
5513 2018-09-02 Gerald Pfeifer <gerald@pfeifer.com>
5515 * doc/standards.texi (Standards): Update Objective-C reference.
5517 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
5519 * doc/install.texi (Prerequisites): Update link for MPC.
5521 2018-09-01 Michael Matz <matz@suse.de>
5523 PR tree-optimization/87074
5524 * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
5525 PHIs for outer-loop uses.
5527 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
5529 * doc/generic.texi (OpenMP): Adjust link to openmp.org.
5530 * doc/invoke.texi (C Dialect Options): Ditto.
5532 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
5534 * doc/install.texi (Prerequisites): Adjust link mpfr.org.
5536 2018-08-31 Richard Biener <rguenther@suse.de>
5538 PR tree-optimization/87168
5539 * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
5540 (rpo_elim::eliminate_avail): When OP was not visited it must
5543 2018-08-31 David Malcolm <dmalcolm@redhat.com>
5545 * tree-vrp.c (copy_value_range): Convert param "from" from
5546 "value_range *" to "const value_range *".
5547 (range_is_null): Likewise for param "vr".
5548 (range_int_cst_p): Likewise.
5549 (range_int_cst_singleton_p): Likewise.
5550 (symbolic_range_p): Likewise.
5551 (value_ranges_intersect_p): Likewise for both params.
5552 (value_range_nonnegative_p): Likewise for param "vr".
5553 (value_range_constant_singleton): Likewise.
5554 (vrp_set_zero_nonzero_bits): Likewise for param "ar".
5555 (extract_range_into_wide_ints): Likewise for param "vr".
5556 (extract_range_from_multiplicative_op): Likewise for params "vr0"
5558 (vrp_can_optimize_bit_op): Likewise.
5559 (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
5561 (extract_range_from_unary_expr): Likewise.
5562 (debug_value_range): Likewise for param "vr".
5563 (value_range::dump): Add "const" qualifier.
5564 (vrp_prop::check_array_ref): Convert local "vr" from
5565 "value_range *" to "const value_range *".
5566 (vrp_prop::check_mem_ref): Likewise.
5567 (vrp_prop::visit_stmt): Likewise for local "old_vr".
5568 (vrp_intersect_ranges_1): Likewise for param "vr_1".
5569 (vrp_intersect_ranges): Likewise.
5570 (simplify_stmt_for_jump_threading): Likewise for local "vr".
5571 (vrp_prop::vrp_finalize): Likewise.
5572 * tree-vrp.h (value_range::dump): Add "const" qualifier.
5573 (vrp_intersect_ranges): Add "const" qualifier to params as above.
5574 (extract_range_from_unary_expr): Likewise.
5575 (value_range_constant_singleton): Likewise.
5576 (symbolic_range_p): Likewise.
5577 (copy_value_range): Likewise.
5578 (extract_range_from_binary_expr_1): Likewise.
5579 (range_int_cst_p): Likewise.
5580 (vrp_set_zero_nonzero_bits): Likewise.
5581 (range_int_cst_singleton_p): Likewise.
5583 2018-08-31 Vlad Lazar <vlad.lazar@arm.com>
5585 * config/aarch64/arm_neon.h (vabsd_s64): New.
5586 (vnegd_s64): Likewise.
5588 2018-08-31 Martin Jambor <mjambor@suse.cz>
5590 * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
5592 2018-08-31 Martin Liska <mliska@suse.cz>
5594 * ipa-icf.c (sem_item::add_type): Use
5595 sem_item::m_type_hash_cache.
5596 * ipa-icf.h: Move the cache from sem_item_optimizer
5599 2018-08-31 Nathan Sidwell <nathan@acm.org>
5601 * doc/extend.texi (Backwards Compatibility): Remove implicit
5602 extern C leeway of () being (...).
5604 2018-08-31 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5606 * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
5608 2018-08-31 Segher Boessenkool <segher@kernel.crashing.org>
5612 * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
5614 2018-08-31 Jakub Jelinek <jakub@redhat.com>
5617 * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
5618 gen_int_mode. Formatting fixes.
5620 2018-08-30 Sandra Loosemore <sandra@codesourcery.com>
5622 * target.def (custom_function_descriptors): Improve documentation.
5623 * doc/tm.texi.in (Trampolines): Expand discussion of function
5624 descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
5625 beginning of the section.
5626 * doc/tm.texi: Regenerated.
5628 2018-08-30 Jose E. Marchesi <jose.marchesi@oracle.com>
5630 * cfg.h (class auto_edge_flag): Spell out the template-id of the
5631 base class in the initializer list. This is a workaround for
5632 building with older GCC.
5633 (class auto_bb_flag): Likewise.
5635 2018-08-30 Aaron Sawdey <acsawdey@linux.ibm.com>
5637 * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
5638 (altivec_vcmpequ<VI_char>_p): Remove star.
5639 * config/rs6000/rs6000-string.c (do_load_for_compare): Support
5641 (expand_strncmp_vec_sequence): New function.
5642 (emit_final_str_compare_vec): New function.
5643 (expand_strn_compare): Add support for vector strncmp.
5644 * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
5645 length specification to bytes.
5646 * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
5647 (vcmpnezb_p): New pattern.
5648 * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
5649 for option -mstring-compare-inline-limit.
5651 2018-08-30 Thiago Macieira <thiago.macieira@intel.com>
5653 * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
5654 (PTA_SKYLAKE): Add PTA_AES.
5655 (PTA_GOLDMONT): Likewise.
5657 2018-08-29 Jan Hubicka <jh@suse.cz>
5660 * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
5661 * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
5663 2018-08-29 Jan Hubicka <jh@suse.cz>
5665 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
5667 (hash_tree): Do not visit TYPE_STUB_DECL.
5668 * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
5669 stream TYPE_STUB_DECL.
5670 * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
5671 * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
5672 after free_lang_data.
5673 (type_in_anonymous_namespace_p): Likewise.
5675 2018-08-29 Jan Hubicka <jh@suse.cz>
5677 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
5678 comment that it has to be even number.
5679 (class sreal): Change m_sig type to int32_t.
5680 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
5681 int64_t for temporary calculations.
5682 (sreal_verify_basics): Drop one bit from minimum and maximum.
5684 2018-08-30 Richard Biener <rguenther@suse.de>
5686 PR tree-optimization/87147
5687 * tree-ssa-sccvn.c (SSA_VISITED): New function.
5688 (visit_phi): When the degenerate result is from the backedge and
5689 we didn't visit its definition yet drop to VARYING.
5690 (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
5692 2018-08-29 Jan Hubicka <jh@suse.cz>
5694 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
5696 (hash_tree): Likewise.
5698 2018-08-29 Jan Hubicka <jh@suse.cz>
5700 * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
5701 and TYPE_NEXT_REF_TO.
5703 2018-08-29 Jan Hubicka <jh@suse.cz>
5705 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
5706 comment that it has to be even number.
5707 (class sreal): Change m_sig type to int32_t.
5708 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
5709 int64_t for temporary calculations.
5710 (sreal_verify_basics): Drop one bit from minimum and maximum.
5712 2018-08-30 Tamar Christina <tamar.christina@arm.com>
5714 * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
5716 2018-08-30 Vlad Lazar <vlad.lazar@arm.com>
5719 * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
5720 if to_add is negative.
5722 2018-08-29 Bernd Edlinger <bernd.edlinger@hotmail.de>
5725 * builtins.c (c_strlen): Improve range checks.
5727 2018-08-29 Martin Sebor <msebor@redhat.com>
5728 Jeff Law <law@redhat.com>
5730 PR tree-optimization/86714
5731 PR tree-optimization/86711
5732 * builtins.c (c_strlen): Add arguments to call to string_constant.
5733 * expr.c (string_constant): Add argument. Detect missing nul
5734 terminator and outermost declaration it's missing in.
5735 * expr.h (string_constant): Add argument.
5736 * fold-const.c (read_from_constant_string): Add arguments to call to
5738 (c_getstr): Likewise.
5739 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
5741 * tree-ssa-strlen.c (get_stridx): Likewise.
5743 2018-08-29 Jan Hubicka <jh@suse.cz>
5745 * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
5746 Do not stream DECL_VINDEX.
5747 * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
5748 * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
5749 (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
5751 2018-08-29 Richard Biener <rguenther@suse.de>
5753 * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
5754 virtual operands that are not default defs to honor region
5756 (rpo_vn_valueize): Remove ineffective code here.
5758 2018-08-29 Richard Biener <rguenther@suse.de>
5760 PR tree-optimization/87132
5761 * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
5762 when skipping defs reachable over backedges.
5764 2018-08-29 Richard Biener <rguenther@suse.de>
5766 * tree-core.h: Document use of deprecated_flag in SSA_NAME.
5767 * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
5768 * tree-into-ssa.c (pass_build_ssa::execute): Initialize
5769 function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
5770 * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
5771 (vn_reference_lookup_3): Remove use of const_parms.
5772 (free_rpo_vn): Do not free const_parms.
5773 (do_rpo_vn): Do not call init_const_parms.
5774 * tree-ssa-alias.c (refs_may_alias_p_1): Honor
5775 SSA_NAME_POINTS_TO_READONLY_MEMORY.
5776 (call_may_clobber_ref_p_1): Likewise.
5778 2018-08-29 Alexander Monakov <amonakov@ispras.ru>
5781 * invoke.texi (Optimization Options): List -ftree-scev-cprop.
5783 (-ftree-scev-cprop): Document.
5785 2018-08-29 Jan Hubicka <jh@suse.cz>
5787 * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
5789 (sreal constructor): Update.
5790 * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
5791 sreal:operator/): Update.
5793 2018-08-29 Martin Liska <mliska@suse.cz>
5795 * tree-switch-conversion.c (switch_conversion::expand):
5796 Strenghten assumption about gswitch statements.
5798 2018-08-29 Richard Biener <rguenther@suse.de>
5800 PR tree-optimization/87117
5801 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
5802 re-value-number released SSA VDEFs.
5804 2018-08-29 Richard Biener <rguenther@suse.de>
5806 PR tree-optimization/87126
5807 * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
5809 2018-08-28 Jim Wilson <jimw@sifive.com>
5811 * config/riscv/pic.md: Rewrite.
5812 * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
5814 * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
5815 (SOFTF, default_load, softload, softstore): New.
5817 2018-08-28 Jeff Law <law@redhat.com>
5819 * fold-const.c (fold_binary_loc): Remove recently added assert.
5821 2018-08-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
5823 * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
5824 to OP parmeter of generated function.
5826 2018-08-28 MCC CS <deswurstes@users.noreply.github.com>
5828 PR tree-optimization/87009
5829 * match.pd: Add boolean optimizations.
5831 2018-08-28 Martin Sebor <msebor@redhat.com>
5834 * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
5835 * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
5836 (pass_walloca::gate): Use it.
5837 (alloca_call_type): Same.
5838 (pass_walloca::execute): Same.
5839 * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
5841 2018-08-28 David Malcolm <dmalcolm@redhat.com>
5843 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
5844 GCC_VERSION for usage of "__gcc_dump_printf__" format from
5847 2018-08-28 Richard Biener <rguenther@suse.de>
5849 PR tree-optimization/87124
5850 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
5851 constants before looking up avail.
5853 2018-08-28 Jakub Jelinek <jakub@redhat.com>
5856 * calls.c (maybe_warn_nonstring_arg): Punt early if
5857 warn_stringop_overflow is zero. Don't call get_range_strlen
5858 on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
5859 Swap comparison operands to have constants on rhs. Only use
5860 lenrng[1] if non-NULL and INTEGER_CST. Don't uselessly
5861 increment lenrng[0].
5863 2018-08-28 Richard Sandiford <richard.sandiford@arm.com>
5865 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
5866 use of tree_to_shwi. Remove duplicated test for the size being
5867 a whole number of bytes.
5869 2018-08-28 Richard Biener <rguenther@suse.de>
5871 PR tree-optimization/87117
5872 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
5873 Handle removed stmt without LHS (GIMPLE_NOP).
5875 2018-08-28 Richard Biener <rguenther@suse.de>
5877 PR tree-optimization/87117
5878 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
5879 void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
5881 2018-08-28 Richard Biener <rguenther@suse.de>
5883 PR tree-optimization/87117
5884 * tree-ssa-pre.c (compute_avail): Do not make expressions
5885 with predicated values available.
5886 (get_expr_value_id): Assert we do not run into predicated value
5889 2018-08-28 Richard Biener <rguenther@suse.de>
5891 PR tree-optimization/87117
5892 * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
5893 get virtual operands.
5894 (get_expr_operands): Handle STRING_CST like other decls.
5896 2018-08-28 Martin Liska <mliska@suse.cz>
5898 * tree.h: Update documentation of fndecl_built_in_p
5902 2018-08-27 Jeff Law <law@redhat.com>
5903 PR tree-optimization/87110
5904 * tree-ssa-dse.c (compute_trims): Handle non-constant
5907 2018-08-27 Martin Sebor <msebor@redhat.com>
5909 PR tree-optimization/86914
5910 * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
5912 2018-08-27 Martin Sebor <msebor@redhat.com>
5914 PR tree-optimization/87112
5915 * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
5916 the type of the bound argument.
5918 2018-08-27 Jeff Law <law@redhat.com>
5920 * tree-ssa-dse.c (compute_trims): Handle case where the reference's
5921 type does not have a TYPE_SIZE_UNIT.
5923 2018-08-27 Steve Ellcey <sellcey@cavium.com>
5925 * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
5926 with include of backend.h.
5928 2018-08-27 Richard Biener <rguenther@suse.de>
5930 PR tree-optimization/86927
5931 * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
5932 use const cond reduction code.
5934 2018-08-27 Alexander Monakov <amonakov@ispras.ru>
5936 PR tree-optimization/85758
5937 * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
5939 2018-08-27 David Malcolm <dmalcolm@redhat.com>
5942 * diagnostic-show-locus.c (class layout_range): Update for
5943 conversion of show_caret_p to a tri-state.
5944 (layout_range::layout_range): Likewise.
5945 (make_range): Likewise.
5946 (layout::maybe_add_location_range): Likewise.
5947 (layout::should_print_annotation_line_p): Don't show annotation
5948 lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
5949 (layout::get_state_at_point): Update for conversion of
5950 show_caret_p to a tri-state. Bail out early for
5951 SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
5952 underlining or source colorization.
5953 (gcc_rich_location::add_location_if_nearby): Update for conversion
5954 of show_caret_p to a tri-state.
5955 (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
5956 (selftest::test_one_liner_fixit_replace_equal_secondary_range):
5958 (selftest::test_one_liner_labels): Likewise.
5959 * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
5960 conversion of show_caret_p to a tri-state.
5961 * pretty-print.c (text_info::set_location): Likewise.
5962 * pretty-print.h (text_info::set_location): Likewise.
5963 * substring-locations.c (format_warning_n_va): Likewise.
5964 * tree-diagnostic.c (default_tree_printer): Likewise.
5965 * tree-pretty-print.c (newline_and_indent): Likewise.
5967 2018-08-27 David Malcolm <dmalcolm@redhat.com>
5970 * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
5971 line above for line-insertion fix-it hints.
5972 (selftest::test_fixit_insert_containing_newline): Update the
5973 expected results, and add a test with line-numbering enabled.
5975 2018-08-27 Martin Liska <mliska@suse.cz>
5978 * sanopt.c (sanitize_rewrite_addressable_params): Ignore
5979 params with DECL_HAS_VALUE_EXPR_P.
5981 2018-08-27 Martin Liska <mliska@suse.cz>
5983 * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
5984 selected expansion strategy.
5986 2018-08-27 Martin Liska <mliska@suse.cz>
5988 * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
5989 * builtins.c (is_builtin_fn): Likewise.
5990 * attribs.c (diag_attr_exclusions): Use new function
5991 fndecl_built_in_p and remove check for FUNCTION_DECL if
5993 (builtin_mathfn_code): Likewise.
5994 (fold_builtin_expect): Likewise.
5995 (fold_call_expr): Likewise.
5996 (fold_builtin_call_array): Likewise.
5997 (fold_call_stmt): Likewise.
5998 (set_builtin_user_assembler_name): Likewise.
5999 (is_simple_builtin): Likewise.
6000 * calls.c (gimple_alloca_call_p): Likewise.
6001 (maybe_warn_nonstring_arg): Likewise.
6002 * cfgexpand.c (expand_call_stmt): Likewise.
6003 * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
6004 (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
6005 (cgraph_node::verify_node): Likewise.
6006 * cgraphclones.c (build_function_decl_skip_args): Likewise.
6007 (cgraph_node::create_clone): Likewise.
6008 * config/arm/arm.c (arm_insert_attributes): Likewise.
6009 * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
6010 * dse.c (scan_insn): Likewise.
6011 * expr.c (expand_expr_real_1): Likewise.
6012 * fold-const.c (operand_equal_p): Likewise.
6013 (fold_binary_loc): Likewise.
6014 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
6015 * gimple-low.c (lower_stmt): Likewise.
6016 * gimple-pretty-print.c (dump_gimple_call): Likewise.
6017 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
6019 * gimple.c (gimple_build_call_from_tree): Likewise.
6020 (gimple_call_builtin_p): Likewise.
6021 (gimple_call_combined_fn): Likewise.
6022 * gimplify.c (gimplify_call_expr): Likewise.
6023 (gimple_boolify): Likewise.
6024 (gimplify_modify_expr): Likewise.
6025 (gimplify_addr_expr): Likewise.
6026 * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
6027 * ipa-cp.c (determine_versionability): Likewise.
6028 * ipa-fnsummary.c (compute_fn_summary): Likewise.
6029 * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
6030 * ipa-split.c (visit_bb): Likewise.
6031 (split_function): Likewise.
6032 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
6033 * lto-cgraph.c (input_node): Likewise.
6034 * lto-streamer-out.c (write_symbol): Likewise.
6035 * omp-low.c (setjmp_or_longjmp_p): Likewise.
6036 (lower_omp_1): Likewise.
6037 * predict.c (strip_predict_hints): Likewise.
6038 * print-tree.c (print_node): Likewise.
6039 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
6040 * trans-mem.c (is_tm_irrevocable): Likewise.
6041 (is_tm_load): Likewise.
6042 (is_tm_simple_load): Likewise.
6043 (is_tm_store): Likewise.
6044 (is_tm_simple_store): Likewise.
6045 (is_tm_abort): Likewise.
6046 (tm_region_init_1): Likewise.
6047 * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
6048 * tree-cfg.c (verify_gimple_call): Likewise.
6049 (move_stmt_r): Likewise.
6050 (stmt_can_terminate_bb_p): Likewise.
6051 * tree-eh.c (lower_eh_constructs_2): Likewise.
6052 * tree-if-conv.c (if_convertible_stmt_p): Likewise.
6053 * tree-inline.c (remap_gimple_stmt): Likewise.
6054 (copy_bb): Likewise.
6055 (estimate_num_insns): Likewise.
6056 (fold_marked_statements): Likewise.
6057 * tree-sra.c (scan_function): Likewise.
6058 * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
6059 (optimize_stack_restore): Likewise.
6060 (pass_fold_builtins::execute): Likewise.
6061 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
6062 (mark_all_reaching_defs_necessary_1): Likewise.
6063 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
6064 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
6065 (pass_forwprop::execute): Likewise.
6066 * tree-ssa-loop-im.c (stmt_cost): Likewise.
6067 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
6068 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
6069 * tree-ssa-strlen.c (get_string_length): Likewise.
6070 * tree-ssa-structalias.c (handle_lhs_call): Likewise.
6071 (find_func_aliases_for_call): Likewise.
6072 * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
6073 * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
6074 * tree-tailcall.c (find_tail_calls): Likewise.
6075 * tree.c (need_assembler_name_p): Likewise.
6076 (free_lang_data_in_decl): Likewise.
6077 (get_call_combined_fn): Likewise.
6078 * ubsan.c (is_ubsan_builtin_p): Likewise.
6079 * varasm.c (incorporeal_function_p): Likewise.
6080 * tree.h (DECL_BUILT_IN): Remove and replace with
6082 (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
6083 (fndecl_built_in_p): New.
6085 2018-08-27 Martin Liska <mliska@suse.cz>
6087 PR tree-optimization/86847
6088 * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
6089 Dump also subtree probability.
6090 (switch_decision_tree::do_jump_if_equal): New function.
6091 (switch_decision_tree::emit_case_nodes): Handle special
6092 situations in balanced tree that can be emitted much simpler.
6093 Fix calculation of probabilities that happen in tree expansion.
6094 * tree-switch-conversion.h (struct cluster): Add
6096 (struct simple_cluster): Likewise.
6097 (struct case_tree_node): Add new function has_child.
6098 (do_jump_if_equal): New.
6100 2018-08-27 Martin Liska <mliska@suse.cz>
6102 * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
6103 Add new argument to bit_test_cluster constructor.
6104 (bit_test_cluster::emit): Set bits really number of values
6106 (bit_test_cluster::hoist_edge_and_branch_if_true): Add
6107 probability argument.
6108 * tree-switch-conversion.h (struct bit_test_cluster):
6109 Add m_handles_entire_switch.
6111 2018-08-27 Martin Liska <mliska@suse.cz>
6113 PR tree-optimization/86702
6114 * tree-switch-conversion.c (jump_table_cluster::emit):
6115 Make probabilities even for values in jump table
6116 according to number of cases handled.
6117 (switch_decision_tree::compute_cases_per_edge): Pass
6118 argument to reset_out_edges_aux function.
6119 (switch_decision_tree::analyze_switch_statement): Likewise.
6120 * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
6123 2018-08-27 Martin Liska <mliska@suse.cz>
6125 * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
6126 cfun argument explicitly.
6127 * gimple-pretty-print.c (dump_gimple_switch): Likewise.
6128 * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
6129 function gimple_switch_default_bb.
6130 (convert_switch_statements):
6132 * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
6133 * stmt.c (label_to_block_fn): Use label_to_block and pass
6134 cfun argument explicitly and use gimple_switch_label_bb.
6135 (expand_case): Likewise.
6136 * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
6137 cfun argument explicitly. Likewise.
6138 (make_edges_bb): Likewise.
6139 (make_cond_expr_edges): Likewise.
6140 (get_cases_for_edge): Likewise.
6141 (make_gimple_switch_edges): Likewise.
6142 (label_to_block_fn): Likewise.
6143 (label_to_block): Likewise.
6144 (make_goto_expr_edges): Likewise.
6145 (make_gimple_asm_edges): Likewise.
6146 (main_block_label): Likewise.
6147 (group_case_labels_stmt): Likewise.
6148 (find_taken_edge_computed_goto): Likewise.
6149 (find_taken_edge_switch_expr): Likewise.
6150 (gimple_verify_flow_info): Likewise.
6151 (gimple_redirect_edge_and_branch): Likewise.
6152 (gimple_switch_label_bb): New function.
6153 (gimple_switch_default_bb): Likewise.
6154 (gimple_switch_edge): Likewise.
6155 (gimple_switch_default_edge): Likewise.
6156 * tree-cfg.h (label_to_block_fn): Remove and replace ...
6157 (label_to_block): ... with this.
6158 (gimple_switch_label_bb): New.
6159 (gimple_switch_default_bb): Likewise.
6160 (gimple_switch_edge): Likewise.
6161 (gimple_switch_default_edge): Likewise.
6162 * tree-cfgcleanup.c (convert_single_case_switch): Use
6163 new gimple functions and pass new argument to label_to_block.
6164 (cleanup_control_flow_bb):
6165 * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
6166 cfun argument explicitly.
6167 (make_eh_edges): Likewise.
6168 (redirect_eh_dispatch_edge): Likewise.
6169 (lower_resx): Likewise.
6170 (lower_eh_dispatch): Likewise.
6171 (maybe_remove_unreachable_handlers): Likewise.
6172 (unsplit_eh): Likewise.
6173 (cleanup_empty_eh): Likewise.
6174 (verify_eh_edges): Likewise.
6175 (verify_eh_dispatch_edge): Likewise.
6176 * tree-ssa-dom.c (record_edge_info): Likewise.
6177 * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
6178 * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
6179 (thread_through_normal_block): Likewise.
6180 * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
6181 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
6182 * tree-switch-conversion.c (switch_conversion::collect): Use new
6184 (switch_conversion::check_final_bb): Likewise.
6185 (switch_conversion::gather_default_values): Pass new argument
6187 (switch_conversion::build_constructors): Likewise.
6188 (switch_decision_tree::compute_cases_per_edge): Use new
6189 gimple_switch_edge function.
6190 (switch_decision_tree::analyze_switch_statement): Pass new argument
6192 (switch_decision_tree::try_switch_expansion): Use
6193 gimple_switch_default_edge.
6194 * tree-vrp.c (find_switch_asserts): Pass new argument
6196 * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
6197 (vr_values::simplify_switch_using_ranges): Likewise.
6199 2018-08-27 Richard Biener <rguenther@suse.de>
6201 * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
6202 * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
6204 * tree-ssa-sccvn.h (struct vn_pval): New structure.
6205 (struct vn_nary_op_s): Add unwind_to member. Add
6206 predicated_values flag and put result into a union together
6207 with a linked list of vn_pval.
6208 (struct vn_ssa_aux): Add name member to make maintaining
6209 a map of SSA name to vn_ssa_aux possible. Remove no longer
6210 needed info, dfsnum, low, visited, on_sccstack, use_processed
6211 and range_info_anti_range_p members.
6212 (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
6213 (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
6215 (vn_valueize): New global.
6216 (vn_context_bb): Likewise.
6217 (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
6218 VN_INFO_PTR_INFO): Remove.
6219 * tree-ssa-sccvn.c: ... (rewrite)
6220 (pass_fre::execute): For -O2+ initialize loops and run
6221 RPO VN in optimistic mode (iterating). For -O1 and -Og
6222 run RPO VN in non-optimistic mode.
6223 * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
6224 (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
6225 * doc/invoke.texi (sccvn-max-scc-size): Remove.
6226 (rpo-vn-max-loop-depth): Document.
6227 * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
6228 when valuezing the VUSE signals we walked out of the region.
6229 * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
6230 (phi_translate): Set VN context block to use for availability
6232 (compute_avail): Likewise.
6233 (pre_valueize): New function.
6234 (pass_pre::execute): Adjust to the RPO VN API.
6236 * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
6237 (propagate_constants_for_unrolling): Remove.
6238 (tree_unroll_loops_completely): Perform value-numbering
6239 on the unrolled bodies loop parent.
6241 2018-08-27 Richard Biener <rguenther@suse.de>
6243 * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
6244 for partial antic compute.
6246 2018-08-27 Jakub Jelinek <jakub@redhat.com>
6248 PR rtl-optimization/87065
6249 * combine.c (simplify_if_then_else): Formatting fix.
6250 (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
6252 (known_cond): Don't return const_true_rtx for vector modes. Use
6253 CONST0_RTX instead of const0_rtx. Formatting fixes.
6255 2018-08-27 Martin Liska <mliska@suse.cz>
6257 PR gcov-profile/87069
6258 * gcov.c (process_file): Record files already processed
6259 and warn about a file being processed multiple times.
6261 2018-08-27 Martin Liska <mliska@suse.cz>
6264 * config/aarch64/aarch64.c (aarch64_override_options_internal):
6265 Set default values for x_aarch64_*_string strings.
6266 * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
6267 prefix. For -mabi do not print '=ABI' in help and use
6268 <option_value> format for -msve-vector-bits and -moverride
6271 2018-08-26 Jeff Law <law@redhat.com>
6273 * config/mips/frame-header-opt.c: Include "backend.h" rather than
6276 2018-08-26 Marek Polacek <polacek@redhat.com>
6278 PR c++/87029, Implement -Wredundant-move.
6279 * doc/invoke.texi: Document -Wredundant-move.
6281 2018-08-25 Martin Sebor <msebor@redhat.com>
6283 PR tree-optimization/87059
6284 * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
6285 to the same type as the other.
6286 * fold-const.c (fold_binary_loc): Assert expectation.
6288 2018-08-25 Iain Sandoe <iain@sandoe.co.uk>
6290 * config/darwin.c (machopic_legitimize_pic_address): Clean up
6291 extraneous parentheses, dead code section and formatting.
6293 2018-08-24 David Malcolm <dmalcolm@redhat.com>
6296 * diagnostic-show-locus.c (layout::layout): Ensure the margin is
6297 wide enough for jumps in the line-numbering to be visible.
6298 (layout::print_gap_in_line_numbering): New member function.
6299 (layout::calculate_line_spans): When using line numbering, merge
6300 line spans that are only 1 line apart.
6301 (diagnostic_show_locus): When printing line numbers, show gaps in
6302 line numbering directly, rather than printing headers.
6303 (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
6304 line-numbering with multiple line spans.
6305 (selftest::test_fixit_insert_containing_newline_2): Add test of
6306 line-numbering, in which the spans are close enough to be merged.
6308 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
6310 * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
6311 to range_includes_zero_p. Do not special case VR_ANTI_RANGE.
6312 * tree-vrp.c (range_is_nonnull): Remove.
6313 (range_includes_zero_p): Accept value_range instead of min/max.
6314 (extract_range_from_binary_expr_1): Do not early bail on
6316 Use range_includes_zero_p instead of range_is_nonnull.
6317 (extract_range_from_unary_expr): Use range_includes_zero_p instead
6318 of range_is_nonnull.
6319 (vrp_meet_1): Pass value_range to range_includes_zero_p. Do not
6320 special case VR_ANTI_RANGE.
6321 (vrp_finalize): Same.
6322 * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
6324 (range_is_nonnull): Remove.
6325 * vr-values.c (vrp_stmt_computes_nonzero): Use
6326 range_includes_zero_p instead of range_is_nonnull.
6327 (extract_range_basic): Pass value_range to range_includes_zero_p
6328 instead of range_is_nonnull.
6330 2018-08-24 Uros Bizjak <ubizjak@gmail.com>
6332 * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
6333 * emit-rtl.h (rtl_data): Remove return_bnd.
6334 * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
6335 * function.c (diddle_return_value): Do not handle crtl->return_bnd.
6336 * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
6337 (POINTER_BOUNDS_MODE): Remove definition.
6338 (make_pointer_bounds_mode): Remove.
6339 (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
6340 * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
6341 (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
6342 * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
6343 * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
6344 * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
6345 * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
6347 * config/i386/i386-modes.def (BND32, BND64): Remove.
6348 * config/i386/i386.c (dbx_register_map): Remove bound registers.
6349 (dbx64_register_map): Ditto.
6350 (svr4_dbx_register_map): Ditto.
6351 (indirect_thunk_bnd_needed): Remove.
6352 (indirect_thunks_bnd_used): Ditto.
6353 (indirect_return_bnd_needed): Ditto.
6354 (indirect_return_via_cx_bnd): Ditto.
6355 (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
6356 (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
6357 (output_indirect_thunk): Ditto. Remove need_prefix argument.
6358 (output_indirect_thunk_function): Remove handling of
6359 indirect_return_bnd_needed, indirect_return_via_cx_bnd,
6360 indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
6361 (ix86_save_reg): Remove handling of crtl->return_bnd.
6362 (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
6363 (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
6364 and UNSPEC_BNDLX_ADDR.
6365 (ix86_output_indirect_branch_via_reg): Remove handling of
6366 indirect_thunk_prefix_bnd.
6367 (ix86_output_indirect_branch_via_push): Ditto.
6368 (ix86_output_function_return): Ditto.
6369 (ix86_output_indirect_function_return): Ditto.
6370 (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
6371 * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
6372 (CALL_USED_REGISTERS): Ditto.
6373 (REG_ALLOC_ORDER): Update for removal of bound registers.
6374 (HI_REGISTER_NAMES): Ditto.
6375 * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
6376 (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
6377 (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
6378 (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
6379 (FIRST_PSEUDO_REG): Update.
6380 (BND): Remove mode iterator.
6381 * config/i386/predicates.md (bnd_mem_operator): Remove.
6383 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
6385 * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
6388 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
6390 * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
6391 the case in which the permute needs only a single element and
6392 repeats for every vector of the result. Extend that case to
6393 handle variable-length vectors.
6394 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
6396 2018-08-24 H.J. Lu <hongjiu.lu@intel.com>
6399 * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
6400 on DW_MACINFO_start_file for -gsplit-dwarf -g3.
6402 2018-08-24 Richard Biener <rguenther@suse.de>
6404 * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
6405 bb_flags_allocated members.
6406 (auto_flag): New RAII class for allocating flags.
6407 (auto_edge_flag): New RAII class for allocating edge flags.
6408 (auto_bb_flag): New RAII class for allocating bb flags.
6409 * cfgloop.c (verify_loop_structure): Allocate temporary edge
6411 * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
6412 in favor of temporarily allocated BB flag.
6413 * hsa-brig.c: Re-order includes.
6414 * hsa-dump.c: Likewise.
6415 * hsa-regalloc.c: Likewise.
6416 * print-rtl.c: Likewise.
6417 * profile-count.c: Likewise.
6419 2018-08-24 Segher Boessenkool <segher@kernel.crashing.org>
6422 * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
6425 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
6428 * wide-int-range.cc (wide_int_range_div): Do not ignore result
6429 from wide_int_range_multiplicative_op.
6431 2018-08-23 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6433 * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
6436 2018-08-23 Giuliano Belinassi <giuliano.belinassi@usp.br>
6438 * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
6441 2018-08-23 Alexander Monakov <amonakov@ispras.ru>
6443 * tree-scalar-evolution.c (final_value_replacement_loop): Dump
6444 full GENERIC expression used for replacement.
6446 2018-08-23 Aldy Hernandez <aldyh@redhat.com>
6448 * tree-vrp.c (abs_extent_range): Remove.
6449 (extract_range_into_wide_ints): Pass wide ints by reference.
6450 (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
6451 Pass wide ints by reference in all calls to
6452 extract_range_into_wide_ints.
6453 * wide-int-range.cc (wide_int_range_div): New.
6454 * wide-int-range.h (wide_int_range_div): New.
6455 (wide_int_range_includes_zero_p): New.
6456 (wide_int_range_zero_p): New.
6458 2018-08-23 Matthew Malcomson <matthew.malcomson@arm.com>
6460 * config/aarch64/aarch64.md (arches): New enum.
6461 (arch): New enum attr.
6462 (arch_enabled): New attr.
6463 (enabled): Now uses arch_enabled only.
6464 (simd, sve, fp16): Removed attribute.
6465 (fp): Attr now defined in terms of 'arch'.
6466 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
6467 *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
6468 <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
6469 <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
6470 attributes into 'arch'.
6471 (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
6472 subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
6473 *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
6474 *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
6475 *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
6476 'simd' attribute into 'arch'.
6477 (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
6478 store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
6479 Convert use of 'fp' attribute to 'arch'.
6480 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
6481 move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
6483 (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
6484 (different modes) Merge 'fp' and 'simd' into 'arch'.
6485 (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
6488 2018-08-23 Segher Boessenkool <segher@kernel.crashing.org>
6490 PR rtl-optimization/87026
6491 * expmed.c (canonicalize_comparison): If we can no longer create
6492 pseudoregisters, don't.
6494 2018-08-23 Richard Earnshaw <rearnsha@arm.com>
6497 * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
6499 * config/arm/arm.c (speculation_barrier_libfunc): New static
6501 (arm_init_libfuncs): Initialize it.
6502 (arm_emit_speculation_barrier): New function.
6503 * config/arm/arm.md (speculation_barrier): Call
6504 arm_emit_speculation_barrier for architectures that do not have
6506 (speculation_barrier_insn): Only match on Armv7 or later.
6508 2018-08-23 Richard Biener <rguenther@suse.de>
6511 * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
6514 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
6516 * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
6517 of single-vector TBLs.
6518 (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
6521 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
6524 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
6525 aarch64_evpc_tbl guard.
6527 2018-08-22 Bernd Edlinger <bernd.edlinger@hotmail.de>
6529 * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
6532 2018-08-22 Martin Sebor <msebor@redhat.com>
6535 * tree-pretty-print.c (pretty_print_string): Add argument.
6536 (dump_generic_node): Call to pretty_print_string with string size.
6538 2018-08-22 Segher Boessenkool <segher@kernel.crashing.org>
6540 PR rtl-optimization/86771
6541 * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
6542 of two SETs into those two SETs, one to be placed at i2, if that SETs
6543 destination is modified between i2 and i3.
6545 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
6547 PR tree-optimization/86725
6548 * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
6550 (vect_analyze_scalar_cycles_1): Check it.
6552 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
6554 PR tree-optimization/86725
6555 * tree-vect-loop.c (vect_is_simple_reduction): When treating
6556 an outer loop phi as a double reduction, make sure that the
6557 single user of the phi result is an inner loop phi.
6559 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
6561 * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
6562 grouped stores with gaps to a strided group.
6564 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
6566 * tree-vect-stmts.c (get_group_load_store_type)
6567 (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
6568 first statement in a group.
6570 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
6572 * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
6573 the sequence used in gcc/gcc.c.
6575 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
6578 * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
6581 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
6583 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
6584 Darwin10-specific unwinder-shim.
6585 * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
6586 * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
6587 New to cater for Darwin10 Rosetta.
6589 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
6591 * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
6592 specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
6594 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
6599 * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
6600 (dwarf2out_switch_text_section): Generate a local label for the
6601 second function sub-section and apply it as the second FDE start
6603 * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
6604 second sub-section start.
6606 2018-08-22 Richard Biener <rguenther@suse.de>
6608 PR tree-optimization/86988
6609 * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
6611 2018-08-22 Richard Biener <rguenther@suse.de>
6613 PR tree-optimization/86945
6614 * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
6616 2018-08-22 Alexandre Oliva <oliva@adacore.com>
6618 * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
6619 a comment about how uses of r2 for .sdata2 come about.
6621 2018-08-22 Alexandre Oliva <aoliva@redhat.com>
6623 * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
6625 2018-08-21 Marek Polacek <polacek@redhat.com>
6627 PR c++/86981, Implement -Wpessimizing-move.
6628 * doc/invoke.texi: Document -Wpessimizing-move.
6630 2018-08-21 Jan Hubicka <jh@suse.cz>
6632 * tree.c (find_decls_types_r): Do not check for redundant typedefs.
6633 * tree.h (is_redundant_typedef): Remove.
6634 * dwarf2out.c (is_redundant_typedef): Turn into static function.
6636 2018-08-21 Jan Hubicka <jh@suse.cz>
6638 * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
6641 2018-08-21 Tamar Christina <tamar.christina@arm.com>
6643 * expmed.c (extract_low_bits): Reject invalid subregs early.
6645 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
6648 * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
6651 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
6653 * config/vxworks.h: Guard vxworks_asm_out_constructor and
6654 vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
6655 * config/vxworks.c: Likewise.
6657 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
6659 * config/vxworks.c: Set targetm.have_ctors_dtors
6660 if HAVE_INITFINI_ARRAY_SUPPORT.
6661 * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
6662 if HAVE_INITFINI_ARRAY_SUPPORT.
6664 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
6666 * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
6667 default search path for VxWorks < 7.
6669 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
6671 * gimple-ssa-sprintf.c (decl_constant_value): Remove.
6672 (get_format_string): Refer to c_getstr.
6674 2018-08-21 Tom de Vries <tdevries@suse.de>
6676 * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
6677 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
6678 (debuginfo_early_stop): Declare.
6679 * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
6680 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
6681 (debuginfo_early_stop): New function.
6682 (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
6683 and debuginfo_early_stop.
6684 * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
6685 * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
6686 (general_init): Call debuginfo_early_init.
6687 (finalize): Call debuginfo_fini.
6688 (do_compile): Call debuginfo_init.
6689 * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
6691 (@item -fdump-debug, @item -fdump-earlydebug): Add.
6693 2018-08-21 Tom de Vries <tdevries@suse.de>
6695 * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
6696 flag_dump_noaddr and flag_dump_unnumbered.
6698 2018-08-21 Aldy Hernandez <aldyh@redhat.com>
6700 * wide-int-range.cc (wide_int_range_abs): New.
6701 (wide_int_range_order_set): Rename from wide_int_range_min_max.
6702 * wide-int-range.h (wide_int_range_abs): New.
6703 (wide_int_range_min_max): New.
6704 * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
6705 case to call wide_int_range_abs.
6706 Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
6707 (extract_range_from_abs_expr): Delete.
6709 2018-08-20 Michael Meissner <meissner@linux.ibm.com>
6712 * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
6713 from 'Y' to 'YZ' to enable the LWAX instruction to be generated
6716 2018-08-20 Nathan Sidwell <nathan@acm.org>
6717 Jeff Law <law@redhat.com>
6719 * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
6720 * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
6722 2018-08-20 David Malcolm <dmalcolm@redhat.com>
6725 * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
6726 (decl_attributes): Likewise.
6727 * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
6729 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
6730 * diagnostic-core.h (class auto_diagnostic_group): New class.
6731 * diagnostic.c (diagnostic_initialize): Initialize the new fields.
6732 (diagnostic_report_diagnostic): Handle the first diagnostics within
6734 (emit_diagnostic): Add auto_diagnostic_group instance.
6736 (inform_n): Likewise.
6737 (warning): Likewise.
6738 (warning_at): Likewise.
6739 (warning_n): Likewise.
6740 (pedwarn): Likewise.
6741 (permerror): Likewise.
6743 (error_n): Likewise.
6744 (error_at): Likewise.
6746 (fatal_error): Likewise.
6747 (internal_error): Likewise.
6748 (internal_error_no_backtrace): Likewise.
6749 (auto_diagnostic_group::auto_diagnostic_group): New ctor.
6750 (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
6751 * diagnostic.h (struct diagnostic_context): Add fields
6752 "diagnostic_group_nesting_depth",
6753 "diagnostic_group_emission_count", "begin_group_cb",
6755 * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
6756 Add auto_diagnostic_group instance(s).
6757 (find_explicit_erroneous_behavior): Likewise.
6758 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
6759 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
6760 * gimplify.c (warn_implicit_fallthrough_r): Likewise.
6761 (gimplify_va_arg_expr): Likewise.
6762 * hsa-gen.c (HSA_SORRY_ATV): Likewise.
6763 (HSA_SORRY_AT): Likewise.
6764 * ipa-devirt.c (compare_virtual_tables): Likewise.
6765 (warn_odr): Likewise.
6766 * multiple_target.c (expand_target_clones): Likewise.
6767 * opts-common.c (cmdline_handle_error): Likewise.
6768 * reginfo.c (globalize_reg): Likewise.
6769 * substring-locations.c (format_warning_n_va): Likewise.
6770 * tree-inline.c (expand_call_inline): Likewise.
6771 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
6772 * tree-ssa-loop-niter.c
6773 (do_warn_aggressive_loop_optimizations): Likewise.
6774 * tree-ssa-uninit.c (warn_uninit): Likewise.
6775 * tree.c (warn_deprecated_use): Likewise.
6777 2018-08-20 H.J. Lu <hongjiu.lu@intel.com>
6780 * config/i386/i386.md (eh_return): Always update EH return
6781 address in word_mode.
6783 2018-08-20 Chung-Lin Tang <cltang@codesourcery.com>
6785 * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
6786 TARGET_SPLIT_COMPLEX_ARG is defined.
6788 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
6790 * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
6792 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
6795 * expr.c (expand_assignment): Assert that bitpos is positive.
6796 (store_field): Likewise
6797 (expand_expr_real_1): Make sure that bitpos is positive.
6798 * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
6801 2018-08-20 Nathan Sidwell <nathan@acm.org>
6803 * Makefile.in (CPP_ID_DATA_H): Delete.
6804 (CPP_INTERNAL_H): Don't add it.
6805 (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
6806 * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
6808 2018-08-20 Richard Biener <rguenther@suse.de>
6810 PR tree-optimization/78655
6811 * tree-vrp.c (extract_range_from_binary_expr_1): Make
6812 pointer + offset nonnull if either operand is nonnull work.
6814 2018-08-20 Tom de Vries <tdevries@suse.de>
6816 * dwarf2out.c (add_scalar_info): Don't add reference to existing die
6817 unless the referenced die describes the added property using
6818 DW_AT_location or DW_AT_const_value. Fall back to exprloc case.
6819 Otherwise, add a DW_AT_location to the referenced die.
6821 2018-08-19 Uros Bizjak <ubizjak@gmail.com>
6824 * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
6825 register_operand when calling ix86_set_reg_reg_cost.
6826 [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
6827 Set *total to 0 for operands that satisfy x86_64_immediate_operand
6828 predicate and to 1 otherwise.
6830 2018-08-18 Iain Sandoe <iain@sandoe.co.uk>
6832 * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
6833 emit a diagnostic that it is not supported and reset the option.
6834 * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
6835 supported and consume the option. (ASM_FINAL_SPEC): New.
6837 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
6839 * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
6842 2018-08-17 Sandra Loosemore <sandra@codesourcery.com>
6844 C-SKY port: Documentation
6846 * doc/extend.texi (C-SKY Function Attributes): New section.
6847 * doc/invoke.texi (Option Summary): Add C-SKY options.
6848 (C-SKY Options): New section.
6849 * doc/md.texi (Machine Constraints): Document C-SKY constraints.
6851 2018-08-17 Jojo <jijie_rong@c-sky.com>
6852 Huibin Wang <huibin_wang@c-sky.com>
6853 Sandra Loosemore <sandra@codesourcery.com>
6854 Chung-Lin Tang <cltang@codesourcery.com>
6856 C-SKY port: Backend implementation
6858 * config/csky/*: New.
6859 * common/config/csky/*: New.
6861 2018-08-17 Jojo <jijie_rong@c-sky.com>
6862 Huibin Wang <huibin_wang@c-sky.com>
6863 Sandra Loosemore <sandra@codesourcery.com>
6864 Chung-Lin Tang <cltang@codesourcery.com>
6865 Andrew Jenner <andrew@codesourcery.com>
6867 C-SKY port: Configury
6869 * config.gcc (csky-*-*): New.
6870 * configure.ac: Add csky to targets for dwarf2 debug_line support.
6871 * configure: Regenerated.
6873 2018-08-17 David Malcolm <dmalcolm@redhat.com>
6875 * dump-context.h: Include "dumpfile.h".
6876 (dump_context::dump_printf_va): Convert final param from va_list
6877 to va_list *. Convert from ATTRIBUTE_PRINTF to
6878 ATTRIBUTE_GCC_DUMP_PRINTF.
6879 (dump_context::dump_printf_loc_va): Likewise.
6880 * dumpfile.c: Include "stringpool.h".
6881 (make_item_for_dump_printf_va): Delete.
6882 (make_item_for_dump_printf): Delete.
6883 (class dump_pretty_printer): New class.
6884 (dump_pretty_printer::dump_pretty_printer): New ctor.
6885 (dump_pretty_printer::emit_items): New member function.
6886 (dump_pretty_printer::emit_any_pending_textual_chunks): New member
6888 (dump_pretty_printer::emit_item): New member function.
6889 (dump_pretty_printer::stash_item): New member function.
6890 (dump_pretty_printer::format_decoder_cb): New member function.
6891 (dump_pretty_printer::decode_format): New member function.
6892 (dump_context::dump_printf_va): Reimplement in terms of
6893 dump_pretty_printer.
6894 (dump_context::dump_printf_loc_va): Convert final param from va_list
6896 (dump_context::begin_scope): Reimplement call to
6897 make_item_for_dump_printf.
6898 (dump_printf): Update for change to dump_printf_va.
6899 (dump_printf_loc): Likewise.
6900 (selftest::test_capture_of_dump_calls): Convert "stmt" from
6901 greturn * to gimple *. Add a test_decl. Add tests of dump_printf
6902 with %T, %E, and %G.
6903 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
6904 (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
6905 ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
6906 (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
6907 ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
6908 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
6909 use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
6910 within a dump_printf_loc call to "%wu".
6911 (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
6912 converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd". Add a
6913 missing space after "=".
6914 * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
6915 call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
6916 * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
6917 convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
6918 * tree-vectorizer.c (try_vectorize_loop_1): Likewise. Remove
6919 duplicate "vectorized" from message.
6921 2018-08-17 Szabolcs Nagy <szabolcs.nagy@arm.com>
6923 * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
6924 polyNxK_t element's TYPE_STRING_FLAG.
6926 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
6928 * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
6929 (they were unnamed before). Fix comments.
6931 2018-08-17 Nathan Sidwell <nathan@acm.org>
6933 * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
6935 2018-08-17 Richard Biener <rguenther@suse.de>
6937 PR tree-optimization/86841
6938 * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
6940 2018-08-17 Martin Liska <mliska@suse.cz>
6942 * common.opt: Remove Warn, Init and Report for options with
6943 Ignore/Deprecated flag. Warning is done automatically for
6945 * config/i386/i386.opt: Likewise.
6946 * config/ia64/ia64.opt: Likewise.
6947 * config/rs6000/rs6000.opt: Likewise.
6948 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
6949 Remove usage of flag_check_pointer_bounds.
6950 * lto-wrapper.c (merge_and_complain): Do not handle
6951 OPT_fcheck_pointer_bounds.
6952 (append_compiler_options): Likewise.
6953 * opt-functions.awk: Do not handle Deprecated.
6954 * optc-gen.awk: Check that Var, Report and Init are not
6955 used for an option with Ignore/Deprecated flag.
6956 * opts-common.c (decode_cmdline_option): Do not report
6958 (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
6960 * opts.h (struct cl_option): Remove cl_deprecated flag.
6961 (CL_ERR_DEPRECATED): Remove error enum value.
6963 2018-08-17 Richard Biener <rguenther@suse.de>
6966 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
6967 across a va-arg-pack using call adjust its return value accordingly.
6969 2018-08-16 Martin Sebor <msebor@redhat.com>
6971 PR tree-optimization/86853
6972 * gimple-ssa-sprintf.c (struct format_result): Rename member.
6973 (struct fmtresult): Add member and initialize it in ctors.
6974 (format_character): Handle %C. Extend range to NUL. Set MAYFAIL.
6975 (format_string): Handle %S the same as %ls. Set MAYFAIL.
6976 (format_directive): Set POSUNDER4K when MAYFAIL is set.
6977 (parse_directive): Handle %C same as %c.
6978 (sprintf_dom_walker::compute_format_length): Adjust.
6979 (is_call_safe): Adjust.
6981 2018-08-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
6983 * builtins.c (c_strlen): Add new parameter eltsize. Use it
6984 for determining how to count the elements.
6985 * builtins.h (c_strlen): Adjust prototype.
6986 * expr.c (string_constant): Add new parameter mem_size.
6987 Set *mem_size appropriately.
6988 * expr.h (string_constant): Adjust protoype.
6989 * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
6990 * gimple-fold.h (get_range_strlen): Adjust prototype.
6991 * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
6992 (format_string): Call get_string_length with eltsize.
6994 2018-08-16 David Malcolm <dmalcolm@redhat.com>
6996 * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
6997 to emit the span, rather than setting it as the prefix.
6999 2018-08-16 David Malcolm <dmalcolm@redhat.com>
7001 * diagnostic-show-locus.c (layout::start_annotation_line): Add
7002 "margin_char" parameter, defaulting to space. Use it in place
7003 of pp_space for the initial part of the margin.
7004 (layout::print_leading_fixits): Use '+' when filling the margin
7005 of line-insertion fix-it hints.
7007 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
7009 * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
7012 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
7014 * config/rs6000/altivec.md: Don't set length attribute to the default
7016 * config/rs6000/darwin.md: Ditto.
7017 * config/rs6000/dfp.md: Ditto.
7018 * config/rs6000/htm.md: Ditto.
7019 * config/rs6000/rs6000.md: Ditto.
7020 * config/rs6000/sync.md: Ditto.
7021 * config/rs6000/vsx.md: Ditto.
7023 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
7025 * config/rs6000/altivec.md: Don't set length attribute to the default
7026 value, for branch instructions.
7027 * config/rs6000/darwin.md: Ditto.
7028 * config/rs6000/rs6000.md: Ditto.
7030 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
7032 * config/rs6000/rs6000.md (length): Always define as const_int 4.
7033 (unnamed conditional branch define_insn): Set length to 4 or 8
7034 depending on offset.
7035 (<bd>_<mode>): Similar, for alternative 0.
7036 (<bd>tf_<mode>): Ditto.
7038 2018-08-16 Tamar Christina <tamar.christina@arm.com>
7040 * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
7042 2018-08-16 Matthew Malcomson <matthew.malcomson@arm.com>
7044 * doc/rtl.texi: Replace old RTX class names with new names.
7047 2018-08-16 Vlad Lazar <vlad.lazar@arm.com>
7049 * expmed.h (canonicalize_comparison): New declaration.
7050 * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
7051 * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
7052 * optabs.c (prepare_cmp_insn): Likewise.
7053 * rtl.h (unsigned_condition_p): New function which checks if a
7054 comparison operator is unsigned.
7056 2018-08-16 Nathan Sidwell <nathan@acm.org>
7058 * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
7059 * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
7061 2018-08-16 Tamar Christina <tamar.christina@arm.com>
7064 * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
7065 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
7066 (mov<mov>): ..this and enable unconditionally.
7068 2018-08-16 Tamar Christina <tamar.christina@arm.com>
7070 * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
7072 2018-08-16 Sam Tebbs <sam.tebbs@arm.com>
7074 * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
7075 (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
7076 "Common" with "Target".
7078 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
7080 * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
7081 * doc/invoke.texi (mmitigate-rop): Remove.
7082 * config/i386/i386.c: Do not include "regrename.h".
7083 (ix86_rop_should_change_byte_p, reg_encoded_number)
7084 (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
7086 (ix86_reorg): Remove call to ix86_mitigate_rop.
7087 * config/i386/i386.md (attr "modrm_class"): Remove.
7088 (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
7089 (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
7090 (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
7092 2018-08-15 Will Schmidt <will_schmidt@vnet.ibm.com>
7094 * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
7095 allow folding of mergeh() and mergel() for the float and double types.
7096 (fold_mergehl_helper): Rework to handle building a permute tree
7099 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
7101 * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
7104 2018-08-15 David Malcolm <dmalcolm@redhat.com>
7106 * common.opt (fdiagnostics-show-labels): New option.
7107 * diagnostic-show-locus.c (class layout_range): Add field
7109 (class layout): Add field "m_show_labels_p".
7110 (layout_range::layout_range): Add param "label" and use it to
7112 (make_range): Pass in NULL for new "label" param of layout_range's
7114 (layout::layout): Initialize m_show_labels_p.
7115 (layout::maybe_add_location_range): Pass in loc_range->m_label
7116 when constructing layout_range instances.
7117 (struct line_label): New struct.
7118 (layout::print_any_labels): New member function.
7119 (layout::print_line): Call it if label-printing is enabled.
7120 (selftest::test_one_liner_labels): New test.
7121 (selftest::test_diagnostic_show_locus_one_liner): Call it.
7122 * diagnostic.c (diagnostic_initialize): Initialize
7123 context->show_labels_p.
7124 * diagnostic.h (struct diagnostic_context): Add field
7126 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
7127 -fno-diagnostics-show-labels.
7128 * dwarf2out.c (gen_producer_string): Add
7129 OPT_fdiagnostics_show_labels to the ignored options.
7130 * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
7132 (gcc_rich_location::maybe_add_expr): Likewise.
7133 * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
7134 label" param, defaulting to NULL.
7135 (gcc_rich_location::add_expr): Add "label" param.
7136 (gcc_rich_location::maybe_add_expr): Likewise.
7137 (class text_range_label): New class.
7138 (class range_label_for_type_mismatch): New class.
7139 * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
7140 of format_warning_va.
7141 (fmtwarn_n): Likewise for new params of format_warning_n_va.
7142 * lto-wrapper.c (merge_and_complain): Add
7143 OPT_fdiagnostics_show_labels to the "pick one setting" options.
7144 (append_compiler_options): Likewise to the dropped options.
7145 (append_diag_options): Likewise to the passed-on options.
7146 * opts.c (common_handle_option): Handle the new option.
7147 * selftest-diagnostic.c
7148 (test_diagnostic_context::test_diagnostic_context): Enable
7150 * substring-locations.c: Include "gcc-rich-location.h".
7151 (format_warning_n_va): Add "fmt_label" and "param_label" params
7152 and use them as appropriate.
7153 (format_warning_va): Add "fmt_label" and "param_label" params,
7154 passing them on to format_warning_n_va.
7155 (format_warning_at_substring): Likewise.
7156 (format_warning_at_substring_n): Likewise.
7157 * substring-locations.h (format_warning_va): Add "fmt_label" and
7158 "param_label" params.
7159 (format_warning_n_va): Likewise.
7160 (format_warning_at_substring): Likewise.
7161 (format_warning_at_substring_n): Likewise.
7162 * toplev.c (general_init): Initialize global_dc->show_labels_p.
7164 2018-08-15 Qing Zhao <qing.zhao@oracle.com>
7167 * builtins.c (expand_builtin_memcmp): Do not expand the call
7168 when overflow is detected.
7170 2018-08-15 Martin Sebor <msebor@redhat.com>
7172 PR tree-optimization/71625
7173 * config/aarch64/aarch64-builtins.c
7174 (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
7176 2018-08-15 Ilya Leoshkevich <iii@linux.ibm.com>
7178 * config/s390/s390.c (s390_reorg): Remove loop.
7180 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
7183 (darwin_function_switched_text_sections): Delete.
7184 * gcc/config/darwin.h
7185 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
7187 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
7190 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
7191 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
7192 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
7194 2018-08-15 Martin Liska <mliska@suse.cz>
7196 PR tree-optimization/86925
7197 * predict.c (expr_expected_value_1): When taking
7198 later predictor, assign also probability.
7199 Use fold_build2_initializer_loc in order to fold
7200 the expression in -frounding-math.
7202 2018-08-14 Allan Sandfeld Jensen <allan.jensen@qt.io>
7204 * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
7206 (expand_vec_perm_1): Try the new method.
7208 2018-08-14 Ilya Leoshkevich <iii@linux.ibm.com>
7211 * lra-lives.c (remove_some_program_points_and_update_live_ranges):
7212 Check whether lra_live_max_point is 0 before dividing.
7214 2018-08-14 Martin Sebor <msebor@redhat.com>
7216 PR tree-optimization/86650
7217 * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
7218 (vrp_prop::check_mem_ref): Same.
7220 2018-08-13 Liu Hao <lh_mouse@126.com>
7222 * pretty-print.c (eat_esc_sequence): Swap the foreground and
7223 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
7224 and clear it thereafter, as it only works for DBCS.
7226 2018-08-13 Liu Hao <lh_mouse@126.com>
7228 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
7229 handle returned by _get_osf_handle().
7231 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
7233 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
7234 for folding vec_perm.
7236 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
7238 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
7239 Add support for gimple-folding of vec_pack() and vec_unpack()
7242 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
7244 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
7245 vec_xst variants to the list.
7246 (rs6000_gimple_fold_builtin): Add support for folding unaligned
7247 vector loads and stores.
7249 2018-08-13 David Edelsohn <dje.gcc@gmail.com>
7251 * config.gcc (rs6000-ibm-aix4.x): Delete.
7252 (rs6000-ibm-aix5.1): Delete.
7253 (rs6000-ibm-aix5.2): Delete.
7254 (rs6000-ibm-aix5.3): Delete.
7255 * config/rs6000/aix43.h: Delete.
7256 * config/rs6000/aix51.h: Delete.
7257 * config/rs6000/aix52.h: Delete.
7258 * config/rs6000/t-aix43: Delete.
7260 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
7262 * config/s390/s390.c (s390_decompose_constant_pool_ref):
7264 (s390_decompose_address): Factor out constant pool ref
7267 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
7269 * config/nds32/nds32-predicates.c
7270 (nds32_can_use_bclr_p): Change return type as bool.
7271 (nds32_can_use_bset_p): Ditto.
7272 (nds32_can_use_btgl_p): Ditto.
7273 (nds32_can_use_bitci_p): Ditto.
7274 * config/nds32/nds32-protos.h
7275 (nds32_can_use_bclr_p): Change declaration.
7276 (nds32_can_use_bset_p): Ditto.
7277 (nds32_can_use_btgl_p): Ditto.
7278 (nds32_can_use_bitci_p): Ditto.
7280 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
7282 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
7283 Support -msched-prolog-epilog option.
7284 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
7286 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
7288 * common/config/nds32/nds32-common.c
7289 (nds32_option_optimization_table): Enalbe -malways-align.
7291 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
7293 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
7295 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
7296 OPT_misr_secure_ case.
7297 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
7298 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
7299 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
7301 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
7302 (nds32_isr_info): New field security_level.
7303 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
7304 * config/nds32/nds32.md (return_internal): Consider critical attribute.
7305 * config/nds32/nds32.opt (misr-secure): New option.
7306 * config/nds32/nds32_init.inc: New file.
7307 * config/nds32/nds32_isr.h: New file.
7309 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
7311 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
7312 Update comment for atomic instructions.
7313 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
7314 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
7316 (atomic_loaddi): Revise fence expansion to only emit fence prior to
7317 load for __ATOMIC_SEQ_CST model.
7318 (atomic_loaddi_1): Remove float register target.
7319 (atomic_storedi): Handle CONST_INT values.
7320 (atomic_storedi_1): Remove float register source. Add special case
7322 (memory_barrier): New expander and insn.
7324 2018-08-11 Jakub Jelinek <jakub@redhat.com>
7326 PR tree-optimization/86835
7327 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
7328 new_stmt after def_gsi, make sure to insert new_square_stmt after
7329 that stmt, not 2 stmts before it.
7331 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
7334 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
7337 2018-08-10 Martin Liska <mliska@suse.cz>
7340 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
7342 * builtins.c (expand_builtin_expect_with_probability):
7344 (expand_builtin_expect_with_probability): New function.
7345 (build_builtin_expect_predicate): Add new argumnet probability
7346 for BUILT_IN_EXPECT_WITH_PROBABILITY.
7347 (fold_builtin_expect):
7350 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
7351 * builtins.h (fold_builtin_expect): Set new argument.
7352 * doc/extend.texi: Document __builtin_expect_with_probability.
7353 * doc/invoke.texi: Likewise.
7354 * gimple-fold.c (gimple_fold_call): Pass new argument.
7355 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
7356 also BUILT_IN_EXPECT_WITH_PROBABILITY.
7357 * predict.c (get_predictor_value): New function.
7358 (expr_expected_value): Add new argument probability. Assume
7359 that predictor and probability are always non-null.
7360 (expr_expected_value_1): Likewise. For __builtin_expect and
7361 __builtin_expect_with_probability set probability. Handle
7362 combination in binary expressions.
7363 (tree_predict_by_opcode): Simplify code by simply calling
7364 get_predictor_value.
7365 (pass_strip_predict_hints::execute): Add handling of
7366 BUILT_IN_EXPECT_WITH_PROBABILITY.
7367 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
7369 * tree.h (DECL_BUILT_IN_P): New function.
7371 2018-08-10 Martin Liska <mliska@suse.cz>
7373 PR tree-optimization/85799
7374 * passes.def: Add argument for pass_strip_predict_hints.
7375 * predict.c (class pass_strip_predict_hints): Add new argument
7377 (strip_predictor_early): New function.
7378 (pass_strip_predict_hints::execute): Call the function to
7380 (strip_predict_hints): New function.
7381 * predict.def: Fix comment.
7383 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
7385 * Makefile.in: Clarify which tm.texi to copy over to assert the
7386 right to grant a GFDL license for all.
7388 2018-08-09 Jeff Law <law@redhat.com>
7390 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
7393 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
7395 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
7398 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
7400 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
7401 reductions for variable-length vectors.
7403 2018-08-09 David Malcolm <dmalcolm@redhat.com>
7406 * common.opt (fdiagnostics-show-line-numbers): New option.
7407 * diagnostic-show-locus.c (class layout): Add fields
7408 "m_show_line_numbers_p" and "m_linenum_width";
7409 (num_digits): New function.
7410 (test_num_digits): New function.
7411 (layout::layout): Initialize new fields. Update m_x_offset
7412 logic to handle any left margin.
7413 (layout::print_source_line): Print line number when requested.
7414 (layout::start_annotation_line): New member function.
7415 (layout::print_annotation_line): Call it.
7416 (layout::print_leading_fixits): Likewise.
7417 (layout::print_trailing_fixits): Likewise. Update calls to
7418 move_to_column for new parameter.
7419 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
7420 it to potentially call start_annotation_line.
7421 (layout::show_ruler): Call start_annotation_line.
7422 (selftest::test_line_numbers_multiline_range): New selftest.
7423 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
7424 and selftest::test_line_numbers_multiline_range.
7425 * diagnostic.c (diagnostic_initialize): Initialize
7426 show_line_numbers_p.
7427 * diagnostic.h (struct diagnostic_context): Add field
7428 "show_line_numbers_p".
7429 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
7430 -fno-diagnostics-show-line-numbers.
7431 * dwarf2out.c (gen_producer_string): Add
7432 OPT_fdiagnostics_show_line_numbers to the ignored options.
7433 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
7434 one setting" options.
7435 (append_compiler_options): Likewise to the dropped options.
7436 (append_diag_options): Likewise to the passed-on options.
7437 * opts.c (common_handle_option): Handle the new option.
7438 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
7440 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
7442 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
7443 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
7444 third argument of type "const signed char" to descriptions of
7445 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
7446 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
7447 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
7448 __builtin_bcdsub_ov functions.
7450 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
7452 PR tree-optimization/86858
7453 * tree-vect-loop.c (vect_is_simple_reduction): Restore
7454 flow_bb_inside_loop_p calls.
7456 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
7458 PR tree-optimization/86871
7459 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
7460 instead of gimple_assign_lhs.
7462 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
7465 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
7466 register constraint to operand 0.
7467 (add<mode>3_carryinC): Likewise.
7468 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
7470 2018-08-09 Martin Liska <mliska@suse.cz>
7473 * common.opt: Remove extra line.
7475 2018-08-09 Martin Liska <mliska@suse.cz>
7477 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
7478 at the end of a line, make first letter capital and end up
7479 a sentence with a dot.
7480 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
7481 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
7482 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
7483 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
7484 (PARAM_MAX_ISL_OPERATIONS): Likewise.
7485 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
7486 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
7487 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
7488 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
7489 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
7490 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
7491 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
7492 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
7493 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
7494 (PARAM_TREE_REASSOC_WIDTH): Likewise.
7495 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
7496 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
7497 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
7499 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
7502 * config/s390/s390.c (s390_option_override_internal): Reduce the
7503 stack-clash-protection-probe-interval param if it would be too big
7506 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
7509 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
7510 * config/m68k/m68k.c (handle_move_double): Don't call
7511 m68k_final_prescan_insn.
7512 (m68k_adjust_decorated_operand): Renamed from
7513 m68k_final_prescan_insn, remove first and third operand and
7515 (print_operand): Call it.
7516 (print_operand_address): Call it.
7518 2018-08-08 Nathan Sidwell <nathan@acm.org>
7520 * diagnostic.c (diagnostic_report_current_module): Use
7521 linemap_included_from & linemap_included_from_linemap.
7523 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
7525 * config/aarch64/aarch64-cores.def: Add phecda core.
7526 * config/aarch64/aarch64-tune.md: Regenerate.
7527 * doc/invoke.texi: Add phecda core.
7529 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
7532 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
7534 * config/s390/s390.md ("movti"): Add more alternatives for
7535 constant to GPR copies.
7537 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
7539 * config/s390/s390.c: Fix whitespace damage throughout the file.
7540 * config/s390/s390.h: Likewise.
7541 * config/s390/tpf.h: Likewise.
7543 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
7545 * config/s390/s390.c (s390_loadrelative_operand_p):
7546 Remove TARGET_CPU_ZARCH usages.
7547 (s390_rtx_costs): Likewise.
7548 (s390_legitimate_constant_p): Likewise.
7549 (s390_cannot_force_const_mem): Likewise.
7550 (legitimate_reload_constant_p): Likewise.
7551 (s390_preferred_reload_class): Likewise.
7552 (legitimize_pic_address): Likewise.
7553 (legitimize_tls_address): Likewise.
7554 (s390_split_branches): Removed.
7555 (s390_add_execute): Removed.
7556 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
7557 (s390_mainpool_start): Likewise.
7558 (s390_mainpool_finish): Likewise.
7559 (s390_mainpool_cancel): Removed.
7560 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
7561 (s390_chunkify_cancel): Likewise.
7562 (s390_return_addr_rtx): Likewise.
7563 (s390_register_info): Remove split_branches_pending_p uages.
7564 (s390_optimize_register_info): Likewise.
7565 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
7566 split_branches_pending_p usages.
7567 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
7568 (s390_load_got): Likewise.
7569 (s390_expand_split_stack_prologue): Likewise.
7570 (output_asm_nops): Likewise.
7571 (s390_function_profiler): Likewise.
7572 (s390_emit_call): Likewise.
7573 (s390_conditional_register_usage): Likewise.
7574 (s390_optimize_prologue): Likewise.
7575 (s390_reorg): Remove TARGET_CPU_ZARCH and
7576 split_branches_pending_p usages.
7577 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
7579 (s390_output_indirect_thunk_function): Likewise.
7580 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
7581 (TARGET_CPU_ZARCH_P): Removed.
7582 (struct machine_function): Remove split_branches_pending_p.
7583 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
7585 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
7587 * common/config/s390/s390-common.c (processor_flags_table):
7589 * config.gcc: Remove with_arch/with_tune support.
7590 * config/s390/2064.md: Remove cpu attribute comparisons.
7591 * config/s390/driver-native.c (s390_host_detect_local_cpu):
7593 * config/s390/linux.h (ASM_SPEC):
7594 Remove -march support.
7595 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
7596 Use a table to get an arch level.
7597 * config/s390/s390-opts.h (enum processor_type):
7599 * config/s390/s390.c
7600 (processor_table): Remove entries, add arch_level values.
7601 (s390_issue_rate): Remove cases.
7602 (s390_option_override): Adjust
7603 s390_option_override_internal() call.
7604 (s390_option_override_internal): Remove deprecation warning.
7605 (s390_valid_target_attribute_tree): Adjust
7606 s390_option_override_internal() call.
7607 * config/s390/s390.h (struct s390_processor):
7608 Share with s390-c.c, add arch_level field.
7609 * config/s390/s390.md:
7610 Remove occurrences in cpu attribute.
7611 * config/s390/s390.opt: Remove -march/-mtune support.
7612 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
7613 * doc/invoke.texi: Remove deprecation warning.
7615 2018-08-08 Luis Machado <luis.machado@linaro.org>
7617 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
7619 (qdf24xx_tunings): Set vector cost structure to
7620 qdf24xx_vector_cost.
7622 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
7623 <register_sextend>: Set to 3.
7625 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
7628 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
7629 * config/aarch64/aarch64-simd.md
7630 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
7631 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
7632 (aarch64_frecpx<mode>): ...this new pattern.
7633 * config/aarch64/aarch64-simd-builtins.def: Remove comment
7634 about aarch64_frecp<FRECP:frecp_suffix><mode>.
7636 2018-08-07 Martin Liska <mliska@suse.cz>
7639 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
7640 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
7641 * predict.def (PRED_MALLOC_NONNULL): New predictor.
7642 * doc/extend.texi: Document that malloc attribute adds
7645 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
7648 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7649 Define to speculation_safe_value_not_needed.
7651 2018-08-06 Jeff Law <law@redhat.com>
7653 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
7654 the vr_values instance to cprop_into_stmt.
7655 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
7656 (cprop_operand): Also query EVRP to determine if OP is a constant.
7658 2018-08-06 Nathan Sidwell <nathan@acm.org>
7660 * diagnostic.c (diagnostic_report_current_module): Reroll
7661 included-at loop. Translate text.
7663 2018-08-06 David Malcolm <dmalcolm@redhat.com>
7665 * function-tests.c (selftest::test_expansion_to_rtl): Call
7666 free_after_compilation.
7668 2018-08-06 Alan Hayward <alan.hayward@arm.com>
7670 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
7672 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
7674 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
7675 loops with memory block operations from getting unrolled.
7677 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
7680 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7681 Define to speculation_safe_value_not_needed.
7683 2018-08-06 Jeff Law <law@redhat.com>
7685 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
7688 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
7691 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
7692 with all enabled __intN types.
7694 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
7696 2018-08-06 Alan Hayward <alan.hayward@arm.com>
7698 * alias.c (record_set): Check for clobber high.
7699 * cfgexpand.c (expand_gimple_stmt): Likewise.
7700 * combine-stack-adj.c (single_set_for_csa): Likewise.
7701 * combine.c (find_single_use_1): Likewise.
7702 (set_nonzero_bits_and_sign_copies): Likewise.
7703 (get_combine_src_dest): Likewise.
7704 (is_parallel_of_n_reg_sets): Likewise.
7705 (try_combine): Likewise.
7706 (record_dead_and_set_regs_1): Likewise.
7707 (reg_dead_at_p_1): Likewise.
7708 (reg_dead_at_p): Likewise.
7709 * dce.c (deletable_insn_p): Likewise.
7710 (mark_nonreg_stores_1): Likewise.
7711 (mark_nonreg_stores_2): Likewise.
7712 * df-scan.c (df_find_hard_reg_defs): Likewise.
7713 (df_uses_record): Likewise.
7714 (df_get_call_refs): Likewise.
7715 * dwarf2out.c (mem_loc_descriptor): Likewise.
7716 * haifa-sched.c (haifa_classify_rtx): Likewise.
7717 * ira-build.c (create_insn_allocnos): Likewise.
7718 * ira-costs.c (scan_one_insn): Likewise.
7719 * ira.c (equiv_init_movable_p): Likewise.
7720 (rtx_moveable_p): Likewise.
7721 (interesting_dest_for_shprep): Likewise.
7722 * jump.c (mark_jump_label_1): Likewise.
7723 * postreload-gcse.c (record_opr_changes): Likewise.
7724 * postreload.c (reload_cse_simplify): Likewise.
7725 (struct reg_use): Add source expr.
7726 (reload_combine): Check for clobber high.
7727 (reload_combine_note_use): Likewise.
7728 (reload_cse_move2add): Likewise.
7729 (move2add_note_store): Likewise.
7730 * print-rtl.c (print_pattern): Likewise.
7731 * recog.c (decode_asm_operands): Likewise.
7732 (store_data_bypass_p): Likewise.
7733 (if_test_bypass_p): Likewise.
7734 * regcprop.c (kill_clobbered_value): Likewise.
7735 (kill_set_value): Likewise.
7736 * reginfo.c (reg_scan_mark_refs): Likewise.
7737 * reload1.c (maybe_fix_stack_asms): Likewise.
7738 (eliminate_regs_1): Likewise.
7739 (elimination_effects): Likewise.
7740 (mark_not_eliminable): Likewise.
7741 (scan_paradoxical_subregs): Likewise.
7742 (forget_old_reloads_1): Likewise.
7743 * reorg.c (find_end_label): Likewise.
7744 (try_merge_delay_insns): Likewise.
7745 (redundant_insn): Likewise.
7746 (own_thread_p): Likewise.
7747 (fill_simple_delay_slots): Likewise.
7748 (fill_slots_from_thread): Likewise.
7749 (dbr_schedule): Likewise.
7750 * resource.c (update_live_status): Likewise.
7751 (mark_referenced_resources): Likewise.
7752 (mark_set_resources): Likewise.
7753 * rtl.c (copy_rtx): Likewise.
7754 * rtlanal.c (reg_referenced_p): Likewise.
7755 (single_set_2): Likewise.
7756 (noop_move_p): Likewise.
7757 (note_stores): Likewise.
7758 * sched-deps.c (sched_analyze_reg): Likewise.
7759 (sched_analyze_insn): Likewise.
7761 2018-08-06 Alan Hayward <alan.hayward@arm.com>
7763 * cse.c (invalidate_reg): New function extracted from...
7764 (invalidate): ...here.
7765 (canonicalize_insn): Check for clobber high.
7766 (invalidate_from_clobbers): invalidate clobber highs.
7767 (invalidate_from_sets_and_clobbers): Likewise.
7768 (count_reg_usage): Check for clobber high.
7769 (insn_live_p): Likewise.
7770 * cselib.c (cselib_expand_value_rtx_1):Likewise.
7771 (cselib_invalidate_regno): Check for clobber in setter.
7772 (cselib_invalidate_rtx): Pass through setter.
7773 (cselib_invalidate_rtx_note_stores):
7774 (cselib_process_insn): Check for clobber high.
7775 * cselib.h (cselib_invalidate_rtx): Add operand.
7777 2018-08-06 Alan Hayward <alan.hayward@arm.com>
7779 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
7780 (mark_not_eliminable): Likewise.
7781 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
7782 * lra-lives.c (process_bb_lives): Check for clobber high.
7783 * lra.c (new_insn_reg): Remember clobber highs.
7784 (collect_non_operand_hard_regs): Check for clobber high.
7785 (lra_set_insn_recog_data): Likewise.
7786 (add_regs_to_insn_regno_info): Likewise.
7787 (lra_update_insn_regno_info): Likewise.
7789 2018-08-06 Alan Hayward <alan.hayward@arm.com>
7791 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
7792 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
7794 2018-08-06 Alan Hayward <alan.hayward@arm.com>
7796 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
7797 (copy_insn_1): Likewise.
7798 (gen_hard_reg_clobber_high): New gen function.
7799 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
7800 * genemit.c (gen_exp): Likewise.
7801 (gen_emit_seq): Pass through info.
7802 (gen_insn): Check for CLOBBER_HIGH.
7803 (gen_expand): Pass through info.
7804 (gen_split): Likewise.
7805 (output_add_clobbers): Likewise.
7806 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
7807 (remove_clobbers): Likewise.
7808 * rtl.h (gen_hard_reg_clobber_high): New declaration.
7810 2018-08-06 Alan Hayward <alan.hayward@arm.com>
7812 * doc/rtl.texi (clobber_high): Add.
7813 (parallel): Add in clobber high
7814 * rtl.c (rtl_check_failed_code3): Add function.
7815 * rtl.def (CLOBBER_HIGH): Add expression.
7816 * rtl.h (RTL_CHECKC3): Add macro.
7817 (rtl_check_failed_code3): Add declaration.
7818 (XC3EXP): Add macro.
7820 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
7823 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
7824 cfun->machine->max_used_stack_alignment if needed.
7826 2018-08-04 Martin Sebor <msebor@redhat.com>
7828 PR tree-optimization/86571
7829 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
7832 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
7835 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7838 2018-08-03 Jeff Law <law@redhat.com>
7841 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7842 Define to speculation_safe_value_not_needed.
7844 2018-08-03 David Malcolm <dmalcolm@redhat.com>
7846 * doc/gcov.texi (-x): Remove duplicate "to".
7847 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
7848 (-Wif-not-aligned): Remove duplicate "is".
7849 (-flto): Remove duplicate "the".
7850 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
7851 duplicate "v5.00.b".
7852 (MSP430 Options): Remove duplicate "and" from the description
7853 of "-mgprel-sec=regexp".
7854 (x86 Options): Remove duplicate copies of "vmldLog102" and
7855 vmlsLog104 from description of "-mveclibabi=type".
7857 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
7859 * internal-fn.h (first_commutative_argument): Declare.
7860 * internal-fn.c (first_commutative_argument): New function.
7861 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
7862 restrictions for pattern statements. Use first_commutative_argument
7863 to look for commutative operands in calls to internal functions.
7865 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
7867 * Makefile.in (wide-int-range.o): New.
7868 * tree-vrp.c: Move all the wide_int_* functions to...
7869 * wide-int-range.cc: ...here.
7870 * tree-vrp.h: Move all the wide_int_* prototypes to...
7871 * wide-int-range.h: ...here.
7873 2018-08-03 Tom de Vries <tdevries@suse.de>
7875 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
7877 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
7878 * except.c (output_function_exception_table): Do early exit if
7879 targetm_common.except_unwind_info (&global_options) == UI_NONE.
7881 2018-08-03 Martin Liska <mliska@suse.cz>
7883 * predict.c (dump_prediction): Change to 2 digits
7886 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
7888 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
7890 (falkor_am_1_gtov_gtov): ... a new insn reservation.
7892 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
7894 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
7895 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
7896 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
7897 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
7898 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
7900 2018-08-02 David Malcolm <dmalcolm@redhat.com>
7902 * diagnostic-show-locus.c (diagnostic_show_locus): Use
7903 pp_take_prefix when saving the existing prefix.
7904 * diagnostic.c (diagnostic_append_note): Likewise.
7905 * langhooks.c (lhd_print_error_function): Likewise.
7906 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
7907 param's type. Free the existing prefix.
7908 (pp_take_prefix): New function.
7909 (pretty_printer::pretty_printer): Drop the prefix parameter.
7910 Rename the length parameter to match the comment.
7911 (pretty_printer::~pretty_printer): Free the prefix.
7912 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
7914 (struct pretty_printer): Drop the "const" from "prefix" field's
7915 type and clarify memory management.
7916 (pp_set_prefix): Drop the "const" from the 2nd param.
7917 (pp_take_prefix): New decl.
7919 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
7921 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
7922 for word_mode_ok here instead of passing as argument.
7923 (expand_block_compare): Change select_block_compare_mode() call.
7924 (expand_strncmp_gpr_sequence): New function.
7925 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
7927 2018-08-02 Jeff Law <law@redhat.com>
7930 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7931 Define to speculation_safe_value_not_needed.
7934 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7935 Define to speculation_safe_value_not_needed.
7937 2018-08-02 Tom de Vries <tdevries@suse.de>
7940 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
7941 function. Return UI_TARGET unconditionally.
7942 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
7943 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
7945 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
7947 * genemit.c (print_overload_test): Fix typo.
7949 2018-08-02 Richard Biener <rguenther@suse.de>
7951 PR tree-optimization/86816
7952 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
7953 which checks for value availability before querying it.
7954 (gvn_uses_equal): Use it.
7955 (same_succ_hash): Likewise.
7956 (gimple_equal_p): Likewise.
7958 2018-08-02 Nick Clifton <nickc@redhat.com>
7961 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7962 Define to speculation_safe_value_not_needed.
7965 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7966 Define to speculation_safe_value_not_needed.
7969 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7970 Define to speculation_safe_value_not_needed.
7973 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7974 Define to speculation_safe_value_not_needed.
7977 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7978 Define to speculation_safe_value_not_needed.
7981 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7982 Define to speculation_safe_value_not_needed.
7985 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7986 Define to speculation_safe_value_not_needed.
7989 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7990 Define to speculation_safe_value_not_needed.
7993 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
7994 speculation_safe_value_not_needed.
7997 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
7998 to speculation_safe_value_not_needed.
8000 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
8002 * doc/md.texi: Expand the documentation of instruction names
8003 to mention port-local uses. Document '@' in pattern names.
8004 * read-md.h (overloaded_instance, overloaded_name): New structs.
8006 (md_reader::handle_overloaded_name): New member function.
8007 (md_reader::get_overloads): Likewise.
8008 (md_reader::m_first_overload): New member variable.
8009 (md_reader::m_next_overload_ptr): Likewise.
8010 (md_reader::m_overloads_htab): Likewise.
8011 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
8012 m_next_overload_ptr and m_overloads_htab.
8013 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
8014 (get_mode_token, get_code_token, get_int_token): New functions.
8015 (map_attr_string): Add an optional argument that passes back
8016 the associated iterator.
8017 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
8018 (md_reader::handle_overloaded_name, add_overload_instance): New
8020 (apply_iterators): Handle '@' names. Report an error if '@'
8021 is used without iterators.
8022 (initialize_iterators): Initialize the new iterator_group fields.
8023 * genopinit.c (handle_overloaded_code_for)
8024 (handle_overloaded_gen): New functions.
8025 (main): Use them to print declarations of maybe_code_for_* and
8026 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
8027 * genemit.c (print_overload_arguments, print_overload_test)
8028 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
8029 (main): Use it to print definitions of maybe_code_for_* and
8030 maybe_gen_* functions.
8031 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
8032 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
8033 instead of explicit mode checks.
8034 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
8035 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
8036 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
8037 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
8038 (aarch64_expand_compare_and_swap): Likewise
8039 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
8040 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
8041 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
8042 (aarch64_constant_pool_reload_icode): Delete.
8043 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
8044 instead of aarch64_constant_pool_reload_icode. Use
8045 code_for_aarch64_reload_mov instead of explicit mode checks.
8046 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
8047 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
8048 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
8049 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
8050 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
8051 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
8052 (aarch64_atomic_load_op_code): Delete.
8053 (aarch64_emit_atomic_load_op): Likewise.
8054 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
8055 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
8056 instead of aarch64_emit_atomic_load_op.
8057 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
8058 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
8059 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
8060 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
8061 character before the pattern name.
8062 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
8063 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
8064 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
8065 (aarch64_frecps<mode>): Likewise.
8066 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
8067 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
8068 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
8069 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
8070 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
8072 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
8074 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
8075 Allow HFmode constants if TARGET_FP_F16INST.
8077 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
8080 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
8081 No longer check last store for clobber of address register.
8083 2018-08-02 Martin Liska <mliska@suse.cz>
8085 PR gcov-profile/86817
8086 * gcov.c (process_all_functions): New function.
8088 (process_file): Move functions processing to
8089 process_all_functions.
8091 2018-08-02 David Malcolm <dmalcolm@redhat.com>
8093 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
8094 "const" to the "gimple *" and "rtx_insn *" parameters.
8095 * dumpfile.h (dump_user_location_t::dump_user_location_t):
8097 (dump_location_t::dump_location_t): Likewise.
8099 2018-08-01 Martin Sebor <msebor@redhat.com>
8101 PR tree-optimization/86650
8102 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
8103 rather than a "gcall *". Directly pass the data of interest
8104 to percent_K_format, rather than building a temporary CALL_EXPR
8106 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
8107 (gimple_fold_builtin_strncat): Adjust.
8108 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
8109 gcall* argument with gimple*.
8110 * gimple-ssa-warn-restrict.c (check_call): Same.
8111 (wrestrict_dom_walker::before_dom_children): Same.
8112 (builtin_access::builtin_access): Same.
8113 (check_bounds_or_overlap): Same
8114 (maybe_diag_overlap): Same.
8115 (maybe_diag_offset_bounds): Same.
8116 * tree-diagnostic.c (default_tree_printer): Move usage of
8117 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
8119 * tree-pretty-print.c (percent_K_format): Add argument.
8120 * tree-pretty-print.h: Add argument.
8121 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
8122 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
8123 (maybe_diag_stxncpy_trunc): Same.
8124 (handle_builtin_stxncpy): Same.
8125 (handle_builtin_strcat): Same.
8127 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
8129 * match.pd: Optimise pointer range checks.
8131 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
8133 PR tree-optimization/86758
8134 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
8135 to remove pattern statements.
8137 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
8139 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
8140 result of dfs_enumerate_from when constructing stmt_vec_infos,
8141 instead of additionally calling get_loop_body.
8143 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
8145 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
8147 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
8148 When creating an iv, assert that the step is not known to be zero.
8149 (vect_setup_realignment): Update call accordingly.
8150 * tree-vect-stmts.c (vectorizable_store): Likewise.
8151 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
8153 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
8155 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
8156 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
8157 (vectorizable_reduction): Likewise.
8158 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8159 (vect_detect_hybrid_slp_stmts): Likewise.
8160 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
8162 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
8164 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
8165 (wide_int_set_zero_nonzero_bits): ...this.
8166 (zero_nonzero_bits_from_vr): Rename to...
8167 (vrp_set_zero_nonzero_bits): ...this.
8168 (extract_range_from_multiplicative_op_1): Abstract wide int
8170 (wide_int_range_multiplicative_op): ...here.
8171 (extract_range_from_binary_expr_1): Extract wide int binary
8172 operations into their own functions.
8173 (wide_int_range_lshift): New.
8174 (wide_int_range_can_optimize_bit_op): New.
8175 (wide_int_range_shift_undefined_p): New.
8176 (wide_int_range_bit_xor): New.
8177 (wide_int_range_bit_ior): New.
8178 (wide_int_range_bit_and): New.
8179 (wide_int_range_trunc_mod): New.
8180 (extract_range_into_wide_ints): New.
8181 (vrp_shift_undefined_p): New.
8182 (extract_range_from_multiplicative_op): New.
8183 (vrp_can_optimize_bit_op): New.
8184 * tree-vrp.h (value_range::dump): New.
8185 (wide_int_range_multiplicative_op): New.
8186 (wide_int_range_lshift):New.
8187 (wide_int_range_shift_undefined_p): New.
8188 (wide_int_range_bit_xor): New.
8189 (wide_int_range_bit_ior): New.
8190 (wide_int_range_bit_and): New.
8191 (wide_int_range_trunc_mod): New.
8192 (zero_nonzero_bits_from_bounds): Rename to...
8193 (wide_int_set_zero_nonzero_bits): ...this.
8194 (zero_nonzero_bits_from_vr): Rename to...
8195 (vrp_set_zero_nonzero_bits): ...this.
8196 (range_easy_mask_min_max): Rename to...
8197 (wide_int_range_can_optimize_bit_op): this.
8198 * vr-values.c (simplify_bit_ops_using_ranges): Rename
8199 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
8201 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
8203 * tree-vectorizer.h (vect_orig_stmt): New function.
8204 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
8205 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
8206 (vect_create_epilog_for_reduction): Likewise.
8207 (vectorizable_live_operation): Likewise.
8208 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
8209 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
8210 * tree-vect-stmts.c (vectorizable_call): Likewise.
8211 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
8213 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
8215 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
8217 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
8218 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
8219 (vect_transform_loop): Likewise.
8220 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
8222 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
8224 * tree-vectorizer.h (vect_schedule_slp): Return void.
8225 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
8226 (vect_schedule_slp): Likewise.
8228 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
8230 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
8232 (vect_transform_loop): Update calls accordingly. Schedule SLP
8233 instances before the main loop, if any exist.
8235 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
8237 PR tree-optimization/86749
8238 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
8239 If the lhs is used in a COND_EXPR, check that it is being used
8240 as the "then" or "else" value.
8242 2018-08-01 Tom de Vries <tdevries@suse.de>
8245 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
8246 speculation_safe_value_not_needed.
8248 2018-08-01 Richard Biener <rguenther@suse.de>
8250 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
8253 2018-08-01 Martin Liska <mliska@suse.cz>
8255 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
8256 format how successful transformation is dumped.
8257 (gimple_mod_pow2_value_transform): Likewise.
8258 (gimple_mod_subtract_transform): Likewise.
8259 (gimple_stringops_transform): Likewise.
8261 2018-08-01 Martin Liska <mliska@suse.cz>
8264 * value-prof.c (interesting_stringop_to_profile_p):
8265 Simplify the code and add BUILT_IN_MEMMOVE.
8266 (gimple_stringops_transform): Likewise.
8268 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
8270 * config/aarch64/aarch64-simd.md
8271 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
8272 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
8273 use GPI iterator instead of SI mode.
8275 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
8277 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
8278 rs6000_speculation_barrier.
8279 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
8280 new barrier pattern name.
8282 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
8284 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
8285 (speculation_barrier): New insn.
8287 2018-08-01 Richard Biener <rguenther@suse.de>
8290 * graphite.h: Include isl/id.h and isl/space.h to allow build
8293 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
8296 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
8297 mode for COFF targets.
8298 * defaults.h (TARGET_COFF): Define.
8299 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
8300 TARGET_COFF): Define.
8301 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
8302 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
8303 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
8305 2018-07-31 Alexandre Oliva <oliva@adacore.com>
8307 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
8308 * gimple-streamer-out.c (output_bb): Save it.
8309 * lto-streamer-in.c (input_struct_function_base): Restore
8310 instance discriminator if available. Create map on demand.
8311 * lto-streamer-out.c (output_struct_function_base): Save it if
8313 * final.c (decl_to_instance_map): Document LTO strategy.
8315 2018-07-31 Alexandre Oliva <oliva@adacore.com>
8316 Olivier Hainque <hainque@adacore.com>
8318 * debug.h (decl_to_instance_map_t): New type.
8319 (decl_to_instance_map): Declare.
8320 (maybe_create_decl_to_instance_map): New inline function.
8321 * final.c (bb_discriminator, last_bb_discriminator): New statics,
8322 to track basic block discriminators.
8323 (final_start_function_1): Initialize them.
8324 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
8326 (decl_to_instance_map): New variable.
8327 (map_decl_to_instance, maybe_set_discriminator): New functions.
8328 (notice_source_line): Set discriminator.
8330 2018-07-31 Ian Lance Taylor <iant@golang.org>
8332 * targhooks.c (default_have_speculation_safe_value): Add
8335 2018-07-31 David Malcolm <dmalcolm@redhat.com>
8337 * dump-context.h: Include "pretty-print.h".
8338 (dump_context::refresh_dumps_are_enabled): New decl.
8339 (dump_context::emit_item): New decl.
8340 (class dump_context): Add fields "m_test_pp" and
8342 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
8343 (temp_dump_context::get_dumped_text): New decl.
8344 (class temp_dump_context): Add field "m_pp".
8345 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
8346 (dump_context::refresh_dumps_are_enabled): ...and add a test for
8348 (set_dump_file): Update for above change.
8349 (set_alt_dump_file): Likewise.
8350 (dump_loc): New overload, taking a pretty_printer *.
8351 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
8352 to any test pretty-printer.
8353 (make_item_for_dump_gimple_stmt): New function, adapted from
8354 optinfo::add_gimple_stmt.
8355 (dump_context::dump_gimple_stmt): Call it, and use the result,
8356 eliminating the direct usage of dump_file and alt_dump_file in
8357 favor of indirectly using them via emit_item.
8358 (make_item_for_dump_gimple_expr): New function, adapted from
8359 optinfo::add_gimple_expr.
8360 (dump_context::dump_gimple_expr): Call it, and use the result,
8361 eliminating the direct usage of dump_file and alt_dump_file in
8362 favor of indirectly using them via emit_item.
8363 (make_item_for_dump_generic_expr): New function, adapted from
8365 (dump_context::dump_generic_expr): Call it, and use the result,
8366 eliminating the direct usage of dump_file and alt_dump_file in
8367 favor of indirectly using them via emit_item.
8368 (make_item_for_dump_printf_va): New function, adapted from
8369 optinfo::add_printf_va.
8370 (make_item_for_dump_printf): New function.
8371 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
8372 and use the result, eliminating the direct usage of dump_file and
8373 alt_dump_file in favor of indirectly using them via emit_item.
8374 (make_item_for_dump_dec): New function.
8375 (dump_context::dump_dec): Call it, and use the result,
8376 eliminating the direct usage of dump_file and alt_dump_file in
8377 favor of indirectly using them via emit_item.
8378 (make_item_for_dump_symtab_node): New function, adapted from
8379 optinfo::add_symtab_node.
8380 (dump_context::dump_symtab_node): Call it, and use the result,
8381 eliminating the direct usage of dump_file and alt_dump_file in
8382 favor of indirectly using them via emit_item.
8383 (dump_context::begin_scope): Reimplement, avoiding direct usage
8384 of dump_file and alt_dump_file in favor of indirectly using them
8386 (dump_context::emit_item): New member function.
8387 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
8388 Set up test pretty-printer on the underlying context. Call
8389 refresh_dumps_are_enabled.
8390 (temp_dump_context::~temp_dump_context): Call
8391 refresh_dumps_are_enabled.
8392 (temp_dump_context::get_dumped_text): New member function.
8393 (selftest::verify_dumped_text): New function.
8394 (ASSERT_DUMPED_TEXT_EQ): New macro.
8395 (selftest::test_capture_of_dump_calls): Run all tests twice, with
8396 and then without optinfo enabled. Add uses of
8397 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
8398 * dumpfile.h: Update comment for the dump_* API.
8399 * optinfo-emit-json.cc
8400 (selftest::test_building_json_from_dump_calls): Update for new
8401 param for temp_dump_context ctor.
8402 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
8403 and "m_owned" field.
8404 (optinfo_item::~optinfo_item): Likewise.
8405 (optinfo::add_item): New member function.
8406 (optinfo::emit): Update comment.
8407 (optinfo::add_string): Delete.
8408 (optinfo::add_printf): Delete.
8409 (optinfo::add_printf_va): Delete.
8410 (optinfo::add_gimple_stmt): Delete.
8411 (optinfo::add_gimple_expr): Delete.
8412 (optinfo::add_tree): Delete.
8413 (optinfo::add_symtab_node): Delete.
8414 (optinfo::add_dec): Delete.
8415 * optinfo.h (class dump_context): New forward decl.
8416 (optinfo::add_item): New decl.
8417 (optinfo::add_string): Delete.
8418 (optinfo::add_printf): Delete.
8419 (optinfo::add_printf_va): Delete.
8420 (optinfo::add_gimple_stmt): Delete.
8421 (optinfo::add_gimple_expr): Delete.
8422 (optinfo::add_tree): Delete.
8423 (optinfo::add_symtab_node): Delete.
8424 (optinfo::add_dec): Delete.
8425 (optinfo::add_poly_int): Delete.
8426 (optinfo_item::optinfo_item): Remove "owned" param.
8427 (class optinfo_item): Remove field "m_owned".
8429 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
8432 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
8433 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
8434 requested variable alignment.
8435 (expand_one_ssa_partition): Likewise.
8436 (expand_one_var): Likewise.
8438 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
8440 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
8441 to speculation_safe_value_not_needed.
8443 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
8445 * targhooks.h (speculation_safe_value_not_needed): New prototype.
8446 * targhooks.c (speculation_safe_value_not_needed): New function.
8447 * target.def (have_speculation_safe_value): Update documentation.
8448 * doc/tm.texi: Regenerated.
8450 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
8452 * config/aarch64/iterators.md (ALLI_TI): New iterator.
8453 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
8455 (despeculate_copy<ALLI:mode>_insn): New insn.
8456 (despeculate_copyti_insn): New insn.
8457 (despeculate_simple<ALLI:mode>): New insn
8458 (despeculate_simpleti): New insn.
8459 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
8461 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
8462 aarch64_speculation_safe_value.
8463 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
8465 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
8467 * config/aarch64/aarch64-speculation.cc: New file.
8468 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
8469 before pass_reorder_blocks.
8470 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
8472 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
8473 X14 and X15 when tracking speculation.
8474 * config/aarch64/aarch64.md (register name constants): Add
8475 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
8476 (unspec): Add UNSPEC_SPECULATION_TRACKER.
8477 (speculation_barrier): New insn attribute.
8478 (cmp<mode>): Allow SP in comparisons.
8479 (speculation_tracker): New insn.
8480 (speculation_barrier): Add speculation_barrier attribute.
8481 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
8482 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
8483 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
8485 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
8487 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
8488 aarch64_track_speculation is true.
8489 (tb<optab><mode>1): Likewise.
8490 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
8491 generate CB[N]Z when tracking speculation.
8492 (aarch64_split_compare_and_swap): Likewise.
8493 (aarch64_split_atomic_op): Likewise.
8495 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
8497 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
8499 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
8501 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
8502 (speculation_barrier): New insn.
8504 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
8506 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
8507 * config/arm/arm.md (speculation_barrier): New expand.
8508 (speculation_barrier_insn): New pattern.
8510 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
8512 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
8513 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
8514 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
8515 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
8517 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
8518 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
8519 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
8520 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
8521 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
8522 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
8523 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
8524 * builtins.c (expand_speculation_safe_value): New function.
8525 (expand_builtin): Call it.
8526 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
8527 * doc/extend.texi: Document __builtin_speculation_safe_value.
8528 * doc/md.texi: Document "speculation_barrier" pattern.
8529 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
8530 TARGET_HAVE_SPECULATION_SAFE_VALUE.
8531 * doc/tm.texi: Regenerated.
8532 * target.def (have_speculation_safe_value, speculation_safe_value): New
8534 * targhooks.c (default_have_speculation_safe_value): New function.
8535 (default_speculation_safe_value): New function.
8536 * targhooks.h (default_have_speculation_safe_value): Add prototype.
8537 (default_speculation_safe_value): Add prototype.
8539 2018-07-31 David Malcolm <dmalcolm@redhat.com>
8541 * dump-context.h (dump_context::dump_loc): New decl.
8542 * dumpfile.c (dump_context::dump_loc): New member function.
8543 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
8544 and dump_gimple_stmt.
8545 (dump_context::dump_gimple_expr_loc): Likewise, using
8547 (dump_context::dump_generic_expr_loc): Likewise, using
8549 (dump_context::dump_printf_loc_va): Likewise, using
8551 (dump_context::begin_scope): Explicitly using the global function
8552 "dump_loc", rather than the member function.
8554 2018-07-31 Martin Sebor <msebor@redhat.com>
8556 PR tree-optimization/86741
8557 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
8559 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
8561 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
8562 depend on whether prefetch instructions will be emitted or not.
8563 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
8564 will be emitted or not.
8565 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
8566 (TARGET_SETMEM_PFD): New macros.
8568 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8570 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
8571 (NULL_STMT_VEC_INFO): Delete.
8572 (stmt_vec_info::operator*): Likewise.
8573 (stmt_vec_info::operator gimple *): Likewise.
8574 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
8575 of NULL_STMT_VEC_INFO.
8576 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
8577 (vect_reassociating_reduction_p): Likewise.
8578 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
8579 (vectorizable_store): Likewise.
8580 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
8581 (vec_info::free_stmt_vec_infos): Likewise.
8583 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8585 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
8586 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
8588 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8590 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
8591 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
8592 (vec_info::free_stmt_vec_info): New private member functions.
8593 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
8594 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
8595 * tree-parloops.c (gather_scalar_reductions): Remove calls to
8596 set_stmt_vec_info_vec and free_stmt_vec_infos.
8597 * tree-vect-loop.c (_loop_vec_info): Remove call to
8598 set_stmt_vec_info_vec.
8599 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
8600 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
8601 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
8602 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
8603 (vec_info::free_stmt_vec_info): ...these new functions. Remove
8604 assignments in {vec_info::,}new_stmt_vec_info that are redundant
8605 with the clearing in the xcalloc.
8606 (stmt_vec_info_vec): Delete.
8607 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
8608 (vectorize_loops): Likewise.
8609 (vec_info::~vec_info): Remove argument from call to
8610 free_stmt_vec_infos.
8611 (vec_info::add_stmt): Remove vinfo argument from call to
8614 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8616 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
8617 rather than a gimple stmt.
8618 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
8619 information for pattern statements when passed the original
8620 statement; instead wait to be passed the pattern statement itself.
8621 Don't call set_vinfo_for_stmt here.
8622 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
8623 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
8624 stmt_vec_infos here.
8625 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
8626 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
8627 stmt_vec_infos entry.
8629 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8631 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
8632 * tree-vectorizer.c (vec_info::replace_stmt): New function.
8633 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
8634 * tree-vect-stmts.c (vectorizable_call): Likewise.
8635 (vectorizable_simd_clone_call): Likewise.
8637 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8639 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
8640 * tree-vectorizer.c (vec_info::remove_stmt): New function.
8641 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
8642 * tree-vect-loop.c (vect_transform_loop): Likewise.
8643 * tree-vect-slp.c (vect_schedule_slp): Likewise.
8644 * tree-vect-stmts.c (vect_remove_stores): Likewise.
8646 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8648 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
8649 (vect_dr_stmt): Delete.
8650 * tree-vectorizer.c (vec_info::lookup_dr): New function.
8651 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
8653 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
8654 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
8655 (vect_verify_datarefs_alignment, vect_peeling_supportable)
8656 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
8657 (vect_analyze_data_refs): Likewise.
8658 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
8660 (vect_find_same_alignment_drs): Likewise.
8661 (vect_slp_analyze_node_dependences): Update calls accordingly.
8662 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
8663 instead of DR_VECT_AUX.
8664 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
8665 of a vector data references. Use vec_info::lookup_dr instead of
8667 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
8668 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
8669 instead of DR_VECT_AUX.
8671 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8673 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
8675 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
8677 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
8678 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
8679 (vect_gen_prolog_loop_niters): Likewise.
8681 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8683 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
8684 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
8685 (vect_known_alignment_in_bytes, vect_dr_behavior)
8686 (vect_get_scalar_dr_size): Take references as dr_vec_infos
8687 instead of data_references. Update calls to other routines for
8688 which the same change has been made.
8689 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
8690 dr_vec_infos instead of stmt_vec_infos.
8691 (vect_analyze_data_ref_dependence): Update call accordingly.
8692 (vect_slp_analyze_data_ref_dependence)
8693 (vect_record_base_alignments): Use DR_VECT_AUX.
8694 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
8695 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
8696 (vector_alignment_reachable_p, vect_get_data_access_cost)
8697 (vect_peeling_supportable, vect_analyze_group_access_1)
8698 (vect_analyze_group_access, vect_analyze_data_ref_access)
8699 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
8700 (vect_compile_time_alias, vect_small_gap_p)
8701 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
8702 (vect_supportable_dr_alignment): Take references as dr_vec_infos
8703 instead of data_references. Update calls to other routines for
8704 which the same change has been made.
8705 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
8706 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
8707 (vect_slp_analyze_and_verify_node_alignment)
8708 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
8709 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
8710 (vect_setup_realignment): Use dr_vec_infos. Update calls after
8712 (_vect_peel_info::dr): Replace with...
8713 (_vect_peel_info::dr_info): ...this new field.
8714 (vect_peeling_hash_get_most_frequent)
8715 (vect_peeling_hash_choose_best_peeling): Update accordingly.
8716 (vect_peeling_hash_get_lowest_cost):
8717 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
8718 routines for which the same change has been made.
8719 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
8721 * tree-vect-loop-manip.c (get_misalign_in_elems)
8722 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
8724 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
8725 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
8726 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
8727 (get_group_load_store_type, get_negative_load_store_type)
8728 (vect_get_data_ptr_increment, vectorizable_store)
8729 (vectorizable_load): Likewise.
8730 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
8731 Update calls to other routines for which the same change has been made.
8733 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8735 * tree-vectorizer.h (vec_info::move_dr): New member function.
8736 (dataref_aux): Rename to...
8737 (dr_vec_info): ...this and add "dr" and "stmt" fields.
8738 (_stmt_vec_info::dr_aux): Update accordingly.
8739 (_stmt_vec_info::data_ref_info): Delete.
8740 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
8741 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
8742 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
8743 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
8745 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
8746 (STMT_VINFO_DR_INFO): New macro.
8747 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
8748 (set_dr_misalignment): Update after rename of dataref_aux.
8749 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
8750 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
8751 initialization of STMT_VINFO_DATA_REF.
8752 * tree-vectorizer.c (vec_info::move_dr): New function.
8753 * tree-vect-patterns.c (vect_recog_bool_pattern)
8754 (vect_recog_mask_conversion_pattern)
8755 (vect_recog_gather_scatter_pattern): Use it.
8756 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
8757 the "dr" and "stmt" fields of dr_vec_info instead of
8758 STMT_VINFO_DATA_REF.
8760 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8762 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
8763 (is_pattern_stmt_p): Use it.
8764 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
8765 on pattern statements.
8767 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8769 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
8770 original stmt as a stmt_vec_info rather than a gimple stmt.
8771 (vect_pattern_recog_1): Take the statement directly as a
8772 stmt_vec_info, rather than via a gimple_stmt_iterator.
8773 Update call to vect_mark_pattern_stmts.
8774 (vect_pattern_recog): Update calls accordingly.
8776 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8778 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
8779 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
8780 a vect_def_type for the first argument.
8781 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
8782 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
8783 operand if it isn't defined by a vectorized statement.
8784 (vect_build_gather_load_calls): Remove the mask_dt argument and
8785 update calls to vect_get_vec_def_for_stmt_copy.
8786 (vectorizable_bswap): Likewise the dt argument.
8787 (vectorizable_call): Update calls to vectorizable_bswap and
8788 vect_get_vec_def_for_stmt_copy.
8789 (vectorizable_simd_clone_call, vectorizable_assignment)
8790 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
8791 (vectorizable_comparison): Update calls to
8792 vect_get_vec_def_for_stmt_copy.
8793 (vectorizable_store): Likewise. Remove now-unnecessary calls to
8795 (vect_get_loop_based_defs): Remove dt argument and update call
8796 to vect_get_vec_def_for_stmt_copy.
8797 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
8798 and vect_get_vec_def_for_stmt_copy.
8799 (vectorizable_load): Update calls to vect_build_gather_load_calls
8800 and vect_get_vec_def_for_stmt_copy.
8801 * tree-vect-loop.c (vect_create_epilog_for_reduction)
8802 (vectorizable_reduction, vectorizable_live_operation): Update calls
8803 to vect_get_vec_def_for_stmt_copy.
8805 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8807 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
8808 and gimple stmt arguments with a stmt_vec_info.
8809 (vect_record_base_alignments): Update calls accordingly.
8810 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
8811 and gimple stmt arguments with a stmt_vec_info.
8812 (vect_build_slp_tree_1): Remove vinfo argument and update call
8813 to vect_record_max_nunits.
8814 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
8815 and vect_record_max_nunits.
8817 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8819 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
8820 file and take a stmt_vec_info instead of a gimple stmt.
8821 (supportable_widening_operation, vect_finish_replace_stmt)
8822 (vect_finish_stmt_generation, vect_get_store_rhs)
8823 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
8824 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
8825 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
8826 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
8827 (vect_create_data_ref_ptr, bump_vector_ptr)
8828 (vect_permute_store_chain, vect_setup_realignment)
8829 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
8830 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
8831 (vectorizable_reduction, vectorizable_induction)
8832 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
8833 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
8834 than gimple stmts as arguments.
8835 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
8836 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
8837 (can_group_stmts_p, vect_check_gather_scatter)
8838 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
8839 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
8840 (vect_permute_load_chain, vect_shift_permute_load_chain)
8841 (vect_transform_grouped_load)
8842 (vect_record_grouped_load_vectors): Likewise.
8843 * tree-vect-loop.c (vect_fixup_reduc_chain)
8844 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
8845 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
8846 (vectorizable_reduction, vectorizable_induction)
8847 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
8848 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
8849 (vect_get_load_store_mask): Likewise.
8850 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
8851 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
8852 * tree-vect-stmts.c (vect_mark_relevant)
8853 (is_simple_and_all_uses_invariant)
8854 (exist_non_indexing_operands_for_use_p, process_use)
8855 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
8856 (vect_get_vec_def_for_operand, vect_get_vec_defs)
8857 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
8858 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
8859 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
8860 (get_negative_load_store_type, get_load_store_type)
8861 (vect_check_load_store_mask, vect_check_store_rhs)
8862 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
8863 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
8864 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
8865 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
8866 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
8867 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
8868 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
8869 (supportable_widening_operation): Likewise.
8871 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8873 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
8874 a stmt_vec_info instead of a gcall.
8875 (vect_check_gather_scatter): Update call accordingly.
8876 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
8878 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
8879 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
8880 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
8881 instead of a gimple stmt.
8882 (vect_transform_loop): Update calls accordingly.
8883 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
8884 stmt_vec_infos instead of gimple stmts.
8885 (vect_analyze_slp_instance): Update use accordingly.
8886 * tree-vect-stmts.c (read_vector_array, write_vector_array)
8887 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
8888 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
8889 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
8890 (vect_gen_widened_results_half, vect_get_loop_based_defs)
8891 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
8892 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
8893 down to subroutines.
8895 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8897 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
8898 of the worklist from a vector of gimple stmts to a vector of
8900 * tree-vect-stmts.c (vect_mark_relevant, process_use)
8901 (vect_mark_stmts_to_be_vectorized): Likewise
8903 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8905 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
8906 statement before passing it to vect_analyze_stmt.
8907 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
8908 the chain of phi vector definitions. Track the exit phi via its
8910 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
8911 STMT_VINFO_REDUC_DEF.
8912 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
8913 stmt_vec_infos to handle the statement chains.
8914 (vect_get_slp_defs): Record the first statement in the node
8915 using a stmt_vec_info.
8916 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
8917 statements here and pass their stmt_vec_info down to subroutines.
8918 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
8919 down to vect_finish_stmt_generation.
8920 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
8921 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
8922 stmt_vec_infos to subroutines.
8923 (vect_remove_stores): Use stmt_vec_infos to handle the statement
8926 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8928 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
8929 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
8930 (vect_permute_store_chain, vect_setup_realignment)
8931 (vect_permute_load_chain, vect_shift_permute_load_chain)
8932 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
8933 stmts internally, and when passing values to other vectorizer routines.
8934 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
8935 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
8936 (vect_analyze_loop_operations, get_initial_def_for_reduction)
8937 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
8938 (vectorizable_reduction, vectorizable_induction)
8939 (vectorizable_live_operation, vect_transform_loop_stmt)
8940 (vect_transform_loop): Likewise.
8941 * tree-vect-patterns.c (vect_reassociating_reduction_p)
8942 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
8943 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
8944 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8945 (vect_slp_analyze_node_operations_1): Likewise.
8946 * tree-vect-stmts.c (vect_mark_relevant, process_use)
8947 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
8948 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
8949 (vect_finish_stmt_generation_1, get_group_load_store_type)
8950 (get_load_store_type, vect_build_gather_load_calls)
8951 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
8952 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
8953 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
8954 (vectorizable_store, vectorizable_load, vectorizable_condition)
8955 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
8956 (supportable_widening_operation): Likewise.
8957 (vect_get_vector_types_for_stmt): Likewise.
8958 * tree-vectorizer.h (vect_dr_behavior): Likewise.
8960 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8962 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
8963 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
8964 (vect_permute_store_chain, vect_permute_load_chain)
8965 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
8966 repeated stmt_vec_info lookups.
8967 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
8968 (vect_update_ivs_after_vectorizer): Likewise.
8969 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
8970 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
8971 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
8972 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8973 (vect_bb_slp_scalar_cost): Likewise.
8974 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
8976 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8978 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
8979 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
8980 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
8981 to get gassigns and gcalls, rather than operating on generc gimple
8983 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
8984 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
8985 (vectorizable_load, vect_analyze_stmt): Likewise.
8986 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
8988 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8990 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
8991 return stmt_vec_infos rather than gimple stmts. Do not accept
8993 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
8995 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
8996 Update use of get_later_stmt.
8997 (vect_get_constant_vectors): Update call accordingly.
8998 (vect_schedule_slp_instance): Likewise
8999 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
9000 (vect_slp_analyze_instance_dependence): Likewise.
9001 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
9003 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9005 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
9006 (stmt_info_for_cost::stmt_info): ...this new field.
9007 (add_stmt_costs): Update accordingly.
9008 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
9009 (vect_get_known_peeling_cost): Likewise.
9010 (vect_estimate_min_profitable_iters): Likewise.
9011 * tree-vect-stmts.c (record_stmt_cost): Likewise.
9013 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9015 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
9016 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
9017 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
9019 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
9021 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9023 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
9024 a gimple stmt to a stmt_vec_info.
9025 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
9027 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9029 * tree-vectorizer.h (vec_info::grouped_stores): Change from
9030 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
9031 (_loop_vec_info::reduction_chains): Likewise.
9032 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
9034 * tree-vect-slp.c (vect_analyze_slp): Likewise.
9036 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9038 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
9039 a gimple stmt to a stmt_vec_info.
9040 (_stmt_vec_info::next_element): Likewise.
9041 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
9042 (vect_slp_analyze_and_verify_node_alignment)
9043 (vect_analyze_group_access_1, vect_analyze_group_access)
9044 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
9045 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
9046 (vect_supportable_dr_alignment): Update accordingly.
9047 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
9048 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
9049 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
9050 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
9051 * tree-vect-slp.c (vect_build_slp_tree_1)
9052 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
9053 (vect_split_slp_store_group, vect_analyze_slp_instance)
9054 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
9055 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
9056 (get_group_load_store_type, get_load_store_type)
9057 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
9058 (vect_transform_stmt, vect_remove_stores): Likewise.
9060 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9062 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
9064 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
9065 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
9066 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
9067 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
9068 (vector_alignment_reachable_p, vect_get_data_access_cost)
9069 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
9070 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
9071 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
9072 (vect_analyze_group_access_1, vect_analyze_group_access)
9073 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
9074 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
9075 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
9076 result of vect_dr_stmt and use the stmt_vec_info instead of
9077 the associated gimple stmt.
9078 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
9079 (vect_gen_prolog_loop_niters): Likewise.
9080 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
9082 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9084 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
9085 to a vec<stmt_vec_info>.
9086 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
9087 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
9089 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
9090 to a vec<stmt_vec_info>.
9091 (bst_traits::value_type, bst_traits::value_type): Likewise.
9092 (bst_traits::hash): Update accordingly.
9093 (vect_get_and_check_slp_defs): Change the stmts parameter from
9094 a vec<gimple *> to a vec<stmt_vec_info>.
9095 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
9096 (vect_build_slp_tree): Likewise.
9097 (vect_build_slp_tree_2): Likewise. Update uses of
9098 SLP_TREE_SCALAR_STMTS.
9099 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
9100 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
9101 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
9102 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
9103 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
9104 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
9105 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
9106 (vect_get_constant_vectors, vect_get_slp_defs)
9107 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
9108 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
9109 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
9110 instead of gimple stmts.
9111 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
9112 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
9113 (vect_slp_analyze_instance_dependence): Update uses of
9114 SLP_TREE_SCALAR_STMTS.
9115 (vect_slp_analyze_and_verify_node_alignment): Likewise.
9116 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
9117 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
9118 (get_initial_defs_for_reduction): Likewise.
9119 (vect_create_epilog_for_reduction): Likewise.
9120 (vectorize_fold_left_reduction): Likewise.
9121 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
9122 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
9123 (can_vectorize_live_stmts): Likewise.
9125 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9127 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
9128 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
9129 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
9131 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
9133 (gather_scalar_reductions): Update after above interface changes.
9134 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
9135 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
9137 (vect_force_simple_reduction): Likewise.
9138 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
9139 LOOP_VINFO_REDUCTIONS.
9140 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9142 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9144 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
9145 a gimple stmt to a stmt_vec_info.
9146 * tree-vect-loop.c (vect_active_double_reduction_p)
9147 (vect_force_simple_reduction, vectorizable_reduction): Update
9150 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9152 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
9153 vec<gimple *> to a vec<stmt_vec_info>.
9154 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
9155 the reduction_phis argument from a vec<gimple *> to a
9157 (vectorizable_reduction): Likewise the phis local variable that
9158 is passed to vect_create_epilog_for_reduction. Update for new type
9159 of SLP_TREE_VEC_STMTS.
9160 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
9161 (vectorizable_live_operation): Likewise.
9162 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
9163 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
9165 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9167 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
9168 a gimple stmt to a stmt_vec_info.
9169 (vectorizable_condition, vectorizable_live_operation)
9170 (vectorizable_reduction, vectorizable_induction): Pass back the
9171 vectorized statement as a stmt_vec_info.
9172 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
9173 use of STMT_VINFO_VEC_STMT.
9174 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
9175 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
9176 as stmt_vec_infos rather than gimple stmts.
9177 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
9179 (vectorizable_live_operation): Likewise.
9180 (vectorizable_reduction, vectorizable_induction): Likewise,
9181 updating use of STMT_VINFO_VEC_STMT.
9182 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
9183 of STMT_VINFO_VEC_STMT.
9184 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
9185 (vectorizable_simd_clone_call, vectorizable_conversion)
9186 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
9187 (vectorizable_store, vectorizable_load, vectorizable_condition)
9188 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
9189 from a gimple stmt to a stmt_vec_info.
9190 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
9191 pointer to a stmt_vec_info to the vectorizable_* routines.
9193 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9195 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
9196 a gimple stmt to a stmt_vec_info.
9197 (is_pattern_stmt_p): Update accordingly.
9198 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
9199 (vect_record_grouped_load_vectors): Likewise.
9200 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
9201 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
9202 (vect_model_reduction_cost): Likewise.
9203 (vect_create_epilog_for_reduction): Likewise.
9204 (vectorizable_reduction, vectorizable_induction): Likewise.
9205 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
9206 Return the stmt_vec_info for the pattern statement.
9207 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
9208 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
9209 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
9210 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
9211 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
9212 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
9213 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
9214 (free_stmt_vec_info, vect_is_simple_use): Likewise.
9216 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9218 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
9219 (vect_finish_stmt_generation): Likewise.
9220 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
9221 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
9222 (vect_build_gather_load_calls): Use the return value of the above
9223 functions instead of a separate call to vinfo_for_stmt. Use narrow
9224 scopes for the input gimple stmt and wider scopes for the associated
9225 stmt_vec_info. Use vec_info::lookup_def when setting these
9226 stmt_vec_infos from an SSA_NAME definition.
9227 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
9228 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
9229 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
9230 (vectorizable_store, vectorizable_load, vectorizable_condition)
9231 (vectorizable_comparison): Likewise.
9232 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
9233 (vectorizable_reduction): Likewise.
9235 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9237 * tree-vectorizer.h (vect_is_simple_use): Add an optional
9238 stmt_vec_info * parameter before the optional gimple **.
9239 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
9240 (process_use, vect_get_vec_def_for_operand_1): Update callers.
9241 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
9242 * tree-vect-loop.c (vectorizable_reduction): Likewise.
9243 (vectorizable_live_operation): Likewise.
9244 * tree-vect-patterns.c (type_conversion_p): Likewise.
9245 (vect_look_through_possible_promotion): Likewise.
9246 (vect_recog_rotate_pattern): Likewise.
9247 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
9249 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9251 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
9252 a typedef to a wrapper class.
9253 (NULL_STMT_VEC_INFO): New macro.
9254 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
9255 (stmt_vec_info::operator*): New function.
9256 (stmt_vec_info::operator gimple *): Likewise.
9257 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
9258 (add_stmt_costs): Likewise.
9259 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
9260 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
9261 (vect_get_known_peeling_cost): Likewise.
9262 (vect_estimate_min_profitable_iters): Likewise.
9263 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
9264 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
9265 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
9266 (vectorizable_store, free_stmt_vec_infos): Likewise.
9267 (new_stmt_vec_info): Change return type of xcalloc to
9270 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9272 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
9273 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
9274 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
9275 a single_imm_use-based sequence.
9276 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
9278 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9280 * tree-vectorizer.h (vec_info::lookup_def): Declare.
9281 * tree-vectorizer.c (vec_info::lookup_def): New function.
9282 * tree-vect-patterns.c (vect_get_internal_def): Use it.
9283 (vect_widened_op_tree): Likewise.
9284 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
9285 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
9286 (vectorizable_reduction): Likewise.
9287 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
9289 (vect_is_slp_reduction): Update calls accordingly. Use
9290 vec_info::lookup_def.
9291 (vect_is_simple_reduction): Likewise
9292 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
9294 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9296 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
9297 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
9298 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
9300 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
9301 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
9302 (vect_update_vf_for_slp, vect_analyze_loop_operations)
9303 (vect_is_slp_reduction, vectorizable_induction)
9304 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
9305 * tree-vect-patterns.c (vect_init_pattern_stmt):
9306 (vect_determine_min_output_precision_1, vect_determine_precisions)
9307 (vect_pattern_recog): Likewise.
9308 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
9309 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
9310 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
9311 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
9312 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
9313 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
9314 info field from a loop to a loop_vec_info.
9316 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9318 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
9319 (vec_info::add_stmt): Declare.
9320 * tree-vectorizer.c (vec_info::add_stmt): New function.
9321 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
9322 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
9323 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
9324 (vectorizable_induction): Likewise.
9325 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
9326 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
9327 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
9328 (vectorizable_load): Likewise.
9329 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
9330 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
9331 (vect_recog_gather_scatter_pattern): Likewise.
9332 (append_pattern_def_seq): Likewise. Remove a check that is
9333 performed by add_stmt itself.
9335 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9337 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
9338 which make_ssa_name was called with new_stmt before new_stmt
9341 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9343 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
9345 (vect_is_slp_reduction): ...here...
9346 (vect_is_simple_reduction): ...and here. Remove repetition of tests
9347 that are already known to be false.
9349 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9351 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
9352 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
9353 STMT_VINFO_NUM_SLP_USES when it's true.
9354 (vect_free_slp_instance): Add a final_p parameter and pass it to
9356 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
9357 (vect_analyze_slp_instance): Likewise.
9358 (vect_slp_analyze_operations): Likewise.
9359 (vect_slp_analyze_bb_1): Likewise.
9360 * tree-vectorizer.c (vec_info): Likewise.
9361 * tree-vect-loop.c (vect_transform_loop): Likewise.
9363 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9365 * tree-vect-loop.c (vectorizable_reduction): Assert that the
9366 function is not called for second and subsequent members of
9369 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9371 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
9372 cases for nested loops from here to ...
9373 (vect_create_epilog_for_reduction): ...here. Only call
9374 vect_is_simple_use for inner-loop reductions.
9376 2018-07-31 Martin Liska <mliska@suse.cz>
9378 PR gcov-profile/85338
9379 PR gcov-profile/85350
9380 PR gcov-profile/85372
9381 * profile.c (struct location_triplet): New.
9382 (struct location_triplet_hash): Likewise.
9383 (output_location): Do not output a BB that
9384 is already recorded for a line.
9385 (branch_prob): Use streamed_locations.
9387 2018-07-31 Martin Liska <mliska@suse.cz>
9389 PR gcov-profile/85370
9390 * coverage.c (coverage_begin_function): Do not mark target
9391 clones as artificial functions.
9393 2018-07-31 Martin Liska <mliska@suse.cz>
9395 PR gcov-profile/83813
9396 PR gcov-profile/84758
9397 PR gcov-profile/85217
9398 PR gcov-profile/85332
9399 * profile.c (branch_prob): Do not record GOTO expressions
9400 for GIMPLE statements which locations are already streamed.
9402 2018-07-31 Olivier Hainque <hainque@adacore.com>
9404 * gcc.c (handle_spec_function): Accept a soft_matched_part
9405 argument, as do_spec_1. Pass it down to ...
9406 (eval_spec_function): Accept a soft_matched_part argument,
9407 and pass it down to ...
9408 (do_spec_2): Accept a soft_matched_part argument, and pass
9409 it down to do_spec_1.
9410 (do_spec_1): Pass soft_matched_part to handle_spec_function.
9411 (handle_braces): Update call to handle_spec_function.
9412 (driver::set_up_specs): Update calls to do_spec_2.
9413 (compare_debug_dump_opt_spec_function): Likewise.
9414 (compare_debug_self_opt_spec_function): Likewise.
9416 2018-07-31 Olivier Hainque <hainque@adacore.com>
9418 * common.opt (nolibc): New option.
9419 * doc/invoke.texi (Link Options): Document it.
9420 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
9421 * config/alpha/linux.h: Likewise.
9422 * config/arc/elf.h: Likewise.
9423 * config/arm/uclinux-elf.h: Likewise.
9424 * config/arm/unknown-elf.h: Likewise.
9425 * config/avr/avrlibc.h: Likewise.
9426 * config/bfin/bfin.h: Likewise.
9427 * config/bfin/linux.h: Likewise.
9428 * config/bfin/uclinux.h: Likewise.
9429 * config/darwin.h: Likewise.
9430 * config/darwin10.h: Likewise.
9431 * config/darwin12.h: Likewise.
9432 * config/gnu-user.h: Likewise.
9433 * config/lm32/uclinux-elf.h: Likewise.
9434 * config/pa/pa-hpux11.h: Likewise.
9435 * config/pa/pa64-hpux.h: Likewise.
9436 * config/sparc/sparc.h: Likewise.
9438 2018-07-31 Olivier Hainque <hainque@adacore.com>
9440 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
9441 undefined variables.
9443 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
9446 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
9449 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
9451 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
9452 terminated string literal.
9454 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
9456 PR rtl-optimization/85160
9457 * combine.c (is_just_move): New function.
9458 (try_combine): Allow combining two instructions into two if neither of
9459 the original instructions was a move.
9461 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
9464 * doc/extend.texi (Global Register Variables): Discourage use of type
9466 (Local Register Variables): Likewise.
9468 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
9470 PR tree-optimization/86506
9471 * hwint.h (ceil_log2): Resync with hwint.c implementation.
9473 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
9476 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
9477 hard_regno, make sure no insn between `from` and `to` clobbers it.
9479 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
9480 Tom de Vries <tdevries@suse.de>
9482 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
9483 (PTX_DEFAULT_RUNTIME_DIM): ... this.
9484 (nvptx_goacc_validate_dims): Set default worker and gang dims to
9485 PTX_DEFAULT_RUNTIME_DIM.
9486 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
9488 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
9490 * config/pa/pa.c (pa_output_addr_vec): Align address table.
9491 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
9492 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
9494 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
9496 * config/rs6000/constraints.md (wG constraint): Delete, no longer
9498 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
9499 predicate to reflect toc fusion has been deleted.
9500 (toc_fusion_mem_raw): Delete, no longer used.
9501 (toc_fusion_mem_wrapped): Likewise.
9502 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
9504 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
9505 Delete, no longer used.
9506 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
9507 meant to be used for toc fusion.
9508 (rs6000_debug_print_mode): Delete toc fusion debugging.
9509 (rs6000_debug_reg_global): Likewise.
9510 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
9511 fusion and secondary reload support that were never used.
9512 (rs6000_option_override_internal): Delete TOC fusion, that was only
9513 partially defined, and it did not work unless you also used the
9515 (rs6000_legitimate_address_p): Delete TOC fusion support.
9516 (rs6000_opt_masks): Likewise.
9517 (fusion_wrap_memory_address): Delete function, no longer used.
9518 (fusion_split_address); Delete TOC fusion support.
9519 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
9520 longer used with toc fusion being deleted.
9521 (TARGET_TOC_FUSION_FP): Likewise.
9522 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
9524 (toc fusion spliter): Delete TOC fusion support.
9525 (toc_fusionload_<mode>): Likewise.
9526 (toc_fusionload_di): Likewise.
9527 (fusion_gpr_load_<mode>): Delete generator function, this insn no
9528 longer needs to be named. Rename predicate to delete TOC fusion.
9529 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
9530 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
9531 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
9532 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
9533 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
9535 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
9537 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
9538 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
9539 __int128 in built-in function prototypes.
9540 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
9541 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
9543 2018-07-27 Martin Sebor <msebor@redhat.com>
9545 PR tree-optimization/86696
9546 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
9547 types, including enums.
9548 (handle_char_store): Be prepared for the above function to fail.
9550 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
9552 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
9553 when optimization level is lower than 2 or optimize for size.
9555 2018-07-26 Martin Sebor <msebor@redhat.com>
9557 PR tree-optimization/86043
9558 PR tree-optimization/86042
9559 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
9560 (get_string_cst_length): Rename...
9561 (get_min_string_length): ...to this. Add argument.
9562 (handle_char_store): Extend to handle multi-character stores by
9564 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
9565 * tree.h (initializer_zerop): Add argument.
9567 2018-07-26 Jakub Jelinek <jakub@redhat.com>
9570 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
9571 declare target to variables if they have always,{to,from,tofrom} map
9574 2018-07-26 Martin Liska <mliska@suse.cz>
9577 * lto-wrapper.c: Add linker_output as prefix
9578 for ltrans_output_file.
9580 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
9582 PR rtl-optimization/85805
9583 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
9584 value for hard registers if that was written in the same mode.
9586 2018-07-26 Martin Liska <mliska@suse.cz>
9588 PR gcov-profile/86536
9589 * gcov.c (format_gcov): Use printf format %.*f directly
9590 and do not handle special values.
9592 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
9594 * common/config/arc/arc-common.c (arc_option_optimization_table):
9595 Update default optimizations for size.
9597 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
9599 * config/arc/arc.md (movsf_insn): Add short instruction selection.
9600 * config/arc/constraints.md (CfZ): New constraint.
9601 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
9602 (subsf3_fpu): Likewise.
9603 (cmpsf_fpu): Likewise.
9604 (cmpsf_fpu_uneq): Likewise.
9606 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
9608 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
9610 (arc_is_uncached_mem_p): uncached applies to both the variable and
9613 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
9615 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
9618 2018-07-25 David Malcolm <dmalcolm@redhat.com>
9620 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
9621 field "m_scopes" from vec to auto_vec.
9623 2018-07-25 Martin Liska <mliska@suse.cz>
9625 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
9628 2018-07-25 Richard Biener <rguenther@suse.de>
9631 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
9632 special wrt context_die late.
9633 (gen_subprogram_die): Re-use DIEs in local scope.
9635 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
9637 PR tree-optimization/86644
9638 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
9640 2018-07-25 Martin Liska <mliska@suse.cz>
9643 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
9644 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
9646 2018-07-25 Martin Liska <mliska@suse.cz>
9649 * params.def: Explain ASan abbreviation and provide
9650 a documentation link.
9652 2018-07-24 Martin Sebor <msebor@redhat.com>
9654 PR tree-optimization/86622
9655 PR tree-optimization/86532
9656 * builtins.h (string_length): Declare.
9657 * builtins.c (c_strlen): Correct handling of non-constant offsets.
9658 (check_access): Be prepared for non-constant length ranges.
9659 (string_length): Make extern.
9660 * expr.c (string_constant): Only handle the minor non-constant
9661 array index. Use string_constant to compute the length of
9662 a generic string constant.
9664 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
9666 PR tree-optimization/86618
9667 * tree-vect-stmts.c (vectorizable_call): Don't take the address
9668 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
9670 2018-07-24 David Malcolm <dmalcolm@redhat.com>
9672 PR tree-optimization/86636
9673 * json.cc (json::object::set): Fix comment. Add assertions.
9674 (json::array::append): Move here from json.h. Add comment and an
9676 (json::string::string): Likewise.
9677 * json.h (json::array::append): Move to json.cc.
9678 (json::string::string): Likewise.
9679 * optinfo-emit-json.cc
9680 (optrecord_json_writer::impl_location_to_json): Assert that we
9681 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
9682 wrapper around it. Expand the location once, rather than three
9684 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
9685 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
9687 (optrecord_json_writer::optinfo_to_json): Likewise, in four
9688 places. Fix some overlong lines.
9690 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
9692 * config/aarch64/aarch64-simd.md
9693 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
9694 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
9695 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
9696 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
9697 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
9698 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
9699 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
9700 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
9701 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
9703 2018-07-24 Jakub Jelinek <jakub@redhat.com>
9706 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
9707 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
9708 and abs_d == d, do the power of two handling if profitable.
9710 2018-07-24 Richard Biener <rguenther@suse.de>
9712 * match.pd: Add BIT_FIELD_REF canonicalizations.
9714 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
9717 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
9719 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
9721 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
9722 terminated STRING_CST object.
9724 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
9726 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
9728 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
9730 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
9731 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
9732 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
9733 the elements into a register.
9734 (rs6000_split_v4si_init_di_reg): Delete.
9735 (rs6000_split_v4si_init): Delete.
9736 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
9737 (vsx_init_v4si): Rewrite as a define_expand.
9739 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
9741 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
9742 zero_extend argument from memory): New.
9744 2018-07-22 Martin Sebor <msebor@redhat.com>
9747 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
9748 diagnosing calls with unknown arguments unless -Walloca-larger-than
9749 is restricted to less than PTRDIFF_MAX bytes.
9751 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
9753 * doc/gcov.texi (Invoking Gcov): Editorial changes.
9755 2018-07-20 David Malcolm <dmalcolm@redhat.com>
9757 * pretty-print.c (text_info::set_location): Remove redundant
9758 "line_table" parameter from call to rich_location::set_range.
9760 2018-07-20 Martin Sebor <msebor@redhat.com>
9763 * builtins.c (expand_builtin_alloca): Adjust.
9764 * calls.c (alloc_max_size): Simplify.
9765 * cgraphunit.c (cgraph_node::expand): Adjust.
9766 * common.opt (larger_than_size, warn_frame_larger_than): Remove
9768 (frame_larger_than_size): Same.
9769 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
9770 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
9772 * doc/invoke.texi (GCC Command Options): Document option arguments.
9773 Explain byte-size arguments and suffixes.
9774 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
9775 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
9776 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
9777 * doc/options.texi (UInteger): Expand.
9778 (Host_Wide_Int, ByteSize): Document new properties.
9779 * final.c (final_start_function_1): Include sizes in an error message.
9780 * function.c (frame_offset_overflow): Same.
9781 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
9782 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
9783 Diagnose unbounded alloca calls only for limits of less than
9785 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
9786 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
9788 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
9789 only for limits of less than PTRDIFF_MAX.
9790 * langhooks-def.h (lhd_handle_option): Change function argument
9792 * langhooks.c (lhd_handle_option): Same.
9793 * langhooks.h (handle_option): Same.
9794 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
9796 (var_type, var_type_struct): Same.
9797 (var_set): Handle ByteSize flag.
9798 * optc-gen.awk: Add comments to output to ease debugging. Make
9799 use of HOST_WIDE_INT where appropriate.
9800 * opts-gen-save.awk: Use %lx to format unsigned long.
9801 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
9802 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
9803 arguments. Parse bytes-size suffixes.
9804 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
9805 (enum_value_to_arg): Same.
9806 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
9807 (handle_option): Adjust.
9808 (generate_option): Change function argument to HOST_WIDE_INT.
9809 (cmdline_handle_error): Adjust.
9810 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
9811 (set_option): Change function argument to HOST_WIDE_INT.
9812 (option_enabled): Handle cl_host_wide_int.
9813 (get_option_state): Handle CLVC_SIZE.
9814 (control_warning_option): Same.
9815 * opts.c (common_handle_option): Change function argument to
9816 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
9817 OPT_Wvla_larger_than_.
9818 * opts.h (enum cl_var_type): Add an enumerator.
9819 * stor-layout.c (layout_decl): Print a more meaningful warning.
9820 * toplev.c (output_stack_usage): Adjust.
9822 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
9824 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
9825 call to inline_expand_builtin_string_cmp.
9826 (expand_builtin_strcmp): Likewise.
9827 (expand_builtin_strncmp): Likewise.
9828 (inline_string_cmp): Delete the last parameter, change char_type_node
9829 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
9831 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
9832 the inlining expansion on target where the type of the call has same or
9833 narrower precision than unsigned char.
9835 2018-07-20 David Malcolm <dmalcolm@redhat.com>
9837 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
9838 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
9839 * common.opt (fsave-optimization-record): New option.
9840 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
9841 * doc/invoke.texi (-fsave-optimization-record): New option.
9842 * dumpfile.c: Include "optinfo-emit-json.h".
9843 (struct kv_pair): Move to coretypes.h.
9844 (optgroup_options): Make non-static.
9845 (dump_context::end_scope): Call
9846 optimization_records_maybe_pop_dump_scope.
9847 * dumpfile.h (optgroup_options): New decl.
9848 * json.cc: New file.
9850 * optinfo-emit-json.cc: New file.
9851 * optinfo-emit-json.h: New file.
9852 * optinfo.cc: Include "optinfo-emit-json.h".
9853 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
9854 (optinfo_enabled_p): Check optimization_records_enabled_p.
9855 (optinfo_wants_inlining_info_p): Likewise.
9856 * optinfo.h: Update comment.
9857 * profile-count.c (profile_quality_as_string): New function.
9858 * profile-count.h (profile_quality_as_string): New decl.
9859 (profile_count::quality): New accessor.
9860 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
9861 and optinfo_emit_json_cc_tests.
9862 * selftest.h (selftest::json_cc_tests): New decl.
9863 (selftest::optinfo_emit_json_cc_tests): New decl.
9864 * toplev.c: Include "optinfo-emit-json.h".
9865 (compile_file): Call optimization_records_finish.
9866 (do_compile): Call optimization_records_start.
9867 * tree-ssa-live.c: Include optinfo.h.
9868 (remove_unused_scope_block_p): Retain inlining information if
9869 optinfo_wants_inlining_info_p returns true.
9871 2018-07-20 Richard Biener <rguenther@suse.de>
9874 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
9875 to cover -flto-partition=none.
9877 2018-07-20 Martin Liska <mliska@suse.cz>
9879 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
9880 (get_decl_source_range): Remove unused function.
9882 2018-07-20 Richard Biener <rguenther@suse.de>
9884 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
9885 (struct vn_phi_s): Likewise.
9886 (struct vn_reference_s): Likewise.
9887 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
9888 for searching the slot of an entry known to be in the hash itself.
9889 (vn_phi_hasher::equal): Likewise.
9890 (vn_reference_hasher::equal): Likewise.
9891 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
9893 (optimistic_info, current_info): Remove, keeping only valid_info.
9894 (vn_reference_lookup_1): Remove fallback lookup.
9895 (vn_reference_lookup_2): Likewise.
9896 (vn_nary_op_lookup_1): Likewise.
9897 (vn_phi_lookup): Likewise.
9898 (vn_nary_build_or_lookup_1): Make sure to not chain the built
9900 (vn_reference_insert): Adjust, chain the inserted hash element
9901 at last_inserted_ref.
9902 (vn_reference_insert_pieces): Likewise.
9903 (visit_reference_op_call): Likewise.
9904 (vn_nary_op_insert_into): Chain the inserted hash element at
9906 (vn_nary_op_insert_pieces): Adjust.
9907 (vn_nary_op_insert): Likewise.
9908 (vn_nary_op_insert_stmt): Likewise.
9909 (vn_phi_insert): Adjust, chain the inserted hash element at
9911 (process_scc): Remove clearing and copying the optimistic
9912 table. Instead remove elements inserted during an optimistic
9913 iteration from the single table we maintain.
9914 (init_scc_vn): Adjust.
9915 (free_scc_vn): Likewise.
9916 (sccvn_dom_walker::record_cond): Likewise.
9917 (sccvn_dom_walker::after_dom_children): Likewise.
9919 2018-07-19 Martin Sebor <msebor@redhat.com>
9921 PR tree-optimization/84047
9922 PR tree-optimization/83776
9923 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
9924 (check_array_bounds): Call it.
9926 2018-07-19 Martin Sebor <msebor@redhat.com>
9928 * align.h (align_flags): Use member initialization.
9930 2018-07-19 David Malcolm <dmalcolm@redhat.com>
9932 * Makefile.in (OBJS): Add optinfo.o.
9933 * coretypes.h (class symtab_node): New forward decl.
9934 (struct cgraph_node): New forward decl.
9935 (class varpool_node): New forward decl.
9936 * dump-context.h: New file.
9937 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
9939 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
9940 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
9941 (set_alt_dump_file): Likewise.
9942 (dump_context::~dump_context): New dtor.
9943 (dump_gimple_stmt): Move implementation to...
9944 (dump_context::dump_gimple_stmt): ...this new member function.
9945 Add the stmt to any pending optinfo, creating one if need be.
9946 (dump_gimple_stmt_loc): Move implementation to...
9947 (dump_context::dump_gimple_stmt_loc): ...this new member function.
9948 Start a new optinfo and add the stmt to it.
9949 (dump_gimple_expr): Move implementation to...
9950 (dump_context::dump_gimple_expr): ...this new member function.
9951 Add the stmt to any pending optinfo, creating one if need be.
9952 (dump_gimple_expr_loc): Move implementation to...
9953 (dump_context::dump_gimple_expr_loc): ...this new member function.
9954 Start a new optinfo and add the stmt to it.
9955 (dump_generic_expr): Move implementation to...
9956 (dump_context::dump_generic_expr): ...this new member function.
9957 Add the tree to any pending optinfo, creating one if need be.
9958 (dump_generic_expr_loc): Move implementation to...
9959 (dump_context::dump_generic_expr_loc): ...this new member
9960 function. Add the tree to any pending optinfo, creating one if
9962 (dump_printf): Move implementation to...
9963 (dump_context::dump_printf_va): ...this new member function. Add
9964 the text to any pending optinfo, creating one if need be.
9965 (dump_printf_loc): Move implementation to...
9966 (dump_context::dump_printf_loc_va): ...this new member function.
9967 Start a new optinfo and add the stmt to it.
9968 (dump_dec): Move implementation to...
9969 (dump_context::dump_dec): ...this new member function. Add the
9970 value to any pending optinfo, creating one if need be.
9971 (dump_context::dump_symtab_node): New member function.
9972 (dump_context::get_scope_depth): New member function.
9973 (dump_context::begin_scope): New member function.
9974 (dump_context::end_scope): New member function.
9975 (dump_context::ensure_pending_optinfo): New member function.
9976 (dump_context::begin_next_optinfo): New member function.
9977 (dump_context::end_any_optinfo): New member function.
9978 (dump_context::s_current): New global.
9979 (dump_context::s_default): New global.
9980 (dump_scope_depth): Delete global.
9981 (dumpfile_ensure_any_optinfo_are_flushed): New function.
9982 (dump_symtab_node): New function.
9983 (get_dump_scope_depth): Reimplement in terms of dump_context.
9984 (dump_begin_scope): Likewise.
9985 (dump_end_scope): Likewise.
9986 (selftest::temp_dump_context::temp_dump_context): New ctor.
9987 (selftest::temp_dump_context::~temp_dump_context): New dtor.
9988 (selftest::verify_item): New function.
9989 (ASSERT_IS_TEXT): New macro.
9990 (ASSERT_IS_TREE): New macro.
9991 (ASSERT_IS_GIMPLE): New macro.
9992 (selftest::test_capture_of_dump_calls): New test.
9993 (selftest::dumpfile_c_tests): Call it.
9994 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
9995 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
9996 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
9997 descriptive comment.
9998 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
9999 (dump_node, dump_bb): Move these unrelated decls.
10000 (class dump_manager): Add leading comment.
10001 * optinfo.cc: New file.
10002 * optinfo.h: New file.
10004 2018-07-19 Michael Collison <michael.collison@arm.com>
10005 Richard Henderson <rth@redhat.com>
10007 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
10008 (subti): Handle op1 zero.
10009 (subvti4, usub4ti4): New.
10010 (*sub<GPI>3_compare1_imm): New.
10011 (sub<GPI>3_carryinCV): New.
10012 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
10013 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
10015 2018-07-19 Michael Collison <michael.collison@arm.com>
10016 Richard Henderson <rth@redhat.com>
10018 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
10019 (addti3): Create simpler code if low part is already known to be 0.
10020 (addvti4, uaddvti4): New.
10021 (*add<GPI>3_compareC_cconly_imm): New.
10022 (*add<GPI>3_compareC_cconly): New.
10023 (*add<GPI>3_compareC_imm): New.
10024 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
10025 handle constants within this pattern..
10026 (*add<GPI>3_compareV_cconly_imm): New.
10027 (*add<GPI>3_compareV_cconly): New.
10028 (*add<GPI>3_compareV_imm): New.
10029 (add<GPI>3_compareV): New.
10030 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
10031 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
10032 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
10033 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
10034 with 'comparison' operator.
10035 (*add<GPI>3_compareV_cconly_imm): Ditto.
10036 (*add<GPI>3_compareV_cconly): Ditto.
10037 (*add<GPI>3_compareV_imm): Ditto.
10038 (add<GPI>3_compareV): Ditto.
10039 (add<mode>3_carryinC): Ditto.
10040 (*add<mode>3_carryinC_zero): Ditto.
10041 (*add<mode>3_carryinC): Ditto.
10042 (add<mode>3_carryinV): Ditto.
10043 (*add<mode>3_carryinV_zero): Ditto.
10044 (*add<mode>3_carryinV): Ditto.
10046 2018-07-19 Michael Collison <michael.collison@arm.com>
10047 Richard Henderson <rth@redhat.com>
10049 * config/aarch64/aarch64-modes.def (CC_V): New.
10050 * config/aarch64/aarch64-protos.h
10051 (aarch64_addti_scratch_regs): Declare
10052 (aarch64_subvti_scratch_regs): Declare.
10053 (aarch64_expand_subvti): Declare.
10054 (aarch64_gen_unlikely_cbranch): Declare
10055 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
10056 for signed overflow using CC_Vmode.
10057 (aarch64_get_condition_code_1): Handle CC_Vmode.
10058 (aarch64_gen_unlikely_cbranch): New function.
10059 (aarch64_addti_scratch_regs): New function.
10060 (aarch64_subvti_scratch_regs): New function.
10061 (aarch64_expand_subvti): New function.
10063 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
10065 * config/aarch64/aarch64-option-extensions.def: New entry for profile
10067 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
10068 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
10071 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
10074 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
10075 address check not strict.
10077 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
10079 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
10081 (store_pair_lanes<mode>): Likewise.
10082 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
10083 enum value 'ADDR_QUERY_LDP_STP_N'.
10084 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
10085 (aarch64_print_address_internal): Add declaration.
10086 (aarch64_print_ldpstp_address): Remove.
10087 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
10088 (aarch64_print_operand): Change printing of 'y'.
10089 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
10090 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
10091 'true' rather than '1'.
10092 * config/aarch64/constraints.md (Uml): Likewise.
10093 (Uml): Rename to Umn.
10096 2018-07-19 Richard Biener <rguenther@suse.de>
10098 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
10100 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
10101 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
10102 (vn_reference_hasher): Likewise.
10103 (struct vn_tables_s): Remove obstack and alloc-pool members.
10104 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
10105 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
10106 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
10107 (vn_reference_insert_pieces): Likewise.
10108 (alloc_vn_nary_op_noinit): Adjust.
10109 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
10110 (vn_phi_eq): Adjust.
10111 (shared_lookup_phiargs): Remove.
10112 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
10113 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
10114 (visit_reference_op_call): Likewise.
10115 (copy_nary, copy_phi, copy_reference): Remove.
10116 (process_scc): Rewind the obstack when iterating. Do not
10117 copy the elements to valid_info but just move them from one
10118 hashtable to the other.
10119 (allocate_vn_table): Adjust.
10120 (free_vn_table): Likewise.
10121 (init_scc_vn): Likewise.
10122 (free_scc_vn): Likewise.
10124 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
10127 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
10128 indirect_return as function type attribute.
10129 (ix86_attribute_table): Change indirect_return to function
10131 * doc/extend.texi: Update indirect_return attribute.
10133 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
10135 * wide-int.h (widest2_int): New.
10136 * gimple-fold.c (arith_overflowed_p): Use it.
10137 * tree.h (widest2_int_cst): New.
10138 * tree-vrp.c (wide_int_binop_overflow): Rename from
10139 vrp_int_const_binop.
10140 Rewrite to work on trees.
10141 (extract_range_from_multiplicative_op_1): Abstract code to...
10142 (wide_int_range_min_max): ...here.
10143 (wide_int_range_cross_product): ...and here.
10144 (extract_range_from_binary_expr_1): Abstract overflow code to...
10145 (wide_int_range_mult_wrapping): ...here.
10146 * tree-vrp.h (wide_int_range_cross_product): New.
10147 (wide_int_range_mult_wrapping): New.
10149 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
10150 Julia Koval <julia.koval@intel.com>
10152 * config/i386/x86-tune-costs.h (skylake_memcpy,
10153 skylake_memset): Replace rep_prefix with unrolling for size 512.
10155 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
10157 PR middle-end/86544
10158 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
10159 comparision with EQ_EXPR in last stmt.
10161 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
10163 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
10164 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
10165 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
10166 previously known as "PowerPC AltiVec Built-in Functions". Move
10167 some material to new subsubsections "PowerPC AltiVec Built-in
10168 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
10170 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
10171 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
10172 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
10173 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
10175 2018-07-18 Richard Biener <rguenther@suse.de>
10177 PR tree-optimization/86557
10178 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
10181 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
10183 * config/s390/s390.c (s390_function_profiler): Generate CFI.
10185 2018-07-17 Jeff Law <law@redhat.com>
10187 * config/arm/arm.c (get_label_padding): Update for recent
10188 changes to label_to_alignment.
10190 PR tree-optimization/86010
10191 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
10193 * config/mips/mips.c (vr4130_align_insns): Update for recent
10194 changes to label_to_alignment.
10196 * config/frv/frv.c (frv_label_align): Update for recent changes
10197 to label_to_alignment.
10199 * config/nios2/nios2.c (nios2_label_align): Update for recent
10200 changes which dropped ALIGN_LABELS_LOG.
10202 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
10204 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
10205 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
10207 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
10209 * config/arc/arc.c (arc_label_align): Use align_labels instead of
10210 deprecated align_labels_log.
10212 2018-07-17 Richard Biener <rguenther@suse.de>
10215 * dwarf2out.c (init_sections_and_labels): Always generate
10216 a debug_line_str_section for early LTO debug.
10217 (dwarf2out_finish): Reset debug_line_str_hash output early.
10218 Bump counter for extra dwarf5 .debug_loc labels to not conflict
10220 (dwarf2out_early_finish): Output debug_line_str.
10222 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
10224 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
10225 index register on z196 or later.
10227 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
10229 * config/s390/s390.c (s390_default_align): Set default function
10231 (s390_override_options_after_change): Call s390_default align.
10232 (s390_option_override_internal): Call s390_default_align.
10233 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
10235 2018-07-17 Jakub Jelinek <jakub@redhat.com>
10237 PR middle-end/86542
10238 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
10239 to _looptemp_ clauses, other than the first two.
10241 2018-07-17 Martin Liska <mliska@suse.cz>
10243 * opts.c: Do not enable OPT_falign_* for -Os.
10245 2018-07-17 Martin Liska <mliska@suse.cz>
10247 * align.h (MAX_CODE_ALIGN): New.
10248 (MAX_CODE_ALIGN_VALUE): New.
10249 * common/config/i386/i386-common.c (ix86_handle_option):
10250 (MAX_CODE_ALIGN): Moved to align.h.
10251 * final.c (MAX_CODE_ALIGN): Likewise.
10252 * opts.c (parse_and_check_align_values):
10253 (MAX_CODE_ALIGN): Likewise.
10254 (MAX_CODE_ALIGN_VALUE): Likewise.
10256 2018-07-17 Martin Liska <mliska@suse.cz>
10258 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
10259 in order to fulfil coding style.
10260 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
10261 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10262 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
10263 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
10264 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
10265 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
10266 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
10267 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
10268 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10270 2018-07-17 Martin Liska <mliska@suse.cz>
10272 * align.h: New file.
10273 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
10275 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
10276 align_flags of label_to_alignment.
10277 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
10279 * config/m68k/m68k.c: Do not use removed align_labels_value and
10281 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
10282 (LOOP_ALIGN): Likewise.
10283 (LABEL_ALIGN): Likewise.
10284 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
10285 Remove not used macro.
10286 (rs6000_loop_align): Change return type to align_flags.
10287 (rs6000_loop_align_max_skip): Remove.
10288 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
10289 Change return type to align_flags.
10290 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
10291 Remove not used macro.
10292 (rs6000_loop_align): Change return type to align_flags.
10293 (rs6000_loop_align_max_skip): Remove.
10294 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
10295 * config/rx/rx-protos.h (rx_align_for_label): Make it
10297 * config/rx/rx.c (rx_align_for_label): Change return type
10299 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
10301 into align_flags class.
10302 (LABEL_ALIGN): Likewise.
10303 (LOOP_ALIGN): Likewise.
10304 * config/s390/s390.c (s390_label_align): Use align_flags
10306 (s390_asm_output_function_label): Likewise.
10307 * config/sh/sh.c (sh_override_options_after_change):
10308 Use align_flags class directly without macros.
10309 (find_barrier): Likewise.
10310 (barrier_align): Likewise.
10311 (sh_loop_align): Likewise.
10312 * config/spu/spu.c (spu_option_override):
10313 Use align_flags_tuple::get_value instead of removed macros.
10314 (spu_sched_init): Likewise.
10315 * config/spu/spu.h (GTY): Likewise.
10316 * config/visium/visium.c (visium_option_override):
10317 Set "8" as default secondary alignment.
10318 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
10319 in order to guarantee secondary alignment of 8.
10320 * coretypes.h: Include align.h header file.
10321 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
10322 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
10323 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
10324 * doc/tm.texi.in: Likewise.
10325 * final.c (struct label_alignment): Remove not used structure.
10326 (LABEL_ALIGN): Change type to align_flags.
10327 (LOOP_ALIGN): Likewise.
10328 (JUMP_ALIGN): Likewise.
10329 (default_loop_align_max_skip): Remove.
10330 (default_label_align_max_skip): Likewise.
10331 (default_jump_align_max_skip): Likewise.
10332 (default_label_align_after_barrier_max_skip):
10333 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
10334 (LABEL_TO_MAX_SKIP): Remove.
10335 (label_to_alignment): Return align_flags type instead of integer.
10336 (label_to_max_skip): Remove.
10337 (align_fuzz): Use align_flags type.
10338 (compute_alignments): Use align_flags type and use align_flags::max
10339 to combine multiple alignments.
10340 (grow_label_align): Grow vec instead of C array.
10341 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
10342 (shorten_branches): Use align_flags type and use align_flags::max
10343 to combine multiple alignments.
10344 (final_scan_insn_1): Remove usage of secondary alignment that comes
10345 from label alignment, but instead use proper secondary alignment
10346 which is computed in grow_label_align.
10347 * flags.h (struct align_flags_tuple): Move to align.h.
10348 (struct align_flags): Likewise.
10349 (state_align_loops): Rename to align_loops.
10350 (state_align_jumps): Rename to align_jumps.
10351 (state_align_labels): Rename to align_labels.
10352 (state_align_functions): Rename to align_functions.
10353 (align_loops_log): Remove.
10354 (align_jumps_log): Remove.
10355 (align_labels_log): Remove.
10356 (align_functions_log): Remove.
10357 (align_loops_max_skip): Remove.
10358 (align_jumps_max_skip): Remove.
10359 (align_labels_max_skip): Remove.
10360 (align_functions_max_skip): Remove.
10361 (align_loops_value): Remove.
10362 (align_jumps_value): Remove.
10363 (align_labels_value): Remove.
10364 (align_functions_value): Remove.
10365 * output.h (label_to_alignment): Change return type to align_flags.
10366 (label_to_max_skip): Remove.
10367 * target.def: Remove loop_align_max_skip, label_align_max_skip,
10368 jump_align_max_skip macros.
10369 * targhooks.h (default_loop_align_max_skip): Remove.
10370 (default_label_align_max_skip): Likewise.
10371 (default_jump_align_max_skip): Likewise.
10372 (default_label_align_after_barrier_max_skip): Remove.
10373 * toplev.c (read_log_maxskip): Use ::normalize function.
10374 (parse_N_M): Remove not used argument and also call ::normalize.
10375 (parse_alignment_opts): Do not pass unused arguments.
10376 * varasm.c (assemble_start_function): Use directly align_functions
10377 instead of removed macros.
10378 * system.h: Do not poison removed macros.
10380 2018-07-17 Jakub Jelinek <jakub@redhat.com>
10382 PR middle-end/86539
10383 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
10384 and cond temporaries don't have reference type if iterator has
10385 pointer type. For init use &for_pre_body instead of pre_p if
10386 for_pre_body is non-empty.
10388 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
10390 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
10391 double-double modes to SFmode directly directly.
10392 (trunc<mode>sf2_fprs): Delete.
10394 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
10396 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
10397 for conversions between IFmode and the decimal floating point modes.
10398 (init_float128_ieee): Use the correct names for conversions between
10399 KFmode and the decimal floating point modes.
10401 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
10403 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
10404 for the conversions between TDmode and IFmode.
10405 (init_float128_ieee): Use more correct names for the conversions
10406 between TDmode and KFmode.
10408 2018-07-16 Jakub Jelinek <jakub@redhat.com>
10410 PR tree-optimization/86526
10411 * builtins.c (expand_builtin_memcmp): Formatting fixes.
10412 (inline_expand_builtin_string_cmp): Likewise.
10413 (inline_string_cmp): Likewise. Use c_readstr instead of
10414 builtin_memcpy_read_str. Add unit_mode temporary.
10416 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
10418 PR middle-end/86528
10419 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
10420 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
10422 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
10424 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
10425 Alphabetize prototypes of built-in functions, separating out
10426 built-in functions that are listed in this section but should be
10427 described elsewhere.
10429 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
10432 * expmed.c (emit_store_flag): Do not emit setcc followed by a
10433 conditional move when trapping comparison was split to a
10434 non-trapping one (and vice versa).
10436 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
10438 * config/s390/s390.c (s390_function_profiler): Generate nops
10439 instead of profiler call sequences.
10440 * config/s390/s390.opt: Add the new option.
10442 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
10444 * config/s390/s390.c (s390_function_profiler): Generate
10445 __mcount_loc section.
10446 * config/s390/s390.opt: Add the new option.
10448 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
10450 * common.opt: Add the new warning.
10451 * config/s390/s390.c (s390_function_profiler): Emit "brasl
10452 %r0,__fentry__" when -mfentry is specified.
10453 (s390_option_override_internal): Disallow -mfentry for 31-bit
10455 * config/s390/s390.opt: Add the new option.
10457 2018-07-16 Richard Biener <rguenther@suse.de>
10460 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
10461 for function-local FUNCTION_DECL and RESULT_DECL immediately.
10463 2018-07-16 Martin Liska <mliska@suse.cz>
10466 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
10469 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
10471 * config/arc/arcHS.md: Update ARCHS scheduling rules.
10473 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
10475 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
10477 * config/arc/arc-cpus.def (hs4x): New cpu.
10479 * config/arc/arc-tables.opt: Regenerate.
10480 * config/arc/arc.c (arc_sched_issue_rate): New function.
10481 (TARGET_SCHED_ISSUE_RATE): Define.
10482 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
10483 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
10485 (attr tune): Add ARCHS4x tune values.
10486 (attr tune_dspmpy): Define.
10487 (*tst): Correct instruction type.
10488 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
10489 * config/arc/arcHS4x.md: New file.
10490 * config/arc/fpu.md: Update instruction type attributes.
10491 * config/arc/t-multilib: Regenerate.
10493 2018-07-16 Tom de Vries <tdevries@suse.de>
10496 * var-tracking.c (vt_initialize): Fix pre_dec handling.
10498 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10500 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
10503 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
10505 PR tree-optimization/86514
10506 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
10507 conversion to a boolean type from a type with greater precision.
10509 2018-07-16 Tom de Vries <tdevries@suse.de>
10511 * var-tracking.c (vt_initialize): Print adjusted insn slim if
10512 dump_flags request TDF_SLIM.
10514 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
10516 * fold-const.c (int_const_binop_1): Abstract...
10517 (wide_int_binop): ...wide int code here.
10518 (poly_int_binop): ...poly int code here.
10519 Abstract the rest of int_const_binop_1 into int_const_binop.
10520 * fold-const.h (wide_int_binop): New.
10521 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
10522 Remove useless PLUS/MINUS_EXPR case.
10523 (zero_nonzero_bits_from_vr): Move wide int code...
10524 (zero_nonzero_bits_from_bounds): ...here.
10525 (extract_range_from_binary_expr_1): Move mask optimization code...
10526 (range_easy_mask_min_max): ...here.
10527 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
10528 (range_easy_mask_min_max): New.
10530 2018-07-15 Jeff Law <law@redhat.com>
10533 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
10536 2018-07-14 Jim Wilson <jimw@sifive.com>
10538 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
10540 2018-07-14 Paul Koning <ni1d@arrl.net>
10542 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
10544 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
10546 * lto-streamer-out.c (copy_function_or_variable): Dump info about
10549 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
10550 Steve Munroe <munroesj52@gmail.com>
10552 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
10553 (_mm_andnot_si128): Likewise.
10554 (_mm_or_si128): Likewise.
10555 (_mm_xor_si128): Likewise.
10557 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
10559 PR middle-end/78809
10560 * builtins.c (expand_builtin_memcmp): Inline the calls first
10561 when result_eq is false.
10562 (expand_builtin_strcmp): Inline the calls first.
10563 (expand_builtin_strncmp): Likewise.
10564 (inline_string_cmp): New routine. Expand a string compare
10565 call by using a sequence of char comparison.
10566 (inline_expand_builtin_string_cmp): New routine. Inline expansion
10567 a call to str(n)cmp/memcmp.
10568 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
10570 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
10572 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
10574 * config/arm/driver-arm.c: Include arm-native.h.
10575 (host_detect_local_cpu): Use auto-generated data tables.
10576 (vendors, arm_cpu_table): Delete. Move part information to ...
10577 * config/arm/arm-cpus.in: ... here.
10578 * config/arm/parsecpu.awk (gen_native): New function.
10579 (vendor, part): New CPU fields.
10580 (END): Add support for building the native CPU detection tables.
10581 * config/arm/t-arm (arm-native.h): Add build rule.
10582 (driver-arm.o): Add dependency on arm-native.h.
10584 2018-07-13 Richard Biener <rguenther@suse.de>
10586 PR middle-end/85974
10587 * match.pd (addr1 - addr2): Allow either of the operand to
10590 2018-07-13 Tom de Vries <tdevries@suse.de>
10592 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
10595 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
10597 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
10598 arrays instead of numbered variables.
10600 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
10602 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
10603 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
10605 2018-07-13 Richard Biener <rguenther@suse.de>
10608 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
10609 instead of get_context_die.
10611 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
10612 Richard Biener <rguenther@suse.de>
10614 PR middle-end/86489
10615 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
10616 that the loop latch destination where phi is defined.
10618 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
10620 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
10621 (riscv_expand_epilogue): Add assertion to check interrupt mode.
10622 (riscv_set_current_function): Extract getting interrupt type to new
10624 (riscv_get_interrupt_type): New function.
10625 (riscv_merge_decl_attributes): New function, checking interrupt type is
10627 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
10629 2018-07-12 Paul Koning <ni1d@arrl.net>
10631 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
10634 2018-07-12 Paul Koning <ni1d@arrl.net>
10636 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
10637 zero reference, add doloop_end instead.
10638 * doc/md.texi (decrement_and_branch_until_zero): Remove.
10639 (Looping patterns): Remove decrement_and_branch_until_zero. Add
10640 detail for doloop_end.
10642 2018-07-12 Martin Sebor <msebor@redhat.com>
10645 * attribs.c (decl_attributes): Reject conflicting attributes before
10646 calling attribute handlers.
10648 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
10650 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
10652 (gcc::dump_manager::get_dump_file_name): likewise.
10653 (dump_begin): Likewise.
10654 * dumpfile.h (dump_begin): Update prototype.
10655 (gcc::dump_manager::get_dump_file_name,
10656 gcc::dump_manager::get_dump_file_name): Update prototype.
10658 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
10660 * internal-fn.h (vectorizable_internal_fn_p): New function.
10661 * tree-vect-slp.c (compatible_calls_p): Likewise.
10662 (vect_build_slp_tree_1): Remove nops argument. Handle calls
10663 to internal functions.
10664 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
10666 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
10668 * fold-const.h (inverse_conditions_p): Declare.
10669 * fold-const.c (inverse_conditions_p): New function.
10670 * match.pd: Use inverse_conditions_p. Add folds of view_converts
10671 that test the inverse condition of a conditional internal function.
10672 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
10673 * internal-fn.c (internal_fn_mask_index): Handle conditional
10674 internal functions.
10675 (vectorized_internal_fn_supported_p): New function.
10676 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
10677 (any_pred_load_store): Replace with...
10678 (need_to_predicate): ...this new variable.
10679 (redundant_ssa_names): New variable.
10680 (ifcvt_can_use_mask_load_store): Move initial checks to...
10681 (ifcvt_can_predicate): ...this new function. Handle tree codes
10682 for which a conditional internal function exists.
10683 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
10684 instead of ifcvt_can_use_mask_load_store. Update after variable
10686 (predicate_load_or_store): New function, split out from
10687 predicate_mem_writes.
10688 (check_redundant_cond_expr): New function.
10689 (value_available_p): Likewise.
10690 (predicate_rhs_code): Likewise.
10691 (predicate_mem_writes): Rename to...
10692 (predicate_statements): ...this. Use predicate_load_or_store
10693 and predicate_rhs_code.
10694 (combine_blocks, tree_if_conversion): Update after above name changes.
10695 (ifcvt_local_dce): Handle redundant_ssa_names.
10696 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
10697 general conditional functions.
10698 * tree-vect-stmts.c (vectorizable_call): Likewise.
10700 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
10701 Alan Hayward <alan.hayward@arm.com>
10702 David Sherwood <david.sherwood@arm.com>
10704 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
10705 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
10706 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
10707 plus and minus and convert them into IFN_COND_FMA-based sequences.
10708 (convert_mult_to_fma): Handle conditional plus and minus.
10710 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
10712 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
10713 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
10714 (cond_fnms_optab): New optabs.
10715 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
10716 internal functions.
10717 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
10718 * internal-fn.h (get_conditional_internal_fn): Declare.
10719 (get_unconditional_internal_fn): Likewise.
10720 * internal-fn.c (cond_ternary_direct): New macro.
10721 (expand_cond_ternary_optab_fn): Likewise.
10722 (direct_cond_ternary_optab_supported_p): Likewise.
10723 (FOR_EACH_COND_FN_PAIR): Likewise.
10724 (get_conditional_internal_fn): New function.
10725 (get_unconditional_internal_fn): Likewise.
10726 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
10727 (gimple_match_op::gimple_match_op): Add a new overload for 5
10729 (gimple_match_op::set_op): Likewise.
10730 (gimple_resimplify5): Declare.
10731 * genmatch.c (decision_tree::gen): Generate simplifications for
10733 * gimple-match-head.c (gimple_simplify): Define an overload for
10734 5 operands. Handle calls with 5 arguments in the top-level overload.
10735 (convert_conditional_op): Handle conversions from unconditional
10736 internal functions to conditional ones.
10737 (gimple_resimplify5): New function.
10738 (build_call_internal): Pass a fifth operand.
10739 (maybe_push_res_to_seq): Likewise.
10740 (try_conditional_simplification): Try converting conditional
10741 internal functions to unconditional internal functions.
10742 Handle 3-operand unconditional forms.
10743 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
10744 Define ternary equivalents of the current rules for binary conditional
10745 internal functions.
10746 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
10747 ternary operations.
10748 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
10749 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
10750 (optab): Handle them.
10751 (SVE_COND_FP_TERNARY): New int iterator.
10752 (sve_fmla_op, sve_fmad_op): New int attributes.
10753 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
10754 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
10755 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
10757 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
10759 * target.def (preferred_else_value): New target hook.
10760 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
10761 * doc/tm.texi: Regenerate.
10762 * targhooks.h (default_preferred_else_value): Declare.
10763 * targhooks.c (default_preferred_else_value): New function.
10764 * internal-fn.h (conditional_internal_fn_code): Declare.
10765 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
10766 (get_conditional_internal_fn): Use it.
10767 (conditional_internal_fn_code): New function.
10768 * gimple-match.h (gimple_match_cond): New struct.
10769 (gimple_match_op): Add a cond member function.
10770 (gimple_match_op::gimple_match_op): Update all forms to take a
10772 * genmatch.c (expr::gen_transform): Use the same condition as res_op
10773 for the suboperation, but don't specify a particular else_value.
10774 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
10775 (visit_nary_op, visit_reference_op_load): Pass
10776 gimple_match_cond::UNCOND to the gimple_match_op constructor.
10777 * gimple-match-head.c: Include tree-eh.h
10778 (convert_conditional_op): New function.
10779 (maybe_resimplify_conditional_op): Likewise.
10780 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
10781 (gimple_resimplify2): Likewise.
10782 (gimple_resimplify3): Likewise.
10783 (gimple_resimplify4): Likewise.
10784 (maybe_push_res_to_seq): Return null for conditional operations.
10785 (try_conditional_simplification): New function.
10786 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
10788 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
10790 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
10792 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
10794 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
10796 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
10798 (hash_tree): Do not hash DECL_FCONTEXT
10799 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
10800 Do not stream DECL_FCONTEXT.
10801 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
10802 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
10804 2018-07-12 Richard Biener <rguenther@suse.de>
10807 * dwarf2out.c (gen_block_die): Only output blocks when they have
10808 at least one !DECL_IGNORED_P variable.
10810 2018-07-12 Richard Biener <rguenther@suse.de>
10813 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
10814 Remove -mieee-fp handling.
10816 2018-07-12 Richard Biener <rguenther@suse.de>
10818 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
10819 left-over from last patch.
10821 2018-07-12 Jakub Jelinek <jakub@redhat.com>
10823 PR tree-optimization/86492
10824 * gimple-ssa-store-merging.c
10825 (imm_store_chain_info::coalesce_immediate_stores): Call
10826 check_no_overlap even for the merge_overlapping case. Formatting fix.
10828 2018-07-12 Richard Biener <rguenther@suse.de>
10830 PR middle-end/86479
10831 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
10832 move possibly trapping operations into the conditional.
10834 2018-07-12 Richard Biener <rguenther@suse.de>
10836 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
10837 (vn_lookup_simplify_result): Remove recursion limit applied
10839 (vn_nary_build_or_lookup_1): Adjust.
10840 (try_to_simplify): Likewise.
10841 * gimple-match-head.c (gimple_resimplify1): Instead apply one
10843 (gimple_resimplify2): Likewise.
10844 (gimple_resimplify3): Likewise.
10845 (gimple_resimplify4): Likewise.
10847 2018-07-11 Jakub Jelinek <jakub@redhat.com>
10849 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
10850 Use __mmask64 type instead of __mmask8 for __M argument.
10851 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
10852 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
10854 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
10855 __mmask16 for __M argument.
10856 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
10857 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
10858 to __mmask16 instead of __mmask8.
10859 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
10860 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
10861 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
10862 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
10863 instead of __mmask16 for __U argument.
10864 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
10865 __mmask16 instead of __mmask8 for __U argument.
10866 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
10868 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
10870 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
10872 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
10874 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
10876 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
10878 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
10880 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
10882 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
10884 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
10886 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
10887 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
10888 return type as well as __M argument type and all casts from __mmask8
10890 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
10891 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
10892 return type as well as __M argument type and all casts from __mmask8
10894 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
10895 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
10896 return type as well as __M argument type and all casts from __mmask8
10898 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
10899 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
10900 return type as well as __M argument type and all casts from __mmask8
10902 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
10903 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
10906 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
10908 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
10909 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
10912 2018-07-11 Paul Koning <ni1d@arrl.net>
10914 * doc/md.texi (define_subst): Document how multiple occurrences of
10915 the same argument in the replacement pattern are handled.
10917 2018-07-11 Paul Koning <ni1d@arrl.net>
10919 * doc/extend.texi (Common Variable Attributes): Move "mode" into
10920 alphabetical order.
10921 (Common Type Attributes): Add "mode" attribute.
10923 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
10925 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
10926 stream DECL_ORIGINAL_TYPE.
10927 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
10928 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
10929 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
10930 Do not walk original type.
10931 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
10933 (write_ts_decl_non_common_tree_pointers): Do not stream
10935 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
10936 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
10938 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
10940 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
10941 thread twice from the same starting edge.
10943 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
10945 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
10946 * gimple.c (gimple_call_nonnull_result_p): ...here...
10947 (gimple_call_nonnull_arg): ...and here.
10948 * gimple.h (gimple_call_nonnull_result_p): New.
10949 (gimple_call_nonnull_arg): New.
10951 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
10953 * config/arm/arm-cpus.in: Move information from fpu field of each
10954 cpu definition to the isa field.
10955 * config/arm/parsecpu.awk (fpu): Delete match rule.
10956 (gen_comm_data): Don't add bits from the CPU's FPU entry.
10958 2018-07-11 Richard Biener <rguenther@suse.de>
10961 * dwarf2out.c (init_sections_and_labels): Use
10962 output_asm_line_debug_info consistently.
10963 (dwarf2out_early_finish): Likewise.
10964 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
10967 2018-07-11 Richard Biener <rguenther@suse.de>
10969 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
10970 Rework father_bb setting in a way to avoid propagating constants
10971 multiple times on a loop body.
10973 2018-07-10 Mark Wielaard <mark@klomp.org>
10976 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
10978 2018-07-10 Richard Biener <rguenther@suse.de>
10980 * hash-map.h (hash_map::iterator::operator*): Return
10981 references to key and value.
10983 2018-07-10 Jakub Jelinek <jakub@redhat.com>
10986 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
10987 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
10988 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
10989 (gimplify_omp_for): For composite loops, move outer
10990 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
10991 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
10992 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
10993 TREE_LIST for both the original class iterator and the "last" helper
10994 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
10995 loop, remember has_decl_expr from outer composite loops for the
10996 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
10998 2018-07-09 Martin Sebor <msebor@redhat.com>
11000 PR middle-end/77357
11001 PR middle-end/86428
11002 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
11003 accessing implicitly initialized array elements.
11004 * expr.c (string_constant): Handle string initializers of
11005 character arrays within aggregates.
11006 * gimple-fold.c (fold_array_ctor_reference): Add argument.
11007 Store element offset. As a special case, handle zero size.
11008 (fold_nonarray_ctor_reference): Same.
11009 (fold_ctor_reference): Add argument. Store subobject offset.
11010 * gimple-fold.h (fold_ctor_reference): Add argument.
11012 2018-07-09 Paul Koning <ni1d@arrl.net>
11014 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
11015 (pdp11_insn_cost): New function.
11016 (pdp11_md_asm_adjust): New function.
11017 (TARGET_INVALID_WITHIN_DOLOOP): Define.
11018 (pdp11_rtx_costs): Update to match machine better.
11019 (output_addr_const_pdp11): Correct format mismatch warnings.
11020 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
11021 * config/pdp11/pdp11.md: General change to add base_cost and/or
11022 length attributes for use by new pdp11_insn_cost function.
11023 (MIN_BRANCH): Correct definition.
11025 (doloop_end): Use standard pattern name for looping pattern.
11026 (doloop_end_nocc): New.
11027 (movsf): Add another constraint alternative.
11028 (zero_extendqihi2): Add constraint alternatives for not in place
11030 (zero_extendhisi2): Remove.
11031 (shift patterns): Add CC handling variants.
11035 (define_peephole2): New peephole to recognize mov that sets CC for
11038 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11040 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
11041 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
11044 2018-07-09 Jakub Jelinek <jakub@redhat.com>
11047 * real.c (real_nextafter): Return true if result is denormal.
11049 2018-07-09 Martin Liska <mliska@suse.cz>
11051 * common.opt: Add back wrongly removed attribute.
11053 2018-07-09 Richard Biener <rguenther@suse.de>
11056 * dwarf2out.c (gen_block_die): For an early generated DIE
11057 always output high/low PC attributes.
11059 2018-07-09 Tom de Vries <tdevries@suse.de>
11061 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
11062 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
11063 onto VAR_DECL with abstract origin.
11065 2018-07-07 Jim Wilson <jimw@sifive.com>
11067 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
11069 2018-07-07 Tom de Vries <tdevries@suse.de>
11071 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
11073 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
11075 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
11077 (combine_bound): Use wide-int overflow calculation instead of
11079 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
11081 * fold-const.c (int_const_binop_2): Same.
11082 (extract_muldiv_1): Same.
11083 (fold_div_compare): Same.
11084 (fold_abs_const): Same.
11086 * poly-int.h (add): Same.
11090 * predict.c (predict_iv_comparison): Same.
11091 * profile-count.c (slow_safe_scale_64bit): Same.
11092 * simplify-rtx.c (simplify_const_binary_operation): Same.
11093 * tree-chrec.c (tree_fold_binomial): Same.
11094 * tree-data-ref.c (split_constant_offset_1): Same.
11095 * tree-if-conv.c (idx_within_array_bound): Same.
11096 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
11097 * tree-ssa-phiopt.c (minmax_replacement): Same.
11098 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
11099 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
11100 * vr-values.c (vr_values::adjust_range_with_scev): Same.
11101 * wide-int.cc (wi::add_large): Same.
11102 (wi::mul_internal): Same.
11103 (wi::sub_large): Same.
11104 (wi::divmod_internal): Same.
11105 * wide-int.h: Change overflow type to overflow_type for neg, add,
11106 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
11107 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
11108 mul_internal, divmod_internal.
11109 (overflow_type): New enum.
11110 (accumulate_overflow): New.
11112 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
11114 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
11115 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
11117 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
11119 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
11120 argument is checked for zero before entering loop, avoid checking again.
11122 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
11124 * gimplify.h (generic_expr_could_trap_p): Set as global function.
11125 * gimplify.h (generic_expr_could_trap_p): Likwise.
11126 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
11128 2018-07-06 Jakub Jelinek <jakub@redhat.com>
11130 PR tree-optimization/86401
11131 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
11132 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
11133 (fold_bit_and_mask): ... here. New helper function for match.pd.
11134 * fold-const.h (fold_bit_and_mask): Declare.
11135 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
11137 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
11140 * target.def (translate_mode_attribute): New hook.
11141 * targhooks.h (default_translate_mode_attribute): Declare.
11142 * targhooks.c (default_translate_mode_attribute): New function.
11143 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
11144 * doc/tm.texi: Regenerate.
11145 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
11146 (rs6000_translate_mode_attribute): New function.
11148 2018-07-06 Paul Koning <ni1d@arrl.net>
11150 * doc/md.texi (define_split): Document DONE and FAIL.
11151 (define_peephole2): Ditto.
11153 2018-07-05 Jeff Law <law@redhat.com>
11155 PR tree-optimization/86010
11156 * tree-ssa-dse.c (compute_trims): More aggressively trim at
11157 both the head and tail of mem* and str* calls.
11159 2018-07-05 Jim Wilson <jimw@sifive.com>
11161 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
11163 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
11165 * config/aarch64/aarch64-simd.md: correct flags text for
11166 MIN_EXPR replacement.
11168 2018-07-05 James Clarke <jrtc27@jrtc27.com>
11170 * configure: Regenerated.
11172 2018-07-05 Carl Love <cel@us.ibm.com>
11174 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
11175 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
11176 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
11177 VSX_BUILTIN_DOUBLEL_V4SF.
11179 2018-07-05 Martin Sebor <msebor@redhat.com>
11182 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
11183 than its domain to compute its the upper bound of a char array.
11185 2018-07-05 Nathan Sidwell <nathan@acm.org>
11187 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
11188 * doc/cpp.texi: Update comment.
11189 * doc/tm.texi: Rebuilt.
11190 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
11191 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
11192 * doc/extend.texi (Backwards Compatibility): Clarify it is system
11193 headers affected by extern "C".
11194 * system.h: Poison NO_IMPLICIT_EXTERN_C.
11195 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
11196 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
11197 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
11198 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
11199 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
11200 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
11201 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
11202 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
11203 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
11204 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
11205 config/sparc/sp64-elf.h, config/spu/spu.h,
11206 config/stormy16/stormy16.h, config/v850/v850.h,
11207 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
11208 define NO_IMPLICIT_EXTERN_C.
11209 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
11211 2018-07-05 Tamar Christina <tamar.christina@arm.com>
11214 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
11215 instead of GET_MODE_SIZE when comparing Units.
11217 2018-07-05 Tamar Christina <tamar.christina@arm.com>
11220 * rtlanal.c (set_noop_p): Constrain on mode change,
11221 include hard-reg-set.h
11223 2018-07-05 Tamar Christina <tamar.christina@arm.com>
11225 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
11227 2018-07-05 Jakub Jelinek <jakub@redhat.com>
11230 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
11233 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
11235 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
11237 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
11240 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
11242 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
11244 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
11246 PR middle-end/86380
11247 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
11249 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
11251 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
11252 neg_*_op* variables.
11254 2018-07-04 Martin Liska <mliska@suse.cz>
11256 * tree-switch-conversion.c: Define
11257 max_ratio_for_speed and max_ratio_for_size constants.
11259 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
11260 Martin Liska <mliska@suse.cz>
11262 PR middle-end/66240
11265 * common.opt: Rename align options with 'str_' prefix.
11266 * common/config/i386/i386-common.c (set_malign_value): New
11268 (ix86_handle_option): Use it to set -falign-* options/
11269 * config/aarch64/aarch64-protos.h (struct tune_params): Change
11270 type from int to string.
11271 * config/aarch64/aarch64.c: Update default values from int
11273 * config/alpha/alpha.c (alpha_override_options_after_change):
11275 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
11276 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
11277 max skip conditionally.
11278 * config/i386/freebsd.h (SUBALIGN_LOG): New.
11279 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
11280 max skip conditionally.
11281 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
11282 max skip conditionally.
11283 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
11284 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
11285 max skip conditionally.
11286 * config/i386/i386.c (struct ptt): Change type from int to
11288 (ix86_default_align): Set default values.
11289 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
11290 max skip conditionally.
11291 * config/i386/iamcu.h (SUBALIGN_LOG): New.
11292 (ASM_OUTPUT_MAX_SKIP_ALIGN):
11293 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
11294 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
11295 max skip conditionally.
11296 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
11297 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
11298 * config/i386/x86-64.h (SUBALIGN_LOG): New.
11299 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
11300 max skip conditionally.
11301 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
11302 * config/ia64/ia64.c (ia64_option_override): Set default values
11303 for alignment options.
11304 * config/m68k/m68k.c: Handle new str_align_* options.
11305 * config/mips/mips.c (mips_set_compression_mode): Change
11307 (mips_option_override): Set default values for options.
11308 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
11310 * config/rs6000/rs6000.c (rs6000_option_override_internal):
11312 * config/rx/rx.c (rx_option_override): Likewise.
11313 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
11314 (LABEL_ALIGN): Use align_labels_log.
11315 (LOOP_ALIGN): Use align_loops_align.
11316 * config/s390/s390.c (s390_asm_output_function_label): Use new
11318 * config/sh/sh.c (sh_override_options_after_change):
11319 Change type of constants.
11320 * config/spu/spu.c (spu_sched_init): Likewise.
11321 * config/sparc/sparc.c (sparc_option_override): Set default
11322 values for options.
11323 * config/visium/visium.c (visium_option_override): Likewise.
11324 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
11325 emit p2align format with last argument if it's not needed.
11326 * doc/invoke.texi: Document extended format of -falign-*.
11327 * final.c: Use align_labels alignment.
11328 * flags.h (struct target_flag_state): Change type to use
11330 (struct align_flags_tuple): New.
11331 (struct align_flags): Likewise.
11332 (align_loops_log): Redefine macro to use new types.
11333 (align_loops_max_skip): Redefine macro to use new types.
11334 (align_jumps_log): Redefine macro to use new types.
11335 (align_jumps_max_skip): Redefine macro to use new types.
11336 (align_labels_log): Redefine macro to use new types.
11337 (align_labels_max_skip): Redefine macro to use new types.
11338 (align_functions_log): Redefine macro to use new types.
11339 (align_loops): Redefine macro to use new types.
11340 (align_jumps): Redefine macro to use new types.
11341 (align_labels): Redefine macro to use new types.
11342 (align_functions): Redefine macro to use new types.
11343 (align_functions_max_skip): Redefine macro to use new types.
11344 (align_loops_value): New macro.
11345 (align_jumps_value): New macro.
11346 (align_labels_value): New macro.
11347 (align_functions_value): New macro.
11348 * function.c (invoke_set_current_function_hook): Propagate
11349 alignment values from flags to global variables default in
11351 * ipa-icf.c (sem_function::equals_wpa): Use
11352 cl_optimization_option_eq instead of memcmp.
11353 * lto-streamer.h (cl_optimization_stream_out): Support streaming
11355 (cl_optimization_stream_in): Likewise.
11356 * optc-save-gen.awk: Support strings in cl_optimization.
11357 * opth-gen.awk: Likewise.
11358 * opts.c (finish_options): Remove error checking of invalid
11360 (MAX_CODE_ALIGN): Remove.
11361 (MAX_CODE_ALIGN_VALUE): Likewise.
11362 (parse_and_check_align_values): New function.
11363 (check_alignment_argument): Likewise.
11364 (common_handle_option): Use check_alignment_argument.
11365 * opts.h (parse_and_check_align_values): Declare.
11366 * toplev.c (init_alignments): Remove.
11367 (read_log_maxskip): New.
11368 (parse_N_M): Likewise.
11369 (parse_alignment_opts): Likewise.
11370 (backend_init_target): Remove usage of init_alignments.
11371 * toplev.h (parse_alignment_opts): Declare.
11372 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
11374 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
11375 * tree.c (cl_option_hasher::equal): New.
11376 * varasm.c: Use new global macros.
11378 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
11380 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
11381 Use a simpler align directive also if MAXSKIP = ALIGN-1.
11382 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11383 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11384 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11385 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
11386 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
11387 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
11388 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
11389 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11390 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11391 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11392 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11394 2018-07-04 Martin Liska <mliska@suse.cz>
11395 Jonathan Wakely <jwakely@redhat.com>
11397 * coverage.c: Use correct type.
11398 * doc/invoke.texi: Language correction.
11400 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
11403 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
11404 ENDBRANCH for non-tail call which may return via indirect branch.
11405 * doc/extend.texi: Document indirect_return attribute.
11407 2018-07-03 Martin Sebor <msebor@redhat.com>
11409 PR tree-optimization/86274
11410 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
11412 (format_floating): Correct handling of infinities and NaNs.
11414 2018-07-03 Martin Sebor <msebor@redhat.com>
11416 * print-tree.c (print_real_cst): New function.
11417 (print_node_brief): Call it.
11418 (print_node): Ditto.
11420 2018-07-03 Jeff Law <law@redhat.com>
11422 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
11423 into a single pattern.
11425 * config/h8300/h8300.md (ors code_iterator): New.
11426 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
11427 a single pattern and single splitter.
11428 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
11429 (iorqi3_1, xorqi3_1): Likewise.
11430 (iorqi3, xorqi3 expanders): Similarly.
11432 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
11433 (movmd_internal) into a single pattern using the P mode iterator.
11434 (movmd splitters): Similarly.
11435 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
11436 (movsd splitters): Similarly.
11438 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
11439 ADDB, ADDW and ADDL into a single ADD attribute which selects the
11440 right table based on the size of the operand.
11441 * config/h8300/h8300.md (length_table): Corresponding changes. All
11442 references to "addb", "addw" and "addl" changed to "add".
11443 (btst patterns): Merge two variants into a single pattern.
11444 (tstqi, tsthi): Likewise.
11445 (addhi3_incdec, addsi3_incdec): Likewise.
11446 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
11447 (mulhi3, mulsi3): Likewise.
11448 (udivhi3, udivsi3): Likewise.
11449 (divhi3, divsi3): Likewise.
11450 (andorqi3, andorhi3, andorsi3): Likewise.
11452 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
11455 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
11456 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
11458 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
11460 PR tree-optimization/85694
11461 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
11462 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
11464 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
11465 (<u>avg<mode>3_ceil): New patterns.
11467 2018-07-03 David Malcolm <dmalcolm@redhat.com>
11469 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
11470 scan-tree-dump directive.
11471 * gcc.dg/vect/slp-perm-2.c: Likewise.
11472 * gcc.dg/vect/slp-perm-3.c: Likewise.
11473 * gcc.dg/vect/slp-perm-5.c: Likewise.
11474 * gcc.dg/vect/slp-perm-6.c: Likewise.
11475 * gcc.dg/vect/slp-perm-7.c: Likewise.
11476 * gcc.dg/vect/slp-perm-8.c: Likewise.
11478 2018-07-03 Marek Polacek <polacek@redhat.com>
11480 PR middle-end/86202
11481 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
11483 2018-07-03 Richard Biener <rguenther@suse.de>
11486 * tree-ssa-structalias.c (find_func_clobbers): Properly
11487 handle indirect calls.
11489 2018-07-03 Jeff Law <law@redhat.com>
11491 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
11492 (shifts): New code iterator.
11493 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
11494 expander. Fix HImode handling on H8/SX.
11495 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
11496 (subqi3, subhi3, subsi3 expanders): Likewise.
11497 (andqi3, andhi3, andsi3 expanders): Likewise.
11498 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
11499 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
11500 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
11501 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
11502 (zero_extendqihi2, zero_extendqisi2): Likewise.
11503 (extendqihi2, extendqisi2): Likewise.
11504 (rotlqi3, rotlhi3, rotlsi3): Likewise.
11505 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
11506 (rotlqi3_1, rotlhi3_1): Likewise.
11507 (logicalhi3_sn, logicalsi3_sn): Likewise.
11508 (logicalhi3, logicalsi3): Likewise.
11510 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
11512 * tree-vect-patterns.c (vect_recog_rotate_pattern)
11513 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
11514 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
11515 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
11516 type to append_pattern_def_seq instead of creating a stmt_vec_info
11518 (build_mask_conversion): Likewise. Remove vinfo argument.
11519 (vect_add_conversion_to_patterm): Likewise, renaming to...
11520 (vect_add_conversion_to_pattern): ...this.
11521 (vect_recog_mask_conversion_pattern): Update call to
11522 build_mask_conversion. Pass the vector type to
11523 append_pattern_def_seq here too.
11524 (vect_recog_gather_scatter_pattern): Update call to
11525 vect_add_conversion_to_pattern.
11527 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
11529 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
11530 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
11531 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
11532 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
11533 STMT_VINFO_PATTERN_DEF_SEQ to null here.
11534 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
11535 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
11536 append_pattern_def_seq instead of new_pattern_def_seq.
11537 (vect_recog_divmod_pattern): Do both of the above.
11538 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
11541 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
11543 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
11544 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
11545 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
11546 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
11547 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
11548 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
11549 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
11550 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
11551 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
11552 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
11553 parameter with a single stmt_vec_info.
11554 (vect_recog_func_ptr): Likewise.
11555 (vect_recog_gather_scatter_pattern): Likewise, folding in...
11556 (vect_try_gather_scatter_pattern): ...this.
11557 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
11558 the stmt_vec_info of the statement to be matched. Don't clear
11559 STMT_VINFO_RELATED_STMT.
11560 (vect_pattern_recog): Update call accordingly.
11562 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
11564 PR tree-optimization/85694
11565 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
11566 (uavgM3_ceil): Document new optabs.
11567 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
11568 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
11570 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
11571 (savg_ceil_optab): New optabs.
11572 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
11573 (vect_vect_recog_func_ptrs): Add it.
11574 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
11575 constant directly from the associated lhs.
11577 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
11579 * tree-vect-patterns.c (vect_split_statement): New function.
11580 (vect_convert_input): Use it to try to split an existing cast.
11582 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
11584 * poly-int.h (print_hex): New function.
11585 * dumpfile.h (dump_dec, dump_hex): Declare.
11586 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
11587 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
11588 min_input_precision, operation_precision and operation_sign.
11589 * tree-vect-patterns.c (vect_get_range_info): New function.
11590 (vect_same_loop_or_bb_p, vect_single_imm_use)
11591 (vect_operation_fits_smaller_type): Delete.
11592 (vect_look_through_possible_promotion): Add an optional
11593 single_use_p parameter.
11594 (vect_recog_over_widening_pattern): Rewrite to use new
11595 stmt_vec_info infomration. Handle one operation at a time.
11596 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
11597 (vect_truncatable_operation_p, vect_set_operation_type)
11598 (vect_set_min_input_precision): New functions.
11599 (vect_determine_min_output_precision_1): Likewise.
11600 (vect_determine_min_output_precision): Likewise.
11601 (vect_determine_precisions_from_range): Likewise.
11602 (vect_determine_precisions_from_users): Likewise.
11603 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
11604 (vect_vect_recog_func_ptrs): Put over_widening first.
11606 (vect_pattern_recog): Call vect_determine_precisions.
11608 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
11610 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
11611 statements that have been replaced by further pattern statements.
11612 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
11614 2018-07-03 Richard Biener <rguenther@suse.de>
11616 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
11617 always set *dt. Dump vectype in vectype overload.
11618 * dumpfile.h (dump_gimple_expr): New function.
11619 (dump_gimple_expr_loc): Likewise.
11620 * dumpfile.c (dump_gimple_expr): New function.
11621 (dump_gimple_expr_loc): Likewise.
11623 2018-07-02 Jeff Law <law@redhat.com>
11625 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
11626 the H8/300, H8/300H and H8/S variants into a single pattern.
11627 (movhi_h8300, movqi_h8300hs): Similarly.
11628 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
11629 (QHI mode iterator): New.
11631 * config/h8300/h8300.md: Remove trailing whitespace.
11633 2018-07-02 Jim Wilson <jimw@sifive.com>
11635 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
11636 instead of emit_insn for interrupt returns.
11637 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
11638 (riscv_sret, riscv_uret): Likewise.
11640 2018-07-02 David Malcolm <dmalcolm@redhat.com>
11642 * pretty-print.c (selftest::test_pp_format): Move save and restore
11643 of quotes to class auto_fix_quotes, and add an instance.
11644 * selftest.c: Include "intl.h".
11645 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
11646 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
11647 * selftest.h (selftest::auto_fix_quotes): New class.
11649 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
11651 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
11652 (aarch64_sve_prepare_conditional_op): Remove.
11653 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
11654 Allow aarch64_simd_reg_or_zero as select operand; remove
11655 the aarch64_sve_prepare_conditional_op call.
11656 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
11657 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
11658 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
11659 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
11660 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
11661 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
11662 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
11663 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
11664 and a splitters to match all of the *_any patterns.
11665 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
11667 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
11668 (SVE_COND_FP_BINARY_REV): Remove.
11669 (sve_int_op_rev, sve_fp_op_rev): New.
11670 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
11671 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
11672 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
11673 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
11674 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
11675 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
11676 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
11677 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
11678 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
11680 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
11681 Remove match_dup 1 from the inner unspec.
11682 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
11684 * config/aarch64/aarch64.md (movprfx): New attr.
11685 (length): Default movprfx to 8.
11686 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
11687 (*madd<SVE_I>, *msub<SVE_I): Likewise.
11688 (*<su>mul<SVE_I>3_highpart): Likewise.
11689 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
11690 (*v<ASHIFT><SVE_I>3): Likewise.
11691 (*<su><MAXMIN><SVE_I>3): Likewise.
11692 (*<su><MAXMIN><SVE_F>3): Likewise.
11693 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
11694 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
11695 (*div<SVE_F>4): Likewise.
11697 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
11699 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
11702 2018-07-02 Richard Biener <rguenther@suse.de>
11704 PR tree-optimization/86363
11705 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
11706 memset argument refers to a non-variable address.
11708 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
11710 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
11711 {PLUS,MINUS}_EXPR code to...
11712 (adjust_symbolic_bound): ...here,
11713 (combine_bound): ...here,
11714 (set_value_range_with_overflow): ...and here.
11716 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
11718 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
11720 (extract_range_from_abs_expr): ...here.
11722 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
11724 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
11725 -fno-omit-frame-pointer when not optimizing.
11727 2018-07-02 Martin Liska <mliska@suse.cz>
11730 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
11731 (propagate_nothrow): Likewise.
11733 2018-07-02 Martin Liska <mliska@suse.cz>
11736 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
11738 2018-07-02 David Malcolm <dmalcolm@redhat.com>
11740 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
11741 function in r262149, changing "loc" param from source_location to
11742 const dump_location_t &.
11743 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
11744 declaration, as above.
11746 2018-07-01 Paul Koning <ni1d@arrl.net>
11748 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
11749 -munit-asm, -mgnu-asm, -mdec-asm.
11750 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
11751 (pdp11_output_labelref): New.
11752 (pdp11_output_def): New.
11753 (pdp11_output_addr_vec_elt): New.
11754 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
11755 %# and %@ format codes.
11756 (pdp11_option_override): New.
11757 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
11758 (pdp11_output_ident): New.
11759 (pdp11_asm_named_section): New.
11760 (pdp11_asm_init_sections): New.
11761 (pdp11_file_start): New.
11762 (pdp11_file_end): New.
11763 (output_ascii): Use .ascii/.asciz for -mdec-asm.
11764 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
11765 %o, like %c but octal.
11766 (pdp11_option_override): New.
11767 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
11769 (DATA_SECTION_ASM_OP): Ditto.
11770 (READONLY_DATA_SECTION_ASM_OP): New.
11771 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
11772 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
11773 (ASM_OUTPUT_LABELREF): Ditto.
11774 (ASM_OUTPUT_DEF): Ditto.
11775 (ASM_OUTPUT_EXTERNAL): New.
11776 (ASM_OUTPUT_SOURCE_FILENAME): New.
11777 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
11778 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
11779 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
11780 %# and %@ format codes.
11781 * config/pdp11/pdp11.opt (mgnu-asm): New.
11782 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
11783 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
11784 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
11786 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
11788 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
11789 dereferencing path[] beyond its length.
11791 (debug_all_paths): New.
11792 (rewire_first_differing_edge): New.
11793 (adjust_paths_after_duplication): New.
11794 (duplicate_thread_path): Call adjust_paths_after_duplication.
11796 (thread_through_all_blocks): Add new argument to
11797 duplicate_thread_path.
11799 2018-06-30 Jim Wilson <jimw@sifive.com>
11801 * config/riscv/predicates.md (p2m1_shift_operand): New.
11802 (high_mask_shift_operand): New.
11803 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
11804 pattern using p2m1_shift_operand.
11805 (lshsi3_zero_extend_3+2): New combiner pattern using
11806 high_mask_shift_operand.
11808 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
11810 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
11812 (vect_recog_rotate_pattern): ...here.
11813 (vect_convert_input): Try to insert casts of invariants in the
11815 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
11816 preheader to be empty.
11818 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
11820 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
11821 vector type. If given, install it in the new statement's
11822 STMT_VINFO_VECTYPE.
11823 (vect_element_precision): New function.
11824 (vect_unpromoted_value): New struct.
11825 (vect_unpromoted_value::vect_unpromoted_value): New function.
11826 (vect_unpromoted_value::set_op): Likewise.
11827 (vect_look_through_possible_promotion): Likewise.
11828 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
11829 (vect_widened_op_tree, vect_convert_input): Likewise.
11830 (vect_convert_inputs, vect_convert_output): Likewise.
11831 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
11832 to handle the optional cast of the multiplication result and
11833 vect_widened_op_tree to detect the widened multiplication itself.
11834 Do not require the input and output of promotion casts to have
11835 the same sign, but base the signedness of the operation on the
11836 input rather than the result. If the pattern includes two
11837 promotions, check that those promotions have the same sign.
11838 Do not restrict the MULT_EXPR handling to a double-width result;
11839 handle quadruple-width results and wider. Use vect_convert_inputs
11840 to convert the inputs to the common type.
11841 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
11842 to handle the optional cast of the ABS result. Also allow a sign
11843 change or a sign extension between the ABS and MINUS.
11844 Use vect_widened_op_tree to detect the widened subtraction and use
11845 vect_convert_inputs to convert the inputs to the common type.
11846 (vect_handle_widen_op_by_const): Delete.
11847 (vect_recog_widen_op_pattern): New function.
11848 (vect_recog_widen_mult_pattern): Use it.
11849 (vect_recog_widen_shift_pattern): Likewise.
11850 (vect_recog_widen_sum_pattern): Use
11851 vect_look_through_possible_promotion to handle the promoted
11854 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
11856 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
11857 the containing gimple_seq *.
11858 * gimple-iterator.h (gsi_for_stmt): Declare it.
11859 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
11860 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
11861 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
11862 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
11863 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
11865 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
11867 (vect_mark_pattern_stmts): ...here. Handle cases in which the
11868 statement being replaced is part of an existing pattern
11869 definition sequence, inserting the new pattern statements before
11871 (vect_pattern_recog_1): Don't return a bool. If the statement
11872 is already part of a pattern, instead apply pattern matching
11873 to the pattern definition statements. Don't clear the
11874 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
11875 (vect_pattern_recog): Don't break after the first match;
11876 continue processing the pattern definition statements instead.
11877 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
11879 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
11881 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
11882 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
11883 (vect_recog_widen_sum_pattern): Use it.
11885 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
11887 * tree-vect-loop.c (vectorizable_reduction): Assert that the
11888 phi is not a pattern statement and has not been replaced by
11889 a pattern statement.
11890 * tree-vect-patterns.c (type_conversion_p): Don't check
11891 STMT_VINFO_IN_PATTERN_P.
11892 (vect_recog_vector_vector_shift_pattern): Likewise.
11893 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
11894 the pattern statement rather than the original statement; check
11895 directly for a WIDEN_MULT_EXPR here.
11896 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
11897 vect_is_simple_use to return the pattern statement rather
11898 than the original statement; use is_pattern_stmt_p to check
11899 for such a pattern statement.
11900 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
11901 to return the pattern statement rather than the original statement;
11902 don't do the same transformation here.
11903 (vect_is_simple_use): If the defining statement has been replaced
11904 by a pattern statement, return the pattern statement instead.
11905 Remove the corresponding (local) transformation from the vectype
11908 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
11910 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
11911 end and default to null.
11912 * tree-vect-loop.c (vect_create_epilog_for_reduction)
11913 (vectorizable_reduction): Update calls accordingly, dropping the
11914 gimple ** argument if the passed-back statement isn't needed.
11915 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
11916 (vect_recog_rotate_pattern): Likewise.
11917 (vect_recog_mask_conversion_pattern): Likewise.
11918 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
11919 (vect_mask_constant_operand_p): Likewise.
11920 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
11921 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
11922 (get_group_load_store_type, get_load_store_type): Likewise.
11923 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
11924 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
11925 (vectorizable_conversion, vectorizable_assignment): Likewise.
11926 (vectorizable_shift, vectorizable_operation): Likewise.
11927 (vectorizable_store, vect_is_simple_cond): Likewise.
11928 (vectorizable_condition, vectorizable_comparison): Likewise.
11929 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
11930 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
11931 and move it to the end. Cope with null def_stmt_outs.
11933 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
11935 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
11937 2018-06-29 Jeff Law <law@redhat.com>
11939 * config/v850/v850.c (v850_legitimate_address_p): Handle large
11940 displacements for TARGET_V850E2V3 and newer.
11941 (TARGET_LRA_P): Remove. Defaults to LRA now.
11942 * config/v850/v850.md (sign23byte_load): Remove.
11943 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
11944 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
11946 2018-06-29 Martin Liska <mliska@suse.cz>
11949 * coverage.c (coverage_init): Mangle full path name.
11950 * doc/invoke.texi: Document the change.
11951 * gcov-io.c (mangle_path): New.
11952 * gcov-io.h (mangle_path): Likewise.
11953 * gcov.c (mangle_name): Use mangle_path for path mangling.
11955 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11957 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
11958 if starting source register is not even.
11960 2018-06-29 Martin Liska <mliska@suse.cz>
11962 PR tree-optimization/86263
11963 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
11964 Make edge redirection.
11966 2018-06-29 David Malcolm <dmalcolm@redhat.com>
11968 * dumpfile.c (dump_loc): Add indentation based on scope depth.
11969 (dump_scope_depth): New variable.
11970 (get_dump_scope_depth): New function.
11971 (dump_begin_scope): New function.
11972 (dump_end_scope): New function.
11973 * dumpfile.h (get_dump_scope_depth): New declaration.
11974 (dump_begin_scope): New declaration.
11975 (dump_end_scope): New declaration.
11976 (class auto_dump_scope): New class.
11977 (AUTO_DUMP_SCOPE): New macro.
11978 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
11981 2018-06-29 Richard Biener <rguenther@suse.de>
11983 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
11984 compute_all_dependences succeeds.
11985 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
11986 exceed --param loop-max-datarefs-for-datadeps.
11988 2018-06-29 Jakub Jelinek <jakub@redhat.com>
11990 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
11992 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
11995 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
11996 alternative 0 in preferred_for_speed attribute.
11998 2018-06-28 Paul Koning <ni1d@arrl.net>
12000 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
12001 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
12002 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
12003 * config/pdp11/pdp11.md: Correct "length" attribute calculation
12004 for shift insn patterns.
12006 2018-06-28 David Malcolm <dmalcolm@redhat.com>
12008 * cgraph.c (cgraph_node::get_body): Replace assignments to
12009 "dump_file" with calls to set_dump_file.
12010 * dumpfile.c (alt_dump_file): Make static, and group with...
12011 (alt_flags): ...this definition.
12012 (dumps_are_enabled): New variable.
12013 (refresh_dumps_are_enabled): New function.
12014 (set_dump_file): New function.
12015 (set_alt_dump_file): New function.
12016 (gcc::dump_manager::dump_start): Replace assignments to
12017 "dump_file" and "alt_dump_file" with calls to set_dump_file and
12019 (gcc::dump_manager::dump_finish): Likewise.
12020 * dumpfile.h (alt_dump_file): Delete decl.
12021 (dumps_are_enabled): New variable decl.
12022 (set_dump_file): New function decl.
12023 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
12025 * tree-nested.c (lower_nested_functions): Replace assignments to
12026 "dump_file" with calls to set_dump_file.
12028 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
12030 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
12031 goto_locus of each outgoing edge of each basic block.
12033 2018-06-28 Richard Biener <rguenther@suse.de>
12035 * dwarf2out.c (decl_scope_table): Remove.
12036 (push_decl_scope): Likewise.
12037 (pop_decl_scope): Likewise.
12038 (gen_type_die_for_member): Do not call push/pop_decl_scope.
12039 (gen_struct_or_union_type_die): Likewise.
12040 (gen_tagged_type_die): Likewise.
12041 (dwarf2out_init): Do not initialize decl_scope_table.
12042 (dwarf2out_c_finalize): Do not free it.
12044 2018-06-28 Richard Biener <rguenther@suse.de>
12046 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
12047 deciding whether to not re-use a DIE.
12049 2018-06-28 Richard Biener <rguenther@suse.de>
12051 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
12052 DW_AT_abstract_origin attribute.
12054 2018-06-28 Martin Liska <mliska@suse.cz>
12056 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
12057 Use newly introduced constants.
12058 * tree-switch-conversion.h (struct jump_table_cluster):
12059 Define max_ratio_for_size and max_ratio_for_speed.
12061 2018-06-28 Martin Liska <mliska@suse.cz>
12063 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
12064 Add new checking assert to catch invalid state.
12065 (jump_table_cluster::can_be_handled): Handle single case
12067 (jump_table_cluster::is_beneficial): Bail out for such case.
12068 (bit_test_cluster::find_bit_tests):
12069 Add new checking assert to catch invalid state.
12070 (bit_test_cluster::can_be_handled): Handle single case
12072 (bit_test_cluster::is_beneficial): Bail out for such case.
12073 (switch_decision_tree::analyze_switch_statement):
12076 2018-06-28 Martin Liska <mliska@suse.cz>
12078 * common.opt: Introduce -completion option.
12079 * gcc.c (driver_handle_option): Handle it.
12080 (driver::main): Print completions if completion
12082 * opt-suggestions.c (option_proposer::get_completions):
12084 (option_proposer::suggest_completion): Likewise.
12085 (option_proposer::find_param_completions): Likewise.
12086 (verify_autocompletions): Likewise.
12087 (test_completion_valid_options): Likewise.
12088 (test_completion_valid_params): Likewise.
12089 (in_completion_p): Likewise.
12090 (empty_completion_p): Likewise.
12091 (test_completion_partial_match): Likewise.
12092 (test_completion_garbage): Likewise.
12093 (opt_proposer_c_tests): Likewise.
12094 * opt-suggestions.h: Declare new functions.
12095 * opts.c (common_handle_option): Handle OPT__completion_.
12096 * selftest-run-tests.c (selftest::run_tests): Add
12097 opt_proposer_c_tests.
12098 * selftest.c (assert_str_startswith): New.
12099 * selftest.h (assert_str_startswith): Likewise.
12100 (opt_proposer_c_tests): New.
12101 (ASSERT_STR_STARTSWITH): Likewise.
12103 2018-06-28 Martin Liska <mliska@suse.cz>
12105 * Makefile.in: Add opt-suggestions.o.
12106 * gcc-main.c: Include opt-suggestions.h.
12107 * gcc.c (driver::driver): Likewise.
12108 (driver::~driver): Remove m_option_suggestions.
12109 (driver::build_option_suggestions): Moved to option_proposer.
12110 (driver::suggest_option): Likewise.
12111 (driver::handle_unrecognized_options): Use option_proposer.
12112 * gcc.h (class driver): Add new memver m_option_proposer.
12113 * opt-suggestions.c: New file.
12114 * opt-suggestions.h: New file.
12116 2018-06-28 Martin Liska <mliska@suse.cz>
12118 * vec.h (class auto_string_vec): New (moved from auto_argvec).
12119 (auto_string_vec::~auto_string_vec): Likewise.
12121 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
12123 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
12124 prevent_decl_creation_for_types fields up and add reset_location field.
12125 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
12126 statement if id->reset_location is true.
12127 (copy_edges_for_bb): Do not set goto_locus on the new edges if
12128 id->reset_location is true.
12129 (copy_phis_for_bb): Force input_location on the arguments if
12130 id->reset_location is true.
12131 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
12132 is set on the function to be inlined.
12134 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
12136 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
12138 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
12140 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
12141 registers for Pmode.
12142 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
12143 hard registers for the clobbered pseudo.
12145 2018-06-27 Paul Koning <ni1d@arrl.net>
12147 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
12148 mutually exclusive options.
12149 * config/pdp11/constraints.md (h): New constraint.
12150 (O): Update definition to match shift code generation.
12151 (D): New constraint.
12152 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
12154 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
12156 (output_jump): Change arguments.
12157 (pdp11_fixed_cc_regs): New function.
12158 (pdp11_cc_mode): Ditto.
12159 (pdp11_expand_shift): Ditto.
12160 (pdp11_assemble_shift): Ditto.
12161 (pdp11_small_shift): Ditto.
12162 (pdp11_branch_cost): Remove.
12163 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
12165 (pdp11_register_move_cost): Update for CC registers.
12166 (pdp11_rtx_costs): Add case for LSHIFTRT.
12167 (pdp11_output_jump): Add CCNZ mode conditional branches.
12168 (notice_update_cc_on_set): Remove.
12169 (pdp11_cc_mode): New function.
12170 (simple_memory_operand): Correct pre/post decrement case.
12171 (no_side_effect_operand): New function.
12172 (pdp11_regno_reg_class): Add CC_REGS class.
12173 (pdp11_fixed_cc_regs): New function.
12174 (pdp11_small_shift): New function.
12175 (pdp11_expand_shift): New function to expand shift insns.
12176 (pdp11_assemble_shift): New function to output shifts.
12177 (pdp11_branch_cost): Remove.
12178 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
12179 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
12180 (WCHAR_TYPE): Ditto.
12181 (PTRDIFF_TYPE): Ditto.
12182 (ADJUST_INSN_LENGTH): New macro.
12183 (FIXED_REGISTERS): Add CC registers.
12184 (CALL_USED_REGISTERS): Ditto.
12185 (reg_class): Ditto.
12186 (REG_CLASS_NAMES): Ditto.
12187 (REG_CLASS_CONTENTS): Ditto.
12188 (SELECT_CC_MODE): Use new function.
12189 (TARGET_FLAGS_REGNUM): New macro.
12190 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
12191 (cc0_reg_rtx): Remove.
12192 (CC_STATUS_MDEP): Remove.
12193 (CC_STATUS_MDEFP_INIT): Remove.
12194 (CC_IN_FPU): Remove.
12195 (NOTICE_UPDATE_CC): Remove.
12196 (REGISTER_NAMES): Add CC registers.
12197 (BRANCH_COST): Change to constant 1.
12198 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
12200 * config/pdp11/pdp11.opt (mbcopy): Remove.
12201 (mbcopy-builtin): Remove.
12202 (mbranch-cheap): Remove.
12203 (mbranch-expensive): Remove.
12204 * config/pdp11/predicates.md (expand_shift_operand): Update to
12205 match shift code generation.
12206 (ccnz_operator): New predicate.
12207 * doc/invoke.texi (PDP-11 Options): Remove deleted options
12208 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
12209 Remove non-existent option -mabshi, -mno-abshi. Document mutually
12211 * doc/md.texi (PDP-11): Document new D and h constraints. Update
12212 description of O constraint.
12214 2018-06-27 Jeff Law <law@redhat.com>
12215 Austin Law <austinklaw@gmail.com>
12217 * config/v850/v850.md (addsi3_set_flags): New pattern.
12218 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
12219 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
12220 (zero_extendhisi2_v850_set_flags): Likewise.
12221 (zero_extendqisi2_v850_set_flags): Likewise.
12222 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
12223 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
12224 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
12226 * config/v850/v850-protos.h (notice_update_cc): Remove.
12227 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
12228 (v850_print_operand): Handle 'D' and "d".
12229 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
12230 Add handling of arithmetic/logical operations compared against zero.
12231 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
12232 Do not look at v850_compare_op, instead get mode from last argument.
12233 (v850_gen_compare): Remove
12234 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
12235 after reload for prologue insns.
12236 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
12238 (construct_save_jarl): Likewise.
12239 (TARGET_FLAGS_REGNUM): Define.
12240 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
12241 (NOTICE_UPDATE_CC): Remove.
12242 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
12243 than cc0. Conditionalize on reload_completed.
12244 (cmpsi_insn, setfcc_insn): Likewise.
12245 (tst1 splitter): Turn into define_and_split which sets the flags
12247 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
12248 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
12249 (cstoresf4, cstoredf4): Clobber the flags.
12250 (cmpsi, cmpsf, cmpdf): Remove expanders.
12251 (setf_insn): Remove pattern.
12252 (addsi3): Turn into define_and_split which clobbers the flags after
12253 reload and a suitable pattern (addsi3_clobber_flags) for use after
12255 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
12256 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
12257 (ashrsi3, ashrsi3_v850e2): Likewise.
12258 (bins): Clobber the flags.
12259 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
12260 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
12261 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
12262 (call_value_internal_short, call_value_internal_long): Likewise.
12263 (callt_save_interrupt, callt_return_interrupt): Likewise.
12264 (save_interrupt, return_interrupt): Likewise.
12265 (callt_save_all_interrupt, save_all_interrupt): Likewise.
12266 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
12267 (restore_all_interrupt, _restore_all_interrupt): Likewise.
12268 (All FP comparisons): Only allow after reload has completed.
12270 (divh, divhu): Tweak output template.
12271 (branch_z_normal, branch_z_invert): Remove
12272 (branch_nz_normal, branch_nz_invert): Likewise.
12273 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
12275 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
12276 * config/v850/v850.c (notice_update_cc): Remove.
12277 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
12278 (CC_NO_CARRY): Likewise.
12279 (NOTICE_UPDATE_CC): Define to nothing.
12280 * config/v850/v850.md: Remove block comment on cc0 handling
12281 Remove "cc" attribute from all patterns. Remove cc_status handling
12282 from all patterns. Minor formatting fixes.
12284 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12286 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
12287 (cortex-a76.cortex-a55): Likewise.
12288 * config/aarch64/aarch64-tune.md: Regenerate.
12289 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
12290 cortex-a76.cortex-a55.
12292 2018-06-27 Jeff Law <law@redhat.com>
12294 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
12295 (MULTILIB_DIRNAMES): Similarly.
12297 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
12299 * gimple.h (gimple_return_retbnd): Delete.
12300 (gimple_return_set_retbnd): Likewise.
12301 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
12302 gimple_return_set_retbnd.
12303 * gimple-pretty-print.c (dump_gimple_return): Remove call to
12304 gimple_return_retbnd and adjust.
12305 * tree-inline.h (struct copy_body_data): Remove retbnd field.
12306 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
12307 Explicitly return NULL in a couple more cases. Move assertion
12308 on debug statements and remove unreachable code.
12309 (reset_debug_binding): Do not test id->retbnd.
12310 (expand_call_inline): Do not set it.
12312 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
12314 * configure.ac: Add --disable-gcov option.
12315 * configure: Regenerate.
12316 * Makefile.in: Honour @enable_gcov@.
12317 * doc/install.texi: Document --disable-gcov.
12319 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12321 * config/arm/arm-cpus.in (cortex-a76): New entry.
12322 (cortex-a76.cortex-a55): Likewise.
12323 * config/arm/arm-tables.opt: Regenerate.
12324 * config/arm/arm-tune.md: Likewise.
12325 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
12326 * doc/invoke.texi (ARM Options): Document cortex-a76 and
12327 cortex-a76.cortex-a55.
12329 2018-06-27 Tamar Christina <tamar.christina@arm.com>
12332 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
12334 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
12336 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
12338 (EPILOGUE_USES): Likewise.
12340 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
12342 * tree-inline.c (remap_location): New function extracted from...
12343 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
12344 (copy_phis_for_bb): ...here. Call remap_location.
12345 (copy_cfg_body): Adjust call to copy_edges_for_bb.
12347 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
12349 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
12350 unaligned vsx for 16B memset.
12352 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
12355 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
12356 ieee128_float_type_node to long_double_type_node unless
12357 TARGET_LONG_DOUBLE_128 is set.
12359 2018-06-26 David Malcolm <dmalcolm@redhat.com>
12361 * cfgloop.c (get_loop_location): Convert return type from
12362 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
12363 by implicit construction from rtx_insn *, and using
12364 dump_user_location_t::from_function_decl for the fallback case.
12365 * cfgloop.h (get_loop_location): Convert return type from
12366 location_t to dump_user_location_t.
12367 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
12368 dump_printf_loc to pass in a dump_location_t rather than a
12369 location_t, via the gimple stmt.
12370 * coverage.c (get_coverage_counts): Update calls to
12371 dump_printf_loc to pass in dump_location_t rather than a
12373 * doc/optinfo.texi (Dump types): Convert example of
12374 dump_printf_loc from taking "locus" to taking "insn". Update
12375 description of the "_loc" calls to cover dump_location_t.
12376 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
12378 (dump_user_location_t::dump_user_location_t): New constructors,
12379 from gimple *stmt and rtx_insn *.
12380 (dump_user_location_t::from_function_decl): New function.
12381 (dump_loc): Make static.
12382 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
12383 const dump_location_t &.
12384 (dump_generic_expr_loc): Delete.
12385 (dump_printf_loc): Convert param "loc" from location_t to
12386 const dump_location_t &.
12387 (selftest::test_impl_location): New function.
12388 (selftest::dumpfile_c_tests): New function.
12389 * dumpfile.h: Include "profile-count.h".
12390 (class dump_user_location_t): New class.
12391 (struct dump_impl_location_t): New struct.
12392 (class dump_location_t): New class.
12393 (dump_printf_loc): Convert 2nd param from source_location to
12394 const dump_location_t &.
12395 (dump_generic_expr_loc): Delete.
12396 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
12397 const dump_location_t &.
12398 * gimple-fold.c (fold_gimple_assign): Update call to
12399 dump_printf_loc to pass in a dump_location_t rather than a
12400 location_t, via the gimple stmt.
12401 (gimple_fold_call): Likewise.
12402 * gimple-loop-interchange.cc
12403 (loop_cand::analyze_iloop_reduction_var): Update for change to
12404 check_reduction_path.
12405 (tree_loop_interchange::interchange): Update for change to
12406 find_loop_location.
12407 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
12408 change in return-type of find_loop_location.
12409 (graphite_regenerate_ast_isl): Likewise.
12410 * graphite-optimize-isl.c (optimize_isl): Likewise.
12411 * graphite.c (graphite_transform_loops): Likewise.
12412 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
12413 pass in a dump_location_t rather than a location_t, via the
12415 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
12416 * ipa.c (walk_polymorphic_call_targets): Likewise.
12417 * loop-unroll.c (report_unroll): Convert "locus" param from
12418 location_t to dump_location_t.
12419 (decide_unrolling): Update for change to get_loop_location's
12421 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
12422 location_t to dump_user_location_t.
12423 (grid_find_single_omp_among_assignments_1): Updates calls to
12424 dump_printf_loc to pass in a dump_location_t rather than a
12425 location_t, via the gimple stmt.
12426 (grid_parallel_clauses_gridifiable): Convert "tloc" from
12427 location_t to dump_location_t. Updates calls to dump_printf_loc
12428 to pass in a dump_location_t rather than a location_t, via the
12430 (grid_inner_loop_gridifiable_p): Likewise.
12431 (grid_dist_follows_simple_pattern): Likewise.
12432 (grid_gfor_follows_tiling_pattern): Likewise.
12433 (grid_target_follows_gridifiable_pattern): Likewise.
12434 (grid_attempt_target_gridification): Convert initialization
12435 of local "grid" from memset to zero-initialization; FIXME: does
12436 this require C++11? Update call to dump_printf_loc to pass in a
12437 optinfo_location rather than a location_t, via the gimple stmt.
12438 * profile.c (read_profile_edge_counts): Updates call to
12439 dump_printf_loc to pass in a dump_location_t rather than a
12441 (compute_branch_probabilities): Likewise.
12442 * selftest-run-tests.c (selftest::run_tests): Call
12444 * selftest.h (dumpfile_c_tests): New decl.
12445 * tree-loop-distribution.c (pass_loop_distribution::execute):
12446 Update for change in return type of find_loop_location.
12447 * tree-parloops.c (parallelize_loops): Likewise.
12448 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
12449 "locus" from location_t to dump_user_location_t.
12450 (canonicalize_loop_induction_variables): Likewise.
12451 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
12452 for change in return type of find_loop_location.
12453 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
12454 to dump_printf_loc to pass in a dump_location_t rather than a
12455 location_t, via the stmt.
12456 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
12458 * tree-vect-loop-manip.c (find_loop_location): Convert return
12459 type from source_location to dump_user_location_t.
12460 (vect_do_peeling): Update for above change.
12461 (vect_loop_versioning): Update for change in type of
12463 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
12464 from location_t to dump_user_location_t.
12465 (vect_estimate_min_profitable_iters): Update for change in type
12467 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
12468 location_t to dump_location_t.
12469 (vect_slp_bb): Update for change in type of vect_location.
12470 * tree-vectorizer.c (vect_location): Convert from source_location
12471 to dump_user_location_t.
12472 (try_vectorize_loop_1): Update for change in vect_location's type.
12473 (vectorize_loops): Likewise.
12474 (increase_alignment): Likewise.
12475 * tree-vectorizer.h (vect_location): Convert from source_location
12476 to dump_user_location_t.
12477 (find_loop_location): Convert return type from source_location to
12478 dump_user_location_t.
12479 (check_reduction_path): Convert 1st param from location_t to
12480 dump_user_location_t.
12481 * value-prof.c (check_counter): Update call to dump_printf_loc to
12482 pass in a dump_user_location_t rather than a location_t; update
12483 call to error_at for change in type of "locus".
12484 (check_ic_target): Update call to dump_printf_loc to
12485 pass in a dump_user_location_t rather than a location_t, via the
12488 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
12490 * config/s390/s390.h (enum processor_flags): Do not use
12491 default tune parameter when -march was specified.
12493 2018-06-26 Jakub Jelinek <jakub@redhat.com>
12496 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
12497 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
12500 2018-06-26 Richard Biener <rguenther@suse.de>
12502 PR tree-optimization/86287
12504 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
12505 (vect_analyze_loop): Initialize n_stmts.
12507 2018-06-26 Richard Biener <rguenther@suse.de>
12509 PR middle-end/86271
12510 * fold-const.c (fold_convertible_p): Pointer extension
12513 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
12516 * dwarf2out.c (loc_list_has_views): Adjust comments.
12517 (dw_loc_list): Split single cross-partition range with
12520 2018-06-25 Jeff Law <law@redhat.com>
12522 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
12523 on -mbig-switch by default.
12525 * config/v850/predicates.md (const_float_1_operand): Fix match_code
12527 (const_float_0_operand): Remove unused predicate.
12528 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
12529 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
12530 (recipsf2): New expander. Original pattern now called
12532 (recipdf2, recipdf2_insn): Similarly.
12533 (rsqrtsf2, rsqrtsf2_insn): Similarly
12534 (rsqrtdf2, rsqrtdf2_insn): Similarly
12536 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
12538 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
12539 Simplify logic for FreeBSD (twice).
12541 2018-06-25 Martin Sebor <msebor@redhat.com>
12543 PR tree-optimization/86204
12544 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
12545 a strnlen result if it's less than the length of the string.
12547 2018-06-25 Martin Sebor <msebor@redhat.com>
12549 PR tree-optimization/85700
12550 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
12551 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
12552 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
12554 2018-06-25 Martin Sebor <msebor@redhat.com>
12556 * doc/extend.texi (Zero-length arrays): Update and clarify.
12558 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
12560 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
12561 added IEEE/IBM long double multilib support on PowerPC little
12562 endian Linux systems.
12563 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
12564 (MULTILIB_DEFAULTS): Likewise.
12565 * config/rs6000/rs6000.c (rs6000_option_override_internal):
12567 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
12568 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
12569 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
12571 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
12573 PR middle-end/86311
12574 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
12575 (REORDER_45): Likewise.
12577 2018-06-25 Jeff Law <law@redhat.com>
12579 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
12580 dividend to 32 bits. Adjust length.
12581 (udivmodhi4): Cleanup output template. Fix length.
12583 2018-06-25 Carl Love <cel@us.ibm.com>
12585 * config/rs6000/vsx.md: Change word selector to prefered location.
12587 2018-06-25 Richard Biener <rguenther@suse.de>
12589 PR tree-optimization/86304
12590 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
12591 epilogue-if-converted loops as well.
12593 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
12595 * lto-section-out.c (lto_begin_section): Do not print section
12596 name for noaddr and unnumbered dumps.
12598 2018-06-25 Richard Biener <rguenther@suse.de>
12600 * tree-vectorizer.h (struct vec_info_shared): New structure
12601 with parts split out from struct vec_info and loop_nest from
12602 struct _loop_vec_info.
12603 (struct vec_info): Adjust accordingly.
12604 (struct _loop_vec_info): Likewise.
12605 (LOOP_VINFO_LOOP_NEST): Adjust.
12606 (LOOP_VINFO_DATAREFS): Likewise.
12607 (LOOP_VINFO_DDRS): Likewise.
12608 (struct _bb_vec_info): Likewise.
12609 (BB_VINFO_DATAREFS): Likewise.
12610 (BB_VINFO_DDRS): Likewise.
12611 (struct _stmt_vec_info): Add dr_aux member.
12612 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
12613 (DR_MISALIGNMENT_UNINITIALIZED): New.
12614 (set_dr_misalignment): Adjust.
12615 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
12616 (vect_analyze_loop): Adjust prototype.
12617 (vect_analyze_loop_form): Likewise.
12618 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
12619 Compute dependences lazily.
12620 (vect_record_base_alignments): Use shared datarefs/ddrs.
12621 (vect_verify_datarefs_alignment): Likewise.
12622 (vect_analyze_data_refs_alignment): Likewise.
12623 (vect_analyze_data_ref_accesses): Likewise.
12624 (vect_analyze_data_refs): Likewise.
12625 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
12626 constructor parameter for shared part.
12627 (vect_analyze_loop_form): Pass in shared part and adjust.
12628 (vect_analyze_loop_2): Pass in storage for the number of
12629 stmts. Move loop nest finding to the caller. Compute
12631 (vect_analyze_loop): Pass in shared part.
12632 (vect_transform_loop): Verify shared datarefs are unchanged.
12633 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
12634 constructor parameter for shared part.
12635 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
12636 (vect_slp_bb): Verify shared datarefs are unchanged before
12638 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
12640 (new_stmt_vec_info): Initialize DR_AUX misalignment to
12641 DR_MISALIGNMENT_UNINITIALIZED.
12642 * tree-vectorizer.c (vec_info::vec_info): Add constructor
12643 parameter for shared part.
12644 (vec_info::~vec_info): Adjust.
12645 (vec_info_shared::vec_info_shared): New.
12646 (vec_info_shared::~vec_info_shared): Likewise.
12647 (vec_info_shared::save_datarefs): Likewise.
12648 (vec_info_shared::check_datarefs): Likewise.
12649 (try_vectorize_loop_1): Construct shared part live for analyses
12650 of a single loop for multiple vector sizes.
12651 * tree-parloops.c (gather_scalar_reductions): Adjust.
12653 2018-06-25 Richard Biener <rguenther@suse.de>
12655 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
12656 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
12657 (vect_analyze_data_refs): Remove similar code from here and
12658 simplify accordingly.
12660 2018-06-25 Richard Biener <rguenther@suse.de>
12662 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
12663 for reverse storage order accesses rather than asserting
12664 they cannot happen here.
12666 2018-06-25 Tom de Vries <tdevries@suse.de>
12669 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
12670 Use data16 instead of .byte for insn prefix.
12672 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
12675 * parser.c (make_char_string_pack): Pass this literal chars
12676 through cpp_interpret_string.
12677 (cp_parser_userdef_numeric_literal): Check the result of
12678 make_char_string_pack.
12680 2018-06-24 Maya Rashish <coypu@sdf.org>
12682 * ginclude/stddef.h: Simplify conditions around avoiding
12683 re-definition of __size_t.
12685 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
12687 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
12688 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
12690 2018-06-22 Maya Rashish <coypu@sdf.org>
12692 * doc/invoke.texi (mno-fancy-math-387): Update for changes
12693 made to OpenBSD and NetBSD through the years.
12695 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
12697 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
12698 behavior of vec_pack (vector double, vector double) to match
12699 behavior of vec_float2 (vector double, vector double).
12701 2018-06-22 Olivier Hainque <hainque@adacore.com>
12703 * gimplify.c (gimplify_function_tree): Prevent creation
12704 of a trampoline for the address of the current function
12705 passed to entry/exit instrumentation hooks.
12707 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
12710 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
12713 2018-06-22 Martin Liska <mliska@suse.cz>
12715 PR tree-optimization/86263
12716 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
12717 Bail out if is_enabled is false.
12718 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
12720 (jump_table_cluster::is_enabled): New function.
12722 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
12724 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
12725 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
12726 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
12727 (lto_input_ts_binfo_tree_pointers): Likewise.
12728 * tree-streamer-out.c (streamer_write_tree_bitfields,
12729 write_ts_binfo_tree_pointers): Likewise.
12730 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
12732 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
12734 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
12736 2018-06-22 Martin Liska <mliska@suse.cz>
12738 * symbol-summary.h (get): Make it pure and inline move
12739 functionality from ::get function.
12740 (get): Remove and inline into ::get and ::get_create.
12741 (get_create): Move code from ::get function.
12743 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12746 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
12747 -x assembler-with-cpp.
12749 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12751 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
12752 _FILE_OFFSET_BITS=64 for C++.
12754 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
12756 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
12757 conversion insn that shows up when pr85657-3.c is compiled using
12758 IEEE 128-bit long double.
12759 (neg<mode>2_internal): Use the correct mode to check whether the
12760 mode is IBM extended.
12761 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
12762 multiply and divide external functions from being created more
12765 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
12767 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
12769 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
12770 the edge can be forwarded.
12771 (cfg_layout_merge_blocks): Likewise.
12773 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
12775 * except.c (finish_eh_generation): Commit edge insertions only after
12776 the EH edges have been redirected from post-landing to landing pads.
12778 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
12780 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
12781 create_tmp_var_for to create the FRAME decl.
12782 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
12784 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
12786 * tree-inline.c (copy_edges_for_bb): Minor tweak.
12787 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
12788 debug statement when resetting its value.
12789 (expand_call_inline): Copy the locus of the call onto the assignment
12790 of the return value, if any. Use local variable in more cases.
12792 2018-06-21 Martin Liska <mliska@suse.cz>
12794 * ipa-pure-const.c (propagate_nothrow): Use
12795 funct_state_summaries->get.
12796 (dump_malloc_lattice): Likewise.
12797 (propagate_malloc): Likewise.
12799 2018-06-21 Richard Biener <rguenther@suse.de>
12801 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
12802 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
12803 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
12805 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
12806 BLOCK_ABSTRACT_ORIGIN unconditionally.
12808 2018-06-21 David Malcolm <dmalcolm@redhat.com>
12810 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
12812 * ipa-reference.c (ipa_reference_c_finalize): Delete
12813 ipa_ref_opt_sum_summaries and set it to NULL.
12815 2018-06-21 Tom de Vries <tdevries@suse.de>
12817 PR tree-optimization/85859
12818 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
12819 test with comment from bb_no_side_effects_p.
12821 2018-06-21 Richard Biener <rguenther@suse.de>
12823 PR tree-optimization/86232
12824 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
12825 max for constant niter.
12827 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
12829 * config/aarch64/aarch64-simd.md
12830 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
12832 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
12834 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
12835 Make opernads of the unspec commutative.
12837 2018-06-21 Richard Biener <rguenther@suse.de>
12839 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
12840 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
12841 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
12842 (vect_analyze_data_ref_dependence): Re-order checks to deal with
12844 (vect_record_base_alignments): Do not record base alignment
12845 for gathers or scatters.
12846 (vect_compute_data_ref_alignment): Drop return value that is always
12847 true. Bail out early for gathers or scatters.
12848 (vect_enhance_data_refs_alignment): Bail out early for gathers
12850 (vect_find_same_alignment_drs): Likewise.
12851 (vect_analyze_data_refs_alignment): Remove dead code.
12852 (vect_slp_analyze_and_verify_node_alignment): Likewise.
12853 (vect_analyze_data_refs): For possible gathers or scatters do
12854 not create an alternate DR, just check their possible validity
12855 and mark them. Adjust DECL_NONALIASED handling to not rely
12856 on DR_BASE_ADDRESS.
12857 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
12858 update inits of gathers or scatters.
12859 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
12860 Also copy gather/scatter flag to pattern vinfo.
12862 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
12864 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
12865 behavior of vec_packsu (vector unsigned long long, vector unsigned
12866 long long) to match behavior of vec_packs with same signature.
12868 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
12869 Thomas Schwinge <thomas@codesourcery.com>
12870 Cesar Philippidis <cesar@codesourcery.com>
12872 * gimplify.c (gimplify_scan_omp_clauses): Add support for
12873 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
12874 (gimplify_adjust_omp_clauses): Likewise.
12875 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
12876 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
12877 (gimplify_omp_target_update): Update handling of acc update and
12879 * omp-low.c (install_var_field): Remove unused parameter
12880 base_pointers_restrict.
12881 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
12882 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
12884 (omp_target_base_pointers_restrict_p): Delete.
12885 (scan_omp_target): Update call to scan_sharing_clauses.
12886 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
12888 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
12889 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
12890 (convert_local_omp_clauses): Likewise.
12891 * tree-pretty-print.c (dump_omp_clause): Likewise.
12892 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
12894 (omp_clause_code_name): Likewise.
12896 2018-06-20 Jakub Jelinek <jakub@redhat.com>
12899 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
12902 PR tree-optimization/86231
12903 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
12904 anti-range don't overwrite *vr0min before using it to compute *vr0max.
12906 2018-06-20 Tom de Vries <tdevries@suse.de>
12908 PR tree-optimization/86097
12909 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
12910 iv type if signedness of iv type is not the same as that of *nit.
12912 2018-06-20 Jakub Jelinek <jakub@redhat.com>
12914 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
12915 EDGE_EH edges, verify they are all EDGE_EH.
12917 2018-06-20 Maya Rashish <coypu@sdf.org>
12919 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
12921 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12923 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
12924 * config/aarch64/aarch64.c (xgene1_tunings): Add
12925 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
12926 (aarch64_mode_valid_for_sched_fusion_p):
12927 Allow 16-byte modes.
12928 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
12929 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
12931 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
12933 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
12934 * config/aarch64/iterators.md (VQ2): New mode iterator.
12936 2018-06-20 Martin Liska <mliska@suse.cz>
12938 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
12939 Change default ratio from 10 to 8.
12941 2018-06-20 Martin Liska <mliska@suse.cz>
12943 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
12945 (bit_test_cluster::find_bit_tests): Likewise.
12946 (switch_decision_tree::analyze_switch_statement): Find clusters.
12947 * tree-switch-conversion.h (struct jump_table_cluster): Document
12950 2018-06-20 Martin Liska <mliska@suse.cz>
12952 * tree-switch-conversion.c (switch_conversion::collect):
12953 Record m_uniq property.
12954 (switch_conversion::expand): Bail out for special conditions.
12955 (group_cluster::~group_cluster): New.
12956 (group_cluster::group_cluster): Likewise.
12957 (group_cluster::dump): Likewise.
12958 (jump_table_cluster::emit): New.
12959 (switch_decision_tree::fix_phi_operands_for_edges): New.
12960 (struct case_node): Remove struct.
12961 (jump_table_cluster::can_be_handled): New.
12962 (case_values_threshold): Moved to header.
12963 (reset_out_edges_aux): Likewise.
12964 (jump_table_cluster::is_beneficial): New.
12965 (bit_test_cluster::can_be_handled): Likewise.
12966 (add_case_node): Remove.
12967 (bit_test_cluster::is_beneficial): New.
12968 (case_bit_test::cmp): New.
12969 (bit_test_cluster::emit): New.
12970 (expand_switch_as_decision_tree_p): Remove.
12971 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
12972 (fix_phi_operands_for_edge): Likewise.
12973 (switch_decision_tree::analyze_switch_statement): New.
12974 (compute_cases_per_edge): Move ...
12975 (switch_decision_tree::compute_cases_per_edge): ... here.
12976 (try_switch_expansion): Likewise.
12977 (switch_decision_tree::try_switch_expansion): Likewise.
12978 (record_phi_operand_mapping): Likewise.
12979 (switch_decision_tree::record_phi_operand_mapping): Likewise.
12980 (emit_case_decision_tree): Likewise.
12981 (switch_decision_tree::emit): Likewise.
12982 (balance_case_nodes): Likewise.
12983 (switch_decision_tree::balance_case_nodes): Likewise.
12984 (dump_case_nodes): Likewise.
12985 (switch_decision_tree::dump_case_nodes): Likewise.
12986 (emit_jump): Likewise.
12987 (switch_decision_tree::emit_jump): Likewise.
12988 (emit_cmp_and_jump_insns): Likewise.
12989 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
12990 (emit_case_nodes): Likewise.
12991 (switch_decision_tree::emit_case_nodes): Likewise.
12992 (conditional_probability): Remove.
12993 * tree-switch-conversion.h (enum cluster_type): New.
12995 (struct cluster): Likewise.
12996 (cluster::cluster): Likewise.
12997 (struct simple_cluster): Likewise.
12998 (simple_cluster::simple_cluster): Likewise.
12999 (struct group_cluster): Likewise.
13000 (struct jump_table_cluster): Likewise.
13001 (struct bit_test_cluster): Likewise.
13002 (struct min_cluster_item): Likewise.
13003 (struct case_tree_node): Likewise.
13004 (case_tree_node::case_tree_node): Likewise.
13005 (jump_table_cluster::case_values_threshold): Likewise.
13006 (struct case_bit_test): Likewise.
13007 (struct switch_decision_tree): Likewise.
13008 (struct switch_conversion): Likewise.
13009 (switch_decision_tree::reset_out_edges_aux): Likewise.
13011 2018-06-20 Martin Liska <mliska@suse.cz>
13013 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
13014 (hoist_edge_and_branch_if_true): Likewise.
13015 (expand_switch_using_bit_tests_p): Likewise.
13016 (struct case_bit_test): Likewise.
13017 (case_bit_test_cmp): Likewise.
13018 (emit_case_bit_tests): Likewise.
13019 (switch_conversion::switch_conversion): New class.
13020 (struct switch_conv_info): Remove old struct.
13021 (collect_switch_conv_info): More to ...
13022 (switch_conversion::collect): ... this.
13023 (check_range): Likewise.
13024 (switch_conversion::check_range): Likewise.
13025 (check_all_empty_except_final): Likewise.
13026 (switch_conversion::check_all_empty_except_final): Likewise.
13027 (check_final_bb): Likewise.
13028 (switch_conversion::check_final_bb): Likewise.
13029 (create_temp_arrays): Likewise.
13030 (switch_conversion::create_temp_arrays): Likewise.
13031 (free_temp_arrays): Likewise.
13032 (gather_default_values): Likewise.
13033 (switch_conversion::gather_default_values): Likewise.
13034 (build_constructors): Likewise.
13035 (switch_conversion::build_constructors): Likewise.
13036 (constructor_contains_same_values_p): Likewise.
13037 (switch_conversion::contains_same_values_p): Likewise.
13038 (array_value_type): Likewise.
13039 (switch_conversion::array_value_type): Likewise.
13040 (build_one_array): Likewise.
13041 (switch_conversion::build_one_array): Likewise.
13042 (build_arrays): Likewise.
13043 (switch_conversion::build_arrays): Likewise.
13044 (gen_def_assigns): Likewise.
13045 (switch_conversion::gen_def_assigns): Likewise.
13046 (prune_bbs): Likewise.
13047 (switch_conversion::prune_bbs): Likewise.
13048 (fix_phi_nodes): Likewise.
13049 (switch_conversion::fix_phi_nodes): Likewise.
13050 (gen_inbound_check): Likewise.
13051 (switch_conversion::gen_inbound_check): Likewise.
13052 (process_switch): Use the newly created class.
13053 (switch_conversion::expand): New.
13054 (switch_conversion::~switch_conversion): New.
13055 * tree-switch-conversion.h: New file.
13057 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
13059 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
13060 tree-vect-patterns.c.
13061 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
13062 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
13063 (vect_recog_sad_pattern): Likewise.
13064 (vect_recog_widen_sum_pattern): Likewise.
13065 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
13066 (vect_recog_widen_shift_pattern): Remove the type_in argument.
13067 (vect_recog_rotate_pattern): Likewise.
13068 (vect_recog_mult_pattern): Likewise.
13069 (vect_recog_vector_vector_shift_pattern): Likewise.
13070 (vect_recog_divmod_pattern): Likewise.
13071 (vect_recog_mixed_size_cond_pattern): Likewise.
13072 (vect_recog_bool_pattern): Likewise.
13073 (vect_recog_mask_conversion_pattern): Likewise.
13074 (vect_try_gather_scatter_pattern): Likewise.
13075 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
13076 (vect_recog_over_widening_pattern): Likewise.
13077 (vect_recog_gather_scatter_pattern): Likewise.
13078 (vect_recog_func_ptr): Move from tree-vectorizer.h
13079 (vect_vect_recog_func_ptrs): Move further down the file.
13080 (vect_recog_func): Likewise. Remove the third argument.
13081 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
13082 (vect_pattern_recog_1): Expect the pattern function to do any
13083 necessary target tests. Also expect it to provide a vector type.
13084 Remove the type_in handling.
13086 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
13088 * tree-vect-patterns.c (vect_pattern_detected): New function.
13089 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
13090 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
13091 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
13092 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
13093 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
13094 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
13095 (vect_recog_mask_conversion_pattern)
13096 (vect_try_gather_scatter_pattern): Likewise.
13098 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
13100 * tree-vect-patterns.c (vect_get_internal_def): New function.
13101 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
13102 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
13103 (search_type_for_mask_1): Use it.
13105 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
13107 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
13108 redundant WIDEN_SUM_EXPR handling.
13109 (vect_recog_sad_pattern): Likewise.
13111 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
13113 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
13114 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
13115 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
13116 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
13117 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
13119 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
13121 * tree-vect-stmts.c (vectorizable_call): Make sure that we
13122 use the stmt_vec_info of the original bb statement for the
13123 new zero assignment, even if the call is part of a pattern.
13125 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
13127 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
13128 that the sequence is attached to the original statement rather
13129 than the pattern statement.
13130 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
13131 PATTERN_DEF_SEQ from the original statement rather than
13132 the main pattern statement.
13133 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
13134 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
13135 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
13137 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
13139 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
13140 definition statements before the early exit for statements that aren't
13142 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
13144 (vect_transform_loop): ...here. Process pattern definition
13145 statements without first checking whether the main pattern
13146 statement is live or relevant.
13148 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
13150 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
13151 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
13153 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
13155 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
13156 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
13157 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
13158 (expand_block_compare): Change select_block_compare_mode call.
13159 (expand_strncmp_align_check): Use new functions, fix comment.
13160 (emit_final_str_compare_gpr): New function.
13161 (expand_strn_compare): Refactor and clean up code.
13162 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
13164 2018-06-19 Tony Reix <tony.reix@atos.com>
13165 Damien Bergamini <damien.bergamini@atos.com>
13166 David Edelsohn <dje.gcc@gmail.com>
13168 * collect2.c (static_obj): New variable.
13169 (static_libs): New variable.
13170 (is_in_list): Uncomment declaration.
13171 (main): Track AIX libraries linked statically.
13172 (is_in_list): Uncomment definition.
13173 (scan_prog_file): Don't add AIX shared libraries initializer
13174 to constructor list if linking statically.
13176 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
13178 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
13180 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
13182 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
13184 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
13187 2018-06-19 Martin Liska <mliska@suse.cz>
13189 * config/i386/i386.c (ix86_can_inline_p): Do not use
13190 ipa_fn_summaries::get_create.
13191 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
13193 (devirtualization_time_bonus): Likewise.
13194 (ipcp_propagate_stage): Likewise.
13195 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
13196 (edge_set_predicate): Likewise.
13197 (evaluate_conditions_for_known_args): Likewise.
13198 (evaluate_properties_for_edge): Likewise.
13199 (ipa_call_summary::reset): Tranform to ...
13200 (ipa_call_summary::~ipa_call_summary): ... this.
13201 (ipa_fn_summary::reset): Transform to ...
13202 (ipa_fn_summary::~ipa_fn_summary): ... this.
13203 (ipa_fn_summary_t::remove): Rename to ...
13204 (ipa_fn_summary_t::remove_callees): ... this.
13205 (ipa_fn_summary_t::duplicate): Use placement new
13206 instead of memory copy.
13207 (ipa_call_summary_t::duplicate): Likewise.
13208 (ipa_call_summary_t::remove): Remove.
13209 (dump_ipa_call_summary): Change get_create to get.
13210 (ipa_dump_fn_summary): Dump only when summary exists.
13211 (analyze_function_body): Use symbol_summary::get instead
13213 (compute_fn_summary): Likewise.
13214 (estimate_edge_devirt_benefit): Likewise.
13215 (estimate_edge_size_and_time): Likewise.
13216 (inline_update_callee_summaries): Likewise.
13217 (remap_edge_change_prob): Likewise.
13218 (remap_edge_summaries): Likewise.
13219 (ipa_merge_fn_summary_after_inlining): Likewise.
13220 (write_ipa_call_summary): Likewise.
13221 (ipa_fn_summary_write): Likewise.
13222 (ipa_free_fn_summary): Likewise.
13223 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
13224 (struct ipa_call_summary): Likewise.
13225 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
13227 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
13228 (estimate_size_after_inlining): Likewise.
13229 (estimate_growth): Likewise.
13230 (growth_likely_positive): Likewise.
13231 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
13232 (inline_call): Likewise.
13233 * ipa-inline.c (caller_growth_limits): Likewise.
13234 (can_inline_edge_p): Likewise.
13235 (can_inline_edge_by_limits_p): Likewise.
13236 (compute_uninlined_call_time): Likewise.
13237 (compute_inlined_call_time): Likewise.
13238 (want_inline_small_function_p): Likewise.
13239 (edge_badness): Likewise.
13240 (update_caller_keys): Likewise.
13241 (update_callee_keys): Likewise.
13242 (inline_small_functions): Likewise.
13243 (inline_to_all_callers_1): Likewise.
13244 (dump_overall_stats): Likewise.
13245 (early_inline_small_functions): Likewise.
13246 (early_inliner): Likewise.
13247 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
13248 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
13249 * ipa-pure-const.c (malloc_candidate_p): Likewise.
13250 * ipa-split.c (execute_split_functions): Likewise.
13251 * symbol-summary.h: Likewise.
13252 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
13254 2018-06-19 Richard Biener <rguenther@suse.de>
13256 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
13257 (vectorize_loops): ... here. Fix dbgcnt handling.
13258 (try_vectorize_loop): Wrap try_vectorize_loop_1.
13260 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
13263 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
13264 ieee128 argument takes up only one (vector) register, not two (floating
13267 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
13269 * gimplify.c (gimplify_init_constructor): Really never clear for an
13270 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
13272 2018-06-19 Richard Biener <rguenther@suse.de>
13274 PR tree-optimization/86179
13275 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
13276 after failed recognition.
13278 2018-06-18 Martin Sebor <msebor@redhat.com>
13280 PR middle-end/85602
13281 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
13282 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
13283 Handle integer subtraction.
13284 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
13285 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
13287 2018-06-18 David Malcolm <dmalcolm@redhat.com>
13289 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
13290 param from rtx to rtx_insn *.
13291 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
13293 (frv_ifcvt_modify_insn): Likwise.
13294 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
13295 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
13296 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
13297 as_a <rtx_insn *> cast to local "unprotected_region" once
13298 it's been established that it's not NULL or pc_rtx.
13299 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
13300 param "sethi" from rtx to rtx_insn *.
13301 (nds32_group_float_insns): Likewise for param "insn".
13302 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
13304 (vax_output_int_subtract): Likewise.
13305 * config/vax/vax.c (vax_output_int_add): Likewise for param
13307 (vax_output_int_subtract): Likewise.
13308 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
13309 (emit_pattern_after): Likewise for param "after".
13310 (emit_insn_after): Likewise.
13311 (emit_jump_insn_after): Likewise.
13312 (emit_call_insn_after): Likewise.
13313 (emit_debug_insn_after): Likewise.
13314 (emit_pattern_before): Likewise for param "before".
13315 (emit_insn_before): Likewise.
13316 (emit_jump_insn_before): Likewise.
13317 * final.c (get_insn_template): Likewise for param "insn", removing
13319 * output.h (get_insn_template): Likewise for 2nd param.
13320 * rtl.h (emit_insn_before): Likewise.
13321 (emit_jump_insn_before): Likewise.
13322 (emit_debug_insn_before_noloc): Likewise.
13323 (emit_insn_after): Likewise.
13324 (emit_jump_insn_after): Likewise.
13325 (emit_call_insn_after): Likewise.
13326 (emit_debug_insn_after): Likewise.
13327 (set_insn_deleted): Likewise for param.
13329 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
13332 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
13333 floating point modes, so that IFmode is numerically greater than
13334 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
13335 to declare the ordering. This prevents IFmode from being
13336 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
13337 machine. Include rs6000-modes.h to share the fractional values
13338 between genmodes* and the rest of the compiler.
13339 (IFmode): Likewise.
13340 (KFmode): Likewise.
13341 (TFmode): Likewise.
13342 * config/rs6000/rs6000-modes.h: New file.
13343 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
13344 meaning of rs6000_long_double_size so that 126..128 selects an
13345 appropriate 128-bit floating point type.
13346 (rs6000_option_override_internal): Likewise.
13347 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
13348 (TARGET_LONG_DOUBLE_128): Change the meaning of
13349 rs6000_long_double_size so that 126..128 selects an appropriate
13350 128-bit floating point type.
13351 (LONG_DOUBLE_TYPE_SIZE): Update comment.
13352 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
13353 source and destination to match the standard usage.
13354 (truncifkf2): Likewise.
13355 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
13356 ISA 2.07 to use an explicit clobber, instead of passing in a
13358 (copysign<mode>3_soft): Likewise.
13360 2018-06-18 David Malcolm <dmalcolm@redhat.com>
13362 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
13363 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
13364 (vect_slp_analyze_instance_dependence): Likewise.
13365 (vect_enhance_data_refs_alignment): Likewise.
13366 (vect_analyze_data_refs_alignment): Likewise.
13367 (vect_slp_analyze_and_verify_instance_alignment
13368 (vect_analyze_data_ref_accesses): Likewise.
13369 (vect_prune_runtime_alias_test_list): Likewise.
13370 (vect_analyze_data_refs): Likewise.
13371 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
13372 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13373 (vect_analyze_scalar_cycles_1): Likewise.
13374 (vect_get_loop_niters): Likewise.
13375 (vect_analyze_loop_form_1): Likewise.
13376 (vect_update_vf_for_slp): Likewise.
13377 (vect_analyze_loop_operations): Likewise.
13378 (vect_analyze_loop): Likewise.
13379 (vectorizable_induction): Likewise.
13380 (vect_transform_loop): Likewise.
13381 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
13382 * tree-vect-slp.c (vect_analyze_slp): Likewise.
13383 (vect_make_slp_decision): Likewise.
13384 (vect_detect_hybrid_slp): Likewise.
13385 (vect_slp_analyze_operations): Likewise.
13386 (vect_slp_bb): Likewise.
13387 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
13388 (vectorizable_bswap): Likewise.
13389 (vectorizable_call): Likewise.
13390 (vectorizable_simd_clone_call): Likewise.
13391 (vectorizable_conversion): Likewise.
13392 (vectorizable_assignment): Likewise.
13393 (vectorizable_shift): Likewise.
13394 (vectorizable_operation): Likewise.
13395 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
13397 2018-06-18 Martin Sebor <msebor@redhat.com>
13399 PR tree-optimization/81384
13400 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
13401 * builtins.c (expand_builtin_strnlen): New function.
13402 (expand_builtin): Call it.
13403 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
13404 * builtins.def (BUILT_IN_STRNLEN): New.
13405 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
13406 Warn for bounds in excess of maximum object size.
13407 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
13408 single-value ranges. Handle strnlen.
13409 (handle_builtin_strlen): Handle strnlen.
13410 (strlen_check_and_optimize_stmt): Same.
13411 * doc/extend.texi (Other Builtins): Document strnlen.
13413 2018-06-18 Maya Rashish <coypu@sdf.org>
13415 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
13416 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
13417 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
13419 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
13421 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
13423 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13425 * tree.c (escaped_string::escape): Replace cast to char * by
13426 const_cast<char *> (unescaped).
13428 2018-06-18 Nick Clifton <nickc@redhat.com>
13431 * tree.c (escaped_string): New class. Converts an unescaped
13432 string into its escaped equivalent.
13433 (warn_deprecated_use): Use the new class to convert the
13434 deprecation message, if present.
13435 (test_escaped_strings): New self test.
13436 (test_c_tests): Add test_escaped_strings.
13437 * doc/extend.texi (deprecated): Add a note that the
13438 deprecation message is affected by the -fmessage-length
13439 option, and that control characters will be escaped.
13440 (#pragma GCC error): Document this pragma.
13441 (#pragma GCC warning): Likewise.
13442 * doc/invoke.texi (-fmessage-length): Document this option's
13443 effect on the #warning and #error preprocessor directives and
13444 the deprecated attribute.
13446 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
13448 * tree.c (decl_value_expr_lookup): Revert latest change.
13449 (decl_value_expr_insert): Likewise.
13451 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
13453 * gimplify.c (nonlocal_vlas): Delete.
13454 (nonlocal_vla_vars): Likewise.
13455 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
13457 (gimplify_body): Do not create and destroy nonlocal_vlas.
13458 * tree-nested.c: Include diagnostic.h.
13459 (use_pointer_in_frame): Tweak.
13460 (lookup_field_for_decl): Add assertion and declare the transformation.
13461 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
13462 internal error when the reference is in a wrong context. Do not
13463 create a debug decl by default.
13464 (note_nonlocal_block_vlas): Delete.
13465 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
13466 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
13467 create a debug decl by default.
13468 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
13469 call is in a wrong context.
13470 (fixup_vla_decls): New function.
13471 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
13472 debug variables were created.
13473 * tree.c (decl_value_expr_lookup): Add checking assertion.
13474 (decl_value_expr_insert): Likewise.
13476 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
13478 PR middle-end/82479
13479 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
13480 * tree-scalar-evolution.c (interpret_expr): Likewise.
13481 (expression_expensive_p): Likewise.
13482 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
13483 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
13484 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
13485 (ssa_defined_by_minus_one_stmt_p): New.
13487 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
13489 PR middle-end/64946
13490 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
13491 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
13492 * dojump.c (do_jump): Likewise.
13493 * expr.c (expand_expr_real_2): Check operand type's sign.
13494 * fold-const.c (const_unop): Handle ABSU_EXPR.
13495 (fold_abs_const): Likewise.
13496 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
13497 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
13498 (strip_sign_op_1): Likesise.
13499 * match.pd: Add new pattern to generate ABSU_EXPR.
13500 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
13501 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
13502 * tree-eh.c (operation_could_trap_helper_p): Likewise.
13503 * tree-inline.c (estimate_operator_cost): Likewise.
13504 * tree-pretty-print.c (dump_generic_node): Likewise.
13505 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
13506 * tree.def (ABSU_EXPR): New.
13508 2018-06-16 Jakub Jelinek <jakub@redhat.com>
13510 PR middle-end/86095
13511 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
13512 documented as preserved for backward compatibility only.
13513 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
13515 PR rtl-optimization/86108
13516 * bb-reorder.c (create_forwarder_block): Renamed to ...
13517 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
13518 jump from new landing pad to the second part.
13519 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
13522 2018-06-15 Jakub Jelinek <jakub@redhat.com>
13524 PR middle-end/85878
13525 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
13526 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
13527 Only call store_expr for halves if the mode is the same.
13529 PR middle-end/86123
13530 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
13531 Fix up comment formatting.
13533 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
13535 * typed-splay-tree.h (typed_splay_tree::remove): New function.
13536 (typed_splay_tree::closure,
13537 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
13538 (typed_splay_tree::typed_splay_tree,
13539 typed_splay_tree::operator =): Declared private.
13540 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
13541 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
13542 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
13543 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
13544 typed_splay_tree::splay_tree_splay,
13545 typed_splay_tree::splay_tree_foreach_helper,
13546 typed_splay_tree::splay_tree_insert,
13547 typed_splay_tree::splay_tree_remove,
13548 typed_splay_tree::splay_tree_lookup,
13549 typed_splay_tree::splay_tree_predecessor,
13550 typed_splay_tree::splay_tree_successor,
13551 typed_splay_tree::splay_tree_min,
13552 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
13553 (typed_splay_tree::root, typed_splay_tree::comp,
13554 typed_splay_tree::delete_key,
13555 typed_splay_tree::delete_value): New data members.
13556 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
13557 typed_splay_tree::remove.
13559 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
13561 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
13562 -mginv and -mno-ginv to the assembler.
13563 * config/mips/mips.opt (-mcrc): New option.
13564 (-mginv): Likewise.
13565 * doc/invoke.text (-mcrc): Document.
13566 (-mginv): Likewise.
13568 2018-06-15 Nick Clifton <nickc@redhat.com>
13571 * tree.c (escaped_string): New class. Converts an unescaped
13572 string into its escaped equivalent.
13573 (warn_deprecated_use): Use the new class to convert the
13574 deprecation message, if present.
13575 (test_escaped_strings): New self test.
13576 (test_c_tests): Add test_escaped_strings.
13577 * doc/extend.texi (deprecated): Add a note that the
13578 deprecation message is affected by the -fmessage-length
13579 option, and that control characters will be escaped.
13580 (#pragma GCC error): Document this pragma.
13581 (#pragma GCC warning): Likewise.
13582 * doc/invoke.texi (-fmessage-length): Document this option's
13583 effect on the #warning and #error preprocessor directives and
13584 the deprecated attribute.
13586 2018-06-15 Richard Biener <rguenther@suse.de>
13588 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
13589 here, also noting vector size used.
13590 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
13591 size used in MSG_OPTIMIZED_LOCATIONS dump.
13592 (pass_slp_vectorize::execute): Adjust.
13594 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
13597 * config/arc/arc.c (arc_return_address_register): Fix
13600 2018-06-15 Richard Biener <rguenther@suse.de>
13602 PR middle-end/86159
13603 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
13604 leave useless conversion stripping to force_gimple_operand_gsi.
13605 (gimplify_build2): Likewise.
13606 (gimplify_build1): Likewise.
13608 2018-06-15 Richard Biener <rguenther@suse.de>
13610 PR middle-end/86076
13611 * tree-cfg.c (move_stmt_op): unshare invariant addresses
13612 before adjusting their block.
13614 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
13616 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
13617 multilibs for *-*-rtems*.
13618 * config/riscv/t-rtems: New file.
13620 2018-06-14 Jakub Jelinek <jakub@redhat.com>
13622 PR middle-end/86122
13623 * match.pd ((A +- CST1) +- CST2): Punt if last resort
13624 unsigned_type_for returns NULL.
13627 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
13628 subregs of multi-word pseudos unless the float mode has word size.
13630 2018-06-14 Richard Biener <rguenther@suse.de>
13632 PR middle-end/86139
13633 * tree-vect-generic.c (build_word_mode_vector_type): Remove
13634 duplicate and harmful type_hash_canon.
13635 * tree.c (type_hash_canon): Assert we didn't find ourselves.
13637 2018-06-14 Richard Biener <rguenther@suse.de>
13640 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
13643 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
13645 * config/rtems.h (STDINT_LONG32): Define.
13647 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
13648 Prachi Godbole <prachi.godbole@imgtec.com>
13650 * config/mips/mips-cpus.def: Define P6600.
13651 * config/mips/mips-tables.opt: Regenerate.
13652 * config/mips/mips.c (mips_ucbranch_type): New enum.
13653 (mips_rtx_cost_data): Add support for P6600.
13654 (mips_issue_rate): Likewise.
13655 (mips_multipass_dfa_lookahead): Likewise.
13656 (mips_avoid_hazard): Likewise.
13657 (mips_reorg_process_insns): Likewise.
13658 (mips_classify_branch_p6600): New function.
13659 * config/mips/mips.h (TUNE_P6600): New define.
13660 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
13661 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
13662 * config/mips/mips.md: Include p6600.md.
13663 (processor): Add p6600.
13664 * config/mips/p6600.md: New file.
13665 * doc/invoke.texi: Add p6600 to supported architectures.
13667 2018-06-13 Martin Sebor <msebor@redhat.com>
13669 PR tree-optimization/86114
13670 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
13672 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
13674 2018-06-13 Richard Biener <rguenther@suse.de>
13676 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
13677 Properly set vector type of the intermediate stmt.
13678 * tree-vect-stmts.c (vectorizable_operation): The destination
13679 var always has vectype_out type.
13681 2018-06-13 Jeff Law <law@redhat.com>
13683 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
13684 integer 0 for argument to print_rtl_with_bb.
13685 (rl78_reorg): Likewise.
13687 2018-06-13 David Malcolm <dmalcolm@redhat.com>
13689 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
13690 from rtx to rtx_insn *.
13691 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
13693 (add_sched_insns_for_speculation): Likewise for local "target",
13694 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
13695 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
13696 from rtx_insn ** to rtx_code_label **.
13697 (reorg_emit_nops): Likewise.
13698 (c6x_reorg): Likewise for local "call_labels".
13699 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
13701 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
13702 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
13703 the loops over LABEL_REFS.
13704 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
13706 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
13707 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
13708 (split_branches): Strengthen local "olabel" from rtx to
13709 rtx_insn *, adding a safe_as_a cast.
13710 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
13712 (add_insn_after): Likewise for first two params.
13713 (add_insn_before): Likewise.
13714 (remove_insn): Likewise for param.
13715 (emit_pattern_before_noloc): Likewise for second and third params.
13716 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
13717 (emit_call_insn_before_noloc): Likewise.
13718 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
13720 (emit_barrier_before): Likewise.
13721 (emit_label_before): Strengthen "label" param from "rtx" to
13722 "rtx_code_label *". Strengthen "before" param from "rtx" to
13724 (emit_insn_after_1): Strengthen "after" param from "rtx" to
13726 (emit_pattern_after_noloc): Likewise.
13727 (emit_insn_after_noloc): Likewise.
13728 (emit_jump_insn_after_noloc): Likewise.
13729 (emit_call_insn_after_noloc): Likewise.
13730 (emit_debug_insn_after_noloc): Likewise.
13731 (emit_barrier_after): Likewise.
13732 (emit_label_after): Likewise for both params.
13733 (emit_pattern_after_setloc): Likewise for "after" param. Convert
13734 "loc" param from "int" to "location_t".
13735 (emit_insn_after_setloc): Likewise.
13736 (emit_jump_insn_after_setloc): Likewise.
13737 (emit_call_insn_after_setloc): Likewise.
13738 (emit_debug_insn_after_setloc): Likewise.
13739 (emit_pattern_before_setloc): Likewise for "before" param. Convert
13740 "loc" param from "int" to "location_t".
13741 (emit_pattern_before): Convert NULL_RTX to NULL.
13742 (emit_insn_before_setloc): Convert "loc" param from "int" to
13744 (emit_jump_insn_before_setloc): Likewise.
13745 (emit_call_insn_before_setloc): Likewise.
13746 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
13747 rtx_insn *. Convert "loc" param from "int" to "location_t".
13748 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
13749 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
13750 Convert 3rd param from "int" to "location_t".
13751 (emit_barrier_before, emit_barrier_after, next_real_insn):
13752 Strengthen param from rtx to rtx_insn *.
13753 (emit_label_before): Strengthen 1st param from "rtx" to
13754 "rtx_code_label *". Strengthen 2nd param from "rtx" to
13756 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
13757 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
13758 Strengthen 2nd param from "rtx" to "rtx_insn *".
13759 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
13760 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
13761 Likewise. Convert 3rd param from "int" to "location_t".
13762 (emit_label_after): Strengthen 1st param from "rtx" to
13763 "rtx_code_label *".
13764 (next_real_insn, remove_insn): Strengthen param from "rtx" to
13766 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
13767 from "rtx" to "rtx_insn *".
13769 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
13771 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
13772 bodies streamed in with -Q.
13773 * dumpfile.c (dump_files): Add lto-stream-out dump file.
13774 * dumpfile.h (tree_dump_index): Add lto_stream_out.
13775 * gimple-streamer-out.c: Include gimple-pretty-print.h
13776 (output_bb): Dump stmts streamed.
13777 * lto-section-out.c: Include print-tree.h
13778 (lto_begin_section): Dump sections created.
13779 (lto_output_decl_index): Dump decl encoded.
13780 * lto-streamer-out.c: Include print-tree.h
13781 (create_output_block): Dump output block created.
13782 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
13783 (output_function): Dump function output.
13784 (output_constructor): Dump constructor streamed.
13785 (write_global_stream): Output indexes encoded.
13786 (produce_asm_for_decls): Dump streams encoded.
13787 * lto-streamer.c (streamer_dump_file): New global var.
13788 * lto-streamer.h (streamer_dump_file): Declare.
13789 * passes.c (ipa_write_summaries): Initialize streamer dump.
13790 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
13793 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
13796 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
13797 offsets for register save directives. Emit a second batch of save
13798 directives, if need be, when the function accesses prior frames.
13800 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
13802 * config/arc/fpu.md (fmasf4): Force operand to register.
13803 (fnmasf4): Likewise.
13805 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
13807 * config/arc/arc-protos.h (arc_pad_return): Remove.
13808 * config/arc/arc.c (machine_function): Remove force_short_suffix
13810 (arc_print_operand): Adjust printing of '&'.
13811 (arc_verify_short): Remove conditional printing of short suffix.
13812 (arc_final_prescan_insn): Remove reference to size_reason.
13813 (pad_return): New function.
13814 (arc_reorg): Call pad_return.
13815 (arc_pad_return): Remove.
13816 (arc_init_machine_status): Remove reference to force_short_suffix.
13817 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
13818 (attr length): When attribute iscompact is true force to 2
13819 regardless; in the case of maybe check if we want to force the
13820 instruction to have 4 bytes length.
13821 (nopv): Change it to generate 4 byte long nop as well.
13822 (blockage): New pattern.
13823 (simple_return): Remove call to arc_pad_return.
13824 (p_return_i): Likewise.
13826 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
13828 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
13830 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
13832 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
13835 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
13837 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
13838 for ARC700 and ARCv2.
13840 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
13843 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
13844 operands[2] instead of operands[1].
13847 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
13849 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
13850 case, check whether the outer register overlaps an unallocatable
13851 register, not just whether it fits the required class.
13853 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
13855 * poly-int.h (can_div_trunc_p): Add new overload in which all values
13857 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
13858 (memrefs_conflict_p): Likewise.
13859 (init_alias_analysis): Likewise.
13860 * cfgexpand.c (expand_debug_expr): Likewise.
13861 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
13862 * cse.c (fold_rtx): Likewise.
13863 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
13864 * expr.c (emit_block_move_hints): Likewise.
13865 (clear_storage_hints, push_block, emit_push_insn): Likewise.
13866 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
13867 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
13868 (emit_group_store): Likewise.
13869 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
13870 to read the PRE/POST_MODIFY increment.
13871 * calls.c (store_one_arg): Use strip_offset.
13872 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
13874 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
13876 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
13877 (simplify_binary_operation_1): Extend CONST_INT handling to
13879 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
13880 than a HOST_WIDE_INT.
13881 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
13883 (adjust_mems, add_stores): Update accodingly.
13884 (vt_canonicalize_addr): Track polynomial offsets.
13885 (emit_note_insn_var_location): Likewise.
13886 (vt_add_function_parameter): Likewise.
13887 (vt_initialize): Likewise.
13889 2018-06-12 Jeff Law <law@redhat.com>
13891 * config.gcc (alpha*-*-freebsd*): Remove.
13892 * config/alpha/freebsd.h: Remove.
13894 2018-06-12 David Malcolm <dmalcolm@redhat.com>
13897 * spellcheck-tree.c (levenshtein_distance): Rename to...
13898 (get_edit_distance): ...this, and update for underlying renaming.
13899 * spellcheck-tree.h (levenshtein_distance): Rename to...
13900 (get_edit_distance): ...this.
13901 * spellcheck.c (levenshtein_distance): Rename to...
13902 (get_edit_distance): ...this. Convert from Levenshtein distance
13903 to Damerau-Levenshtein distance by supporting transpositions of
13904 adjacent characters. Rename "v1" to "v_next" and "v0" to
13906 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
13907 (selftest::test_edit_distance_unit_test_oneway): ...this, and
13908 update for underlying renaming.
13909 (selftest::levenshtein_distance_unit_test): Rename to...
13910 (selftest::test_get_edit_distance_unit): ...this, and update for
13911 underlying renaming.
13912 (selftest::test_find_closest_string): Add example from PR 69968
13913 where transposition helps
13914 (selftest::test_metric_conditions): Update for renaming.
13915 (selftest::test_metric_conditions): Likewise.
13916 (selftest::spellcheck_c_tests): Likewise.
13917 * spellcheck.h (levenshtein_distance): Rename both overloads to...
13918 (get_edit_distance): ...this.
13919 (best_match::consider): Update for renaming.
13921 2018-06-12 Martin Sebor <msebor@redhat.com>
13923 PR tree-optimization/85259
13924 * builtins.c (compute_objsize): Handle constant offsets.
13925 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
13926 true iff a warning has been issued.
13927 * gimple.h (gimple_nonartificial_location): New function.
13928 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
13929 gimple_nonartificial_location and handle -Wno-system-headers.
13930 (handle_builtin_stxncpy): Same.
13932 2018-06-12 Martin Sebor <msebor@redhat.com>
13935 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
13937 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
13939 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
13940 BUILTIN_VEC_XST entries for pointer to double and long long.
13942 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
13945 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
13947 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
13950 2018-06-12 Martin Liska <mliska@suse.cz>
13952 * doc/options.texi: Document IntegerRange.
13954 2018-06-12 Martin Liska <mliska@suse.cz>
13956 * config/i386/i386.opt: Make MPX-related options as Deprecated.
13957 * opt-functions.awk: Handle Deprecated flag.
13958 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
13960 (read_cmdline_option): Report warning for a deprecated option.
13961 * opts.h (struct cl_option): Add new field cl_deprecated.
13962 (CL_ERR_DEPRECATED): New.
13964 2018-06-12 Martin Liska <mliska@suse.cz>
13966 * doc/options.texi: Document Deprecated option flag.
13968 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
13970 * config/arc/arc-arch.h (arc_extras): New enum.
13971 (arc_cpu_t):Add field extra.
13972 (arc_cpu_types): Consider the extras.
13973 * config/arc/arc-cpus.def: Add extras info.
13974 * config/arc/arc-opts.h (processor_type): Consider extra field.
13975 * config/arc/arc.c (arc_override_options): Handle extra field.
13977 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
13979 * config/arc/arc-arch.h: Update ARC_OPTX macro.
13980 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
13982 * config/arc/arc.c (arc_init): Update pic warning.
13983 (irq_range): Update irq range parsing warnings.
13984 (arc_override_options): Update various warning messages.
13985 (arc_handle_aux_attribute): Likewise.
13987 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
13989 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
13991 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
13993 * doc/sourcebuild.texi: Document usage of line number 0 in verify
13994 compiler messages directives.
13996 2018-06-12 Matthew Fortune <mfortune@gmail.com>
13998 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
13999 * config/mips/mips-tables.opt: Regenerate.
14000 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
14002 * doc/invoke.texi: Document -march=i6500.
14004 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
14006 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
14007 (i6400_gpmul): Add cpu_unit.
14008 (i6400_gpdiv): Likewise.
14009 (i6400_msa_add_d): Update reservations.
14010 (i6400_msa_int_add) Likewise.
14011 (i6400_msa_short_logic3) Likewise.
14012 (i6400_msa_short_logic2) Likewise.
14013 (i6400_msa_short_logic) Likewise.
14014 (i6400_msa_move) Likewise.
14015 (i6400_msa_cmp) Likewise.
14016 (i6400_msa_short_float2) Likewise.
14017 (i6400_msa_div_d) Likewise.
14018 (i6400_msa_long_logic1) Likewise.
14019 (i6400_msa_long_logic2) Likewise.
14020 (i6400_msa_mult) Likewise.
14021 (i6400_msa_long_float2) Likewise.
14022 (i6400_msa_long_float4) Likewise.
14023 (i6400_msa_long_float5) Likewise.
14024 (i6400_msa_long_float8) Likewise.
14025 (i6400_fpu_fadd): Include frint type.
14026 (i6400_fpu_store): New define_insn_reservation.
14027 (i6400_fpu_load): Likewise.
14028 (i6400_fpu_move): Likewise.
14029 (i6400_fpu_fcmp): Likewise.
14030 (i6400_fpu_fmadd): Likewise.
14031 (i6400_int_mult): Include imul3nc type and update reservation.
14032 (i6400_int_div): Include idiv3 type and update reservation.
14033 (i6400_int_load): Update to check type not move_type.
14034 (i6400_int_store): Likewise.
14035 (i6400_int_prefetch): Set zero latency.
14037 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
14039 * gcc.c: Document new %@{...} sequence.
14040 (LINK_COMMAND_SPEC): Use it for the -L switches.
14041 (cpp_unique_options): Use it for the -I switches.
14042 (at_file_argbuf): New global variable.
14043 (in_at_file): Likewise.
14044 (alloc_args): Create at_file_argbuf.
14045 (clear_args): Truncate at_file_argbuf.
14046 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
14047 (open_at_file): New function.
14048 (close_at_file): Likewise.
14049 (create_at_file): Delete.
14050 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
14053 (validate_switches_from_spec): Deal with %@{...} sequence.
14054 (validate_switches): Likewise.
14055 (driver::finalize): Call clear_args.
14057 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
14059 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
14061 2018-06-11 Martin Sebor <msebor@redhat.com>
14063 * doc/invoke.texi (-Wall): List -Wc++17-compat.
14064 (Wno-class-memaccess): Add @opindex.
14065 (Wno-templates, Wno-multiple-inheritance): Same.
14066 (Wno-virtual-inheritance, Wno-namespaces): Same.
14067 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
14068 (Wno-format-overflow, Wno-format-truncation): Same.
14069 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
14070 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
14071 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
14072 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
14073 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
14074 (Wno-misspelled-isr): Same.
14076 2018-06-11 Martin Sebor <msebor@redhat.com>
14078 * PR tree-optimization/86083
14079 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
14081 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
14083 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
14085 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
14088 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
14089 on the correct operand.
14090 (*movdi_internal64): Ditto.
14092 2018-06-11 Martin Liska <mliska@suse.cz>
14094 PR tree-optimization/86089
14095 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
14097 2018-06-11 Julia Koval <julia.koval@intel.com>
14099 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
14100 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
14101 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
14103 2018-06-11 Olivier Hainque <hainque@adacore.com>
14105 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
14106 for Ada with strict dwarf2.
14108 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
14111 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
14114 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
14116 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
14118 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
14120 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
14123 2018-06-08 Martin Liska <mliska@suse.cz>
14125 * tree-cfg.h (debug_function): Fix argument type to match
14128 2018-06-08 Martin Liska <mliska@suse.cz>
14130 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
14131 Remove usage of MPX-related (and removed) fields.
14132 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
14134 2018-06-08 David Malcolm <dmalcolm@redhat.com>
14136 * cfg.c (debug): Use TDF_NONE rather than 0.
14137 * cfghooks.c (debug): Likewise.
14138 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
14139 (struct dump_option_value_info): Convert to...
14140 (struct kv_pair): ...this template type.
14141 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
14143 (optinfo_verbosity_options): Likewise.
14144 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
14146 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
14147 than int for "optgroup_flags" param.
14148 (dump_generic_expr_loc): Use dump_flags_t rather than int for
14150 (dump_dec): Likewise.
14151 (dump_finish): Use TDF_NONE rather than 0.
14152 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
14153 rather than int for "optgroup_flags" param. Use TDF_NONE rather
14154 than 0. Update for change to option_ptr.
14155 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
14156 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
14157 0. Update for changes to optinfo_verbosity_options and
14159 (opt_info_switch_p): Convert optgroup_flags from int to
14161 (dump_basic_block): Use dump_flags_t rather than int
14162 for "dump_kind" param.
14163 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
14164 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
14165 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
14166 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
14167 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
14168 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
14169 TDF_NONE): Convert from macros to...
14170 (enum dump_flag): ...this new enum.
14171 (dump_flags_t): Update to use enum.
14172 (operator|, operator&, operator~, operator|=, operator&=):
14173 Implement for dump_flags_t.
14174 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
14175 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
14176 Convert from macros to...
14177 (enum optgroup_flag): ...this new enum.
14178 (optgroup_flags_t): New typedef.
14179 (operator|, operator|=): Implement for optgroup_flags_t.
14180 (struct dump_file_info): Convert field "alt_flags" to
14181 dump_flags_t. Convert field "optgroup_flags" to
14183 (dump_basic_block): Use dump_flags_t rather than int for param.
14184 (dump_generic_expr_loc): Likewise.
14185 (dump_dec): Likewise.
14186 (dump_register): Convert param "optgroup_flags" to
14188 (opt_info_enable_passes): Likewise.
14189 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
14191 * gimple-pretty-print.c (debug): Likewise.
14192 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
14193 (merged_store_group::apply_stores): Likewise.
14194 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
14195 * gimple.c (verify_gimple_pp): Likewise.
14196 * graphite-poly.c (print_pbb_body): Likewise.
14197 * passes.c (pass_manager::register_one_dump_file): Convert
14198 local "optgroup_flags" to optgroup_flags_t.
14199 * print-tree.c (print_node): Use TDF_NONE rather than 0.
14201 (debug_body): Likewise.
14202 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
14203 to optgroup_flags_t.
14204 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
14206 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
14207 (convert_mult_to_fma): Likewise.
14208 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
14209 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
14210 * tree-vect-data-refs.c (dump_lower_bound): Convert param
14211 "dump_kind" to dump_flags_t.
14213 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
14215 * config/rs6000/rs6000.c (min, max): Delete.
14217 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
14219 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
14220 -mabi=spe and -mabi=no-spe.
14222 2018-06-08 Martin Liska <mliska@suse.cz>
14224 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
14225 where we expect an existing summary.
14227 2018-06-08 Martin Liska <mliska@suse.cz>
14229 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
14230 * ipa-inline.h (estimate_edge_growth): Likewise.
14232 2018-06-08 Martin Liska <mliska@suse.cz>
14234 * cgraph.c (function_version_hasher::hash): Use
14235 cgraph_node::get_uid ().
14236 (function_version_hasher::equal):
14237 * cgraph.h (cgraph_node::get_uid): New method.
14238 * ipa-inline.c (update_caller_keys): Use
14239 cgraph_node::get_uid ().
14240 (update_callee_keys): Likewise.
14241 * ipa-utils.c (searchc): Likewise.
14242 (ipa_reduced_postorder): Likewise.
14243 * lto-cgraph.c (input_node): Likewise.
14244 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
14245 * symbol-summary.h (symtab_insertion): Likewise.
14246 (symtab_removal): Likewise.
14247 (symtab_duplication): Likewise.
14248 * tree-pretty-print.c (dump_function_header): Likewise.
14249 * tree-sra.c (convert_callers_for_node): Likewise.
14251 2018-06-08 Martin Liska <mliska@suse.cz>
14253 * cgraph.c (symbol_table::create_edge): Always assign a new
14255 (symbol_table::free_edge): Do not recycle numbers.
14256 * cgraph.h (cgraph_edge::get): New method.
14257 * symbol-summary.h (symtab_removal): Use it.
14258 (symtab_duplication): Likewise.
14259 (call_summary::hashable_uid): Remove.
14261 2018-06-08 Martin Liska <mliska@suse.cz>
14263 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
14264 (initialize_growth_caches): Remove.
14265 (free_growth_caches): Likewise.
14266 (do_estimate_edge_time): Use edge_growth_cache.
14267 (do_estimate_edge_size): Likewise.
14268 (do_estimate_edge_hints): Likewise.
14269 * ipa-inline.c (reset_edge_caches): Likewise.
14270 (recursive_inlining): Likewise.
14271 (inline_small_functions): Likewise.
14272 * ipa-inline.h (initialize_growth_caches): Remove.
14273 (estimate_edge_size): Likewise.
14274 (estimate_edge_time): Likewise.
14275 (estimate_edge_hints): Likewise.
14276 (reset_edge_growth_cache): Likewise.
14277 * symbol-summary.h (call_summary::remove): New method.
14279 2018-06-08 Martin Liska <mliska@suse.cz>
14281 * ipa-cp.c (class edge_clone_summary): New summary.
14282 (grow_edge_clone_vectors): Remove.
14283 (ipcp_edge_duplication_hook): Remove.
14284 (class edge_clone_summary_t): New call_summary class.
14285 (ipcp_edge_removal_hook): Remove.
14286 (edge_clone_summary_t::duplicate): New function.
14287 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
14288 (create_specialized_node): Likewise.
14289 (ipcp_driver): Initialize edge_clone_summaries and do not
14292 2018-06-08 Martin Liska <mliska@suse.cz>
14294 * symbol-summary.h (get): New function.
14295 (call_summary::m_initialize_when_cloning): New class member.
14297 2018-06-08 Martin Liska <mliska@suse.cz>
14299 * cgraph.c (cgraph_node::remove): Do not recycle uid.
14300 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
14301 (symbol_table::allocate_cgraph_symbol): Do not set uid.
14302 * passes.c (uid_hash_t): Record removed_nodes by their uids.
14303 (remove_cgraph_node_from_order): Use the removed_nodes set.
14304 (do_per_function_toporder): Likwise.
14305 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
14306 instead of summary_uid.
14307 (symtab_removal): Likewise.
14308 (symtab_duplication): Likewise.
14310 2018-06-08 Martin Liska <mliska@suse.cz>
14312 * ipa-cp.c (ipcp_store_bits_results): Use
14313 ipcp_transformation_sum.
14314 (ipcp_store_vr_results): Likewise.
14315 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
14317 (ipcp_transformation_initialize): ... this.
14318 (ipa_set_node_agg_value_chain):
14319 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
14320 (write_ipcp_transformation_info): Likewise.
14321 (read_ipcp_transformation_info): Likewise.
14322 (ipcp_update_bits): Likewise.
14323 (ipcp_update_vr): Likewise.
14324 (ipcp_transform_function): Likewise.
14325 * ipa-prop.h: Rename ipcp_transformation_summary to
14326 ipcp_transformation.
14327 (class ipcp_transformation_t): New function summary.
14328 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
14329 (ipa_get_agg_replacements_for_node): Likewise.
14331 2018-06-08 Martin Liska <mliska@suse.cz>
14333 * ipa-pure-const.c (struct funct_state_d): Do it class instead
14335 (class funct_state_summary_t): New function_summary class.
14336 (has_function_state): Remove.
14337 (get_function_state): Likewise.
14338 (set_function_state): Likewise.
14339 (add_new_function): Likewise.
14340 (funct_state_summary_t::insert): New function.
14341 (duplicate_node_data): Remove.
14342 (remove_node_data): Remove.
14343 (funct_state_summary_t::duplicate): New function.
14344 (register_hooks): Create new funct_state_summaries.
14345 (pure_const_generate_summary): Use it.
14346 (pure_const_write_summary): Likewise.
14347 (pure_const_read_summary): Likewise.
14348 (propagate_pure_const): Likewise.
14349 (propagate_nothrow): Likewise.
14350 (dump_malloc_lattice): Likewise.
14351 (propagate_malloc): Likewise.
14352 (execute): Do not register hooks, just remove summary
14354 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
14357 2018-06-08 Martin Liska <mliska@suse.cz>
14359 * ipa-reference.c (remove_node_data): Remove.
14360 (duplicate_node_data): Likewise.
14361 (class ipa_ref_var_info_summary_t): New class.
14362 (class ipa_ref_opt_summary_t): Likewise.
14363 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
14364 (get_reference_optimization_summary): Use
14365 ipa_ref_opt_sum_summaries.
14366 (set_reference_vars_info): Remove.
14367 (set_reference_optimization_summary): Likewise.
14368 (ipa_init): Create summaries.
14369 (init_function_info): Use function summary.
14370 (ipa_ref_opt_summary_t::duplicate): New function.
14371 (ipa_ref_opt_summary_t::remove): New function.
14372 (get_read_write_all_from_node): Fix GNU coding style.
14373 (propagate): Use function summary.
14374 (write_node_summary_p): Fix GNU coding style.
14375 (stream_out_bitmap): Likewise.
14376 (ipa_reference_read_optimization_summary): Use function summary.
14377 (ipa_reference_c_finalize): Do not release hooks.
14379 2018-06-08 Martin Liska <mliska@suse.cz>
14381 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
14382 (analyze_function_body): Extract multiple calls of get_create.
14383 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
14384 * ipa-inline.c (recursive_inlining): Use ::get method.
14385 * ipa-inline.h (estimate_edge_growth): Likewise.
14387 2018-06-08 Martin Liska <mliska@suse.cz>
14389 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
14391 (hsa_function_summary::hsa_function_summary): Use the new enum
14393 (hsa_gpu_implementation_p): Use hsa_summaries::get.
14394 * hsa-gen.c (hsa_get_host_function): Likewise.
14395 (get_brig_function_name): Likewise.
14396 * ipa-hsa.c (process_hsa_functions): Likewise.
14397 (ipa_hsa_write_summary): Likewise.
14398 * symbol-summary.h (symtab_duplication): Use ::get function/
14399 (get): New function.
14401 2018-06-08 Martin Liska <mliska@suse.cz>
14403 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
14405 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
14406 (hsa_register_kernel): Likewise.
14407 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
14408 * hsa-gen.c (hsa_get_host_function): Likewise.
14409 (get_brig_function_name): Likewise.
14410 (generate_hsa): Likewise.
14411 (pass_gen_hsail::execute): Likewise.
14412 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
14413 (devirtualization_time_bonus): Likewise.
14414 (ipcp_propagate_stage): Likewise.
14415 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
14416 (edge_set_predicate): Likewise.
14417 (evaluate_conditions_for_known_args): Likewise.
14418 (evaluate_properties_for_edge): Likewise.
14419 (ipa_fn_summary::reset): Likewise.
14420 (ipa_fn_summary_t::duplicate): Likewise.
14421 (dump_ipa_call_summary): Likewise.
14422 (ipa_dump_fn_summary): Likewise.
14423 (analyze_function_body): Likewise.
14424 (compute_fn_summary): Likewise.
14425 (estimate_edge_devirt_benefit): Likewise.
14426 (estimate_edge_size_and_time): Likewise.
14427 (estimate_calls_size_and_time): Likewise.
14428 (estimate_node_size_and_time): Likewise.
14429 (inline_update_callee_summaries): Likewise.
14430 (remap_edge_change_prob): Likewise.
14431 (remap_edge_summaries): Likewise.
14432 (ipa_merge_fn_summary_after_inlining): Likewise.
14433 (ipa_update_overall_fn_summary): Likewise.
14434 (read_ipa_call_summary): Likewise.
14435 (inline_read_section): Likewise.
14436 (write_ipa_call_summary): Likewise.
14437 (ipa_fn_summary_write): Likewise.
14438 (ipa_free_fn_summary): Likewise.
14439 * ipa-hsa.c (process_hsa_functions): Likewise.
14440 (ipa_hsa_write_summary): Likewise.
14441 (ipa_hsa_read_section): Likewise.
14442 * ipa-icf.c (sem_function::merge): Likewise.
14443 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
14444 (do_estimate_edge_time): Likewise.
14445 (estimate_size_after_inlining): Likewise.
14446 (estimate_growth): Likewise.
14447 (growth_likely_positive): Likewise.
14448 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
14449 (inline_call): Likewise.
14450 * ipa-inline.c (caller_growth_limits): Likewise.
14451 (can_inline_edge_p): Likewise.
14452 (can_inline_edge_by_limits_p): Likewise.
14453 (compute_uninlined_call_time): Likewise.
14454 (compute_inlined_call_time): Likewise.
14455 (want_inline_small_function_p): Likewise.
14456 (edge_badness): Likewise.
14457 (update_caller_keys): Likewise.
14458 (update_callee_keys): Likewise.
14459 (recursive_inlining): Likewise.
14460 (inline_small_functions): Likewise.
14461 (inline_to_all_callers_1): Likewise.
14462 (dump_overall_stats): Likewise.
14463 (early_inline_small_functions): Likewise.
14464 (early_inliner): Likewise.
14465 * ipa-inline.h (estimate_edge_growth): Likewise.
14466 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
14467 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
14468 * ipa-prop.h (IPA_NODE_REF): Likewise.
14469 (IPA_EDGE_REF): Likewise.
14470 * ipa-pure-const.c (malloc_candidate_p): Likewise.
14471 (propagate_malloc): Likewise.
14472 * ipa-split.c (execute_split_functions): Likewise.
14473 * symbol-summary.h: Rename get to get_create.
14475 (get_create): Likewise.
14476 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
14478 2018-06-08 Martin Liska <mliska@suse.cz>
14480 * symbol-summary.h (release): Move definition out of class
14482 (symtab_removal): Likewise.
14483 (symtab_duplication): Likewise.
14485 2018-06-08 Martin Liska <mliska@suse.cz>
14487 * symbol-summary.h (function_summary): Move constructor
14488 implementation out of class declaration.
14489 (release): Likewise.
14490 (symtab_insertion): Likewise.
14491 (symtab_removal): Likewise.
14492 (symtab_duplication): Likewise.
14495 2018-06-08 Martin Liska <mliska@suse.cz>
14497 * Makefile.in: Remove support for MPX (macros, related functions,
14498 fields in cgraph_node, ...).
14499 * builtin-types.def (BT_BND): Likewise.
14500 (BT_FN_BND_CONST_PTR): Likewise.
14501 (BT_FN_CONST_PTR_BND): Likewise.
14502 (BT_FN_VOID_PTR_BND): Likewise.
14503 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
14504 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
14505 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
14506 (expand_builtin_mempcpy_with_bounds): Likewise.
14507 (expand_builtin_memset_with_bounds): Likewise.
14508 (expand_builtin_memset_args): Likewise.
14509 (std_expand_builtin_va_start): Likewise.
14510 (expand_builtin): Likewise.
14511 (expand_builtin_with_bounds): Likewise.
14512 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
14513 (DEF_LIB_BUILTIN_CHKP): Likewise.
14514 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
14515 (DEF_CHKP_BUILTIN): Likewise.
14516 (BUILT_IN_MEMCPY): Likewise.
14517 (BUILT_IN_MEMMOVE): Likewise.
14518 (BUILT_IN_MEMPCPY): Likewise.
14519 (BUILT_IN_MEMSET): Likewise.
14520 (BUILT_IN_STPCPY): Likewise.
14521 (BUILT_IN_STRCAT): Likewise.
14522 (BUILT_IN_STRCHR): Likewise.
14523 (BUILT_IN_STRCPY): Likewise.
14524 (BUILT_IN_STRLEN): Likewise.
14525 (BUILT_IN_MEMCPY_CHK): Likewise.
14526 (BUILT_IN_MEMMOVE_CHK): Likewise.
14527 (BUILT_IN_MEMPCPY_CHK): Likewise.
14528 (BUILT_IN_MEMSET_CHK): Likewise.
14529 (BUILT_IN_STPCPY_CHK): Likewise.
14530 (BUILT_IN_STRCAT_CHK): Likewise.
14531 (BUILT_IN_STRCPY_CHK): Likewise.
14532 * calls.c (store_bounds): Likewise.
14533 (emit_call_1): Likewise.
14534 (special_function_p): Likewise.
14535 (maybe_warn_nonstring_arg): Likewise.
14536 (initialize_argument_information): Likewise.
14537 (finalize_must_preallocate): Likewise.
14538 (compute_argument_addresses): Likewise.
14539 (expand_call): Likewise.
14540 * cfgexpand.c (expand_call_stmt): Likewise.
14541 (expand_return): Likewise.
14542 (expand_gimple_stmt_1): Likewise.
14543 (pass_expand::execute): Likewise.
14544 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
14545 (cgraph_node::remove): Likewise.
14546 (cgraph_node::dump): Likewise.
14547 (cgraph_node::verify_node): Likewise.
14548 * cgraph.h (chkp_function_instrumented_p): Likewise.
14549 (symtab_node::get_alias_target): Likewise.
14550 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
14551 (cgraph_local_p): Likewise.
14552 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
14553 (cgraph_edge::rebuild_references): Likewise.
14554 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
14555 (walk_polymorphic_call_targets): Likewise.
14556 (cgraph_node::expand_thunk): Likewise.
14557 (symbol_table::output_weakrefs): Likewise.
14558 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
14559 (ix86_handle_option): Likewise.
14560 * config/i386/constraints.md: Likewise.
14561 * config/i386/i386-builtin-types.def (BND): Likewise.
14565 * config/i386/i386-builtin.def (BDESC_END): Likewise.
14566 (BDESC_FIRST): Likewise.
14568 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
14569 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
14570 * config/i386/i386.c (enum reg_class): Likewise.
14571 (ix86_target_string): Likewise.
14572 (ix86_option_override_internal): Likewise.
14573 (ix86_conditional_register_usage): Likewise.
14574 (ix86_valid_target_attribute_inner_p): Likewise.
14575 (ix86_set_indirect_branch_type): Likewise.
14576 (ix86_set_current_function): Likewise.
14577 (ix86_function_arg_regno_p): Likewise.
14578 (init_cumulative_args): Likewise.
14579 (ix86_function_arg_advance): Likewise.
14580 (ix86_function_arg): Likewise.
14581 (ix86_pass_by_reference): Likewise.
14582 (ix86_function_value_regno_p): Likewise.
14583 (ix86_function_value_1): Likewise.
14584 (ix86_function_value_bounds): Likewise.
14585 (ix86_return_in_memory): Likewise.
14586 (ix86_setup_incoming_vararg_bounds): Likewise.
14587 (ix86_va_start): Likewise.
14588 (indirect_thunk_need_prefix): Likewise.
14589 (print_reg): Likewise.
14590 (ix86_print_operand): Likewise.
14591 (ix86_expand_call): Likewise.
14592 (ix86_output_function_return): Likewise.
14593 (reg_encoded_number): Likewise.
14594 (BDESC_VERIFYS): Likewise.
14595 (ix86_init_mpx_builtins): Likewise.
14596 (ix86_init_builtins): Likewise.
14597 (ix86_emit_cmove): Likewise.
14598 (ix86_emit_move_max): Likewise.
14599 (ix86_expand_builtin): Likewise.
14600 (ix86_builtin_mpx_function): Likewise.
14601 (ix86_get_arg_address_for_bt): Likewise.
14602 (ix86_load_bounds): Likewise.
14603 (ix86_store_bounds): Likewise.
14604 (ix86_load_returned_bounds): Likewise.
14605 (ix86_store_returned_bounds): Likewise.
14606 (ix86_class_likely_spilled_p): Likewise.
14607 (ix86_hard_regno_mode_ok): Likewise.
14608 (x86_order_regs_for_local_alloc): Likewise.
14609 (ix86_mitigate_rop): Likewise.
14610 (ix86_bnd_prefixed_insn_p): Likewise.
14611 (ix86_mpx_bound_mode): Likewise.
14612 (ix86_make_bounds_constant): Likewise.
14613 (ix86_initialize_bounds): Likewise.
14614 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
14615 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
14616 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
14617 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
14618 (TARGET_CHKP_BOUND_MODE): Likewise.
14619 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
14620 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
14621 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
14622 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
14623 * config/i386/i386.h (TARGET_MPX): Likewise.
14624 (TARGET_MPX_P): Likewise.
14625 (VALID_BND_REG_MODE): Likewise.
14626 (FIRST_BND_REG): Likewise.
14627 (LAST_BND_REG): Likewise.
14628 (enum reg_class): Likewise.
14629 (BND_REG_P): Likewise.
14630 (BND_REGNO_P): Likewise.
14631 (BNDmode): Likewise.
14632 (ADJUST_INSN_LENGTH): Likewise.
14633 * config/i386/i386.md: Likewise.
14634 * config/i386/i386.opt: Likewise.
14635 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
14636 (defined): Likewise.
14637 (LINK_MPX): Likewise.
14638 (MPX_SPEC): Likewise.
14639 (LIBMPX_SPEC): Likewise.
14640 (LIBMPXWRAPPERS_SPEC): Likewise.
14641 (CHKP_SPEC): Likewise.
14642 * config/i386/predicates.md: Likewise.
14643 * dbxout.c (dbxout_type): Likewise.
14644 * doc/extend.texi: Likewise.
14645 * doc/invoke.texi: Likewise.
14646 * doc/md.texi: Likewise.
14647 * doc/tm.texi: Likewise.
14648 * doc/tm.texi.in: Likewise.
14649 * dwarf2out.c (is_base_type): Likewise.
14650 (gen_formal_types_die): Likewise.
14651 (gen_subprogram_die): Likewise.
14652 (gen_type_die_with_usage): Likewise.
14653 (gen_decl_die): Likewise.
14654 (dwarf2out_late_global_decl): Likewise.
14655 * expr.c (expand_assignment): Likewise.
14656 (emit_storent_insn): Likewise.
14657 (store_expr_with_bounds): Likewise.
14658 (store_expr): Likewise.
14659 (expand_expr_real_1): Likewise.
14660 * expr.h (store_expr_with_bounds): Likewise.
14661 * function.c (use_register_for_decl): Likewise.
14662 (struct bounds_parm_data): Likewise.
14663 (assign_parms_augmented_arg_list): Likewise.
14664 (assign_parm_find_entry_rtl): Likewise.
14665 (assign_parm_is_stack_parm): Likewise.
14666 (assign_parm_load_bounds): Likewise.
14667 (assign_bounds): Likewise.
14668 (assign_parms): Likewise.
14669 (expand_function_start): Likewise.
14670 * gcc.c (CHKP_SPEC): Likewise.
14671 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
14672 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
14673 (wrestrict_dom_walker::check_call): Likewise.
14674 * gimple.c (gimple_build_call_from_tree): Likewise.
14675 * gimple.h (enum gf_mask): Likewise.
14676 (gimple_call_with_bounds_p): Likewise.
14677 (gimple_call_set_with_bounds): Likewise.
14678 * gimplify.c (gimplify_init_constructor): Likewise.
14679 * ipa-cp.c (initialize_node_lattices): Likewise.
14680 (propagate_constants_across_call): Likewise.
14681 (find_more_scalar_values_for_callers_subset): Likewise.
14682 * ipa-hsa.c (process_hsa_functions): Likewise.
14683 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
14684 * ipa-icf.c (sem_function::merge): Likewise.
14685 * ipa-inline.c (early_inliner): Likewise.
14686 * ipa-pure-const.c (warn_function_noreturn): Likewise.
14687 (warn_function_cold): Likewise.
14688 (propagate_pure_const): Likewise.
14689 * ipa-ref.h (enum GTY): Likewise.
14690 * ipa-split.c (find_retbnd): Likewise.
14691 (consider_split): Likewise.
14692 (split_function): Likewise.
14693 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
14694 * ipa.c (walk_polymorphic_call_targets): Likewise.
14695 (symbol_table::remove_unreachable_nodes): Likewise.
14696 (process_references): Likewise.
14697 (cgraph_build_static_cdtor_1): Likewise.
14698 * lto-cgraph.c (lto_output_node): Likewise.
14699 (output_refs): Likewise.
14700 (compute_ltrans_boundary): Likewise.
14701 (input_overwrite_node): Likewise.
14702 (input_node): Likewise.
14703 (input_cgraph_1): Likewise.
14704 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
14705 * passes.c (pass_manager::execute_early_local_passes): Likewise.
14706 (class pass_chkp_instrumentation_passes): Likewise.
14707 (make_pass_chkp_instrumentation_passes): Likewise.
14708 * passes.def: Likewise.
14709 * rtl.h (struct GTY): Likewise.
14710 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
14711 * stor-layout.c (layout_type): Likewise.
14712 * symtab.c: Likewise.
14713 * target.def: Likewise.
14714 * targhooks.c (default_chkp_bound_type): Likewise.
14715 (default_chkp_bound_mode): Likewise.
14716 (default_builtin_chkp_function): Likewise.
14717 (default_chkp_function_value_bounds): Likewise.
14718 (default_chkp_make_bounds_constant): Likewise.
14719 (default_chkp_initialize_bounds): Likewise.
14720 * targhooks.h (default_chkp_bound_type): Likewise.
14721 (default_chkp_bound_mode): Likewise.
14722 (default_builtin_chkp_function): Likewise.
14723 (default_chkp_function_value_bounds): Likewise.
14724 (default_chkp_make_bounds_constant): Likewise.
14725 (default_chkp_initialize_bounds): Likewise.
14726 * toplev.c (compile_file): Likewise.
14727 (process_options): Likewise.
14728 * tree-core.h (DEF_BUILTIN): Likewise.
14729 (DEF_BUILTIN_CHKP): Likewise.
14730 * tree-inline.c (declare_return_variable): Likewise.
14731 (remap_gimple_stmt): Likewise.
14732 (copy_bb): Likewise.
14733 (initialize_inlined_parameters): Likewise.
14734 (expand_call_inline): Likewise.
14735 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
14736 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
14737 (make_pass_ipa_chkp_produce_thunks): Likewise.
14738 (make_pass_chkp): Likewise.
14739 (make_pass_chkp_opt): Likewise.
14740 (make_pass_chkp_instrumentation_passes): Likewise.
14741 * tree-pretty-print.c (dump_generic_node): Likewise.
14742 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
14743 * tree-ssa-dce.c (propagate_necessity): Likewise.
14744 (eliminate_unnecessary_stmts): Likewise.
14745 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
14746 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
14747 * tree-ssa-sccvn.h: Likewise.
14748 * tree-ssa-strlen.c (get_string_length): Likewise.
14749 (valid_builtin_call): Likewise.
14750 (adjust_last_stmt): Likewise.
14751 (handle_builtin_strchr): Likewise.
14752 (handle_builtin_strcpy): Likewise.
14753 (handle_builtin_stxncpy): Likewise.
14754 (handle_builtin_memcpy): Likewise.
14755 (handle_builtin_strcat): Likewise.
14756 (strlen_check_and_optimize_stmt): Likewise.
14757 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
14758 * tree-streamer-in.c: Likewise.
14759 * tree-streamer.c (record_common_node): Likewise.
14760 * tree.c (tree_code_size): Likewise.
14761 (wide_int_to_tree_1): Likewise.
14762 (type_contains_placeholder_1): Likewise.
14763 (build_common_tree_nodes): Likewise.
14764 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
14765 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
14766 (POINTER_BOUNDS_P): Likewise.
14767 (BOUNDED_TYPE_P): Likewise.
14768 (BOUNDED_P): Likewise.
14769 (CALL_WITH_BOUNDS_P): Likewise.
14770 (pointer_bounds_type_node): Likewise.
14771 * value-prof.c (gimple_ic): Likewise.
14772 * var-tracking.c (vt_add_function_parameters): Likewise.
14773 * varasm.c (make_decl_rtl): Likewise.
14774 (assemble_start_function): Likewise.
14775 (output_constant): Likewise.
14776 (maybe_assemble_visibility): Likewise.
14777 * varpool.c (ctor_for_folding): Likewise.
14778 * chkp-builtins.def: Remove.
14779 * ipa-chkp.c: Remove.
14780 * ipa-chkp.h: Remove.
14781 * rtl-chkp.c: Remove.
14782 * rtl-chkp.h: Remove.
14783 * tree-chkp-opt.c: Remove.
14784 * tree-chkp.c: Remove.
14785 * tree-chkp.h: Remove.
14787 2018-06-07 Carl Love <cel@us.ibm.com>
14789 * config/rs6000/vsx.md (vextract_fp_from_shorth,
14790 vextract_fp_from_shortl): Add BE support.
14792 2018-06-07 Paul Koning <ni1d@arrl.net>
14794 * compare-elim.c (try_merge_compare): Don't merge compare if
14795 address contains a side effect.
14796 (try_eliminate_compare): Likewise.
14798 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
14800 * config.gcc: Support "tremont".
14801 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
14802 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14804 * config/i386/i386.c (m_TREMONT): Define.
14805 (processor_target_table): Add "tremont".
14806 (PTA_TREMONT): Define.
14807 (ix86_lea_outperforms): Add TARGET_TREMONT.
14808 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
14809 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
14810 and M_INTEL_GOLDMONT_PLUS.
14811 (fold_builtin_cpu): Add "tremont".
14812 (ix86_add_stmt_cost): Add TARGET_TREMONT.
14813 (ix86_option_override_internal): Add "tremont".
14814 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
14815 (processor_type): Add PROCESSOR_TREMONT.
14816 * config/i386/x86-tune.def: Add m_TREMONT.
14817 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
14819 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
14821 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
14822 symbol defined for msp430i* devices to be lower case.
14824 2018-06-07 Richard Biener <rguenther@suse.de>
14826 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
14827 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
14828 Properly wrap signed arithmetic if overflow wraps.
14830 2018-06-07 Jakub Jelinek <jakub@redhat.com>
14832 PR tree-optimization/69615
14833 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
14834 of a cast from a same precision integral SSA_NAME in a bb dominated
14835 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
14836 cast to utype if rhs2 has already a compatible type.
14838 2018-06-07 Richard Biener <rguenther@suse.de>
14840 PR tree-optimization/85935
14841 * graphite-scop-detection.c (find_params_in_bb): Analyze
14842 condition operands with respect to the correct loop. Assert
14843 the analysis doesn't fail.
14845 2018-06-04 Carl Love <cel@us.ibm.com>
14847 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
14848 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
14849 as it is slightly cheaper.
14850 (first_match_or_eos_index_<mode>):
14851 Calculate index using natural element order.
14852 (first_match_index_<mode>):
14853 Calculate index using natural element order.
14854 (first_match_or_eos_index_<mode>):
14855 Calculate index using natural order.
14856 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
14857 for BE and LE modes.
14858 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
14859 P9V_BUILTIN_VCLZLSBB_V16QI.
14860 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
14863 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
14865 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
14866 indentation and line wrap for many prototypes. Add missing
14867 @smallexample directives around block of prototypes for vec_xl and
14870 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
14872 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
14873 track if we pass or return IEEE 128-bit floating point.
14874 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
14875 C++ mangling that is compatible with GCC 8.1.
14876 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
14877 (init_cumulative_args): Note if we pass or return IEEE 128-bit
14878 floating point types.
14879 (rs6000_function_arg_advance_1): Likewise.
14880 (rs6000_mangle_type): Optionally generate mangled names that match
14881 what GCC 8.1 generated for IEEE 128-bit floating point types.
14882 (rs6000_globalize_decl_name): If we have an external function that
14883 passes or returns IEEE 128-bit types, generate a weak reference
14884 from the mangled name used in GCC 8.1 to the current mangled
14886 (rs6000_init_builtins): Make __ibm128 use the long double type if
14887 long double is IBM extended double. Make __float128 use the long
14888 double type if long double is IEEE 128-bit.
14891 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
14892 macro for __ibm128 built-in functions.
14893 (PACK_IF): Add __ibm128 pack/unpack functions.
14894 (UNPACK_IF): Likewise.
14895 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
14896 enable long double built-in functions if long double is IEEE
14897 128-bit floating point.
14898 (rs6000_invalid_builtin): Update long double built-in function
14900 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
14901 functions, adjust the built-in function to use the long double
14902 built-in function if __ibm128 and long double are the same type.
14903 * doc/extend.texi (PowerPC builtins): Update documention for
14904 __builtin_{,un}pack_longdouble. Add documentation for
14905 __builtin_{,un}pack_ibm128.
14907 2018-06-06 Jim Wilson <jimw@sifive.com>
14909 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
14910 (struct machine_function): New field interrupt_mode.
14911 (riscv_handle_type_attribute): New function. Add forward declaration.
14912 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
14913 (riscv_expand_epilogue): Check interrupt_mode field.
14914 (riscv_set_current_function): Check interrupt attribute args and
14915 set interrupt_mode field.
14916 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
14917 (riscv_sret, riscv_uret): New.
14918 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
14919 new arguments to interrupt attribute.
14921 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
14924 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
14925 Don't handle -mcpu=power8 if -mpower9-vector is also used.
14927 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
14929 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
14930 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
14931 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
14932 several redundant entries.
14934 2018-06-06 David Malcolm <dmalcolm@redhat.com>
14936 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
14937 type from "rtx" to "rtx_insn *".
14938 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
14939 for local "call_insn", removing cast.
14940 (ix86_expand_call): Likewise, introducing a "call_insn" local.
14942 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
14944 PR tree-optimization/86066
14945 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
14946 for BIT_INSERT_EXPR stores.
14948 2018-06-06 Richard Biener <rguenther@suse.de>
14950 PR tree-optimization/86062
14951 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
14952 component refs ontop
14953 of to be offsetted base.
14955 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
14957 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
14958 to be static and remove check on interrupt attribute name.
14960 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
14962 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
14963 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
14965 2018-06-05 Steve Ellcey <sellcey@cavium.com>
14968 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
14970 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
14971 Remove second argument, change how error is called.
14972 (aarch64_layout_arg): Remove second argument from
14973 aarch64_err_no_fpadvsimd call.
14974 (aarch64_init_cumulative_args): Ditto.
14975 (aarch64_gimplify_va_arg_expr): Ditto.
14976 * config/aarch64/aarch64.md (mov<mode>): Ditto.
14978 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
14980 * config/i386/i386.md (simple_return_indirect_internal): New expander.
14981 (*simple_return_indirect_internal<mode>): Rename from
14982 simple_return_indirect_internal. Use W mode iterator.
14983 (rstorssp): New expander.
14984 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
14985 (clrssbsy): New expander.
14986 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
14988 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
14990 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
14992 (cmse_check_pointed_object): Likewise.
14994 2018-06-05 Martin Liska <mliska@suse.cz>
14996 PR gcov-profile/47618
14997 * doc/invoke.texi: Document how -fprofile-dir format
15000 2018-06-05 Richard Biener <rguenther@suse.de>
15002 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
15003 removal pretend DOM info isn't available so we do not update
15004 it and only remove edges, not dominated blocks. Actually free
15005 DOM info in case we removed something. Remove unreachable blocks.
15006 (mfb_keep_latches): Work with either DOM info or marked backedges.
15007 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
15008 first. Mark backedges if DOM info isn't available.
15009 (Re-)compute DOM info after cleanup_control_flow_pre.
15011 2018-06-05 Richard Biener <rguenther@suse.de>
15013 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
15014 (locus_discrim_hasher::hash): Adjust.
15015 (locus_discrim_hasher::equal): Likewise.
15016 (next_discriminator_for_locus): Work on line directly.
15017 (same_line_p): Pass in expanded locus1 as well.
15018 (assign_discriminators): Avoid redundant location expansions.
15020 2018-06-05 Richard Biener <rguenther@suse.de>
15022 PR tree-optimization/86046
15023 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
15024 if required after clearing TREE_ADDRESSABLE.
15026 2018-06-05 Richard Biener <rguenther@suse.de>
15028 PR tree-optimization/86047
15029 * tree-ssa-loop.c (for_each_index): Glob handling of all
15030 decls and constants and really handle all of them.
15032 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15035 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
15036 qualifier_void_pointer and qualifier_const_void_pointer.
15037 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
15038 (arm_init_builtins): Handle the above.
15039 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
15040 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
15041 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
15044 2018-06-05 Martin Liska <mliska@suse.cz>
15046 * auto-profile.c (read_autofdo_file): Do not use
15047 gcov_ctr_summary struct.
15048 (afdo_callsite_hot_enough_for_early_inline): Likewise.
15049 * coverage.c (struct counts_entry): Likewise.
15050 (read_counts_file): Read just single summary entry.
15051 (get_coverage_counts): Use gcov_summary struct.
15052 * coverage.h (get_coverage_counts): Likewise.
15053 * gcov-dump.c (dump_working_sets): Likewise.
15054 (tag_summary): Dump just single summary.
15055 * gcov-io.c (gcov_write_summary): Write just histogram
15057 (gcov_read_summary): Read just single summary.
15058 (compute_working_sets): Use gcov_summary struct.
15059 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
15060 of GCOV_COUNTERS_SUMMABLE.
15061 (GCOV_COUNTERS_SUMMABLE): Remove.
15062 (GCOV_FIRST_VALUE_COUNTER): Replace with
15063 GCOV_COUNTER_V_INTERVAL.
15064 (struct gcov_ctr_summary): Remove.
15065 (struct gcov_summary): Directly use fields of former
15067 (compute_working_sets): Use gcov_summary struct.
15068 * gcov.c (read_count_file): Do not use ctrs fields.
15069 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
15071 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
15073 * profile.c: Likewise.
15074 * profile.h: Likewise.
15076 2018-06-05 Martin Liska <mliska@suse.cz>
15078 PR gcov-profile/84846
15079 * gcov.c (output_lines): Print working directory only
15080 in intermediate format.
15082 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
15084 * config/s390/s390-builtin-types.def: Add void function type.
15085 * config/s390/s390-builtins.def: Use the function type for the
15088 2018-06-04 Jim Wilson <jimw@sifive.com>
15090 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
15092 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
15093 and maybe_eh_return. Change regno to unsigned int. Use new args to
15094 handle EH_RETURN_DATA_REGNO registers properly.
15095 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
15096 (riscv_expand_epilogue): Update comment. Change argument name and
15097 type. Update code to use new name and type. Pass new args to
15098 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
15100 * config/riscv/riscv.md (NORMAL_RETURN): New.
15101 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
15102 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
15103 (eh_return): Call gen_eh_return_internal and emit barrier.
15104 (eh_return_internal): Call riscv_expand_epilogue.
15106 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
15108 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
15109 bit_insertion field and declare can_be_merged_into method.
15110 (merged_store_group::can_be_merged_into): New method.
15111 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
15112 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
15113 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
15115 2018-06-04 Richard Biener <rguenther@suse.de>
15117 PR tree-optimization/85955
15118 * builtins.c (fold_builtin_sincos): Convert pointers to
15119 destination to appropriate type before dereferencing.
15121 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
15123 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
15125 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
15127 * expr.c (expand_expr_real_1): Force the operand into memory if
15128 its TYPE_MODE is BLKmode and if there is no integer mode for
15129 the number of bits being extracted.
15131 2018-06-04 Jakub Jelinek <jakub@redhat.com>
15135 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
15136 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
15138 2018-06-04 Richard Biener <rguenther@suse.de>
15140 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
15141 (cleanup_tree_cfg_noloop): ... single caller. Do
15142 start_recording_case_labels later.
15144 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
15146 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
15147 to _IMMINTRIN_H_INCLUDED.
15148 * config/i386/pconfigintrin.h: Ditto.
15149 * config/i386/waitpkgintrin.h: Ditto.
15150 * config/i386/immintrin.h: Add includes for sgxintrin.h,
15151 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
15152 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
15153 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
15154 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
15155 waitpkgintrin.h and cldemoteintrin.h.
15157 2018-06-04 Richard Biener <rguenther@suse.de>
15159 PR tree-optimization/86038
15160 * tracer.c (find_best_successor): Check probability for
15161 being initialized, bail out if not.
15163 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
15166 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
15167 of bits to ignore when comparing architectures.
15169 2018-06-04 Jakub Jelinek <jakub@redhat.com>
15171 PR tree-optimization/69615
15172 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
15173 maximum or minimum of the type, try to merge it also as if
15174 range1 is + [-, x - 1] or + [x + 1, -].
15177 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
15179 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
15181 PR tree-optimization/86034
15182 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
15183 the unsigned bitfield type in a bit insertion sequence if it does not
15184 have a larger precision than the bitfield size.
15185 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
15187 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
15189 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
15191 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
15193 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
15194 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
15195 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
15196 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
15198 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
15200 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
15201 Disable -fdelete-null-pointer-checks for ELF toolchain.
15203 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
15204 Kito Cheng <kito.cheng@gmail.com>
15206 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
15207 (nds32le-*-*, nds32be-*-*): Integrate checking process.
15208 (nds32*-*-*): Add glibc and uclibc conditions.
15209 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
15210 (TARGET_EXCEPT_UNWIND_INFO): Define.
15211 * config/nds32/elf.h: New file.
15212 * config/nds32/linux.h: New file.
15213 * config/nds32/nds32-elf.opt: New file.
15214 * config/nds32/nds32-linux.opt: New file.
15215 * config/nds32/nds32-fp-as-gp.c
15216 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
15217 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
15219 (nds32_asm_file_end): Ditto.
15220 (nds32_print_operand): Ditto.
15221 (nds32_insert_attributes): Ditto.
15222 (nds32_init_libfuncs): New function.
15223 (TARGET_HAVE_TLS): Define.
15224 (TARGET_INIT_LIBFUNCS): Define.
15225 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
15227 (TARGET_ELF): Apply different mcmodel setting.
15228 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
15229 been migrated into elf.h and linux.h files.
15230 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
15231 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
15232 (mcmodel): The content has been migrated into nds32-elf.opt and
15233 nds32-linux.opt files.
15234 * config/nds32/t-elf: New file.
15235 * config/nds32/t-linux: New file.
15237 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
15238 Shiva Chen <shiva0217@gmail.com>
15240 * config/nds32/constants.md (unspec_volatile_element): Add
15241 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
15242 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
15244 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
15245 (make_pass_nds32_fp_as_gp): Declare.
15246 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
15248 (nds32_asm_function_end_prologue): Remove unused asm output.
15249 (nds32_asm_function_begin_epilogue): Remove unused asm output.
15250 (nds32_asm_file_start): Output necessary fp_as_gp information.
15251 (nds32_option_override): Adjust register usage.
15252 (nds32_expand_prologue): Consider fp_as_gp situation.
15253 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
15254 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
15257 (simple_return): Ditto.
15258 (omit_fp_begin): Output special directive for fp_as_gp.
15259 (omit_fp_end): Output special directive for fp_as_gp.
15260 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
15261 mforbid-fp-as-gp): New options.
15263 2018-06-01 Mark Wielaard <mark@klomp.org>
15265 * dwarf2out.c (dwarf2out_finish): Remove generation of
15266 DW_AT_loclists_base.
15268 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
15270 * gimple-ssa-store-merging.c: Include gimple-fold.h.
15271 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
15272 (struct merged_store_group): Add bit_insertion field.
15273 (dump_char_array): Use standard hexadecimal format.
15274 (merged_store_group::merged_store_group): Set bit_insertion to false.
15275 (merged_store_group::apply_stores): Use optimal buffer size. Deal
15276 with BIT_INSERT_EXPR stores. Move up code updating the mask and
15277 also print the mask in the dump file.
15278 (pass_store_merging::gate): Minor tweak.
15279 (imm_store_chain_info::coalesce_immediate): Fix wrong association
15280 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
15281 stores with INTEGER_CST stores.
15282 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
15283 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
15284 and use it throughout. Generate bit insertion sequences if need be.
15285 (pass_store_merging::process_store): Remove redundant condition.
15286 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
15288 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
15290 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
15291 the 128-bit floating point types. Fix function comment.
15293 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15295 * config/aarch64/aarch64-simd.md
15296 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
15298 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
15301 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
15303 PR tree-optimization/85989
15304 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
15306 (backprop::intersect_uses): Check it when deciding whether this
15307 is a backedge reference.
15308 (backprop::process_block): Add each phi to m_visited_phis
15309 after visiting it, then clear it at the end.
15311 2018-06-01 Richard Biener <rguenther@suse.de>
15313 * tree-vectorizer.h (vect_dr_stmt): New function.
15314 (vect_get_load_cost): Adjust.
15315 (vect_get_store_cost): Likewise.
15316 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
15317 Use vect_dr_stmt instead of DR_SMTT.
15318 (vect_record_base_alignments): Likewise.
15319 (vect_calculate_target_alignment): Likewise.
15320 (vect_compute_data_ref_alignment): Likewise and make static.
15321 (vect_update_misalignment_for_peel): Likewise.
15322 (vect_verify_datarefs_alignment): Likewise.
15323 (vector_alignment_reachable_p): Likewise.
15324 (vect_get_data_access_cost): Likewise. Pass down
15325 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
15326 (vect_get_peeling_costs_all_drs): Likewise.
15327 (vect_peeling_hash_get_lowest_cost): Likewise.
15328 (vect_enhance_data_refs_alignment): Likewise.
15329 (vect_find_same_alignment_drs): Likewise.
15330 (vect_analyze_data_refs_alignment): Likewise.
15331 (vect_analyze_group_access_1): Likewise.
15332 (vect_analyze_group_access): Likewise.
15333 (vect_analyze_data_ref_access): Likewise.
15334 (vect_analyze_data_ref_accesses): Likewise.
15335 (vect_vfa_segment_size): Likewise.
15336 (vect_small_gap_p): Likewise.
15337 (vectorizable_with_step_bound_p): Likewise.
15338 (vect_prune_runtime_alias_test_list): Likewise.
15339 (vect_analyze_data_refs): Likewise.
15340 (vect_supportable_dr_alignment): Likewise.
15341 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
15342 (vect_gen_prolog_loop_niters): Likewise.
15343 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
15344 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
15346 (vect_recog_mask_conversion_pattern): Likewise.
15347 (vect_try_gather_scatter_pattern): Likewise.
15348 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
15349 to vect_get_store_cost.
15350 (vect_get_store_cost): Get stmt_info instead of DR.
15351 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
15352 (vect_get_load_cost): Get stmt_info instead of DR.
15354 2018-06-01 Richard Biener <rguenther@suse.de>
15356 PR middle-end/86017
15357 * gimple-fold.c (var_decl_component_p): Also allow offsetted
15358 vars wrapped in MEM_REFs.
15360 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
15362 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
15363 Fix subreg tests so that we only return a choice between
15364 GENERAL_REGS and FP_REGS if the original classes included both.
15366 2018-06-01 Richard Biener <rguenther@suse.de>
15369 * tree-ssa-structalias.c (get_function_part_constraint):
15370 Handle NULL fi->decl.
15371 (find_func_aliases_for_call): Properly handle indirect
15372 fi from direct call.
15373 (find_func_clobbers): Likewise.
15374 (ipa_pta_execute): Likewise.
15375 (create_variable_info_for): For functions that are ifunc_resolver
15376 resolve to a varinfo that contains the result of the resolver call.
15377 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
15380 2018-05-31 Michael Collison <michael.collison@arm.com>
15382 * config/aarch64/aarch64.md:
15383 (*fix_to_zero_extenddfdi2): New pattern.
15384 * gcc.target/aarch64/fix_extend1.c: New testcase.
15386 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
15388 PR middle-end/78809
15389 PR middle-end/83026
15390 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
15391 and BUILT_IN_STRNCMP_EQ.
15392 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
15393 BUILT_IN_STRNCMP_EQ.
15394 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
15395 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
15396 (gimple_fold_builtin): Likewise.
15397 * tree-ssa-strlen.c (compute_string_length): New function.
15398 (determine_min_obsize): New function.
15399 (handle_builtin_string_cmp): New function to handle calls to
15400 string compare functions.
15401 (strlen_optimize_stmt): Add handling to builtin string compare
15403 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
15404 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
15405 * tree.c (build_common_builtin_nodes): Add new defines of
15406 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
15408 2018-05-31 Jakub Jelinek <jakub@redhat.com>
15411 * bb-reorder.c (pass_partition_blocks::gate): Return false for
15412 functions with naked attribute.
15414 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
15416 * config/i386/sse.md (avx_vec_concat<mode>):
15417 Substitute concat_tg_mode mode attribute with xtg_mode.
15418 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
15419 (concat_tg_mode): Remove mode attribute.
15421 2018-05-31 Martin Sebor <msebor@redhat.com>
15424 * calls.c (alloc_max_size): Correct a logic error/typo.
15425 Treat excessive arguments as infinite. Warn for invalid arguments.
15426 * doc/invoke.texi (-Walloc-size-larger-than): Update.
15428 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
15431 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
15432 and movx for Haswell.
15434 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
15435 Cesar Philippidis <cesar@codesourcery.com>
15437 PR middle-end/85879
15438 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
15439 when emitting error on private/firstprivate reductions.
15440 * omp-low.c (lower_omp_target): Avoid reference-type processing
15441 on pointers for firstprivate clause.
15443 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
15445 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
15448 * config/aarch64/aarch64-simd.md
15449 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
15450 (aarch64_ld1_x3_<mode>): Likewise
15451 (aarch64_st1x2<VALLDIF:mode>): Likewise
15452 (aarch64_st1_x2_<mode>): Likewise
15453 (aarch64_st1x3<VALLDIF:mode>): Likewise
15454 (aarch64_st1_x3_<mode>): Likewise
15455 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
15456 (vld1_s8_x3): Likewise.
15457 (vld1_u16_x3): Likewise.
15458 (vld1_s16_x3): Likewise.
15459 (vld1_u32_x3): Likewise.
15460 (vld1_s32_x3): Likewise.
15461 (vld1_u64_x3): Likewise.
15462 (vld1_s64_x3): Likewise.
15463 (vld1_f16_x3): Likewise.
15464 (vld1_f32_x3): Likewise.
15465 (vld1_f64_x3): Likewise.
15466 (vld1_p8_x3): Likewise.
15467 (vld1_p16_x3): Likewise.
15468 (vld1_p64_x3): Likewise.
15469 (vld1q_u8_x3): Likewise.
15470 (vld1q_s8_x3): Likewise.
15471 (vld1q_u16_x3): Likewise.
15472 (vld1q_s16_x3): Likewise.
15473 (vld1q_u32_x3): Likewise.
15474 (vld1q_s32_x3): Likewise.
15475 (vld1q_u64_x3): Likewise.
15476 (vld1q_s64_x3): Likewise.
15477 (vld1q_f16_x3): Likewise.
15478 (vld1q_f32_x3): Likewise.
15479 (vld1q_f64_x3): Likewise.
15480 (vld1q_p8_x3): Likewise.
15481 (vld1q_p16_x3): Likewise.
15482 (vld1q_p64_x3): Likewise.
15483 (vst1_s64_x2): Likewise.
15484 (vst1_u64_x2): Likewise.
15485 (vst1_f64_x2): Likewise.
15486 (vst1_s8_x2): Likewise.
15487 (vst1_p8_x2): Likewise.
15488 (vst1_s16_x2): Likewise.
15489 (vst1_p16_x2): Likewise.
15490 (vst1_s32_x2): Likewise.
15491 (vst1_u8_x2): Likewise.
15492 (vst1_u16_x2): Likewise.
15493 (vst1_u32_x2): Likewise.
15494 (vst1_f16_x2): Likewise.
15495 (vst1_f32_x2): Likewise.
15496 (vst1_p64_x2): Likewise.
15497 (vst1q_s8_x2): Likewise.
15498 (vst1q_p8_x2): Likewise.
15499 (vst1q_s16_x2): Likewise.
15500 (vst1q_p16_x2): Likewise.
15501 (vst1q_s32_x2): Likewise.
15502 (vst1q_s64_x2): Likewise.
15503 (vst1q_u8_x2): Likewise.
15504 (vst1q_u16_x2): Likewise.
15505 (vst1q_u32_x2): Likewise.
15506 (vst1q_u64_x2): Likewise.
15507 (vst1q_f16_x2): Likewise.
15508 (vst1q_f32_x2): Likewise.
15509 (vst1q_f64_x2): Likewise.
15510 (vst1q_p64_x2): Likewise.
15511 (vst1_s64_x3): Likewise.
15512 (vst1_u64_x3): Likewise.
15513 (vst1_f64_x3): Likewise.
15514 (vst1_s8_x3): Likewise.
15515 (vst1_p8_x3): Likewise.
15516 (vst1_s16_x3): Likewise.
15517 (vst1_p16_x3): Likewise.
15518 (vst1_s32_x3): Likewise.
15519 (vst1_u8_x3): Likewise.
15520 (vst1_u16_x3): Likewise.
15521 (vst1_u32_x3): Likewise.
15522 (vst1_f16_x3): Likewise.
15523 (vst1_f32_x3): Likewise.
15524 (vst1_p64_x3): Likewise.
15525 (vst1q_s8_x3): Likewise.
15526 (vst1q_p8_x3): Likewise.
15527 (vst1q_s16_x3): Likewise.
15528 (vst1q_p16_x3): Likewise.
15529 (vst1q_s32_x3): Likewise.
15530 (vst1q_s64_x3): Likewise.
15531 (vst1q_u8_x3): Likewise.
15532 (vst1q_u16_x3): Likewise.
15533 (vst1q_u32_x3): Likewise.
15534 (vst1q_u64_x3): Likewise.
15535 (vst1q_f16_x3): Likewise.
15536 (vst1q_f32_x3): Likewise.
15537 (vst1q_f64_x3): Likewise.
15538 (vst1q_p64_x3): Likewise.
15540 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
15542 * config/msp430/msp430.c (msp430_output_labelref): Prepend
15543 user_label_prefix to name.
15545 * tree-core.h: Update comment about the format of NAME string
15546 passed to handler in attribute_spec.
15548 * config/msp430/msp430.md: Remove erroneous subreg expression from
15549 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
15550 zero_extend{q,h}isi2.
15552 2018-05-30 Borislav Petkov <bp@suse.de>
15554 * doc/extend.texi: Document some architecture specific
15555 constraints and sort entries.
15557 2018-05-30 Martin Sebor <msebor@redhat.com>
15559 PR middle-end/85369
15560 * builtins.c (expand_builtin_stpcpy_1): New function.
15561 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
15562 only if the former succeeds.
15564 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
15566 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
15569 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
15571 * doc/invoke.texi (-flinker-output): Document
15573 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
15575 * passes.c (ipa_write_summaries): Only modify statements if body
15577 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
15578 incrementally linking.
15579 (ipa_passes): Likewise.
15580 * lto-cgraph.c (lto_output_node): When incrementally linking do not
15581 pass down resolution info.
15582 * common.opt (flag_incremental_link): Update info.
15583 * gcc.c (plugin specs): Turn flinker-output=* to
15584 -plugin-opt=-linker-output-known
15585 * toplev.c (compile_file): Also cut compilation when doing incremental
15587 * flag-types. (enum lto_partition_model): Add
15588 LTO_LINKER_OUTPUT_NOLTOREL.
15589 (invoke.texi): Add -flinker-output docs.
15590 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
15591 link same way as WPA; do not stream in dead initializers.
15593 * dwarf2out.c (dwarf2out_die_ref_for_decl,
15594 darf2out_register_external_decl): Support incremental link.
15596 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
15598 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
15600 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
15602 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
15603 it down to simple_object_copy_lto_debug_sections.
15604 (run_gcc): Determine incremental LTO link time and configure
15605 lto1 into non-wpa mode, disable renaming of debug sections.
15607 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
15609 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
15610 descriptions of various incorrectly documented functions.
15612 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
15615 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
15616 address check not strict.
15618 2018-05-30 Richard Biener <rguenther@suse.de>
15620 PR tree-optimization/85964
15621 * tracer.c (better_p): Drop initialized count check, we only
15622 call the function with initialized counts now.
15623 (find_best_successor): Do find a best edge if one
15624 has uninitialized count.
15625 (find_best_predecessor): Likewise. Do BB frequency check only
15626 if count is initialized.
15628 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
15630 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
15631 (aarch64_ldrstr_offset_compare): New.
15632 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
15633 load/store orderings.
15634 (aarch64_gen_adjusted_ldpstp): Likewise.
15636 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
15638 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
15639 Check for subset of GENERAL_REGS and FP_REGS.
15640 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
15643 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
15645 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
15646 and wi::to_poly_offset. Add the current offset and then check
15647 whether the sum fits, rather than using an unchecked addition of
15648 a checked term. Check for a shwi rather than a uhwi.
15649 * expr.c (get_bit_range): Use tree_to_poly_uint64.
15650 (store_constructor): Use poly_int_tree_p.
15651 (expand_expr_real_1): Likewise.
15652 * function.c (assign_temp): Likewise.
15653 * fold-const.c (const_binop): Use poly_int_tree_p and
15654 wi::to_poly_offset.
15655 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
15657 * ipa-icf-gimple.c (func_checker::compare_operand): Use
15658 to_poly_offset for MEM offsets.
15659 * ipa-icf.c (sem_variable::equals): Likewise.
15660 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
15661 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
15662 wi::to_poly_offset for BIT_FIELD_REF offsets.
15663 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
15664 wi::to_poly_offset.
15665 * var-tracking.c (emit_note_insn_var_location): Use
15666 tree_to_poly_uint64.
15668 2018-05-29 Jim Wilson <jimw@sifive.com>
15670 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
15672 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
15675 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
15676 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
15678 (sse4_1_round<mode>2): Use nonimmediate_operand
15679 for operand 1 predicate.
15681 2018-05-29 Martin Sebor <msebor@redhat.com>
15682 Richard Biener <rguenther@suse.de>
15685 * calls.c (get_size_range): Call determine_value_range instead
15686 of get_value_range..
15687 * tree-vrp.h (determine_value_range): Declared new function.
15688 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
15690 2018-05-29 Richard Biener <rguenther@suse.de>
15692 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
15693 sure to use non-pattern stmts for get_earlier_stmt arguments.
15694 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
15695 called on pattern stmts.
15696 (get_later_stmt): Likewise.
15698 2018-05-29 Martin Liska <mliska@suse.cz>
15700 PR gcov-profile/85759
15701 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
15704 2018-05-29 Jakub Jelinek <jakub@redhat.com>
15706 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
15708 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
15712 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
15713 VEC_PACK_FLOAT_EXPR): New tree codes.
15714 * tree-pretty-print.c (op_code_prio): Handle
15715 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
15716 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
15717 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
15718 * tree-inline.c (estimate_operator_cost): Likewise.
15719 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
15720 * fold-const.c (const_binop): Likewise.
15721 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
15722 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
15723 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
15724 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
15725 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
15726 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
15727 * expr.c (expand_expr_real_2): Likewise.
15728 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
15729 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
15730 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
15732 * optabs.c (expand_widen_pattern_expr): For
15733 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
15734 sign from result type rather than operand's type.
15735 (expand_binop_directly): For vec_packu_float_optab and
15736 vec_packs_float_optab allow result type to be different from operand's
15738 * optabs-tree.c (optab_for_tree_code): Handle
15739 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
15740 VEC_PACK_FLOAT_EXPR. Formatting fixes.
15741 * tree-vect-generic.c (expand_vector_operations_1): Handle
15742 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
15743 VEC_PACK_FLOAT_EXPR.
15744 * tree-vect-stmts.c (supportable_widening_operation): Handle
15746 (supportable_narrowing_operation): Handle FLOAT_EXPR.
15747 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
15748 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
15749 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
15750 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
15752 (vec_pack<floatprefix>_float_<mode>): New expander.
15753 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
15755 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
15756 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
15757 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
15758 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
15759 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
15761 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
15762 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
15763 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
15764 VEC_PACK_FLOAT_EXPR): Document.
15766 2018-05-29 Richard Biener <rguenther@suse.de>
15768 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
15770 (stmt_vec_info_vec): Make pointer.
15771 (init_stmt_vec_info_vec): Remove.
15772 (free_stmt_vec_info_vec): Likewise.
15773 (set_stmt_vec_info_vec): New function.
15774 (free_stmt_vec_infos): Likewise.
15775 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
15776 (set_vinfo_for_stmt): Likewise.
15777 (get_earlier_stmt): Likewise.
15778 (get_later_stmt): Likewise.
15779 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
15780 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
15781 (vec_info::~vec_info): Free stmt_vec_infos.
15782 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
15783 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
15784 (pass_slp_vectorize::execute): Likewise.
15785 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
15786 (free_stmt_vec_info_vec): Likewise.
15787 (set_stmt_vec_info_vec): New function.
15788 (free_stmt_vec_infos): Likewise.
15789 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
15790 the global stmt_vec_info_vec.
15791 * tree-parloops.c (gather_scalar_reductions): Use
15792 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
15795 2018-05-29 Richard Biener <rguenther@suse.de>
15797 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
15799 2018-05-29 Martin Liska <mliska@suse.cz>
15800 David Malcolm <dmalcolm@redhat.com>
15802 * vec.c (test_reverse): New.
15803 (vec_c_tests): Add new test.
15804 * vec.h (vl_ptr>::reverse): New function.
15806 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
15808 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
15810 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
15813 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
15815 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
15817 2018-05-28 Richard Biener <rguenther@suse.de>
15819 PR tree-optimization/85933
15820 * tree-vect-data-refs.c (vect_record_base_alignments): Only
15821 look at stmts marked as vectorizable.
15823 2018-05-28 Richard Biener <rguenther@suse.de>
15825 PR tree-optimization/85934
15826 * tree-vect-generic.c (expand_vector_operations_1): Hoist
15827 vector boolean check before scalar optimization.
15829 2018-05-28 Jakub Jelinek <jakub@redhat.com>
15831 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
15834 2018-05-28 Mark Wielaard <mark@klomp.org>
15836 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
15837 if it is an expression containing a minus sign.
15839 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
15841 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
15843 2018-05-27 Paul Koning <ni1d@arrl.net>
15845 * config/pdp11/pdp11.md (truncsihi2): Remove.
15847 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
15848 Chung-Ju Wu <jasonwucj@gmail.com>
15850 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
15852 (unaligned_store_dw): Ditto.
15853 * config/nds32/nds32-memory-manipulation.c
15854 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
15855 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
15856 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
15857 (emit_setmem_word_loop): Rename to ...
15858 (emit_setmem_doubleword_loop): ... this.
15859 (nds32_gen_dup_4_byte_to_word_value): New function.
15860 (nds32_gen_dup_8_byte_to_double_word_value): New function.
15861 (nds32_expand_setmem_loop): Refine implementation.
15862 (nds32_expand_setmem_loop_v3m): Ditto.
15863 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
15866 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
15868 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
15870 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
15872 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
15873 (nds32_init_machine_status): Initialize machine->attr_naked_p and
15874 machine->attr_no_prologue_p.
15875 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
15876 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
15877 (nds32_expand_epilogue): Consider attr_naked_p.
15878 (nds32_expand_epilogue_v3pop): Likewise.
15879 (nds32_can_use_return_insn): Likewise.
15880 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
15881 attr_no_prologue_p fields.
15882 * config/nds32/nds32.opt (mret-in-naked-func): New option.
15884 2018-05-27 Jakub Jelinek <jakub@redhat.com>
15887 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
15889 * config/i386/sse.md
15890 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
15892 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
15894 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
15896 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
15898 (*<floatsuffix>floatv2div2sf2): Rename to ...
15899 (*float<floatunssuffix>v2div2sf2): ... this.
15900 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
15901 (float<floatunssuffix>v2div2sf2_mask): ... this.
15902 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
15903 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
15904 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
15906 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
15908 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
15910 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
15912 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
15914 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
15916 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
15917 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
15918 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
15919 gen_ufix_truncv8dfv8si2.
15920 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
15921 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
15922 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
15923 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
15924 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
15925 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
15926 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
15927 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
15929 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
15933 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
15935 2018-05-25 Jim Wilson <jimw@sifive.com>
15937 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
15938 * config/riscv/riscv.c (struct machine_function): Add
15939 interrupt_handler_p and attribute_checked_p fields.
15940 (riscv_attribute_table): Add interrupt.
15941 (riscv_interrupt_type_p): New.
15942 (riscv_save_reg_p): Save extra regs for interrupt handler.
15943 (riscv_use_save_libcall): Return false for interrupt handler.
15944 (riscv_first_stack_step): Add forward declaration.
15945 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
15946 for interrupt handler with large frame. Use it for saved reg list.
15947 (riscv_expand_prologue): Move flag_stack_usage_info support to
15948 eliminate duplication.
15949 (riscv_expand_epilogue): Generate mret for interrupt handler.
15950 (riscv_epilogue_uses): New.
15951 (riscv_can_use_return_insn): Return false for interrupt handler.
15952 (riscv_function_ok_for_sibcall): Likewise.
15953 (riscv_set_current_function): Add interrupt handler support.
15954 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
15955 * config/riscv/riscv.md (UNSPECV_MRET): New.
15957 (riscv_frflags, riscv_fsflags): Use tab after opcode.
15959 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
15961 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
15963 PR tree-optimization/85712
15964 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
15965 this candidate has already been replaced in-situ by a copy.
15967 2018-05-25 Jason Merrill <jason@redhat.com>
15969 PR c++/80485 - inline function non-zero address.
15970 * symtab.c (nonzero_address): Check DECL_COMDAT.
15972 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
15975 * config/alpha/alpha.md (ashlsi3): New insn pattern.
15976 (*ashlsi_se): Rename from *ashldi_se. Define as sign
15977 extension of SImode operation. Use const123_operand predicate.
15978 (*saddsi_1): Remove.
15979 (*saddl_se_1): Ditto.
15980 (*ssubsi_1): Ditto.
15981 (*ssubl_se_1): Ditto.
15982 * config/alpha/predicates.md (const123_operand): New predicate.
15983 * config/alpha/constraints.md (P): Use IN_RANGE.
15985 2018-05-25 Richard Biener <rguenther@suse.de>
15987 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
15989 (ref_maybe_used_by_stmt_p): Likewise.
15990 (stmt_may_clobber_ref_p): Likewise.
15991 (stmt_may_clobber_ref_p_1): Likewise.
15992 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
15994 (ref_maybe_used_by_stmt_p): Likewise.
15995 (stmt_may_clobber_ref_p): Likewise.
15996 (stmt_may_clobber_ref_p_1): Likewise.
15997 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
15998 the alias oracle to disambiguate DRs with stmts DR analysis
16000 (vect_analyze_data_refs): Do not give up on not analyzable
16001 DRs for BB vectorization. Remove code truncating the dataref
16004 2018-05-25 Jakub Jelinek <jakub@redhat.com>
16007 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
16008 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
16009 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
16011 2018-05-25 Richard Biener <rguenther@suse.de>
16013 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
16014 function, combining stmt data ref gathering and fatal analysis
16016 (vect_analyze_data_refs): Remove now redudnant code and simplify.
16017 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
16018 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
16019 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
16020 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
16022 2018-05-25 Bin Cheng <bin.cheng@arm.com>
16024 PR tree-optimization/85720
16025 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
16026 SCC if all partitions are builtins.
16027 (version_loop_by_alias_check): New parameter. Generate cancelable
16028 runtime alias check if all partitions are builtins.
16029 (distribute_loop): Update call to above function.
16031 2018-05-25 Bin Cheng <bin.cheng@arm.com>
16033 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
16034 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
16035 (parm_default_def_partition_arg): Ditto.
16036 (set_parm_default_def_partition): Ditto.
16037 (get_parm_default_def_partitions): Ditto and make it static.
16038 (get_undefined_value_partitions): Ditto and make it static.
16039 (remove_ssa_form): Refactor call to init_var_map here.
16040 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
16041 computation for loop region.
16042 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
16043 (register_default_def): Delete.
16044 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
16045 (parm_default_def_partition_arg): Ditto.
16046 (set_parm_default_def_partition): Ditto.
16047 (get_parm_default_def_partitions): Ditto and make it static.
16048 (get_undefined_value_partitions): Ditto and make it static.
16049 (coalesce_with_default, coalesce_with_default): Update comment.
16050 (create_coalesce_list_for_region): New func factored out from
16051 create_outofssa_var_map.
16052 (populate_coalesce_list_for_outofssa): New func factored out from
16053 create_outofssa_var_map and coalesce_ssa_name.
16054 (create_outofssa_var_map): Delete.
16055 (coalesce_ssa_name): Refactor to support live range computation.
16056 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
16057 (get_parm_default_def_partitions): Delete.
16058 (get_undefined_value_partitions): Ditto.
16059 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
16060 computation for loop region.
16061 (new_tree_live_info, loe_visit_block): Ditto.
16062 (live_worklist, set_var_live_on_entry): Ditto.
16063 (calculate_live_on_exit, verify_live_on_entry): Ditto.
16064 * tree-ssa-live.h (struct _var_map): New fields.
16065 (init_var_map): Change decl.
16066 (region_contains_p): New.
16068 2018-05-25 Bin Cheng <bin.cheng@arm.com>
16070 * tree-ssa-live.h (live_merge_and_clear): Delete.
16072 2018-05-25 Richard Biener <rguenther@suse.de>
16075 * tree-dump.c (dequeue_and_dump): Remove access to removed
16076 operand 2 of a SWITCH_EXPR.
16078 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
16080 * doc/sourcebuild.texi (vect_double_cond_arith): Include
16081 multiplication and division.
16082 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
16083 (cond_udiv@var{m}, cond_umod@var{m}): Document.
16084 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
16085 (cond_udiv_optab, cond_umod_optab): New optabs.
16086 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
16087 (IFN_COND_RDIV): New internal functions.
16088 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
16089 TRUNC_MOD_EXPR and RDIV_EXPR.
16090 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
16091 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
16093 (SVE_INT_BINARY): Include mult.
16094 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
16095 (optab, sve_int_op): Handle mult.
16096 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
16098 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
16099 for SVE_INT_BINARY_SD.
16101 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
16103 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
16104 (optab, sve_int_op): Handle div and udiv.
16105 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
16106 for SVE_INT_BINARY_SD.
16107 (*<optab><mode>3): New insn for the same.
16109 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
16111 * tree-vect-patterns.c: Include predict.h.
16112 (vect_recog_divmod_pattern): Restrict check for division support
16113 to when optimizing for size.
16115 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
16117 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
16118 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
16119 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
16120 (gimple_match_op::set_op): Likewise.
16121 (gimple_resimplify4): Declare.
16122 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
16123 (expr::gen_transform): Likewise.
16124 (decision_tree::gen): Generate a simplification routine for 4 operands.
16125 * gimple-match-head.c (gimple_simplify): Add an overload for
16126 4 operands. In the top-level function, handle up to 4 call
16127 arguments and call gimple_resimplify4.
16128 (gimple_resimplify4): New function.
16129 (build_call_internal): Pass a fourth operand.
16130 (maybe_push_to_seq): Likewise.
16131 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
16132 Fold VEC_COND_EXPRs of an operation and a default value into
16133 an IFN_COND_* function if possible.
16134 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
16136 (SVE_COND_FP_BINARY): Include them.
16137 (optab, sve_fp_op): Handle them.
16138 (SVE_INT_BINARY_REV): New code iterator.
16139 (SVE_COND_FP_BINARY_REV): New int iterator.
16140 (commutative): New int attribute.
16141 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
16143 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
16145 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
16146 (*cond_<optab><mode>): New patterns for reversed operands.
16148 2018-05-25 Richard Biener <rguenther@suse.de>
16150 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
16151 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
16152 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
16153 (STMT_VINFO_GROUPED_ACCESS): Adjust.
16154 * tree-vect-data-refs.c (everywhere): Adjust users.
16155 * tree-vect-loop.c (everywhere): Likewise.
16156 * tree-vect-slp.c (everywhere): Likewise.
16157 * tree-vect-stmts.c (everywhere): Likewise.
16158 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
16160 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16162 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
16164 (gcc_cv_as_section_exclude): ... this.
16165 Try Solaris as #exclude syntax.
16166 * configure: Regenerate.
16167 * config.in: Regenerate.
16168 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
16170 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
16171 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
16173 * varasm.c (default_elf_asm_named_section): Don't check if
16174 HAVE_GAS_SECTION_EXCLUDE is defined.
16176 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
16178 * doc/md.texi: Update the documentation of the cond_* optabs
16179 to mention the new final operand. Fix GET_MODE_NUNITS call.
16180 Describe the scalar case too.
16181 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
16182 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
16184 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
16185 (get_conditional_internal_fn): Update comment.
16186 * tree-vect-loop.c (vectorizable_reduction): Pass the original
16187 accumulator value as a final argument to conditional functions.
16188 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
16189 a define_expand and add an "else" operand. Assert for now that
16190 the else operand is equal to operand 2. Use SVE_INT_BINARY and
16191 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
16192 (*cond_<optab><mode>): New patterns.
16193 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
16194 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
16195 (UNSPEC_COND_EOR): Delete.
16196 (optab): Remove associated mappings.
16197 (SVE_INT_BINARY): New code iterator.
16198 (sve_int_op): Remove int attribute and add "minus" to the code
16200 (SVE_COND_INT_OP): Delete.
16201 (SVE_COND_FP_OP): Rename to...
16202 (SVE_COND_FP_BINARY): ...this.
16204 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
16206 * optabs.c (can_reuse_operands_p): New function.
16207 (maybe_legitimize_operands): Try to reuse the results for
16210 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
16212 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
16213 Add {q} suffix to insn mnemonic.
16215 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
16217 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
16218 (msp430_warn_func_return): New.
16220 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
16222 * fold-const.c (tree_nonzero_bits): New function.
16223 * fold-const.h (tree_nonzero_bits): Likewise.
16224 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
16225 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
16227 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
16231 * varasm.c (assemble_alias): Check ifunc_resolver only on
16234 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
16237 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
16238 when memory input operand is handled.
16240 2018-05-24 Luis Machado <luis.machado@linaro.org>
16242 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
16244 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
16246 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
16248 * match.pd: Delay FMA folds until after vectorization.
16250 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
16253 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
16254 address check not strict.
16256 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
16258 * gimple-match.h (gimple_match_op): New class.
16259 (mprts_hook): Replace parameters with a gimple_match_op *.
16260 (maybe_build_generic_op): Likewise.
16261 (gimple_simplified_result_is_gimple_val): Replace parameters with
16262 a const gimple_match_op *.
16263 (gimple_simplify): Replace code_helper * and tree * parameters with
16264 a gimple_match_op * parameter.
16265 (gimple_resimplify1): Replace code_helper *, tree and tree *
16266 parameters with a gimple_match_op * parameter.
16267 (gimple_resimplify2): Likewise.
16268 (gimple_resimplify3): Likewise.
16269 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
16270 parameters with a gimple_match_op * parameter.
16271 * gimple-match-head.c (gimple_simplify): Change prototypes of
16272 auto-generated functions to take a gimple_match_op * instead of
16273 separate code_helper * and tree * parameters. Make the same
16274 change in the top-level overload and update calls to the
16275 gimple_resimplify routines. Update calls to the auto-generated
16276 functions and to maybe_push_res_to_seq in the publicly-facing
16277 operation-specific gimple_simplify overloads.
16278 (gimple_match_op::MAX_NUM_OPS): Define.
16279 (gimple_resimplify1): Replace rcode and ops with a single res_op
16280 parameter. Update call to gimple_simplify.
16281 (gimple_resimplify2): Likewise.
16282 (gimple_resimplify3): Likewise.
16283 (mprts_hook): Replace parameters with a gimple_match_op *.
16284 (maybe_build_generic_op): Likewise.
16285 (build_call_internal): Replace type, nargs and ops with
16286 a gimple_match_op *.
16287 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
16288 with a single gimple_match_op *. Update calls to mprts_hook,
16289 build_call_internal and gimple_simplified_result_is_gimple_val.
16290 Factor out code that is common to the tree_code and combined_fn cases.
16291 * genmatch.c (expr::gen_transform): Replace tem_code and
16292 tem_ops with a gimple_match_op called tem_op. Update calls
16293 to the gimple_resimplify functions and maybe_push_res_to_seq.
16294 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
16295 res_ops. Update call to the gimple_resimplify functions.
16296 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
16297 (decision_tree::gen): Make the functions take a gimple_match_op *
16298 called res_op instead of separate res_code and res_ops parameters.
16299 Update call accordingly.
16300 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
16301 and ops with a single res_op parameter. Update calls to
16302 maybe_build_generic_op and maybe_push_res_to_seq.
16303 (fold_stmt_1): Update calls to gimple_simplify and
16304 replace_stmt_with_simplification.
16305 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
16306 and gimple_simplified_result_is_gimple_val.
16307 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
16309 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
16310 with a gimple_match_op *.
16311 (vn_nary_build_or_lookup): Likewise. Update call to
16312 vn_nary_build_or_lookup_1.
16313 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
16314 gimple_match_op *. Update calls to the gimple_resimplify routines
16315 and to gimple_simplified_result_is_gimple_val.
16316 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
16317 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
16318 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
16319 (visit_nary_op): Likewise.
16320 (visit_reference_op_load): Likewise.
16322 2018-05-23 Luis Machado <luis.machado@linaro.org>
16324 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
16325 modifier for printing the step amount.
16327 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
16330 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
16333 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
16335 * doc/sourcebuild.texi (Endianness): New subsubsection.
16337 2018-05-23 Luis Machado <luis.machado@linaro.org>
16339 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
16340 <prefetch_dynamic_strides>: New const bool field.
16341 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
16342 prefetch_dynamic_strides.
16343 (exynosm1_prefetch_tune): Likewise.
16344 (thunderxt88_prefetch_tune): Likewise.
16345 (thunderx_prefetch_tune): Likewise.
16346 (thunderx2t99_prefetch_tune): Likewise.
16347 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
16349 (aarch64_override_options_internal): Update to set
16350 PARAM_PREFETCH_DYNAMIC_STRIDES.
16351 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
16352 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
16353 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
16354 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
16355 prefetch-dynamic-strides setting.
16357 2018-05-23 Luis Machado <luis.machado@linaro.org>
16359 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
16360 <minimum_stride>: New const int field.
16361 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
16362 minimum_stride field defaulting to -1.
16363 (exynosm1_prefetch_tune): Likewise.
16364 (thunderxt88_prefetch_tune): Likewise.
16365 (thunderx_prefetch_tune): Likewise.
16366 (thunderx2t99_prefetch_tune): Likewise.
16367 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
16368 <default_opt_level>: Set to 3.
16369 (aarch64_override_options_internal): Update to set
16370 PARAM_PREFETCH_MINIMUM_STRIDE.
16371 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
16372 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
16373 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
16374 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
16375 stride is constant and is below the minimum stride threshold.
16377 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16379 * config/arm/arm-cpus.in (mode26): Delete.
16380 (armv4): Delete mode26 reference.
16381 * config/arm/arm.c (arm_configure_build_target): Delete use of
16384 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
16386 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
16388 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
16389 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
16391 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
16392 (floatunsdidf2): Ditto.
16394 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
16396 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
16397 (fixuns_trunc<mode>si2_avx512f): Ditto.
16398 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
16399 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
16400 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
16402 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
16404 PR rtl-optimization/79985
16405 * df-scan.c (df_insn_refs_collect): Remove special case for
16406 global registers and asm statements.
16408 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
16410 * extend.texi (Global Register Variables): Rewrite the bullet list.
16411 Note that the register is available for allocation. Note that access
16412 via inline asm must use constraints. Add note about async-signal
16413 handlers. Remove paragraph about automagic register selection.
16415 2018-05-23 Richard Biener <rguenther@suse.de>
16417 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
16418 of fixed offset from memset VN.
16420 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
16422 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
16423 first_interp field.
16424 (alloc_cand_and_find_basis): Initialize first_interp field.
16425 (slsr_process_mul): Modify first_interp field.
16426 (slsr_process_add): Likewise.
16427 (slsr_process_cast): Modify first_interp field for each new
16429 (slsr_process_copy): Likewise.
16430 (dump_candidate): Dump first_interp field.
16431 (replace_mult_candidate): Process all interpretations, not just
16433 (replace_rhs_if_not_dup): Likewise.
16434 (replace_one_candidate): Likewise.
16436 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
16438 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
16440 (aarch64_needs_frame_chain): New function.
16441 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
16443 2018-05-23 Sudakshina Das <sudi.das@arm.com>
16446 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
16447 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
16448 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
16449 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
16450 as true for strict-align.
16451 (aarch64_can_inline_p): Perform checks even when callee has no
16452 attributes to check for strict alignment.
16453 * doc/extend.texi (AArch64 Function Attributes): Document
16455 * doc/invoke.texi: (AArch64 Options): Likewise.
16457 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
16459 PR tree-optimization/85853
16460 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
16461 the handling of the root of the node to...
16462 (vect_slp_analyze_node_operations_1): ...this new function,
16463 and run the whole thing with the child nodes' def types
16464 set according to their SLP node's def type.
16466 2018-05-23 Richard Biener <rguenther@suse.de>
16468 PR middle-end/85874
16469 * tree-data-ref.c (create_runtime_alias_checks): Defer
16470 and ignore overflow warnings.
16472 2018-05-23 Yury Gribov <tetra2005@gmail.com>
16474 PR tree-optimization/85822
16475 * tree-vrp.c (is_masked_range_test): Fix handling of negative
16478 2018-05-23 Richard Biener <rguenther@suse.de>
16480 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
16481 memset constants via native_interpret_expr.
16483 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
16486 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
16488 (cgraph_node::create_alias): Likewise.
16489 (cgraph_node::get_availability): Check ifunc_resolver instead
16490 of looking up ifunc attribute.
16491 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
16492 * varasm.c (do_assemble_alias): Likewise.
16493 (assemble_alias): Likewise.
16494 (default_binds_local_p_3): Likewise.
16495 * cgraph.h (cgraph_node): Add ifunc_resolver.
16496 (cgraph_node::only_called_directly_or_aliased_p): Return false
16497 for IFUNC resolver.
16498 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
16500 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
16501 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
16502 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
16503 instead of looking up ifunc attribute.
16505 2018-05-22 Luis Machado <luis.machado@linaro.org>
16507 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
16509 2018-05-22 Martin Sebor <msebor@redhat.com>
16511 PR middle-end/85359
16512 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
16513 only when expasion succeeds.
16514 (expand_builtin_strcmp): Same.
16515 (expand_builtin_strncmp): Same.
16517 2018-05-22 Martin Sebor <msebor@redhat.com>
16519 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
16521 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
16522 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16524 * config/aarch64/aarch64-ldpstp.md: Replace uses of
16525 aarch64_mem_pair_operand with memory_operand and delete operand swapping
16527 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
16528 Add check for legitimate_address.
16529 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
16530 (aarch64_swap_ldrstr_operands): New.
16531 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
16534 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
16535 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16537 * config/aarch64/aarch64.md: New patterns to generate stp
16539 (store_pair_sw, store_pair_dw): New patterns to generate stp for
16540 single words and double words.
16541 (load_pair_sw, load_pair_dw): Likewise.
16542 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
16544 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
16546 * config/aarch64/aarch64-ldpstp.md: Modify peephole
16547 for different mode ldpstp and add peephole for merged zero stores.
16548 Likewise for loads.
16549 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
16551 (aarch64_gen_store_pair): Rename calls to match new patterns.
16552 (aarch64_gen_load_pair): Rename calls to match new patterns.
16553 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
16554 (load_pair<DREG:mode><DREG2:mode>): ... This.
16555 (store_pair<mode>): Rename to...
16556 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
16557 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
16558 New mode iterators.
16559 (V_INT_EQUIV): Handle SImode.
16560 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
16563 2018-05-22 Martin Sebor <msebor@redhat.com>
16566 * calls.c (maybe_warn_nonstring_arg): Use string length to set
16567 or ajust the presumed bound on an operation to avoid unnecessary
16570 2018-05-22 Martin Sebor <msebor@redhat.com>
16572 PR tree-optimization/85826
16573 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
16574 assuming that a DECL necesarily has a constant size.
16576 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
16578 PR middle-end/85862
16579 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
16581 2018-05-22 Richard Biener <rguenther@suse.de>
16583 PR tree-optimization/85834
16584 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
16585 non-constant and non-zero memset arguments.
16587 2018-05-22 Martin Liska <mliska@suse.cz>
16590 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
16592 2018-05-22 Richard Biener <rguenther@suse.de>
16594 PR tree-optimization/85863
16595 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
16596 comparisons when vectype is specified.
16597 (vectorizable_condition): Do not specify vectype for
16598 vect_is_simple_cond when SLP vectorizing.
16600 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
16603 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
16604 define __ibm128 as long double.
16605 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
16606 as a distinct type when IEEE 128-bit support is enabled.
16607 (init_float128_ieee): Fix up conversions between IFmode and IEEE
16608 128-bit types to use the correct functions.
16609 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
16610 convert between 128-bit floating point types that have different
16611 modes but the same representation, instead of using gen_lowpart to
16613 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
16615 (IFKF_reg): New attributes to give the register constraints for
16617 (extend<mode>tf2_internal): New insns to mark an explicit
16618 conversion between 128-bit floating point types that have a
16619 different mode but share the same representation.
16621 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
16623 PR tree-optimization/85814
16624 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
16625 a null return from get_strinfo when unsharing the next
16626 strinfo in the chain.
16628 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
16631 * varasm.c (weak_finish): Clean up weak_decls.
16633 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16635 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
16636 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
16637 UNSPEC_UADALP values.
16638 * config/aarch64/iterators.md (ABAL): New int iterator.
16641 (sur): Add mappings for the above.
16642 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
16644 (aarch64_<sur>abal<mode>_4): Likewise.
16645 (aarch64_<sur>adalp<mode>_3): Likewise.
16646 (<sur>sadv16qi): New define_expand.
16648 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
16650 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
16651 (*movdf_internal): Ditto.
16652 (*rcpsf2_sse): Ditto.
16653 (*rsqrtsf2_sse): Ditto.
16654 (*sqrt<mode>2_sse): Ditto.
16656 2018-05-21 Tamar Christina <tamar.christina@arm.com>
16658 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
16660 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
16661 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
16662 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
16663 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
16665 * config/aarch64/arm_neon.h: Likewise.
16666 * config/aarch64/iterators.md (VQ_I): New.
16668 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
16670 * config.gcc: Add arc/t-multilib-linux to tmake_file for
16672 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
16675 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
16677 * config/nds32/constraints.md (S): New constraint.
16678 * config/nds32/nds32.md (call_internal): Use constraint S.
16679 (call_value_internal): Likewise.
16680 (sibcall_internal): Likewise.
16681 (sibcall_value_internal): Likewise.
16683 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
16684 Chung-Ju Wu <jasonwucj@gmail.com>
16686 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
16687 into consideration.
16689 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
16690 Chung-Ju Wu <jasonwucj@gmail.com>
16692 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
16693 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
16694 (nds32_rtx_costs_impl): Simplify.
16695 (nds32_address_cost_impl): Simplify.
16696 (nds32_init_rtx_costs): New function.
16697 (nds32_rtx_costs_speed_prefer): Likewise.
16698 (nds32_rtx_costs_size_prefer): Likewise.
16699 (nds32_address_cost_speed_prefer): Likewise.
16700 (nds32_address_cost_speed_fwprop): Likewise.
16701 (nds32_address_cost_size_prefer): Likewise.
16702 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
16703 * config/nds32/nds32.c (nds32_option_override): Use
16704 nds32_init_rtx_costs function.
16706 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
16708 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
16709 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
16710 (TARGET_PIPELINE_N8): Likewise.
16711 (TARGET_PIPELINE_N10): Likewise.
16712 (TARGET_PIPELINE_N13): Likewise.
16713 (TARGET_PIPELINE_GRAYWOLF): Likewise.
16715 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
16717 * config/nds32/nds32-fpu.md: Update copyright year.
16719 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
16721 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
16723 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
16725 * config/nds32/nds32.c
16726 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
16727 * config/nds32/nds32.opt (minline-asm-r15): New option.
16729 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
16731 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
16733 * config/nds32/nds32.md (abssi2): New pattern.
16735 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
16737 * config/i386/i386.md (rex64namesuffix): New mode attribute.
16738 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
16739 Merge insn pattern from sse_cvtsi2ss<round_name> and
16740 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
16741 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
16742 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
16743 using SWI48 mode iterator.
16744 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
16745 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
16746 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
16747 pattern from sse_cvttss2si<round_saeonly_name>
16748 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
16749 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
16750 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
16751 using SWI48 mode iterator.
16752 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
16753 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
16754 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
16755 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
16756 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
16757 using SWI48 mode iterator.
16758 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
16759 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
16760 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
16761 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
16762 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
16763 SWI48 mode iterator.
16764 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
16765 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
16766 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
16767 pattern from sse_cvttsd2si<round_saeonly_name>
16768 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
16770 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
16772 * config/nds32/nds32-md-auxiliary.c
16773 (nds32_valid_smw_lwm_base_p): Refine.
16774 (nds32_output_smw_single_word): Refine.
16775 (nds32_output_smw_double_word): New.
16776 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
16778 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
16780 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
16781 (nds32_output_stack_pop): Refine.
16782 (nds32_expand_unaligned_load): Refine.
16783 (nds32_expand_unaligned_store): Refine.
16785 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
16786 Chung-Ju Wu <jasonwucj@gmail.com>
16788 * config/nds32/constants.md: Add TP_REGNUM constant.
16789 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
16790 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
16792 * config/nds32/nds32-doubleword.md: Consider flag_pic.
16793 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
16794 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
16795 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
16796 and PIC code generation.
16797 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
16799 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
16801 * config/nds32/nds32.md: Support TLS and PIC.
16802 * config/nds32/nds32.c: Support TLS and PIC.
16803 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
16804 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
16807 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
16809 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
16810 mode with E_ prefix.
16812 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
16813 Chung-Ju Wu <jasonwucj@gmail.com>
16815 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
16816 * config/nds32/nds32-md-auxiliary.c
16817 (symbolic_reference_mentioned_p): New.
16818 (nds32_legitimize_ict_address): New.
16819 (nds32_expand_ict_move): New.
16820 (nds32_indirect_call_referenced_p): New.
16821 (nds32_symbol_binds_local_p): Delete.
16822 (nds32_long_call_p): Modify.
16823 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
16824 * config/nds32/nds32-protos.h
16825 (symbolic_reference_mentioned_p): Declare.
16826 (nds32_legitimize_ict_address): Declare.
16827 (nds32_expand_ict_move): Declare.
16828 (nds32_indirect_call_referenced_p): Declare.
16829 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
16830 (nds32_relax_group): Use nds32_ict_const_p as condition.
16831 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
16832 (nds32_asm_file_start): Output ict_model directive in asm code.
16833 (nds32_legitimate_address_p): Consider indirect call.
16834 (nds32_print_operand): Consider indirect call.
16835 (nds32_print_operand_address): Consider indirect call.
16836 (nds32_insert_attributes): Handle "indirect_call" attribute.
16837 (TARGET_LEGITIMATE_ADDRESS_P): Define.
16838 (TARGET_LEGITIMATE_CONSTANT_P): Define.
16839 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
16840 (TARGET_DELEGITIMIZE_ADDRESS): Define.
16841 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
16842 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
16843 (TARGET_ICT_MODEL_SMALL): Define.
16844 (TARGET_ICT_MODEL_LARGE): Define.
16845 * config/nds32/nds32.md (movsi): Consider ict model.
16846 (call, call_value): Consider ict model.
16847 (sibcall, sibcall_value): Consider ict model.
16848 * config/nds32/nds32.opt (mict-model): New option.
16849 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
16852 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
16853 Monk Chiang <sh.chiang04@gmail.com>
16854 Jim Wilson <jimw@sifive.com>
16856 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
16857 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
16858 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
16859 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
16860 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
16861 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
16862 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
16863 compute save_libcall_adjustment properly.
16864 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
16865 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
16866 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
16867 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
16868 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
16869 (ABI_SPEC): Handle mabi=ilp32e.
16870 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
16871 (RVE): Add RVE mask.
16872 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
16873 <-march>: Add rv32e as an example.
16875 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
16878 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
16879 (intra_create_variable_infos): Handle C++ constructors.
16881 2018-05-18 Martin Liska <mliska@suse.cz>
16883 * passes.def: Remove a redundant pass.
16885 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
16888 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
16890 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16892 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
16894 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
16896 (armv2, armv2a, armv3, armv3m): Delete architectures.
16897 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
16898 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
16899 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
16901 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
16902 (*mulsidi3adddi): Likewise.
16903 (mulsidi3): Likewise.
16904 (*mulsidi3_nov6): Likewise.
16905 (umulsidi3): Likewise.
16906 (umulsidi3_nov6): Likewise.
16907 (umaddsidi4): Likewise.
16908 (*umulsidi3adddi): Likewise.
16909 (smulsi3_highpart): Likewise.
16910 (*smulsi3_highpart_nov6): Likewise.
16911 (umulsi3_highpart): Likewise.
16912 (*umulsi3_highpart_nov6): Likewise.
16913 * config/arm/arm.h (arm_arch3m): Delete.
16914 * config/arm/arm.c (arm_arch3m): Delete.
16915 (arm_option_override_internal): Update armv3-related comment.
16916 (arm_configure_build_target): Delete use of isa_bit_mode32.
16917 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
16918 (arm_rtx_costs_internal): Delete check of arm_arch3m.
16919 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
16920 (mulsa3): Likewise.
16921 (mulusa3): Likewise.
16922 * config/arm/arm-protos.h (arm_arch3m): Delete.
16923 * config/arm/arm-tables.opt: Regenerate.
16924 * config/arm/arm-tune.md: Likewise.
16925 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
16926 deleted architectures.
16928 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16930 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
16931 (armv5t, armv5te): New features.
16932 (ARMv5, ARMv5e): Delete fgroups.
16933 (ARMv5t, ARMv5te): Adjust for above changes.
16934 (ARMv6m): Likewise.
16935 (armv5, armv5e): Delete arches.
16936 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
16938 (*call_reg_arm): Likewise.
16939 (*call_value_reg_armv5): Likewise.
16940 (*call_value_reg_arm): Likewise.
16941 (*call_symbol): Likewise.
16942 (*call_value_symbol): Likewise.
16943 (*sibcall_insn): Likewise.
16944 (*sibcall_value_insn): Likewise.
16945 (clzsi2): Likewise.
16946 (prefetch): Likewise.
16947 (define_split and define_peephole2 dependent on arm_arch5):
16949 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
16951 (TARGET_ARM_QBIT): Likewise.
16952 (TARGET_DSP_MULTIPLY): Likewise.
16953 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
16954 (arm_arch5, arm_arch5e): Delete.
16955 (arm_arch5t, arm_arch5te): Declare.
16956 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
16957 (arm_arch5t): Declare.
16958 (arm_option_reconfigure_globals): Update for the above.
16959 (arm_options_perform_arch_sanity_checks): Update comment, replace
16960 use of arm_arch5 with arm_arch5t.
16961 (use_return_insn): Likewise.
16962 (arm_emit_call_insn): Likewise.
16963 (output_return_instruction): Likewise.
16964 (arm_final_prescan_insn): Likewise.
16965 (arm_coproc_builtin_available): Likewise.
16966 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
16967 arm_arch5e with arm_arch5t and arm_arch5te.
16968 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
16969 (arm_arch5t, arm_arch5te): Declare.
16970 * config/arm/arm-tables.opt: Regenerate.
16971 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
16972 * config/arm/t-multilib: Likewise.
16973 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
16974 instead of arm_arch5.
16975 (*call_reg_thumb1): Likewise.
16976 (*call_value_reg_thumb1_v5): Likewise.
16977 (*call_value_reg_thumb1): Likewise.
16978 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
16980 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
16982 2018-05-18 Martin Liska <mliska@suse.cz>
16984 PR gcov-profile/84846
16985 * doc/gcov.texi: Document -t option of gcov tool.
16987 2018-05-18 Martin Liska <mliska@suse.cz>
16989 PR gcov-profile/84846
16990 * gcov.c (print_usage): Add new -t option.
16991 (process_args): Handle the option.
16992 (generate_results): Use stdout as output when requested by
16995 2018-05-18 Martin Liska <mliska@suse.cz>
16997 PR gcov-profile/84846
16998 * coverage.c (coverage_init): Write PWD to .gcno file.
16999 * doc/gcov.texi: Document how working directory is printed.
17000 * gcov-dump.c (dump_gcov_file): Print PWD.
17001 * gcov.c (output_intermediate_file): Likewise.
17002 (read_graph_file): Read PWD string.
17003 (output_lines): Print PWD.
17005 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17007 PR middle-end/85817
17008 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
17009 for retval and return false if all args to phi are zero.
17011 2018-05-18 Richard Biener <rguenther@suse.de>
17013 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
17015 (evrp_dom_walker::before_dom_children): Call it.
17017 2018-05-18 Richard Biener <rguenther@suse.de>
17019 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
17020 results when processing array refs with variable index.
17022 2018-05-18 Toon Moene <toon@moene.org>
17024 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
17025 directly after that of -floop-interchange. Indicate that both
17026 options are enabled by default when specifying -O3.
17028 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17030 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
17031 iterator. Delete separate integer-mode vec_set<mode> expander.
17032 (aarch64_simd_vec_setv2di): Delete.
17033 (vec_setv2di): Delete.
17034 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
17035 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
17036 the "w, r" alternative.
17038 2018-05-18 Martin Liska <mliska@suse.cz>
17040 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
17041 * tree-pass.h (make_pass_lower_switch_O0): New function.
17042 * tree-switch-conversion.c (node_has_low_bound): Remove.
17043 (node_has_high_bound): Likewise.
17044 (node_is_bounded): Likewise.
17045 (class pass_lower_switch): Make it a template type and create
17047 (pass_lower_switch::execute): Add template argument.
17048 (make_pass_lower_switch): New function.
17049 (make_pass_lower_switch_O0): New function.
17050 (do_jump_if_equal): Remove.
17051 (emit_case_nodes): Simplify to just handle all 3 cases and leave
17052 all the hard work to tree optimization passes.
17054 2018-05-18 Martin Liska <mliska@suse.cz>
17056 * dbgcnt.c (limit_low): Renamed from limit.
17057 (limit_high): New variable.
17058 (dbg_cnt_is_enabled): Check for upper limit.
17059 (dbg_cnt): Adjust dumping.
17060 (dbg_cnt_set_limit_by_index): Add new argument for high
17062 (dbg_cnt_set_limit_by_name): Likewise.
17063 (dbg_cnt_process_single_pair): Parse new format.
17064 (dbg_cnt_process_opt): Use strtok.
17065 (dbg_cnt_list_all_counters): Remove 'value' and add
17067 * doc/invoke.texi: Document changes.
17069 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
17071 * doc/sourcebuild.texi (scalar_all_fma): Document.
17072 * tree.def (FMA_EXPR): Delete.
17073 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
17074 * internal-fn.c (ternary_direct): New macro.
17075 (expand_ternary_optab_fn): Likewise.
17076 (direct_ternary_optab_supported_p): Likewise.
17077 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
17078 * builtins.c (fold_builtin_fma): Delete.
17079 (fold_builtin_3): Don't call it.
17080 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
17081 * expr.c (expand_expr_real_2): Likewise.
17082 * fold-const.c (operand_equal_p): Likewise.
17083 (fold_ternary_loc): Likewise.
17084 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
17085 * gimple.c (DEFTREECODE): Likewise.
17086 * gimplify.c (gimplify_expr): Likewise.
17087 * optabs-tree.c (optab_for_tree_code): Likewise.
17088 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
17089 * tree-eh.c (operation_could_trap_p): Likewise.
17090 (stmt_could_throw_1_p): Likewise.
17091 * tree-inline.c (estimate_operator_cost): Likewise.
17092 * tree-pretty-print.c (dump_generic_node): Likewise.
17093 (op_code_prio): Likewise.
17094 * tree-ssa-loop-im.c (stmt_cost): Likewise.
17095 * tree-ssa-operands.c (get_expr_operands): Likewise.
17096 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
17097 * fold-const-call.h (fold_fma): Delete.
17098 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
17099 CFN_FNMA and CFN_FNMS.
17100 (fold_fma): Delete.
17101 * genmatch.c (combined_fn): New enum.
17102 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
17103 (commutative_op): New function.
17104 (commutate): Use it. Handle more than 2 operands.
17105 (dt_operand::gen_gimple_expr): Use commutative_op.
17106 (parser::parse_expr): Allow :c to be used with non-binary
17107 operators if the commutative operand is known.
17108 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
17109 CFN_FMS, CFN_FNMA and CFN_FNMS.
17110 (backprop::process_assign_use): Remove FMA_EXPR handling.
17111 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
17112 (gen_hsa_fma): New function.
17113 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
17114 IFN_FNMA and IFN_FNMS.
17115 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
17116 * gimple-fold.h (follow_all_ssa_edges): Declare.
17117 * gimple-fold.c (follow_all_ssa_edges): New function.
17118 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
17119 gimple_build interface and use follow_all_ssa_edges to fold the result.
17120 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
17121 instead of checking for optabs directly.
17122 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
17123 rather than FMA_EXPRs.
17124 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
17125 call to IFN_FMA instead of an FMA_EXPR.
17127 2018-05-17 Jim Wilson <jimw@sifive.com>
17129 * expr.c (do_tablejump): When converting index to Pmode, if we have a
17130 sign extended promoted subreg, and the range does not have the sign bit
17131 set, then do a sign extend.
17133 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
17134 test, check for sign extended subreg and/or constant operands, and
17135 do a sign extend in that case.
17137 2018-05-17 Steve Ellcey <sellcey@cavium.com>
17139 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
17140 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
17142 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
17143 Change logics_shift_reg to logics_shift_imm.
17144 (thunderx2t99_fp_loadpair_basic): Delete.
17145 (thunderx2t99_fp_storepair_basic): Delete.
17146 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
17147 (thunderx2t99_asimd_polynomial): Delete.
17148 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
17149 and neon_fp_mul_d_scalar_q.
17150 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
17151 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
17152 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
17153 (thunderx2t99_asimd_lut): Add missing tbl types.
17154 (thunderx2t99_asimd_ext): Delete.
17155 (thunderx2t99_asimd_load1_1_mult): Delete.
17156 (thunderx2t99_asimd_load1_2_mult): Delete.
17157 (thunderx2t99_asimd_load1_ldp): New.
17158 (thunderx2t99_asimd_load1): New.
17159 (thunderx2t99_asimd_load2): Add missing *load2* types.
17160 (thunderx2t99_asimd_load3): New.
17161 (thunderx2t99_asimd_load4): New.
17162 (thunderx2t99_asimd_store1_1_mult): Delete.
17163 (thunderx2t99_asimd_store1_2_mult): Delete.
17164 (thunderx2t99_asimd_store2_mult): Delete.
17165 (thunderx2t99_asimd_store2_onelane): Delete.
17166 (thunderx2t99_asimd_store_stp): New.
17167 (thunderx2t99_asimd_store1): New.
17168 (thunderx2t99_asimd_store2): New.
17169 (thunderx2t99_asimd_store3): New.
17170 (thunderx2t99_asimd_store4): New.
17172 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
17174 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
17175 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
17177 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
17178 Segher Boessenkool <segher@kernel.crashing.org>
17181 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
17184 2018-05-17 Richard Biener <rguenther@suse.de>
17186 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
17187 for pruning loop and prune defs feeding only already visited PHIs.
17189 2018-05-17 Richard Biener <rguenther@suse.de>
17191 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
17193 2018-05-17 Bin Cheng <bin.cheng@arm.com>
17194 Richard Biener <rguenther@suse.de>
17196 PR tree-optimization/85793
17197 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
17198 for VMAT_ELEMENTWISE.
17200 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
17202 * internal-fn.h (lookup_internal_fn): Declare
17203 * internal-fn.c (lookup_internal_fn): New function.
17204 * gimple.c (gimple_build_call_from_tree): Handle calls to
17205 internal functions.
17206 * gimple-pretty-print.c (dump_gimple_call): Print "." before
17207 internal function names.
17208 * tree-pretty-print.c (dump_generic_node): Likewise.
17209 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
17211 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
17213 * gimple-fold.h (gimple_build): Make the function forms take
17214 combined_fn rather than built_in_function.
17215 (gimple_simplify): Likewise.
17216 * gimple-match-head.c (gimple_simplify): Likewise.
17217 * gimple-fold.c (gimple_build): Likewise.
17218 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
17219 rather than gimple_build_call_internal.
17220 (get_initial_defs_for_reduction): Likewise.
17221 (vect_create_epilog_for_reduction): Likewise.
17222 (vectorizable_live_operation): Likewise.
17224 2018-05-17 Martin Liska <mliska@suse.cz>
17226 * gimple-ssa-sprintf.c (format_directive): Do not use
17227 space in between 'G_' and '('.
17229 2018-05-17 Jakub Jelinek <jakub@redhat.com>
17232 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
17233 even if the mask is not all ones.
17236 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
17238 (ix86_gimple_fold_builtin): Likewise.
17241 * config/i386/i386.c: Include tree-vector-builder.h.
17242 (ix86_vector_shift_count): New function.
17243 (ix86_fold_builtin): Fold shift builtins by scalar count.
17244 (ix86_gimple_fold_builtin): Likewise.
17246 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
17247 _mm512_setzero): New intrinsics.
17249 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
17250 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17252 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
17253 code generation for cases where splatting a value is not useful.
17254 * simplify-rtx.c (simplify_ternary_operation): Simplify
17255 vec_merge across a vec_duplicate and a paradoxical subreg forming
17256 a vector mode to a vec_concat.
17258 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
17260 * config.gcc: Support "goldmont-plus".
17261 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
17263 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17264 PROCESSOR_GOLDMONT_PLUS.
17265 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
17266 (processor_target_table): Add "goldmont-plus".
17267 (PTA_GOLDMONT_PLUS): Define.
17268 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
17269 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
17270 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
17271 (fold_builtin_cpu): Add "goldmont-plus".
17272 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
17273 (ix86_option_override_internal): Add "goldmont-plus".
17274 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
17275 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
17276 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
17277 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
17279 2018-05-17 Richard Biener <rguenther@suse.de>
17281 PR tree-optimization/85757
17282 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
17283 remove defs that only feed that PHI from further processing.
17285 2018-05-16 Jim Wilson <jimw@sifive.com>
17287 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
17289 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
17291 2018-05-16 Mark Wielaard <mark@klomp.org>
17293 * dwarf2out.c (count_index_strings): New function.
17294 (output_indirect_strings): Call count_index_strings and generate
17295 header for dwarf_version >= 5.
17297 2018-05-16 Mark Wielaard <mark@klomp.org>
17299 * dwarf2out.c (dwarf_FORM): New function.
17300 (set_indirect_string): Use dwarf_FORM.
17301 (reset_indirect_string): Likewise.
17302 (size_of_die): Likewise.
17303 (value_format): Likewise.
17304 (output_die): Likewise.
17305 (add_skeleton_AT_string): Likewise.
17306 (output_macinfo_op): Likewise.
17307 (index_string): Likewise.
17308 (output_index_string_offset): Likewise.
17309 (output_index_string): Likewise.
17310 (count_index_strings): Likewise.
17312 2018-05-16 Carl Love <cel@us.ibm.com>
17314 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
17315 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
17317 2018-05-16 Martin Jambor <mjambor@suse.cz>
17319 * ipa-prop.c (ipa_free_all_edge_args): Remove.
17320 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
17322 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
17324 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
17325 (fnma<mode>4): Likewise.
17326 (fms<mode>4): Likewise.
17327 (fnms<mode>4): Likewise.
17328 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
17329 (aarch64_fnma<mode>4): Likewise.
17330 (aarch64_fms<mode>4): Likewise.
17331 (aarch64_fnms<mode>4): Likewise.
17332 (aarch64_fnmadd<mode>4): Likewise.
17334 2018-05-16 Jason Merrill <jason@redhat.com>
17336 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
17338 2018-05-16 Richard Biener <rguenther@suse.de>
17340 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
17341 (dump_stmt_cost): Declare.
17342 (add_stmt_cost): Dump cost we add.
17343 (add_stmt_costs): New function.
17344 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
17345 No longer exported.
17346 (vect_analyze_stmt): Adjust prototype.
17347 (vectorizable_condition): Likewise.
17348 (vectorizable_live_operation): Likewise.
17349 (vectorizable_reduction): Likewise.
17350 (vectorizable_induction): Likewise.
17351 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
17352 cost vector to pass to vectorizable_ and record afterwards.
17353 (vect_model_reduction_cost): Take cost vector argument and adjust.
17354 (vect_model_induction_cost): Likewise.
17355 (vectorizable_reduction): Likewise.
17356 (vectorizable_induction): Likewise.
17357 (vectorizable_live_operation): Likewise.
17358 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
17359 SLP_TREE_NUMBER_OF_VEC_STMTS.
17360 (vect_analyze_slp_cost_1): Remove.
17361 (vect_analyze_slp_cost): Likewise.
17362 (vect_slp_analyze_node_operations): Take visited args and
17363 a target cost vector. Avoid processing already visited stmt sets.
17364 (vect_slp_analyze_operations): Use a local cost vector to gather
17365 costs and register those of non-discarded instances.
17366 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
17367 (vect_schedule_slp_instance): Remove copying of
17368 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
17370 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
17371 adding cost. Record cost entry location.
17372 (vect_prologue_cost_for_slp_op): Function to compute cost of
17373 a constant or invariant generated for SLP vect in the prologue,
17374 split out from vect_analyze_slp_cost_1.
17375 (vect_model_simple_cost): Make static. Adjust for SLP costing.
17376 (vect_model_promotion_demotion_cost): Likewise.
17377 (vect_model_store_cost): Likewise, make static.
17378 (vect_model_load_cost): Likewise.
17379 (vectorizable_bswap): Add cost vector arg and adjust.
17380 (vectorizable_call): Likewise.
17381 (vectorizable_simd_clone_call): Likewise.
17382 (vectorizable_conversion): Likewise.
17383 (vectorizable_assignment): Likewise.
17384 (vectorizable_shift): Likewise.
17385 (vectorizable_operation): Likewise.
17386 (vectorizable_store): Likewise.
17387 (vectorizable_load): Likewise.
17388 (vectorizable_condition): Likewise.
17389 (vectorizable_comparison): Likewise.
17390 (can_vectorize_live_stmts): Likewise.
17391 (vect_analyze_stmt): Likewise.
17392 (vect_transform_stmt): Adjust calls to vectorizable_*.
17393 * tree-vectorizer.c: Include gimple-pretty-print.h.
17394 (dump_stmt_cost): New function.
17396 2018-05-16 Richard Biener <rguenther@suse.de>
17398 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
17399 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
17400 * tree-ssa-dse.c: Include tree-ssa-loop.h.
17401 (check_name): New callback.
17402 (dse_classify_store): Track cycles via a visited bitmap of PHI
17403 defs and simplify handling of in-loop and across loop dead stores
17404 and properly fail for loop-variant refs. Handle byte-tracking with
17405 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
17408 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
17410 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
17411 (vect_get_mask_type_for_stmt): Likewise.
17412 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
17414 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
17415 to determine the statement's vector type and the vector type that
17416 should be used for calculating nunits. Deal with cases in which
17417 the type has to be deferred.
17418 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
17419 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
17420 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
17421 (vect_determine_vf_for_stmt): New functions, split out from...
17422 (vect_determine_vectorization_factor): ...here.
17423 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
17424 (vect_get_mask_type_for_stmt): New functions, split out from
17425 vect_determine_vectorization_factor.
17427 2018-05-16 Richard Biener <rguenther@suse.de>
17429 * tree-cfg.c (verify_gimple_assign_ternary): Properly
17430 verify the [VEC_]COND_EXPR embedded comparison.
17432 2018-05-15 Martin Sebor <msebor@redhat.com>
17434 PR tree-optimization/85753
17435 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
17436 RECORD_TYPE in addition to ARRAY_TYPE.
17438 2018-05-15 Martin Sebor <msebor@redhat.com>
17440 PR middle-end/85643
17441 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
17443 2018-05-15 Richard Biener <rguenther@suse.de>
17445 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
17446 add by_clobber_p one. Change algorithm to collect all defs
17447 representing uses we need to walk and try reducing them to
17448 a single one before failing.
17449 (dse_dom_walker::dse_optimize_stmt): Adjust.
17451 2018-05-13 Mark Wielaard <mark@klomp.org>
17453 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
17454 (size_of_loc_descr): Likewise.
17455 (output_loc_operands): Likewise.
17456 (output_loc_operands_raw): Likewise.
17457 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
17458 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
17459 (hash_loc_operands): Likewise.
17460 (compare_loc_operands): Likewise.
17462 2018-05-14 Mark Wielaard <mark@klomp.org>
17464 * dwarf2out.c (count_index_addrs): New function.
17465 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
17467 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17469 PR tree-optimization/83648
17470 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
17471 return value as malloc candidate.
17473 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17476 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
17477 param as true in call to suggest_attribute.
17479 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
17481 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
17482 -mreadonly-in-sdata.
17484 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17486 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
17488 (aarch64_crypto_aesd_fused): Likewise.
17490 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
17492 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
17493 (movsi_aarch64): Likewise.
17494 (load_pairsi): Likewise.
17495 (load_pairdi): Likewise.
17496 (store_pairsi): Likewise.
17497 (store_pairdi): Likewise.
17498 (load_pairsf): Likewise.
17499 (load_pairdf): Likewise.
17500 (store_pairsf): Likewise.
17501 (store_pairdf): Likewise.
17502 (zero_extend): Likewise.
17503 (trunc): Swap alternatives.
17504 (fcvt_target): Add '?' to prefer w over r.
17506 2018-05-14 Jakub Jelinek <jakub@redhat.com>
17509 * config/i386/i386.md: Disallow non-commutative arithmetics in
17510 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
17511 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
17512 in the peephole2 before it.
17514 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
17516 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
17517 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
17518 (ix86_handle_option): Handle -mcldemote.
17519 * config.gcc: New header.
17520 * config/i386/cldemoteintrin.h: New file.
17521 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
17522 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
17524 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17525 OPTION_MASK_ISA_CLDEMOTE.
17526 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
17527 (ix86_valid_target_attribute_inner_p): Ditto.
17528 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
17529 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
17530 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
17531 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
17532 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
17534 * config/i386/i386.opt: Add -mcldemote.
17535 * config/i386/x86intrin.h: New header.
17536 * doc/invoke.texi: Add -mcldemote.
17538 2018-05-14 Richard Biener <rguenther@suse.de>
17540 * doc/match-and-simplify.texi: Adjust :s documentation.
17542 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
17544 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
17545 intended memcpy size.
17546 (REORDER_45): Likewise.
17548 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
17550 * sort.cc: New file.
17551 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
17552 * vec.c (qsort_chk): Use gcc_qsort.
17553 * Makefile.in (OBJS-libcommon): Add sort.o.
17554 (build/sort.o): New target. Use it...
17555 (BUILD_RTL): ... here, and...
17556 (build/gencfn-macros): ... here, and...
17557 (build/genmatch): ... here.
17559 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
17560 Chung-Ju Wu <jasonwucj@gmail.com>
17562 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
17563 * config/nds32/nds32-graywolf.md: New file.
17564 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
17565 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
17567 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
17568 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
17569 * config/nds32/nds32.md (pipeline_model): Add graywolf.
17570 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
17571 * config/nds32/pipelines.md: Include n15 settings.
17573 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
17574 Chung-Ju Wu <jasonwucj@gmail.com>
17576 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
17577 * config/nds32/nds32-n13.md: New file.
17578 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
17579 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
17581 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
17582 * config/nds32/nds32.md (pipeline_model): Add n13.
17583 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
17584 * config/nds32/pipelines.md: Include n13 settings.
17586 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
17587 Chung-Ju Wu <jasonwucj@gmail.com>
17589 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
17590 * config/nds32/nds32-n10.md: New file.
17591 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
17592 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
17594 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
17595 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
17596 * config/nds32/nds32.md (pipeline_model): Add n10.
17597 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
17598 * config/nds32/pipelines.md: Include n10 settings.
17600 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
17601 Kito Cheng <kito.cheng@gmail.com>
17602 Chung-Ju Wu <jasonwucj@gmail.com>
17604 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
17605 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
17606 Add enum values for DSP extension instructions.
17607 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
17609 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
17610 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
17611 New code iterators.
17612 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
17613 * config/nds32/nds32-dspext.md: New file for DSP implementation.
17614 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
17615 * config/nds32/nds32-intrinsic.md: Likewise.
17616 * config/nds32/nds32_intrinsic.h: Likewise.
17617 * config/nds32/nds32-md-auxiliary.c: Likewise.
17618 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
17619 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
17620 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
17621 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
17622 * config/nds32/nds32-protos.h: New declarations for DSP extension.
17623 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
17624 TYPE_DMAC in switch statement.
17625 * config/nds32/nds32.c: New checking and implementation for DSP
17626 extension instructions.
17627 * config/nds32/nds32.h: Likewise.
17628 * config/nds32/nds32.md: Likewise.
17629 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
17630 * config/nds32/predicates.md: Implement new predicates for DSP
17633 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
17635 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
17636 Reformat alternatives and attributes so it is easier to identify
17637 which constraints/attributes go with which instruction.
17638 (mov<mode>_hardfloat32, FMOVE64): Likewise.
17639 (mov<mode>_softfloat32, FMOVE64): Likewise.
17640 (mov<mode>_hardfloat64, FMOVE64): Likewise.
17641 (mov<mode>_softfloat64, FMOVE64): Likewise.
17643 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
17645 * doc/extend.texi (PowerPC Built-in Functions): Rename this
17647 (Basic PowerPC Built-in Functions): The new name of the
17648 subsection previously known as "PowerPC Built-in Functions".
17649 (Basic PowerPC Built-in Functions Available on all Configurations):
17651 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
17652 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
17653 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
17654 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
17656 2018-05-11 Martin Jambor <mjambor@suse.cz>
17659 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
17662 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
17665 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
17667 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
17669 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
17670 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
17671 (ix86_handle_option): Handle -mwaitpkg.
17672 * config.gcc: New header.
17673 * config/i386/cpuid.h (bit_WAITPKG): New bit.
17674 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
17675 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
17677 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17678 OPTION_MASK_ISA_WAITPKG.
17679 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
17680 (ix86_option_override_internal): Add PTA_WAITPKG.
17681 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
17682 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
17683 IX86_BUILTIN_TPAUSE.
17684 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
17685 __builtin_ia32_umwait and __builtin_ia32_tpause.
17686 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
17687 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
17688 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
17689 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
17690 UNSPECV_TPAUSE): New.
17691 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
17692 * config/i386/i386.opt: Add -mwaitpkg.
17693 * config/i386/waitpkgintrin.h: New file.
17694 * config/i386/x86intrin.h: New header.
17695 * doc/invoke.texi: Add -mwaitpkg.
17697 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
17700 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
17702 (cortex-m0): Use armv6s-m isa.
17703 (cortex-m0plus): Likewise.
17704 (cortex-m1): Likewise.
17705 (cortex-m0.small-multiply): Likewise.
17706 (cortex-m0plus.small-multiply): Likewise.
17707 (cortex-m1.small-multiply): Likewise.
17709 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
17710 Jakub Jelinek <jakub@redhat.com>
17712 PR tree-optimization/85692
17713 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
17714 source permute as well.
17716 2018-05-11 Martin Liska <mliska@suse.cz>
17719 * doc/extend.texi: Document LLVM style format for no_sanitize
17722 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
17724 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
17725 mode_supports_vsx_dform_quad to mode_supports_dq_form.
17726 (mode_supports_vsx_dform_quad): Likewise.
17727 (mode_supports_vmx_dform): Move these functions to be next to the
17728 other mode_supports functions.
17729 (mode_supports_dq_form): Likewise.
17730 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
17731 mode_supports_dq_form.
17732 (reg_offset_addressing_ok_p): Likewise.
17733 (offsettable_ok_by_alignment): Likewise.
17734 (rs6000_legitimate_offset_address_p): Likewise.
17735 (legitimate_lo_sum_address_p): Likewise.
17736 (rs6000_legitimize_address): Likewise.
17737 (rs6000_legitimize_reload_address): Likewise.
17738 (rs6000_secondary_reload_inner): Likewise.
17739 (rs6000_preferred_reload_class): Likewise.
17740 (rs6000_output_move_128bit): Likewise.
17742 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
17744 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
17745 Generate SImode target register for null target.
17746 <case IX86_BUILTIN_XGETBV>: Ditto.
17747 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
17748 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
17750 2018-05-10 Carl Love <cel@us.ibm.com>
17752 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
17753 dcbtt and dcbtstt if operands[2] is 0.
17755 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
17758 * config/i386/sse.md (usadv64qi): New expander.
17760 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
17762 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
17763 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
17764 -maltivec=be support.
17765 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
17766 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
17767 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
17768 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
17769 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
17770 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
17771 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
17772 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
17773 altivec_vsumsws): Adjust.
17774 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
17775 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
17776 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
17778 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
17779 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
17780 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
17781 (altivec_lve<VI_char>x): Delete expand.
17782 (*altivec_lve<VI_char>x_internal): Rename to...
17783 (altivec_lve<VI_char>x): ... this.
17784 (altivec_lvxl_<mode>): Delete expand.
17785 (*altivec_lvxl_<mode>_internal): Rename to ...
17786 (altivec_lvxl_<mode>): ... this.
17787 (altivec_stvxl_<mode>): Delete expand.
17788 (*altivec_stvxl_<mode>_internal): Rename to ...
17789 (altivec_stvxl_<mode>): ... this.
17790 (altivec_stve<VI_char>x): Delete expand.
17791 (*altivec_stve<VI_char>x_internal): Rename to ...
17792 (altivec_stve<VI_char>x): ... this.
17793 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
17794 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
17795 reduc_plus_scal_<mode>): Adjust.
17796 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
17798 (rs6000_cpu_cpp_builtins): Adjust.
17799 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
17800 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
17801 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
17802 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
17803 -maltivec=be support.
17804 (rs6000_split_vec_extract_var): Adjust.
17805 (rs6000_split_v4si_init): Adjust.
17806 (swap_selector_for_mode): Delete.
17807 (altivec_expand_lvx_be, altivec_expand_stvx_be,
17808 altivec_expand_stvex_be): Delete.
17809 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
17810 -maltivec=be support.
17811 (rs6000_gimple_fold_builtin): Ditto.
17812 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
17814 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
17815 (TARGET_DIRECT_MOVE_64BIT): Adjust.
17816 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
17817 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
17818 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
17819 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
17820 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
17821 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
17822 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
17823 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
17824 anonymous split): Adjust.
17825 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
17826 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
17828 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
17830 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
17831 when --with-gxx-include-dir is also specified.
17832 * configure: Regenerate.
17834 2018-05-09 Jim Wilson <jimw@sifive.com>
17837 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
17838 * config/riscv/t-withmultilib: New.
17839 * config/riscv/withmultilib.h: New.
17840 * doc/install.texi: Document RISC-V --with-multilib-list support.
17842 2018-05-09 Richard Biener <rguenther@suse.de>
17844 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
17846 (vect_bb_vectorization_profitable_p): Adjust. Compute
17847 actual scalar cost using the cost vector and the add_stmt_cost
17850 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
17852 PR rtl-optimization/85645
17853 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
17854 in the REG_CFA_REGISTER note for LR, don't leave it empty.
17856 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
17858 PR rtl-optimization/85645
17859 * shrink-wrap.c (spread_components): Return a boolean saying if
17860 anything was changed.
17861 (try_shrink_wrapping_separate): Iterate spread_components until
17862 nothing changes anymore.
17864 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
17866 PR rtl-optimization/85645
17867 * regrename.c (build_def_use): Also kill the chains that include the
17868 destination of a REG_CFA_REGISTER note.
17870 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
17872 PR rtl-optimization/85645
17873 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
17874 insn that has a REG_CFA_REGISTER note.
17876 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
17878 * cfgexpand.c (expand_clobber): New function.
17879 (expand_gimple_stmt_1): Use it.
17880 * tree-vect-stmts.c (vect_clobber_variable): New function,
17882 (vectorizable_simd_clone_call): ...here.
17883 (vectorizable_store): Emit a clobber either side of an
17884 IFN_STORE_LANES sequence.
17885 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
17887 2018-05-09 Tom de Vries <tom@codesourcery.com>
17890 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
17891 (define_insn "trap_if_false"): Add exit after trap.
17893 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
17895 PR rtl-optimization/85638
17896 * bb-reorder.c: Include common/common-target.h.
17897 (create_forwarder_block): New function extracted from...
17898 (fix_up_crossing_landing_pad): ...here. Rename into...
17899 (dw2_fix_up_crossing_landing_pad): ...this.
17900 (sjlj_fix_up_crossing_landing_pad): New function.
17901 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
17902 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
17903 from both partitions and exit the loop after one iteration.
17905 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
17908 * doc/extend.texi (PowerPC Built-in Functions): Rename this
17910 (Basic PowerPC Built-in Functions): The new name of the
17911 subsection previously known as "PowerPC Built-in Functions".
17912 (Basic PowerPC Built-in Functions Available on all Configurations):
17914 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
17916 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
17918 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
17920 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
17923 2018-05-08 Jim Wilson <jimw@sifive.com>
17925 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
17926 (LD_EMUL_SUFFIX): New.
17927 (LINK_SPEC): Use it.
17929 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
17931 * doc/extend.texi (PowerPC Built-in Functions): Rename this
17933 (Basic PowerPC Built-in Functions): The new name of the
17934 subsection previously known as "PowerPC Built-in Functions".
17935 (Basic PowerPC Built-in Functions Available on all Configurations):
17937 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
17939 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
17941 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
17943 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
17946 2018-05-08 Jakub Jelinek <jakub@redhat.com>
17949 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
17950 after cmpelim optimization.
17952 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
17954 * config.gcc: Support "goldmont".
17955 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
17956 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17957 PROCESSOR_GOLDMONT.
17958 * config/i386/i386.c (m_GOLDMONT): Define.
17959 (processor_target_table): Add "goldmont".
17960 (PTA_GOLDMONT): Define.
17961 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
17962 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
17963 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
17964 (fold_builtin_cpu): Add "goldmont".
17965 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
17966 (ix86_option_override_internal): Add "goldmont".
17967 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
17968 (processor_type): Add PROCESSOR_GOLDMONT.
17969 * config/i386/i386.md: Add CPU "glm".
17970 * config/i386/glm.md: New file.
17971 * config/i386/x86-tune.def: Add m_GOLDMONT.
17972 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
17974 2018-05-08 Jakub Jelinek <jakub@redhat.com>
17977 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
17979 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
17980 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
17981 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
17984 * config/i386/i386.c (ix86_fold_builtin): Handle
17985 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
17988 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
17989 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
17991 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
17994 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
17995 (check_arch): Likewise.
17996 (check_fpu): Return the result rather than printing it.
17997 (end arch): Fix operator precedence.
17998 (end cpu): Likewise.
17999 (END): Print the result from check_fpu.
18001 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
18002 Alan Hayward <alan.hayward@arm.com>
18003 David Sherwood <david.sherwood@arm.com>
18005 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
18006 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
18007 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
18008 (*fcmuo<mode>_and): New patterns.
18010 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
18012 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
18013 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
18014 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
18015 (cmp_op, sve_imm_con): New code attributes.
18016 (SVE_COND_INT_CMP, imm_con): Delete.
18017 (cmp_op): Remove above unspecs from int attribute.
18018 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
18020 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
18021 comparison-specific unspecs.
18022 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
18023 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
18024 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
18025 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
18026 (*vec_fcm<cmp_op><mode>): Rename to...
18027 (*fcm<cmp_op><mode>): ...this and adjust likewise.
18028 (*vec_fcmuo<mode>): Rename to...
18029 (*fcmuo<mode>): ...this and adjust likewise.
18030 (*pred_fcm<cmp_op><mode>): New pattern.
18031 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
18032 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
18034 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
18036 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
18037 (aarch64_emit_sve_predicated_cond): New function.
18038 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
18039 (aarch64_emit_unspec_cond_or): Replace with...
18040 (aarch64_emit_sve_or_conds): ...this new function. Use
18041 aarch64_emit_sve_ptrue_op for the individual comparisons and
18042 aarch64_emit_binop to OR them together.
18043 (aarch64_emit_inverted_unspec_cond): Replace with...
18044 (aarch64_emit_sve_inverted_cond): ...this new function. Use
18045 aarch64_emit_sve_ptrue_op for the comparison and
18046 aarch64_emit_unop to invert the result.
18047 (aarch64_expand_sve_vec_cmp_float): Update after the above
18048 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
18050 2018-05-07 Nathan Sidwell <nathan@acm.org>
18052 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
18053 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
18054 (Backwards Compatibility): Likewise.
18056 2018-05-07 Luis Machado <luis.machado@linaro.org>
18060 2018-05-07 Luis Machado <luis.machado@linaro.org>
18062 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
18063 <prefetch_dynamic_strides>: New const bool field.
18064 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
18065 prefetch_dynamic_strides.
18066 (exynosm1_prefetch_tune): Likewise.
18067 (thunderxt88_prefetch_tune): Likewise.
18068 (thunderx_prefetch_tune): Likewise.
18069 (thunderx2t99_prefetch_tune): Likewise.
18070 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
18072 (aarch64_override_options_internal): Update to set
18073 PARAM_PREFETCH_DYNAMIC_STRIDES.
18074 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
18075 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
18076 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
18077 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
18078 prefetch-dynamic-strides setting.
18080 2018-05-07 Luis Machado <luis.machado@linaro.org>
18082 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
18083 <minimum_stride>: New const int field.
18084 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
18085 minimum_stride field.
18086 (exynosm1_prefetch_tune): Likewise.
18087 (thunderxt88_prefetch_tune): Likewise.
18088 (thunderx_prefetch_tune): Likewise.
18089 (thunderx2t99_prefetch_tune): Likewise.
18090 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
18091 (aarch64_override_options_internal): Update to set
18092 PARAM_PREFETCH_MINIMUM_STRIDE.
18093 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
18094 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
18095 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
18096 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
18097 stride is constant and is below the minimum stride threshold.
18099 2018-05-07 Luis Machado <luis.machado@linaro.org>
18101 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
18104 2018-05-07 Luis Machado <luis.machado@linaro.org>
18106 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
18107 <prefetch_dynamic_strides>: New const bool field.
18108 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
18109 prefetch_dynamic_strides.
18110 (exynosm1_prefetch_tune): Likewise.
18111 (thunderxt88_prefetch_tune): Likewise.
18112 (thunderx_prefetch_tune): Likewise.
18113 (thunderx2t99_prefetch_tune): Likewise.
18114 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
18116 (aarch64_override_options_internal): Update to set
18117 PARAM_PREFETCH_DYNAMIC_STRIDES.
18118 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
18119 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
18120 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
18121 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
18122 prefetch-dynamic-strides setting.
18124 2018-05-07 Luis Machado <luis.machado@linaro.org>
18126 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
18127 <minimum_stride>: New const int field.
18128 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
18129 minimum_stride field.
18130 (exynosm1_prefetch_tune): Likewise.
18131 (thunderxt88_prefetch_tune): Likewise.
18132 (thunderx_prefetch_tune): Likewise.
18133 (thunderx2t99_prefetch_tune): Likewise.
18134 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
18135 (aarch64_override_options_internal): Update to set
18136 PARAM_PREFETCH_MINIMUM_STRIDE.
18137 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
18138 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
18139 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
18140 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
18141 stride is constant and is below the minimum stride threshold.
18143 2018-05-06 Jakub Jelinek <jakub@redhat.com>
18146 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
18147 the type is addressable. Don't force op into register if it has
18150 2018-05-05 Roland McGrath <mcgrathr@google.com>
18153 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
18154 any section for which we don't know a specific type it should have,
18155 regardless of name. Previously this was done only for the exact
18156 names ".init_array", ".fini_array", and ".preinit_array".
18157 (default_elf_asm_named_section): Add comment about
18158 relationship with default_section_type_flags and SECTION_NOTYPE.
18159 (get_section): Don't consider it a type conflict if one side has
18160 SECTION_NOTYPE and the other doesn't, as long as neither has the
18161 SECTION_BSS et al used in the default_section_type_flags logic.
18163 2018-05-05 Tom de Vries <tom@codesourcery.com>
18166 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
18167 (workaround_barsyncs): New function.
18168 (nvptx_reorg): Use workaround_barsyncs.
18169 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
18170 (define_expand "nvptx_membar_cta"): New define_expand.
18171 (define_insn "*nvptx_membar_cta"): New insn.
18173 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
18175 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
18176 To improve optimization opportunities.
18177 * builtin-types.def: The new needed builtin types for the above.
18179 2018-05-04 Richard Biener <rguenther@suse.de>
18181 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
18182 * gimple-ssa-store-merging.c
18183 (imm_store_chain_info::output_merged_store): Remove redundant create,
18184 release split_store vector contents on failure.
18185 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
18186 scalar stmt vector on cache hit.
18188 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
18190 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
18192 * config.gcc (powerpc-xilinx-eabi*): Remove.
18193 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
18195 (fusion_addis_mem_combo_load): Ditto.
18196 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
18198 (rs6000_cpu_cpp_builtins): Ditto.
18199 * config/rs6000/rs6000-linux.c
18200 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
18201 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
18202 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
18204 (rs6000_setup_reg_addr_masks): Ditto.
18205 (rs6000_init_hard_regno_mode_ok): Ditto.
18206 (rs6000_option_override_internal): Ditto.
18207 (legitimate_lo_sum_address_p): Ditto.
18208 (rs6000_legitimize_address): Ditto.
18209 (rs6000_legitimize_reload_address): Ditto.
18210 (rs6000_legitimate_address_p): Ditto.
18211 (abi_v4_pass_in_fpr): Ditto.
18212 (setup_incoming_varargs): Ditto.
18213 (rs6000_gimplify_va_arg): Ditto.
18214 (rs6000_split_multireg_move): Ditto.
18215 (rs6000_savres_strategy): Ditto.
18216 (rs6000_emit_prologue_components): Ditto.
18217 (rs6000_emit_epilogue_components): Ditto.
18218 (rs6000_emit_prologue): Ditto.
18219 (rs6000_emit_epilogue): Ditto.
18220 (rs6000_elf_file_end): Ditto.
18221 (rs6000_function_value): Ditto.
18222 (rs6000_libcall_value): Ditto.
18223 * config/rs6000/rs6000.h: Ditto.
18224 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
18225 (TARGET_MINMAX): ... this. New.
18226 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
18227 * config/rs6000/rs6000.md: Remove Xilinx FP support.
18228 (*movsi_internal1_single): Delete.
18229 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
18230 mfpu=, mxilinx-fpu): Delete.
18231 * config/rs6000/singlefp.h: Delete.
18232 * config/rs6000/sysv4.h: Remove Xilinx FP support.
18233 * config/rs6000/t-rs6000: Ditto.
18234 * config/rs6000/t-xilinx: Delete.
18235 * config/rs6000/titan.md: Adjust for fp_type removal.
18236 * config/rs6000/vsx.md: Remove Xilinx FP support.
18237 (VStype_simple): Delete.
18238 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
18239 * config/rs6000/xfpu.h: Delete.
18240 * config/rs6000/xfpu.md: Delete.
18241 * config/rs6000/xilinx.h: Delete.
18242 * config/rs6000/xilinx.opt: Delete.
18243 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
18244 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
18246 2018-05-04 Tom de Vries <tom@codesourcery.com>
18249 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
18252 2018-05-04 Richard Biener <rguenther@suse.de>
18254 PR middle-end/85627
18255 * tree-complex.c (update_complex_assignment): We are always in SSA form.
18256 (expand_complex_div_wide): Likewise.
18257 (expand_complex_operations_1): Likewise.
18258 (expand_complex_libcall): Preserve EH info of the original stmt.
18259 (tree_lower_complex): Handle removed blocks.
18260 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
18261 on complex multiplication and division libcall builtins.
18263 2018-05-04 Richard Biener <rguenther@suse.de>
18265 PR middle-end/85574
18266 * fold-const.c (negate_expr_p): Restrict negation of operand
18267 zero of a division to when we know that can happen without
18269 (fold_negate_expr_1): Likewise.
18271 2018-05-04 Jakub Jelinek <jakub@redhat.com>
18274 * real.h (real_nextafter): Declare.
18275 * real.c (real_nextafter): New function.
18276 * fold-const-call.c (fold_const_nextafter): New function.
18277 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
18278 CASE_CFN_NEXTTOWARD.
18279 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
18280 even when arg1_mode is different from arg0_mode.
18282 2018-05-03 Nathan Sidwell <nathan@acm.org>
18284 * doc/extend.texi (Deprecated Features): Remove
18285 -ffriend-injection.
18286 (Backwards Compatibility): Likewise.
18287 * doc/invoke.texi (C++ Language Options): Likewise.
18288 (C++ Dialect Options): Likewise.
18290 2018-05-03 Jakub Jelinek <jakub@redhat.com>
18293 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
18294 _mm512_mask_mullox_epi64): New intrinsics.
18296 2018-05-03 Tom de Vries <tom@codesourcery.com>
18299 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
18300 dump files): Add offload-tree.
18302 2018-05-03 Richard Biener <rguenther@suse.de>
18304 PR tree-optimization/85615
18305 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
18306 to loops not nested in BBs loop father to avoid creating multi-entry
18309 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18311 PR tree-optimization/70291
18312 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
18313 arguments. Change return type to tree. Emit libcall as a new
18314 statement rather than replacing existing one when inplace_p is true.
18315 (expand_complex_multiplication_components): New function.
18316 (expand_complex_multiplication): Expand floating-point complex
18317 multiplication using the above.
18318 (expand_complex_division): Rename inner_type parameter to type.
18319 Update expand_complex_libcall call-site.
18320 (expand_complex_operations_1): Update expand_complex_multiplication
18321 and expand_complex_division call-sites.
18323 2018-05-02 Jakub Jelinek <jakub@redhat.com>
18326 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
18327 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
18328 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
18329 the highest significant bit of the shift count mask is clear. In
18330 check whether and[sq]i3 is needed verify that all significant bits
18331 of the shift count other than the highest are set.
18333 2018-05-02 Tom de Vries <tom@codesourcery.com>
18336 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
18337 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
18338 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
18339 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
18340 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
18341 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
18342 __builtin_goacc_parlevel_size.
18344 2018-05-02 Richard Biener <rguenther@suse.de>
18346 PR tree-optimization/85597
18347 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
18348 do not use split vect_get_vec_defs call but call vect_get_slp_defs
18351 2018-05-02 Tom de Vries <tom@codesourcery.com>
18354 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
18355 dump files): Add ltrans-tree.
18357 2018-05-02 Tom de Vries <tom@codesourcery.com>
18360 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
18361 dump files): Add wpa-ipa.
18363 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
18365 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
18366 powerpc*-*-linux*paired* target.
18367 * config/rs6000/750cl.h: Delete.
18368 * config/rs6000/paired.h: Delete.
18369 * config/rs6000/paired.md: Delete.
18370 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
18372 * config/rs6000/rs6000-builtin.def: Remove paired float support.
18373 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
18374 comment. Remove paired float support.
18375 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
18376 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
18378 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
18379 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
18381 * config/rs6000/rs6000.c: Remove paired float support.
18382 (paired_expand_vector_init, paired_expand_vector_move,
18383 paired_emit_vector_compare, paired_emit_vector_cond_expr,
18384 (paired_expand_lv_builtin, paired_expand_stv_builtin,
18385 paired_expand_builtin, paired_expand_predicate_builtin,
18386 paired_init_builtins): Delete.
18387 * config/rs6000/rs6000.h: Remove paired float support.
18388 * config/rs6000/rs6000.md: Remove paired float support.
18389 (move_from_CR_ov_bit): Delete.
18390 * config/rs6000/rs6000.opt (mpaired): Delete.
18391 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
18392 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
18394 2018-05-02 Richard Biener <rguenther@suse.de>
18396 PR middle-end/85567
18397 * gimplify.c (gimplify_save_expr): When in SSA form allow
18398 SAVE_EXPRs to compute to SSA vars.
18400 2018-05-02 Jakub Jelinek <jakub@redhat.com>
18403 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
18404 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
18405 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
18406 clobber operands[2], instead use a new pseudo. Formatting fixes.
18408 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
18410 PR tree-optimization/85586
18411 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
18412 exit early for statements in the same group if the accesses are
18415 2018-05-02 Tom de Vries <tom@codesourcery.com>
18418 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
18421 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
18423 PR tree-optimization/85143
18424 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
18426 2018-05-01 Tom de Vries <tom@codesourcery.com>
18429 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
18430 not found" error message.
18432 2018-05-01 Tom de Vries <tom@codesourcery.com>
18435 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
18436 * vec.c (test_ordered_remove_if): New function.
18437 (vec_c_tests): Call test_ordered_remove_if.
18438 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
18439 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
18440 * tree-vect-patterns.c (vect_pattern_recog_1): Use
18441 VEC_ORDERED_REMOVE_IF.
18443 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18445 PR tree-optimization/82665
18446 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
18447 pointer subtraction where arguments come from a memchr call.
18449 2018-05-01 Jakub Jelinek <jakub@redhat.com>
18451 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
18452 --push-state --as-needed and --pop-state instead of --as-needed and
18453 --no-as-needed if ld supports it.
18454 * configure: Regenerated.
18457 * doc/install.texi2html: Replace _002d with - and _002a with * in
18458 generated html files using sed.
18460 2018-04-30 David Malcolm <dmalcolm@redhat.com>
18463 * gcc-rich-location.c (blank_line_before_p): New function.
18464 (use_new_line): New function.
18465 (gcc_rich_location::add_fixit_insert_formatted): New function.
18466 * gcc-rich-location.h
18467 (gcc_rich_location::add_fixit_insert_formatted): New function.
18469 2018-04-30 David Malcolm <dmalcolm@redhat.com>
18471 * selftest.c (assert_streq): Rename "expected" and "actual" to
18472 "val1" and "val2". Extend NULL-handling to cover both inputs
18473 symmetrically, while still requiring both to be non-NULL for a pass.
18474 * selftest.h (assert_streq): Rename "expected" and "actual" to
18476 (ASSERT_EQ): Likewise.
18477 (ASSERT_EQ_AT): Likewise.
18478 (ASSERT_KNOWN_EQ): Likewise.
18479 (ASSERT_KNOWN_EQ_AT): Likewise.
18480 (ASSERT_NE): Likewise.
18481 (ASSERT_MAYBE_NE): Likewise.
18482 (ASSERT_MAYBE_NE_AT): Likewise.
18483 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
18484 the assertion to pass.
18485 (ASSERT_STREQ_AT): Likewise.
18487 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
18489 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
18490 interaction with -pie.
18492 2018-04-30 David Malcolm <dmalcolm@redhat.com>
18494 * selftest.h: Fix alphabetization of per-source-file selftest
18497 2018-04-30 Jason Merrill <jason@redhat.com>
18499 PR c++/61982 - dead stores to destroyed objects.
18500 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
18503 2018-04-30 Jason Merrill <jason@redhat.com>
18505 * tree.c (build_clobber): New.
18506 * tree.h: Declare it.
18507 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
18509 2018-04-30 David Malcolm <dmalcolm@redhat.com>
18511 * diagnostic-show-locus.c (layout::layout): Update for
18512 location_get_source_line returning a char_span.
18513 (struct char_span): Move to input.h.
18514 (struct correction): Update for fields in char_span becoming
18516 (struct source_line): Update for location_get_source_line
18517 returning a char_span.
18518 (layout::print_line): Likewise.
18519 * edit-context.c (edited_file::print_content): Likewise.
18520 (edited_file::print_diff_hunk): Likewise.
18521 (edited_file::print_run_of_changed_lines): Likewise.
18522 (edited_file::get_num_lines): Likewise.
18523 (edited_line::edited_line): Likewise.
18524 * final.c (asm_show_source): Likewise.
18525 * input.c (location_get_source_line): Convert return type
18526 from const char * to char_span, losing the final "line_len"
18528 (dump_location_info): Update for the above.
18529 (get_substring_ranges_for_loc): Likewise. Use a char_span
18530 when handling the literal within the line.
18531 (test_reading_source_line): Update for location_get_source_line
18532 returning a char_span.
18533 * input.h (class char_span): Move here from
18534 diagnostic-show-locus.c, converting from a struct to a class.
18535 Make data members private.
18536 (char_span::operator bool): New.
18537 (char_span::length): New.
18538 (char_span::get_buffer): New.
18539 (char_span::operator[]): New.
18540 (char_span::subspan): Make const.
18541 (char_span::xstrdup): New.
18542 (location_get_source_line): Convert return type from const char *
18543 to char_span, losing the final "line_size" param.
18545 2018-04-30 Jan Hubicka <jh@suse.cz>
18547 * lto-wrapper.c (ltrans_priorities): New static var.
18548 (cmp_priority): New.
18549 (run_gcc): Read priorities and if doing parallel build order
18550 the Makefile by them.
18552 2018-04-30 David Malcolm <dmalcolm@redhat.com>
18554 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
18556 2018-04-30 Richard Biener <rguenther@suse.de>
18558 * tree-cfg.c (verify_address): Remove base argument, add
18559 flag whether to check TREE_ADDRESSABLE and do that.
18560 (verify_expr): Remove.
18561 (verify_types_in_gimple_reference): Add pieces from verify_expr.
18562 (verify_gimple_assign_single): Likewise.
18563 (verify_gimple_switch): Likewise.
18564 (verify_expr_location_1): Dereference tp once. Add (disabled)
18565 piece from verify_expr.
18566 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
18568 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
18570 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
18572 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
18574 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
18575 (small_data_pattern): Likewise.
18576 (arc_rewrite_small_data): Likewise.
18577 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
18578 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
18579 (get_symbol_alignment): New function.
18580 (legitimate_small_data_address_p): Likewise.
18581 (legitimate_scaled_address): Update, call
18582 legitimate_small_data_address_p.
18583 (output_sdata): New static variable.
18584 (arc_print_operand): Update how we handle small data operands.
18585 (arc_print_operand_address): Likewise.
18586 (arc_legitimate_address_p): Update, use
18587 legitimate_small_data_address_p.
18588 (arc_rewrite_small_data_p): Remove.
18589 (arc_rewrite_small_data_1): Likewise.
18590 (arc_rewrite_small_data): Likewise.
18591 (small_data_pattern): Likewise.
18592 (compact_sda_memory_operand): Update to use
18593 legitimate_small_data_address_p and get_symbol_alignment.
18594 (prepare_move_operands): Don't rewite sdata pattern.
18595 (prepare_extend_operands): Remove.
18596 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
18598 (zero_extendqisi2): Likewise.
18599 (zero_extendhisi2): Likewise.
18600 (extendqihi2): Likewise.
18601 (extendqisi2): Likewise.
18602 (extendhisi2): Likewise.
18603 (addsi3): Likewise.
18604 (subsi3): Likewise.
18605 (andsi3): Likewise.
18606 * config/arc/constraints.md (Usd): Change it to memory constraint.
18608 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
18610 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
18611 as source of std instructions.
18612 * config/arc/arc.md (movsi_insn): Update pattern predicate to
18613 allow 6-bit constants as source for store instructions.
18614 (movdi_insn): Update instruction pattern to allow 6-bit constants
18615 as source for store instructions.
18617 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
18619 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
18621 2018-04-30 Nathan Sidwell <nathan@acm.org>
18622 Sandra Loosemore <sandra@codesourcery.com>
18624 * dumpfile.c (dump_open): Allow '-' for stdout.
18625 * doc/invoke.texi (Developer Options): Document dump filename
18626 determination early. Document stdin/stdout selection.
18628 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
18630 Microblaze Target: PIC data text relative
18632 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
18633 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
18635 * config/microblaze/microblaze.h (microblaze_constant_address_p):
18636 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
18637 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
18638 New addressing mode for data-text relative position indepenedent code.
18639 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
18640 'ADDRESS_SYMBOLIC_TXT_REL'.
18641 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
18642 (microblaze_legitimate_pic_operand): Exclude function calls from
18643 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
18644 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
18646 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
18647 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
18648 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
18649 for 'address + offset'.
18650 (microblaze_expand_prologue): Add new function prologue call for
18652 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
18653 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
18654 table in case of TARGET_PIC_DATA_TEXT_REL.
18655 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
18656 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
18657 Add new macros 'UNSPEC_TEXT',
18658 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
18659 + exclude function calls from 'UNSPEC_PLT' in case of data text
18661 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
18662 new target hook for generating address diff vector tables in case of
18664 * doc/tm.texi : Regenerate.
18665 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
18666 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
18667 of addr diff vector generation.
18668 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
18669 target hook definition.
18670 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
18671 Add default function for generate_pic_addr_diff_vec -> flag_pic.
18672 * doc/invoke.texi (Add new pic option): Add new microblaze pic
18673 option for data text relative.
18675 2018-04-30 Richard Biener <rguenther@suse.de>
18677 * tree-chrec.h (evolution_function_is_constant_p): Remove
18679 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
18681 2018-04-30 Richard Biener <rguenther@suse.de>
18684 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
18686 2018-04-30 Richard Biener <rguenther@suse.de>
18688 PR tree-optimization/28364
18689 PR tree-optimization/85275
18690 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
18691 copying first exit test.
18693 2018-04-28 Mark Wielaard <mark@klomp.org>
18695 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
18696 dwarf_version >= 5.
18697 (dwarf_AT): Handle DW_AT_addr_base.
18698 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
18700 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
18703 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
18704 (*ashl<dwi>3_doubleword_mask_1): Ditto.
18705 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
18706 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
18708 2018-04-28 Richard Biener <rguenther@suse.de>
18710 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
18711 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
18712 to reflect use. Only add interesting stmts.
18714 2018-04-27 Martin Jambor <mjambor@suse.cz>
18717 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
18718 the jump function allows for passing through aggregate values.
18720 2018-04-27 David Malcolm <dmalcolm@redhat.com>
18722 * input.h (in_system_header_at): Convert from macro to inline
18724 (from_macro_expansion_at): Likewise.
18725 (from_macro_definition_at): Likewise.
18727 2018-04-27 Jeff Law <law@redhat.com>
18729 * config.gcc: Mark tile* targets as deprecated/obsolete.
18731 2018-04-27 Richard Biener <rguenther@suse.de>
18733 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
18736 2018-04-27 Richard Biener <rguenther@suse.de>
18738 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
18740 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
18742 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
18743 with Yd constraint. Set "preferred_for_speed" attribute from
18744 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
18745 with Yd constraint.
18746 (*movdi_internal): Ditto.
18747 (movti_interunit splitters): Remove
18748 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
18749 (movdi_interunit splitters): Ditto.
18750 * config/i386/constraints.md (Ye): Remove.
18751 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
18753 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18756 * config/aarch64/constraints.md (Usg): Limit to 31.
18757 (Usj): Limit to 63.
18759 2018-04-27 Jakub Jelinek <jakub@redhat.com>
18761 PR tree-optimization/85529
18762 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
18763 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
18764 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
18765 zero extension or masking of the MSB bit.
18766 (optimize_range_tests): Add FIRST_BB argument, pass it through
18767 to optimize_range_tests_var_bound.
18768 (maybe_optimize_range_tests, reassociate_bb): Adjust
18769 optimize_range_tests callers.
18771 2018-04-26 Richard Biener <rguenther@suse.de>
18772 Jakub Jelinek <jakub@redhat.com>
18774 * cgraph.h (symbol_table): Just declare debug method here.
18775 * symtab.c (symbol_table::debug): Define.
18777 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
18779 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
18781 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
18783 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
18784 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
18785 (*movdi_internal): Substitute Yi and Yj constraint with x
18786 and Ym and Yn constraint with y constraint. Update "isa"
18787 attribute and set "preferred_for_speed" attribute from
18788 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
18789 (*movsi_internal): Ditto.
18790 (*movdf_internal): Ditto.
18791 (*movsf_internal): Ditto.
18792 (*zero_extendsidi2): Ditto.
18793 * config/i386/sse.md (vec_set<mode>_0): Ditto.
18794 (sse2_loadld): Ditto.
18795 (*vec_extract<ssevecmodelower>_0): Ditto.
18796 (*vec_extractv4si_0_zext_sse4): Ditto.
18797 (vec_concatv2di): Ditto.
18798 (*vec_dup<mode>): Ditto.
18799 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
18800 * config/i386/constraints.md (Yi): Remove.
18805 2018-04-26 Nathan Sidwell <nathan@acm.org>
18807 * dumpfile.c (dump_open): New.
18808 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
18809 (dump_finish): Detect stdio/stderr by value not name.
18811 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
18813 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
18815 2018-04-26 Tom de Vries <tom@codesourcery.com>
18818 * config/nvptx/nvptx.c (verify_neutering_jumps)
18819 (verify_neutering_labels): New function
18820 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
18822 2018-04-26 Tom de Vries <tom@codesourcery.com>
18825 * config/nvptx/nvptx.c (needs_neutering_p): New function.
18826 (nvptx_single): Use needs_neutering_p to skip over insns that do not
18829 2018-04-26 Richard Biener <rguenther@suse.de>
18830 Tom de Vries <tom@codesourcery.com>
18833 * lto-streamer-out.c (output_function): Fixup loops if required to match
18834 discovery done in the reader.
18836 2018-04-26 Richard Biener <rguenther@suse.de>
18838 PR tree-optimization/85116
18839 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
18840 have a loop exit from the single latch predecessor. Remove
18841 case of header with just condition.
18842 (ch_base::copy_headers): Exclude infinite loops from any
18844 (pass_ch::execute): Record exits.
18846 2018-04-26 Richard Biener <rguenther@suse.de>
18848 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
18849 prologue cost vector and pass it to vect_get_load_cost.
18850 (vect_get_peeling_costs_all_drs): Likewise.
18851 (vect_peeling_hash_get_lowest_cost): Likewise.
18852 (vect_enhance_data_refs_alignment): Likewise.
18854 2018-04-26 Richard Biener <rguenther@suse.de>
18856 PR middle-end/85450
18857 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
18858 checking of integer<->pointer conversions.
18859 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
18860 sign-/zero-extending pointer types.
18861 (expand_omp_for_static_chunk): Likewise.
18863 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
18864 Jean Lee <xiaoyur347@gmail.com>
18866 * config/mips/mips.c (mips_asan_shadow_offset): New function.
18867 (TARGET_ASAN_SHADOW_OFFSET): Define.
18868 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
18869 true for -fsanitize=address.
18871 2018-04-25 Mark Wielaard <mark@klomp.org>
18873 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
18876 2018-04-25 Jakub Jelinek <jakub@redhat.com>
18878 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
18879 than "alu", remove explicit "memory" and "imm_disp" attributes.
18880 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
18882 PR middle-end/85414
18883 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
18884 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
18885 gen_lowpart_no_emit.
18887 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
18890 * config/i386/i386.c (ix86_expand_builtin): Change memory
18891 operand to XI, extend p0 to Pmode.
18892 * config/i386/i386.md: Change unspec volatile and operand
18893 1 mode to XI, change operand 0 mode to P.
18895 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
18897 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
18898 GET_MODE_MASK before any checking.
18899 (nds32_can_use_bset_p): Likewise.
18900 (nds32_can_use_btgl_p): Likewise.
18902 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
18904 * config/nds32/nds32-doubleword.md: New define_split pattern for
18905 illegal register number.
18907 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
18909 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
18911 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
18913 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
18915 2018-04-25 Richard Biener <rguenther@suse.de>
18917 * lto-streamer.h (LTO_major_version): Bump to 8.
18919 2018-04-25 Jakub Jelinek <jakub@redhat.com>
18921 * BASE-VER: Set to 9.0.0.
18923 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
18925 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
18926 in __abskf2 and __powikf2.
18928 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18931 * config/aarch64/constraints.md (Usg, Usj): New constraints.
18932 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
18933 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
18934 Use the above on operand 2. Reindent.
18935 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
18937 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
18940 * common/config/i386/i386-common.c (ix86_handle_option): Don't
18942 * config/i386/i386.opt (mcet): Removed.
18943 * doc/install.texi: Remove -mcet documentation.
18944 * doc/invoke.texi: Likewise.
18946 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
18949 * doc/install.texi: Remove -mcet from bootstrap-cet.
18951 2018-04-24 Jakub Jelinek <jakub@redhat.com>
18954 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
18955 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
18959 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
18960 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
18961 containing a CONST_VECTOR.
18963 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
18965 * doc/install.texi: Update newlib dependency for nvptx.
18967 2018-04-24 Jakub Jelinek <jakub@redhat.com>
18970 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
18971 instead of INTVAL when shifting x left.
18973 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
18975 PR tree-optimization/85478
18976 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
18977 vect_grouped_store_supported for single element vectors.
18979 2018-04-24 Richard Biener <rguenther@suse.de>
18982 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
18983 load cost increase to the case of non-constant step.
18985 2018-04-24 Jakub Jelinek <jakub@redhat.com>
18988 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
18989 destination if any_malformed_asm.
18991 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
18993 PR middle-end/85496
18994 * expr.c (store_field): In the bitfield case, if the value comes from
18995 a function call and is returned in registers by means of a PARALLEL,
18996 do not change the mode of the temporary unless BLKmode and VOIDmode.
18998 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
19000 PR rtl-optimization/85423
19001 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
19002 dependencies to debug insns when the previous insn is non-debug.
19004 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
19006 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
19007 enums into a single definition.
19008 (fls): Fix predicates and printing.
19011 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
19013 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
19014 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
19015 and short u6 immediate.
19016 (check_if_valid_sleep_operand): Remove.
19017 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
19019 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
19021 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
19022 flag_always_save_lp condition.
19023 * config/nds32/nds32.opt (malways-save-lp): New option.
19025 2018-04-22 Shiva Chen <shiva0217@gmail.com>
19027 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
19028 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
19029 * config/nds32/nds32.h
19030 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
19031 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
19033 2018-04-22 Shiva Chen <shiva0217@gmail.com>
19035 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
19036 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
19038 2018-04-22 Shiva Chen <shiva0217@gmail.com>
19039 Chung-Ju Wu <jasonwucj@gmail.com>
19041 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
19043 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
19044 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
19046 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
19048 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
19050 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
19052 * config/nds32/nds32-protos.h (nds32_data_alignment,
19053 nds32_local_alignment): Declare.
19054 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
19055 nds32_local_alignment): New functions.
19056 (TARGET_CONSTANT_ALIGNMENT): Define.
19057 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
19059 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
19061 * config/nds32/nds32.c
19062 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
19063 (TARGET_MODES_TIEABLE_P): Likewise.
19065 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
19067 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
19068 level Ofast and Og.
19070 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
19071 Chung-Ju Wu <jasonwucj@gmail.com>
19073 * config/nds32/constants.md (unspec_volatile_element): Add enum values
19074 for unaligned access.
19075 * config/nds32/nds32-intrinsic.c: Implementation of expanding
19077 * config/nds32/nds32-intrinsic.md: Likewise.
19078 * config/nds32/nds32_intrinsic.h: Likewise.
19079 * config/nds32/nds32.h (nds32_builtins): Likewise.
19080 * config/nds32/nds32.opt (munaligned-access): New option.
19081 * config/nds32/nds32.c (nds32_asm_file_start): Display
19082 flag_unaligned_access status.
19084 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
19086 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
19087 -mno-relax is present.
19088 * config/riscv/linux.h (LINK_SPEC): Ditto.
19090 2018-04-20 Martin Sebor <msebor@redhat.com>
19093 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
19095 (gimple_fold_builtin_stxcpy_chk): Same.
19096 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
19098 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
19101 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
19102 __powikf2 when long double is IEEE 128-bit.
19104 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
19106 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
19107 step to make sure stack always aligned.
19109 2018-04-20 Carl Love <cel@us.ibm.com>
19112 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
19113 size check for arg0.
19115 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
19116 Tom de Vries <tom@codesourcery.com>
19119 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
19120 Emit insns for calls too.
19121 (nvptx_find_par): Always look for worker-level predecessor insn.
19122 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
19124 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
19125 (nvptx_process_pars): Propagate frames for calls.
19127 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
19130 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
19132 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
19133 (ix86_handle_option): Don't handle OPT_mibt.
19134 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
19136 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
19138 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
19139 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
19140 (ix86_target_macros): Define __CET__ with flag_cf_protection
19141 for -fcf-protection.
19142 * config/i386/i386.c (isa2_opts): Remove -mibt.
19143 * config/i386/i386.h (TARGET_IBT): Removed.
19144 (TARGET_IBT_P): Likewise.
19145 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
19146 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
19147 * config/i386/i386.opt (mcet): Update help message.
19148 (mshstk): Likewise.
19150 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
19151 -mcet as an alias for -mshstk.
19153 2018-04-20 Richard Biener <rguenther@suse.de>
19155 PR middle-end/85475
19156 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
19157 complexity by forcing a single use of the multiply operand.
19159 2018-04-20 Martin Jambor <mjambor@suse.cz>
19162 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
19163 recursion dependency to only apply to non-clones.
19165 2018-04-20 Martin Jambor <mjambor@suse.cz>
19168 * ipa-cp.c (create_specialized_node): Check that clones of
19169 self-recursive edges exist during IPA-CP.
19171 2018-04-19 Toon Moene <toon@moene.org>
19173 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
19176 2018-04-19 Jakub Jelinek <jakub@redhat.com>
19178 PR tree-optimization/85467
19179 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
19180 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
19181 VECTOR_CST element to type.
19183 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
19186 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
19187 * config/i386/i386.md (builtin_setjmp_setup): Removed.
19188 (builtin_longjmp): Likewise.
19189 (save_stack_nonlocal): New pattern.
19190 (restore_stack_nonlocal): Likewise.
19192 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
19195 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
19196 Replace ASM_OUTPUT_LABEL with fprintf.
19198 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
19201 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
19202 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
19203 * config/i386/i386-c.c (ix86_target_macros_internal): Also
19204 define __IBT__ and __SHSTK__ for -fcf-protection.
19205 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
19207 (ix86_trampoline_init): Likewise.
19208 (x86_output_mi_thunk): Likewise.
19209 (ix86_notrack_prefixed_insn_p): Likewise.
19210 (ix86_option_override_internal): Don't disallow -fcf-protection.
19211 * config/i386/i386.md (rdssp<mode>): Also enable for
19213 (incssp<mode>): Likewise.
19214 (nop_endbr): Likewise.
19215 * config/i386/i386.opt (mcet): Change help message to built-in
19218 (mshstk): Likewise.
19219 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
19220 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
19221 enable CET built-in functions.
19223 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
19225 * common/config/i386/i386-common.c
19226 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
19227 OPTION_MASK_ISA_MOVDIRI_UNSET,
19228 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
19229 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
19230 * config.gcc (movdirintrin.h): New header.
19231 * config/i386/cpuid.h (bit_MOVDIRI,
19232 bit_MOVDIR64B): New bits.
19233 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
19235 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
19236 (VOID, PVOID, PCVOID)): New function types.
19237 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
19238 __builtin_ia32_directstoreu_u64,
19239 __builtin_ia32_movdir64b): New builtins.
19240 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
19241 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
19243 (ix86_valid_target_attribute_inner_p): Ditto.
19244 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
19245 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
19246 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
19247 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
19248 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
19249 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
19250 (movdiri<mode>, movdir64b_<mode>): New.
19251 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
19252 * config/i386/immintrin.h: Include movdirintrin.h.
19253 * config/i386/movdirintrin.h: New file.
19254 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
19256 2018-04-19 Richard Biener <rguenther@suse.de>
19258 PR middle-end/85455
19259 * cfg.c (clear_bb_flags): When loop state says we have
19260 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
19262 2018-04-19 Richard Biener <rguenther@suse.de>
19264 PR tree-optimization/84737
19265 * tree-vect-data-refs.c (vect_copy_ref_info): New function
19266 copying restrict info.
19267 (vect_setup_realignment): Use it.
19268 * tree-vectorizer.h (vect_copy_ref_info): Declare.
19269 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
19270 the first DR to all generated stores.
19271 (vectorizable_load): Likewise for loads.
19273 2018-04-19 Jakub Jelinek <jakub@redhat.com>
19275 PR tree-optimization/85446
19276 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
19277 the integral and pointer types to have the same precision.
19279 * doc/install.texi: Document --disable-cet being the default and
19282 2018-04-18 Martin Liska <mliska@suse.cz>
19284 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
19287 2018-04-18 Martin Liska <mliska@suse.cz>
19290 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
19293 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
19294 arguments if they are comparable.
19296 2018-04-18 Martin Liska <mliska@suse.cz>
19299 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
19302 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
19305 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
19308 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
19309 ENDBR after calling __morestack.
19311 2018-04-18 David Malcolm <dmalcolm@redhat.com>
19314 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
19315 by using gcc_base_ver to generate a gcc_driver_version, and use
19316 it when generating GCC_DRIVER_NAME.
19317 * configure: Regenerate.
19319 2018-04-18 Jakub Jelinek <jakub@redhat.com>
19322 * config.gcc: Obsolete powerpc*-*-*spe*.
19324 2018-04-17 Jakub Jelinek <jakub@redhat.com>
19327 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
19328 (stabstr_D): Change type of unum from unsigned int to
19329 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
19332 2018-04-17 Jim Wilson <jimw@sifive.com>
19335 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
19336 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
19337 Set arg_pointer_offset after using pretend_args_size.
19339 2018-04-17 Jakub Jelinek <jakub@redhat.com>
19341 PR rtl-optimization/85431
19342 * dse.c (record_store): Ignore zero width stores.
19345 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
19346 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
19347 __builtin_stack_restore rather than after it.
19348 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
19349 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
19350 argument instead of virtual_dynamic_stack_rtx.
19352 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
19354 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
19356 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
19357 Add note to error message to explain internal mapping of overloaded
19358 built-in function name to non-overloaded built-in function name.
19359 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
19362 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
19365 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
19366 where the inputs overlap with the output.
19368 2018-04-17 Jakub Jelinek <jakub@redhat.com>
19370 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
19371 (=v, v) alternative and explicit "memory" attribute.
19372 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
19373 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
19375 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
19376 "sselog1" type instead of "sselog".
19377 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
19378 "sselog". Remove explicit "memory" attribute.
19379 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
19380 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
19382 (vec_extract_hi_v32hi): Merge all alternatives into one, use
19383 "sselog1" type instead of "sselog". Remove explicit "memory"
19385 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
19386 use "sselog1" type instead of "sselog". Remove explicit "memory"
19388 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
19389 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
19391 (vec_extract_hi_v64qi): Merge all alternatives into one, use
19392 "sselog1" type instead of "sselog". Remove explicit "memory"
19394 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
19395 use "sselog1" type instead of "sselog". Remove explicit "memory"
19399 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
19401 PR middle-end/85414
19402 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
19405 2018-04-17 Martin Jambor <mjambor@suse.cz>
19408 * ipa-cp.c (create_specialized_node): Call
19409 expand_all_artificial_thunks if necessary.
19411 2018-04-17 Martin Liska <mliska@suse.cz>
19414 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
19415 in message, remote space in between '_G' and '('.
19417 2018-04-17 Jakub Jelinek <jakub@redhat.com>
19420 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
19421 avx512f_vmcmp<mode>3<round_saeonly_name>,
19422 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
19423 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
19424 avx512f_rndscale<mode><round_saeonly_name>,
19425 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
19426 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
19427 Use %<iptr>2 instead of %2 for -masm=intel.
19428 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
19429 avx512f_vcvttss2usi<round_saeonly_name>,
19430 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
19432 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
19433 avx512f_vcvttsd2usi<round_saeonly_name>,
19434 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
19435 Use %q1 instead of %1 for -masm=intel.
19436 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
19437 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
19438 of %3 for -masm=intel.
19439 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
19441 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
19443 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
19445 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
19446 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
19448 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
19450 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
19451 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
19452 %g1 and one with %0 and %1.
19453 (avx512er_vmrcp28<mode><round_saeonly_name>,
19454 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
19455 %1 for -masm=intel.
19456 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
19457 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
19458 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
19459 of %0 and %{%4%} for -masm=intel.
19460 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
19461 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
19462 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
19463 order of %0 and %{%5%}%{z%} for -masm=intel.
19465 2018-04-17 Jan Hubicka <jh@suse.cz>
19468 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
19470 2018-04-17 Martin Liska <mliska@suse.cz>
19473 * multiple_target.c (create_dispatcher_calls): Set apostrophes
19474 for target_clone error message. Make default implementation
19475 clone to be a local declaration.
19476 (separate_attrs): Add new argument and check for an empty
19478 (expand_target_clones): Handle it.
19479 (ipa_target_clone): Make redirection just for target_clones
19482 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
19483 Tom de Vries <tom@codesourcery.com>
19485 PR middle-end/84955
19486 * omp-expand.c (expand_oacc_for): Add dummy false branch for
19487 tiled basic blocks without omp continue statements.
19489 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
19492 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
19493 vec_extract expression as having side effects to make sure it gets
19496 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
19499 * config/i386/i386.c (get_builtin_code_for_version): Check
19502 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
19505 * config.gcc: Support "skylake".
19506 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19508 * config/i386/i386.c (m_SKYLAKE): Define.
19509 (processor_target_table): Add "skylake".
19510 (ix86_option_override_internal): Add "skylake".
19511 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
19512 PROCESSOR_CANNONLAKE.
19513 (get_builtin_code_for_version): Fix priority for
19514 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
19515 PROCESSOR_SKYLAKE-AVX512.
19516 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
19517 (processor_type): Add PROCESSOR_SKYLAKE.
19519 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
19520 Jason Merrill <jason@redhat.com>
19523 * convert.c (convert_to_integer_1): Use direct recursion for
19524 enumeral types and types with a precision less than the number
19525 of bits in their mode.
19527 2018-04-16 Julia Koval <julia.koval@intel.com>
19530 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
19531 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
19533 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
19536 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
19537 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
19538 and -mno-direct-move.
19540 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
19543 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
19544 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
19545 Ensure negative shifts result in {0}.
19547 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
19549 PR rtl-optimization/79916
19550 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
19551 regs (if any) to define how to gnerate SD moves when LRA is in
19554 2018-04-13 Jakub Jelinek <jakub@redhat.com>
19556 PR rtl-optimization/85393
19557 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
19558 * except.c (expand_dw2_landing_pad_for_region): Make static.
19559 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
19560 a label and unconditional jump to old_bb, rather than
19561 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
19564 PR rtl-optimization/85376
19565 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
19566 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
19567 instead of a specific value.
19569 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
19570 Bin Cheng <bin.cheng@arm.com>
19572 PR tree-optimization/82965
19573 PR tree-optimization/83991
19574 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
19575 by_profile_only parameter.
19576 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
19577 information if the loop was predicted to iterate too many times.
19578 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
19580 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
19583 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
19586 2018-04-13 Martin Liska <mliska@suse.cz>
19587 Jakub Jelinek <jakub@redhat.com>
19589 PR middle-end/81657
19590 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
19591 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
19592 * builtins.c (expand_builtin_memory_copy_args): Use
19593 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
19594 handle dest_addr == pc_rtx.
19596 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
19599 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
19600 asked to not generate direct moves.
19601 (fix_trunc<mode>si2_stfiwx): Similar.
19602 (fix_trunc<mode>si2_internal): Similar.
19604 2018-04-12 Jakub Jelinek <jakub@redhat.com>
19607 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
19608 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
19609 lookup if dest in some wider mode is known to be const0_rtx and
19610 if so, record permanent equivalence for it to be ZERO_EXTEND of
19611 the narrower mode destination.
19613 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
19615 * lto-streamer-out.c (output_function): Revert 259346.
19616 * omp-expand.c (expand_oacc_for): Likewise.
19618 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
19620 PR rtl-optimization/85354
19621 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
19622 * sel-sched.c (sel_global_init): ... here.
19624 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
19627 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
19628 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
19629 mode for PE-COFF targets.
19630 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
19631 (i386_pe_asm_lto_end): Likewise.
19632 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
19633 (TARGET_ASM_LTO_END): Likewise.
19634 * config/i386/winnt.c (saved_debug_info_level): New static variable.
19635 (i386_pe_asm_lto_start): New function.
19636 (i386_pe_asm_lto_end): Likewise.
19638 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
19639 Richard Biener <rguenther@suse.de>
19641 PR middle-end/84955
19642 * lto-streamer-out.c (output_function): Fix CFG loop state before
19644 * omp-expand.c (expand_oacc_for): Handle calls to internal
19645 functions like regular functions.
19647 2018-04-12 Richard Biener <rguenther@suse.de>
19650 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
19651 for the early LTO debug to properly generate references to it
19652 during DIE emission. Do not re-use that for the skeleton for
19654 (dwarf2out_early_finish): Likewise.
19656 2018-04-12 Jakub Jelinek <jakub@redhat.com>
19659 * config/i386/sse.md
19660 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
19661 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
19662 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
19663 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
19664 and output is a reg, avoid creating invalid lowpart subreg, but
19665 instead split into a 512-bit move. Don't split if not AVX512VL,
19666 input is xmm16+ reg and output is a mem.
19667 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
19668 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
19669 xmm16+ reg and output is a mem.
19671 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
19673 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
19674 also for flag_dwarf2_cfi_asm.
19676 2018-04-12 Jakub Jelinek <jakub@redhat.com>
19678 PR rtl-optimization/85342
19679 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
19680 a bool scalar var inside of the loop instead. Don't try to update
19681 recog_data.operand after failed apply_change_group.
19683 2018-04-12 Tom de Vries <tom@codesourcery.com>
19686 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
19687 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
19688 array with flexible array member as array without given dimension.
19689 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
19690 argument for undefined param to true.
19692 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
19695 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
19696 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
19697 from PowerPC section.
19698 * config/rs6000/sysv4.opt (mcall-): Improve help text.
19699 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
19700 help text that is too long.
19701 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
19702 help text that is too long.
19703 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
19704 help text that is too long.
19706 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
19708 * config/alpha/alpha.md (stack_probe_internal): Rename
19709 from "probe_stack". Update all callers.
19711 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
19713 PR rtl-optimization/84566
19714 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
19715 sched_macro_fuse_insns.
19717 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
19720 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
19721 (compute_block_dependences): ... from here.
19723 2018-04-11 Jakub Jelinek <jakub@redhat.com>
19725 PR tree-optimization/85331
19726 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
19727 from int to HOST_WIDE_INT.
19729 2018-04-11 Martin Jambor <mjambor@suse.cz>
19732 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
19733 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
19734 not the same as the source val.
19735 (cgraph_edge_brings_value_p): New parameter.
19736 (gather_edges_for_value): Pass destination value to
19737 cgraph_edge_brings_value_p.
19738 (perhaps_add_new_callers): Likewise.
19739 (get_info_about_necessary_edges): Likewise and exclude values brought
19740 only by self-recursive edges.
19741 (create_specialized_node): Redirect only clones of self-calling edges.
19742 (+self_recursive_pass_through_p): New function.
19743 (find_more_scalar_values_for_callers_subset): Use it.
19744 (find_aggregate_values_for_callers_subset): Likewise.
19745 (known_aggs_to_agg_replacement_list): Removed.
19746 (decide_whether_version_node): Re-calculate known constants for all
19747 remaining context clones.
19749 2018-04-11 Richard Biener <rguenther@suse.de>
19752 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
19753 from early DWARF output.
19754 (dwarf2out_early_finish): Output line info unconditionally into
19755 early DWARF and add reference to it.
19757 2018-04-11 Jakub Jelinek <jakub@redhat.com>
19760 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
19761 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
19762 other than V2DFmode using iptr mode attribute.
19763 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
19765 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
19767 PR rtl-optimization/84659
19768 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
19770 2018-04-11 Jakub Jelinek <jakub@redhat.com>
19773 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
19775 (output_loc_list): Pass address of a dummy size variable even in the
19776 locview handling loop.
19777 (index_location_lists): Add comment on why skip_loc_list_entry can't
19780 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
19783 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
19786 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
19789 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
19790 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
19791 and -mstring-compare-inline-limit.
19793 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
19796 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
19797 for stack clash protection in a register whenever we need it to be in
19800 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
19802 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
19803 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
19805 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
19808 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
19810 (mlong-double-): Ditto.
19811 * config/rs6000/sysv4.opt (msdata=): Ditto.
19812 (mtls-size=): Ditto.
19814 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
19816 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
19817 erroneous entries for
19818 "vector int vec_ldl (int, long int *)", and
19819 "vector unsigned int vec_ldl (int, unsigned long int *)".
19820 Add comments and entries for
19821 "vector bool char vec_ldl (int, bool char *)",
19822 "vector bool short vec_ldl (int, bool short *)",
19823 "vector bool int vec_ldl (int, bool int *)",
19824 "vector bool long long vec_ldl (int, bool long long *)",
19825 "vector pixel vec_ldl (int, pixel *)",
19826 "vector long long vec_ldl (int, long long *)",
19827 "vector unsigned long long vec_ldl (int, unsigned long long *)".
19828 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
19829 type tree bool_long_long_type_node and correct definition of
19830 bool_V2DI_type_node to make reference to this new type tree.
19831 (rs6000_mangle_type): Replace erroneous reference to
19832 bool_long_type_node with bool_long_long_type_node.
19833 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
19834 comments to emphasize sign distinctions for char and int types and
19835 replace RS6000_BTI_bool_long constant with
19836 RS6000_BTI_bool_long_long constant. Also add comment to restrict
19837 use of RS6000_BTI_pixel.
19838 (bool_long_type_node): Remove this macro definition.
19839 (bool_long_long_type_node): New macro definition
19841 2018-04-10 Jakub Jelinek <jakub@redhat.com>
19843 PR rtl-optimization/85300
19844 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
19845 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
19846 simplify_unary_operation fails.
19848 2018-04-10 Martin Liska <mliska@suse.cz>
19850 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
19851 cgraph_edge and ipa_ref.
19853 2018-04-10 Jakub Jelinek <jakub@redhat.com>
19857 * config/i386/sse.md
19858 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
19859 computation of the VEC_MERGE selector from mask.
19860 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
19861 Fix decoding of the VEC_MERGE selector into mask.
19863 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
19865 PR tree-optimization/85286
19866 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
19868 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
19870 * final.c (final_1): Set insn_last_address as well as
19871 insn_current_address.
19873 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19876 * explow.c (emit_stack_probe): Call validize_mem on memory location
19877 before passing it to gen_probe_stack. Create address operand and
19878 legitimize it for the probe_stack_address case.
19880 2018-04-09 Jan Hubicka <jh@suse.cz>
19883 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
19884 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
19885 * tree.c (free_lang_data_in_type): Fix handling of binfos;
19887 (free_lang_data): Rebuild type inheritance graph.
19889 2018-04-09 Martin Sebor <msebor@redhat.com>
19891 * invoke.texi (-finline-small-functions): Mention other optimization
19893 (-findirect-inlining, -fpartial-inlining): Same.
19894 (-finline-functions-called-once): Same.
19895 (-freorder-blocks-and-partition): Same.
19897 2018-04-09 Jan Hubicka <jh@suse.cz>
19900 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
19901 jumps; choose last target that matches the criteria (i.e.
19902 no partition changes for non-crossing jumps).
19903 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
19904 support for redirecting crossing jumps to non-crossing.
19906 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
19908 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
19909 also for naked functions.
19911 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
19913 * config/arc/arc.md (add_shift): New pattern.
19914 (add_shift2): Likewise.
19915 (sub_shift): Likewise.
19916 (sub_shift_cmp0_noout): Likewise.
19917 (compare_si_ashiftsi): Likewise.
19918 (xbfu_cmp0_noout): New combine pattern.
19919 (xbfu_cmp0"): Likewise.
19920 (movsi_set_cc_insn): Place the predicable variant first.
19921 (commutative_binary_cmp0_noout): Remove clobber.
19922 (commutative_binary_cmp0): New pattern.
19923 (noncommutative_binary_cmp0): Likewise.
19924 (noncommutative_binary_cmp0_noout): Likewise.
19925 (noncommutative_binary_comparison_result_used): Removed.
19926 (rsub_cmp0): New pattern.
19927 (rsub_cmp0_noout): Likewise.
19928 (extzvsi): Changed, keep only meaningful variants.
19929 (SQH, SEZ): New iterators.
19930 (SQH_postfix): New mode attribute.
19931 (SEZ_prefix): New code attribute.
19932 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
19933 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
19934 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
19935 of numerical value.
19936 (noncommutative_operator): Check the availability of barrel
19939 2018-04-09 Richard Biener <rguenther@suse.de>
19941 PR tree-optimization/85284
19942 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
19943 Only use the niter constraining form of simple_iv when the exit
19944 is always executed.
19946 2018-04-09 Tom de Vries <tom@codesourcery.com>
19949 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
19950 (define_expand "*memory_barrier"): New define_expand.
19951 (define_insn "memory_barrier"): New insn.
19953 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
19955 PR rtl-optimization/80463
19956 PR rtl-optimization/83972
19957 PR rtl-optimization/83480
19959 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
19960 correct producer for the insn.
19961 (tidy_control_flow): Fixup seqnos in case of debug insns.
19963 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
19965 PR rtl-optimization/83913
19967 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
19968 different sched-times when merging exprs.
19970 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
19972 PR rtl-optimization/83962
19974 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
19975 tidy_fallthru_edge and tidy_control_flow.
19977 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
19979 PR rtl-optimization/83530
19981 * sel-sched.c (force_next_insn): New global variable.
19982 (remove_insn_for_debug): When force_next_insn is true, also leave only
19983 next insn in the ready list.
19984 (sel_sched_region): When the region wasn't scheduled, make another pass
19985 over it with force_next_insn set to 1.
19987 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
19989 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
19991 * config/nds32/constants.md (unspec_volatile_element): Add enum values
19992 for interrupt control.
19993 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
19994 functions for interrupt control.
19995 * config/nds32/nds32-intrinsic.md: Likewise.
19996 * config/nds32/nds32_intrinsic.h: Likewise.
19997 * config/nds32/nds32.h (nds32_builtins): Likewise.
19999 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
20001 * config/nds32/nds32.c (nds32_init_machine_status,
20002 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
20003 strict_aligned_p field.
20004 (nds32_expand_to_rtl_hook): New function.
20005 (TARGET_EXPAND_TO_RTL_HOOK): Define.
20006 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
20008 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
20009 Chung-Ju Wu <jasonwucj@gmail.com>
20011 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
20012 * config/nds32/nds32-n7.md: New file.
20013 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
20014 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
20016 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
20017 * config/nds32/nds32.md (pipeline_model): Add n7.
20018 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
20019 * config/nds32/pipelines.md: Include n7 settings.
20021 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
20022 Chung-Ju Wu <jasonwucj@gmail.com>
20024 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
20025 * config/nds32/nds32-e8.md: New file.
20026 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
20027 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
20029 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
20030 * config/nds32/nds32.md (pipeline_model): Add e8.
20031 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
20032 * config/nds32/pipelines.md: Include e8 settings.
20034 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
20035 Chung-Ju Wu <jasonwucj@gmail.com>
20037 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
20038 * config/nds32/nds32-n8.md: New file.
20039 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
20040 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
20042 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
20043 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
20044 * config/nds32/nds32.md (pipeline_model): Add n8.
20045 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
20046 * config/nds32/pipelines.md: Include n8 settings.
20048 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
20049 Chung-Ju Wu <jasonwucj@gmail.com>
20051 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
20052 * config/nds32/nds32-n9-2r1w.md: New file.
20053 * config/nds32/nds32-n9-3r2w.md: New file.
20054 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
20055 nds32_register_ports): New or modify for cpu n9.
20056 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
20058 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
20059 * config/nds32/nds32-utils.c: New file.
20060 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
20061 TARGET_MUL_SLOW): Define.
20062 * config/nds32/nds32.md (pipeline_model): New attribute.
20063 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
20064 New options that support cpu n9.
20065 * config/nds32/pipelines.md: Include n9 settings.
20066 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
20068 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
20070 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
20071 information if necessary.
20072 (output_cond_branch_compare_zero): Likewise.
20073 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
20074 (nds32_target_alignment): Refine for alignment.
20075 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
20076 (FUNCTION_BOUNDARY): Modify.
20077 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
20079 * config/nds32/nds32.opt (malways-align, malign-functions): New.
20081 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
20083 * config/nds32/constants.md (unspec_volatile_element): Add values for
20084 TLB operation and data prefetch.
20085 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
20086 functions for TLB operation and data prefetch.
20087 * config/nds32/nds32-intrinsic.md: Likewise.
20088 * config/nds32/nds32_intrinsic.h: Likewise.
20089 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
20090 (nds32_print_operand): Likewise.
20091 * config/nds32/nds32.h (nds32_builtins): Likewise.
20093 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
20094 Andrew Pinski <pinsika@gcc.gnu.org>
20096 PR middle-end/82976
20097 * match.pd: Use constant_boolean_node of correct type instead of
20098 boolean_true_node or boolean_false_node for simplifying
20099 pointer comparisons to zero.
20101 2018-04-07 Jakub Jelinek <jakub@redhat.com>
20103 PR tree-optimization/80021
20104 * tree.c (verify_type_variant): Make error call in verify_variant_match
20105 translatable and remove final full stop.
20107 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
20109 * config/nds32/constants.md (unspec_volatile_element): Add
20110 UNSPEC_VOLATILE_EH_RETURN.
20111 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
20112 nds32_output_stack_pop): Support dwarf exception handling process.
20113 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
20114 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
20115 exception handling process.
20116 (nds32_compute_stack_frame): Likewise.
20117 (nds32_return_addr_rtx): Likewise.
20118 (nds32_initial_elimination_offset): Likewise.
20119 (nds32_expand_prologue): Likewise.
20120 (nds32_expand_epilogue): Likewise.
20121 (nds32_dynamic_chain_address): New function.
20122 * config/nds32/nds32.h (machine_function): Add fields for dwarf
20123 exception handling.
20124 (DYNAMIC_CHAIN_ADDRESS): Define.
20125 (EH_RETURN_DATA_REGNO): Define.
20126 (EH_RETURN_STACKADJ_RTX): Define.
20127 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
20128 patterns for dwarf exception handling.
20130 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
20132 * config/nds32/nds32.h: Clean up obsolete macros.
20134 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
20136 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
20137 Add enum values for particular instructions.
20138 * config/nds32/nds32-intrinsic.c: Implementation of expanding
20139 particular intrinsic functions.
20140 * config/nds32/nds32-intrinsic.md: Likewise.
20141 * config/nds32/nds32_intrinsic.h: Likewise.
20142 * config/nds32/nds32.h (nds32_builtins): Likewise.
20143 * config/nds32/nds32.md (type): Add pbsad and pbsada.
20144 (btst, ave): New patterns for particular instructions.
20146 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
20148 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
20149 Add enum values for atomic load/store and memory sync.
20150 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
20152 * config/nds32/nds32-intrinsic.md: Likewise.
20153 * config/nds32/nds32_intrinsic.h: Likewise.
20154 * config/nds32/nds32.h (nds32_builtins): Likewise.
20156 2018-04-07 Jakub Jelinek <jakub@redhat.com>
20158 PR tree-optimization/85257
20159 * fold-const.c (native_encode_vector): If not all elts could fit
20160 and off is -1, return 0 rather than offset.
20161 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
20162 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
20163 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
20164 adjust buffer in native_interpret_expr call.
20166 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
20168 * config/nds32/constants.md (unspec_volatile_element): Add cache
20169 control enum values.
20170 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
20171 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
20172 * config/nds32/nds32.c (nds32_cctl_names): New.
20173 (nds32_print_operand): Handle cache control register names.
20174 * config/nds32/nds32.h (nds32_builtins): New enum values.
20175 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
20177 * config/nds32/nds32.md (type): Add mmu.
20178 * config/nds32/pipelines.md (simple_insn): Add mmu.
20180 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
20182 * config/nds32/nds32.md (type): Remove call.
20183 * config/nds32/pipelines.md (simple_insn): Likewise.
20185 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
20187 * config/nds32/constants.md (unspec_volatile_element): Add
20188 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
20189 UNSPEC_VOLATILE_FMFCFG.
20190 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
20191 description for fmfcfg and fmfcsr.
20192 (bdesc_1arg): Add fmtcsr.
20193 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
20194 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
20195 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
20196 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
20197 unspec_fmfcfg): New patterns.
20198 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
20199 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
20200 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
20201 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
20202 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
20203 __nds32__fmfcfg): Define.
20205 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
20207 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
20208 intrinsic register names.
20209 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
20210 intrinsic register enum values and macros.
20212 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
20214 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
20215 for load/store addressing form.
20216 (nds32_print_operand_address): Likewise.
20218 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
20221 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
20222 based on LABEL_REF. Remove useless assertion.
20223 (pic_address_needs_scratch): Fix formatting.
20224 (sparc_legitimize_pic_address): Minor tweaks.
20225 (sparc_delegitimize_address): Adjust assertion accordingly.
20226 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
20227 into symbolic_operand.
20228 (movsi_high_pic_label_ref): Likewise.
20229 (movsi_lo_sum_pic_label_ref): Likewise.
20230 (movdi_pic_label_ref): Likewise.
20231 (movdi_high_pic_label_ref): Likewise.
20232 (movdi_lo_sum_pic_label_ref): Likewise.
20234 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
20236 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
20237 custom LIB_SPEC setup.
20239 2018-04-06 Ruslan Bukin <br@bsdpad.com>
20240 Kito Cheng <kito.cheng@gmail.com>
20242 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
20243 * config/riscv/freebsd.h: New.
20245 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
20247 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
20248 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
20251 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
20252 Kito Cheng <kito.cheng@gmail.com>
20254 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
20255 nds32_output_call, nds32_symbol_binds_local_p): New functions.
20256 * config/nds32/nds32-protos.h (nds32_output_call,
20257 nds32_output_return): Declare.
20258 * config/nds32/nds32.md: Refine all the call and return patterns.
20260 2018-04-06 Jakub Jelinek <jakub@redhat.com>
20263 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
20264 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
20266 PR rtl-optimization/84872
20267 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
20268 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
20269 EDGE_CROSSING edge.
20271 2018-04-06 Tamar Christina <tamar.christina@arm.com>
20273 * expr.c (copy_blkmode_to_reg): Revert 254862.
20274 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
20276 2018-04-06 Richard Biener <rguenther@suse.de>
20278 PR middle-end/85244
20279 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
20280 after seeing a component reference with an adjacent field. Treat
20281 refs to arrays at struct end of external decls similar to
20282 refs to unconstrained commons.
20284 2018-04-06 Jakub Jelinek <jakub@redhat.com>
20287 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
20288 look through SAVE_EXPRs with non-side-effects argument. Adjust
20290 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
20293 2018-04-06 Richard Biener <rguenther@suse.de>
20295 PR middle-end/85180
20296 * alias.c (find_base_term): New wrapper around find_base_term
20297 unwinding CSELIB_VAL_PTR changes.
20298 (find_base_term): Do not restore CSELIB_VAL_PTR during the
20301 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
20303 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
20305 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
20306 constant definitions.
20307 ("nop"): lr 0,0 -> nopr r0
20308 ("nop_lr0", "nop_lr1"): New insn definitions.
20310 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
20312 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
20313 NDS32_V3PUSH_AVAILABLE_P macro.
20315 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
20316 Chung-Ju Wu <jasonwucj@gmail.com>
20318 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
20319 (nds32*-*-*): Add float and fpu_config into supported_defaults.
20320 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
20321 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
20322 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
20323 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
20324 * config/nds32/constraints.md: New constraints and checking for hard
20325 float configuration.
20326 * config/nds32/iterators.md: New mode iterator and attribute for hard
20327 float configuration.
20328 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
20330 * config/nds32/nds32-fpu.md: New file.
20331 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
20332 deal with hard float code generation.
20333 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
20335 (abi_type, float_reg_number): New enum type.
20336 * config/nds32/nds32-predicates.c: New predicates for hard float.
20337 * config/nds32/nds32-protos.h: Declare functions for hard float.
20338 * config/nds32/nds32.c: Implementation for hard float configuration.
20339 * config/nds32/nds32.h: Definitions for hard float configuration.
20340 * config/nds32/nds32.md: Include hard float machine description and
20341 modify patterns for hard float configuration.
20342 * config/nds32/nds32.opt: New options for hard float configuration.
20343 * config/nds32/predicates.md: New predicates for hard float
20346 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
20348 * common/config/nds32/nds32-common.c
20349 (nds32_option_optimization_table): Enable -mreleax-hint by default.
20351 2018-04-05 Jakub Jelinek <jakub@redhat.com>
20353 PR middle-end/85195
20354 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
20355 CONSTRUCTOR_ELT (ctor, ...)->value.
20357 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
20360 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
20362 2018-04-05 Tom de Vries <tom@codesourcery.com>
20365 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
20368 2018-04-05 Shiva Chen <shiva0217@gmail.com>
20369 Kito Cheng <kito.cheng@gmail.com>
20371 * config/nds32/constraints.md (U33): Fine-tune checking condition.
20372 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
20373 * config/nds32/nds32.h (nds32_16bit_address_type): Add
20374 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
20376 2018-04-05 Shiva Chen <shiva0217@gmail.com>
20377 Kito Cheng <kito.cheng@gmail.com>
20379 * config/nds32/constraints.md (Ufe): New memory constraint.
20380 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
20381 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
20382 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
20384 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
20385 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
20387 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
20389 * config/nds32/nds32.md: Use optimize_size in the condition for
20390 alu-shift instructions.
20392 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
20394 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
20396 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
20398 * config/nds32/nds32.md (negsi2): Refine pattern.
20400 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
20401 Chung-Ju Wu <jasonwucj@gmail.com>
20403 * config/nds32/iterators.md (shift_rotate): New code iterator.
20404 (shift): New code attribute.
20405 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
20406 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
20407 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
20408 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
20409 bit-wise operations.
20410 (andsi3, *andsi3): Ditto.
20411 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
20412 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
20413 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
20414 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
20415 nds32_ior_operand, nds32_xor_operand): New predicates.
20417 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
20419 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
20420 (addsi3, subsi3): ... this.
20422 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
20424 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
20426 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
20428 * config/nds32/nds32.md: Adjust indention.
20430 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
20432 * config/nds32/nds32.md (feature): New attribute.
20434 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
20436 * config/nds32/nds32.md (subtype): New attribute.
20438 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
20441 * config/arm/arm-builtins.c (arm_expand_builtin): Change
20442 expansion to perform a bitwise AND of the argument followed by a
20443 boolean negation of the result.
20445 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
20447 PR rtl-optimization/84878
20448 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
20449 the basic block. Assert the use reference is not artificial and that
20450 it has an associated insn.
20452 2018-04-04 Michael Matz <matz@suse.de>
20454 * builtins.c (compute_objsize): Pass correct operand
20455 to array_at_struct_end_p.
20457 2018-04-04 Richard Biener <rguenther@suse.de>
20460 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
20461 from contexts for DINFO_LEVEL_TERSE and below.
20463 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
20465 * config/nds32/nds32-doubleword.md (move_<mode>): Require
20466 resiter_operand condition.
20467 * config/nds32/nds32.md (*move<mode>): Ditto.
20469 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
20470 Monk Chiang <sh.chiang04@gmail.com>
20472 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
20474 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
20476 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
20478 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
20479 Kito Cheng <kito.cheng@gmail.com>
20481 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
20482 nds32_cond_code_str, output_cond_branch,
20483 output_cond_branch_compare_zero, nds32_expand_cbranch,
20484 nds32_expand_cstore, nds32_expand_movcc,
20485 nds32_output_cbranchsi4_equality_zero,
20486 nds32_output_cbranchsi4_equality_reg,
20487 nds32_output_cbranchsi4_equality_reg_or_const_int,
20488 nds32_output_cbranchsi4_greater_less_zero: New functions.
20489 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
20490 nds32_expand_cstore, nds32_expand_movcc,
20491 nds32_output_cbranchsi4_equality_zero,
20492 nds32_output_cbranchsi4_equality_reg,
20493 nds32_output_cbranchsi4_equality_reg_or_const_int,
20494 nds32_output_cbranchsi4_greater_less_zero): Declare.
20495 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
20496 nds32_rimm11s_operand): New predicates.
20497 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
20498 * config/nds32/nds32.md: Rewrite all the branch and conditional move
20501 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
20503 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
20504 * config/nds32/nds32.md: Ditto.
20505 * config/nds32/pipelines.md: Ditto.
20507 2018-04-04 Richard Biener <rguenther@suse.de>
20509 PR tree-optimization/85168
20510 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
20511 propagating abnormals.
20513 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
20515 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
20517 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
20518 Kito Cheng <kito.cheng@gmail.com>
20520 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
20521 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
20522 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
20523 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
20524 * config/nds32/nds32.md (sibcall_internal): New.
20525 (sibcall_register): Remove.
20526 (sibcall_immediate): Remove.
20527 (sibcall_value_internal): New.
20528 (sibcall_value_register): Remove.
20529 (sibcall_value_immediate): Remove.
20530 * config/nds32/predicates.md (nds32_general_register_operand): New.
20531 (nds32_call_address_operand): New.
20533 2018-04-03 Jakub Jelinek <jakub@redhat.com>
20535 PR rtl-optimization/85167
20536 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
20537 bb_defs if *split_p, instead preinitialize it to NULL.
20539 PR tree-optimization/85156
20540 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
20541 evaluating the argument multiple times.
20543 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
20545 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
20547 (_mm_cvtpd_ps): Likewise.
20548 (_mm_cvttpd_epi32): Likewise.
20549 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
20550 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
20551 vector, pixel, and bool following altivec.h include.
20553 2018-04-03 Martin Sebor <msebor@redhat.com>
20555 * doc/extend.texi (Common Function Attributes): Clarify.
20556 (const attribute): Likewise.
20557 (pure attribute): Likewise.
20559 2018-04-03 Jakub Jelinek <jakub@redhat.com>
20562 * config/i386/i386.c (ix86_expand_vector_set): Use
20563 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
20565 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
20567 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
20568 instructions when changing rounding bits to preserve precision bits
20569 in the x87 control word.
20571 2018-04-03 Martin Liska <mliska@suse.cz>
20573 PR tree-optimization/82491
20574 * rtl.h (strip_offset_and_add): Replace += suboffset with
20575 poly_uint64 () + suboffset.
20577 2018-03-29 Martin Liska <mliska@suse.cz>
20578 Martin Jambor <mjambor@suse.cz>
20581 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
20582 param_type is not an integral or pointer type.
20584 2018-04-03 Richard Biener <rguenther@suse.de>
20586 * sese.h (recompute_all_dominators): Remove.
20588 2018-04-02 Martin Sebor <msebor@redhat.com>
20590 * doc/invoke.texi (-Wrestrict): Fix typos.
20592 2018-04-02 Jim Wilson <jimw@sifive.com>
20594 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
20595 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
20596 (<optab>di3, <optab>si3_extend): Likewise.
20597 (<optab>si3_mask, <optab>si3_mask_1): New.
20598 (<optab>di3_mask, <optab>di3_mask_1): New.
20599 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
20600 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
20601 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
20603 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
20605 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
20608 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
20610 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
20611 (nds32_canonicalize_comparison): New function.
20613 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
20614 Kito Cheng <kito.cheng@gmail.com>
20615 Kuan-Lin Chen <kuanlinchentw@gmail.com>
20617 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
20618 * config/nds32/constants.md (unspec_volatile_element): Add
20619 UNSPEC_VOLATILE_RELAX_GROUP.
20620 * config/nds32/nds32-relax-opt.c: New file.
20621 * config/nds32/nds32-predicates.c
20622 (nds32_symbol_load_store_p): New function.
20623 * config/nds32/nds32-protos.h
20624 (nds32_symbol_load_store_p): Declare function.
20625 (make_pass_nds32_relax_opt): Declare new rtl pass function.
20626 * config/nds32/nds32.c
20627 (nds32_register_pass): New function to register pass.
20628 (nds32_register_passes): New function to register passes.
20629 * config/nds32/nds32.md (relax_group): New pattern.
20630 * config/nds32/nds32.opt (mrelax-hint): New option.
20631 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
20633 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
20635 * config/nds32/t-nds32: Modify files dependency.
20637 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
20639 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
20640 (PROFILE_HOOK): Define its implementation.
20642 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
20644 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
20645 type and 32-bit size.
20647 2018-04-01 Jakub Jelinek <jakub@redhat.com>
20649 PR middle-end/85090
20650 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
20651 (V_128_256): New mode iterator.
20652 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
20653 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
20654 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
20656 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
20657 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
20659 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
20662 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
20663 NaN inputs correctly.
20665 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
20668 * config/rs6000/vsx.md (??r): New mode attribute.
20669 (*vsx_mov<mode>_64bit): Use it.
20670 (*vsx_mov<mode>_32bit): Likewise.
20672 2018-03-30 Martin Sebor <msebor@redhat.com>
20674 PR tree-optimization/84818
20675 * builtins.c (check_access): Use warning_n.
20677 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
20680 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
20682 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
20685 2018-03-30 Julia Koval <julia.koval@intel.com>
20688 * x86-tune.def (movx, partial_reg_dependency): Enable for
20691 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
20693 PR inline-asm/84985
20694 * lra-constraints.c (process_alt_operands): Move setting
20695 this_alternative_matches below.
20697 2018-03-29 Martin Liska <mliska@suse.cz>
20700 * doc/invoke.texi: Document how LTO works with debug info.
20701 Describe auto-load support of binutils. Mention 'x86-64'
20702 as valid option value of -march option.
20704 2018-03-29 Jakub Jelinek <jakub@redhat.com>
20706 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
20709 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
20710 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
20711 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
20714 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
20717 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
20718 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
20719 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
20720 for RS6000_BTM_POWERPC64.
20721 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
20722 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
20723 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
20726 (DIVDEU): Likewise.
20728 2018-03-28 Carl Love <cel@us.ibm.com>
20731 2017-09-27 Carl Love <cel@us.ibm.com>
20733 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
20734 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
20735 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
20738 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20740 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
20741 instead of __vector bool.
20742 (_mm_max_pu8): Likewise.
20743 (_mm_min_pi16): Likewise.
20745 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
20748 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
20749 (DIVWEUO): Likewise.
20750 (DIVDEO): Likewise.
20751 (DIVDEUO): Likewise.
20752 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
20753 DIVWEUO and DIVDEUO.
20754 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
20755 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
20756 (div_extend): Likewise.
20757 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
20759 (__builtin_divweuo): Likewise.
20760 (__builtin_divdeo): Likewise.
20761 (__builtin_divdeuo): Likewise.
20763 2018-03-28 Jakub Jelinek <jakub@redhat.com>
20766 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
20767 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
20769 PR tree-optimization/82004
20770 * gimple-match-head.c (optimize_pow_to_exp): New function.
20771 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
20772 Don't fold to exp if optimize_pow_to_exp is false.
20774 2018-03-28 Martin Liska <mliska@suse.cz>
20777 * calls.c (initialize_argument_information): Fix trailing space.
20778 * common.opt: Fix typo and provide better explanation for
20779 -fsanitize-coverage option.
20780 * config/i386/i386.opt: Fix typo.
20782 2018-03-28 Jakub Jelinek <jakub@redhat.com>
20783 Martin Liska <mliska@suse.cz>
20786 * gimplify.c (asan_poison_variable): Don't do the check for
20787 gimplify_omp_ctxp here.
20788 (gimplify_decl_expr): Do it here.
20789 (gimplify_target_expr): Likewise.
20791 2018-03-28 Martin Liska <mliska@suse.cz>
20794 * config/i386/i386.c (ix86_function_arg_advance): Do not call
20795 chkp_type_bounds_count if MPX is not enabled.
20797 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
20799 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
20801 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
20804 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
20805 function to create the function decl for complex long double
20806 multiply and divide for -mabi=ieeelongdouble.
20807 (init_float128_ieee): Call it.
20809 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
20812 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
20813 -fcf-protection=branch -mibt.
20814 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
20816 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
20819 * config/arm/arm.c (arm_valid_symbolic_address): Handle
20820 arm_word_relocations.
20822 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
20825 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
20826 extern array declarations.
20828 2018-03-27 Richard Biener <rguenther@suse.de>
20830 PR middle-end/84067
20831 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
20832 explicit single_use checks.
20834 2018-03-27 Richard Biener <rguenther@suse.de>
20836 PR tree-optimization/85082
20837 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
20840 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
20842 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
20843 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
20844 Turn on fasynchronous-unwind-tables and funwind-tables.
20846 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
20849 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
20850 (*bmi_blsr_<mode>_ccz): Ditto.
20852 2018-03-26 Tom de Vries <tom@codesourcery.com>
20854 PR tree-optimization/85063
20855 * omp-general.c (offloading_function_p): New function. Factor out
20857 * omp-offload.c (pass_omp_target_link::gate): ... here.
20858 * omp-general.h (offloading_function_p): Declare.
20859 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
20860 with attribute omp declare target for offloading functions.
20862 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
20864 PR tree-optimization/84005
20865 * tree-data-ref.h (get_base_for_alignment): Declare.
20866 * tree-data-ref.c (get_base_for_alignment_1): New function.
20867 (get_base_for_alignment): Likewise.
20868 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
20869 get_base_for_alignment to find a suitable base object, instead
20870 of always using drb->base_address.
20872 2018-03-23 Jakub Jelinek <jakub@redhat.com>
20874 PR inline-asm/85022
20875 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
20876 known size by default.
20878 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
20880 PR inline-asm/85030
20881 * lra-constraints.c (process_alt_operands): Don't match BLKmode
20882 and non BLKmode operands.
20884 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20887 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
20888 Clean up attributes.
20890 2018-03-23 Richard Biener <rguenther@suse.de>
20893 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
20894 we are going to emit early debug for LTO.
20896 2018-03-23 Jakub Jelinek <jakub@redhat.com>
20898 PR inline-asm/85034
20899 * function.c (match_asm_constraints_1): Don't optimize if input
20900 doesn't satisfy general_operand predicate for output's mode.
20902 PR inline-asm/85022
20903 * alias.c (write_dependence_p): Don't require for x_canonicalized
20904 non-VOIDmode if x has VOIDmode.
20907 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
20908 just don't try to optimize it rather than assert it never happens.
20910 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
20912 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
20913 macro expansions for definition of ST_INTERNAL_<mode> and
20914 LD_INTERNAL_<mode> builtins.
20915 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
20917 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
20919 (altivec_expand_st_builtin): Likewise.
20920 (altivec_expand_builtin): Remove calls to deleted functions.
20921 (rs6000_address_for_altivec): Delete this function.
20922 * config/rs6000/vector.md: Remove expands for
20923 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
20925 2018-03-22 Sudakshina Das <sudi.das@arm.com>
20928 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
20929 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
20930 re-computing once computed.
20931 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
20932 (arm_init_machine_status): Initialize
20933 machine->static_chain_stack_bytes.
20935 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
20938 * doc/extend.texi: Add four new prototypes for vec_ld.
20939 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
20940 definitions for more logical presentation.
20941 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
20942 entries for V1TI variants of __builtin_altivec_ld builtin.
20943 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
20944 handling of V1TI variant of LVX icode pattern.
20945 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
20946 (rs6000_gimple_fold_builtin): Likewise.
20947 (altivec_init_builtins): Add code to define
20948 __builtin_altivec_lvx_v1ti function.
20950 2018-03-22 Jakub Jelinek <jakub@redhat.com>
20952 PR inline-asm/84941
20953 * function.c (match_asm_constraints_1): Don't do the optimization
20954 if input isn't a REG, SUBREG, MEM or constant.
20956 2018-03-22 Tom de Vries <tom@codesourcery.com>
20958 PR tree-optimization/84956
20959 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
20960 bb_has_abnormal_pred.
20962 2018-03-22 Jakub Jelinek <jakub@redhat.com>
20965 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
20966 DECL_INITIAL (decl) to decl at the end.
20967 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
20968 adjust the comment.
20970 2018-03-21 Joseph Myers <joseph@codesourcery.com>
20972 * doc/extend.texi (__builtin_tgmath): Document when complex
20973 integer types are treated as _Complex _Float64.
20975 2018-03-21 Tom de Vries <tom@codesourcery.com>
20977 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
20979 2018-03-21 Jakub Jelinek <jakub@redhat.com>
20981 PR tree-optimization/84960
20982 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
20983 if it is ENTRY block, move them into single succ of ENTRY in that case.
20985 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
20987 PR tree-optimization/84811
20988 * poly-int.h (poly_span_traits): Remove the T3 parameter and
20989 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
20990 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
20991 (known_subrange_p): Update accordingly. Cast each value involved
20992 in the size comparison, rather than casting the result of the
20995 2018-03-21 Jakub Jelinek <jakub@redhat.com>
20997 PR tree-optimization/84982
20998 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
20999 by flipping the least significant bit rather than all bits from
21000 bitpos to bitpos + bitsize - 1.
21002 2018-03-21 Nathan Sidwell <nathan@acm.org>
21004 * doc/extend.texi (Deprecated Features): Remove mention of
21005 long-deleted deprecations.
21007 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
21010 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
21011 * configure: Regenerate.
21013 2018-03-21 Tom de Vries <tom@codesourcery.com>
21015 PR tree-optimization/83126
21016 * tree-parloops.c (num_phis): New function.
21017 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
21019 2018-03-21 Nathan Sidwell <nathan@acm.org>
21021 * doc/extend.texi (Deprecated Features): Update deprecated flags,
21022 mention anon-struct/union members and trailing attributes.
21024 2018-03-21 Bin Cheng <bin.cheng@arm.com>
21026 PR tree-optimization/84969
21027 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
21028 builtin memset partitions if they set different rhs values.
21030 2018-03-21 Jakub Jelinek <jakub@redhat.com>
21032 PR rtl-optimization/84989
21033 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
21034 VEC_DUPLICATE with scalar result mode.
21036 2018-03-21 Martin Liska <mliska@suse.cz>
21039 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
21040 not intended return statement.
21042 2018-03-21 Martin Liska <mliska@suse.cz>
21045 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
21046 (chkp_find_bound_slots_1): Limit number of iterations.
21048 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
21051 * Minor grammar fixes for x86 options.
21053 2018-03-20 Jakub Jelinek <jakub@redhat.com>
21056 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
21057 holding REG_CFA_RESTORE notes, instead turn them into a USE.
21059 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
21062 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
21063 (altivec_lvx_<mode>_1op): Likewise.
21064 (altivec_stvx_<mode>_2op): Likewise.
21065 (altivec_stvx_<mode>_1op): Likewise.
21066 (altivec_lvx_<VM2:mode>): New define_expand.
21067 (altivec_stvx_<VM2:mode>): Likewise.
21068 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
21069 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
21070 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
21071 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
21072 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
21073 (rs6000_gen_lvx): Likewise.
21074 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
21075 (altivec_expand_stv_builtin): Likewise.
21076 (altivec_expand_builtin): Likewise.
21077 * config/rs6000/vector.md: Likewise.
21079 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21082 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
21085 2018-03-20 Richard Biener <rguenther@suse.de>
21088 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
21089 sign-conversions as zero, fall back to standard scalar_stmt
21092 2018-03-20 Martin Liska <mliska@suse.cz>
21095 * predict.c (rebuild_frequencies): Handle case when we have
21096 PROFILE_ABSENT, but flag_guess_branch_prob is false.
21098 2018-03-20 Jakub Jelinek <jakub@redhat.com>
21101 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
21102 flag_section_anchors.
21103 * varasm.c (use_blocks_for_decl_p): Remove hack for
21104 dw2_force_const_mem.
21107 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
21109 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
21110 be created, use lowpart_subreg of operands[0] rather than operands[0]
21112 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
21113 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
21114 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
21115 and n constraint instead of aarch64_shift_imm_di and Usd.
21116 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
21117 (*aarch64_<optab>_reg_minus<mode>3): ... this.
21119 2018-03-20 Sudakshina Das <sudi.das@arm.com>
21122 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
21123 to favor GPR over NEON registers.
21124 (<shift>di3_neon): Likewise.
21126 2018-03-20 Tom de Vries <tom@codesourcery.com>
21129 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
21130 (nvptx_process_pars): Emit bar.sync asap and alap.
21132 2018-03-20 Tom de Vries <tom@codesourcery.com>
21135 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
21136 seen_label if seen_label is already set.
21138 2018-03-20 Jakub Jelinek <jakub@redhat.com>
21141 * config/i386/i386.c (fold_builtin_cpu): For features above 31
21142 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
21143 Use 1U instead of 1. Formatting fixes.
21146 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
21147 instead of TREE_TYPE (s1) for the return value.
21149 2018-03-19 Jakub Jelinek <jakub@redhat.com>
21151 PR tree-optimization/84946
21152 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
21153 bitsize + bitsize in poly_uint64 rather than poly_int64.
21156 * dwarf2asm.c: Include fold-const.c.
21157 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
21158 of decl rather than decl itself.
21160 PR rtl-optimization/84643
21161 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
21163 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
21166 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
21167 calling assemble_variable.
21169 2018-03-19 Sudakshina Das <sudi.das@arm.com>
21172 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
21173 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
21175 2018-03-19 Jim Wilson <jimw@sifive.com>
21178 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
21179 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
21180 (riscv_first_stack_step): Likewise.
21181 (riscv_option_override): Use STACK_BOUNDARY instead of
21182 MIN_STACK_BOUNDARY.
21183 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
21184 MIN_STACK_BOUNDARY.
21185 (BIGGEST_ALIGNMENT): Set to 128.
21186 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
21187 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
21190 2018-03-19 Richard Biener <rguenther@suse.de>
21192 PR tree-optimization/84933
21193 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
21194 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
21196 2018-03-19 Richard Biener <rguenther@suse.de>
21198 PR tree-optimization/84859
21199 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
21200 (cond_if_else_store_replacement): Perform sinking operation on
21201 single-store BBs regardless of MAX_STORES_TO_SINK setting.
21202 Generalize what a BB with a single eligible store is.
21204 2018-03-19 Richard Biener <rguenther@suse.de>
21206 PR tree-optimization/84929
21207 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
21208 chrec_is_positive against non-chrec arg.
21210 2018-03-19 Tamar Christina <tamar.christina@arm.com>
21213 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
21215 2018-03-18 Martin Liska <mliska@suse.cz>
21217 PR rtl-optimization/84635
21218 * regrename.c (build_def_use): Use matches_mode only when
21221 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
21223 PR tree-optimization/84913
21224 * tree-vect-loop.c (vectorizable_reduction): Don't try to
21225 vectorize chains of COND_EXPRs.
21227 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
21229 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
21231 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
21233 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
21235 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
21237 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
21239 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
21240 Kito Cheng <kito.cheng@gmail.com>
21242 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
21243 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
21244 (nds32_adjust_reg_alloc_order): New function.
21245 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
21247 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
21249 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
21250 nds32_print_operand, nds32_print_operand_address): Use
21251 HOST_WIDE_INT_PRINT_DEC instead.
21253 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
21255 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
21257 2018-03-17 Jakub Jelinek <jakub@redhat.com>
21260 * config/i386/i386.c (initial_ix86_tune_features,
21261 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
21262 unsigned long long.
21263 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
21264 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
21265 rather than 1u << ix86_tune. Formatting fix.
21266 (ix86_option_override_internal): Change ix86_arch_mask from
21267 unsigned int to unsigned HOST_WIDE_INT, initialize to
21268 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
21269 (ix86_function_specific_restore): Likewise.
21271 2018-03-16 Jakub Jelinek <jakub@redhat.com>
21274 * postreload.c (reload_combine_recognize_pattern): Perform
21275 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
21276 truncate_int_for_mode the result for the destination's mode.
21279 * hsa-gen.c (mem_type_for_type): Fix comment typo.
21280 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
21282 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
21285 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
21288 * lra-assigns.c (lra_split_hard_reg_for): Don't use
21289 regno_allocno_class_array and sorted_pseudos.
21290 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
21291 insns where regno is used.
21293 2018-03-16 Martin Liska <mliska@suse.cz>
21296 * multiple_target.c (create_dispatcher_calls): Redirect
21297 reference in the symbol table.
21299 2018-03-16 Martin Liska <mliska@suse.cz>
21302 * multiple_target.c (create_dispatcher_calls): Redirect also
21305 2018-03-16 Jakub Jelinek <jakub@redhat.com>
21309 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
21310 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
21311 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
21313 2018-03-16 Julia Koval <julia.koval@intel.com>
21315 * doc/invoke.texi (Skylake Server): Add CLWB.
21316 Cannonlake): Remove CLWB.
21318 2018-03-16 Jakub Jelinek <jakub@redhat.com>
21320 PR tree-optimization/84841
21321 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
21323 (FLOAT_ONE_CONST_TYPE): Define.
21324 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
21325 (sort_by_operand_rank): Put entries with higher constant_type last
21326 rather than first to match comments.
21328 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
21330 * config/nios2/nios2.md (movsi_internal): Fix thinko in
21333 2018-03-15 Jakub Jelinek <jakub@redhat.com>
21336 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
21337 check and use address of target always.
21339 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
21342 * config/i386/i386.c (indirect_thunk_needed): Update comments.
21343 (indirect_thunk_bnd_needed): Likewise.
21344 (indirect_thunks_used): Likewise.
21345 (indirect_thunks_bnd_used): Likewise.
21346 (indirect_return_needed): New.
21347 (indirect_return_bnd_needed): Likewise.
21348 (output_indirect_thunk_function): Add a bool argument for
21350 (output_indirect_thunk_function): Don't generate alias for
21351 function return thunk.
21352 (ix86_code_end): Call output_indirect_thunk_function to generate
21353 function return thunks.
21354 (ix86_output_function_return): Set indirect_return_bnd_needed
21355 and indirect_return_needed instead of indirect_thunk_bnd_needed
21356 and indirect_thunk_needed.
21358 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
21360 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
21361 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
21362 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
21364 2018-03-15 David Malcolm <dmalcolm@redhat.com>
21365 Paul Hua <paul.hua.gm@gmail.com>
21368 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
21370 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
21372 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
21373 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
21374 resp. SFmode cases.
21376 2018-03-15 Tamar Christina <tamar.christina@arm.com>
21379 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
21380 instead of GET_MODE_SIZE when comparing Units.
21382 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
21385 * varasm.c (hash_section): Return an unchangeble hash value
21386 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
21387 Return !aarch64_can_use_per_function_literal_pools_p ().
21389 2018-03-15 Jakub Jelinek <jakub@redhat.com>
21392 * optabs.c (emit_conditional_move): Pass address of cmode's copy
21393 rather than address of cmode as last argument to prepare_cmp_insn.
21395 2018-03-15 Julia Koval <julia.koval@intel.com>
21397 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
21398 F_AVX512VNNI, F_AVX512BITALG): New.
21400 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
21403 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
21404 insn for floating-point loads and stores.
21406 2018-03-14 Carl Love <cel@us.ibm.com>
21408 * config/rs6000/rs6000-c.c: Add macro definitions for
21409 ALTIVEC_BUILTIN_VEC_PERMXOR.
21410 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
21411 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
21412 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
21413 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
21415 * config/doc/extend.texi: Add prototypes for vec_permxor.
21417 2018-03-14 David Malcolm <dmalcolm@redhat.com>
21420 * diagnostic-show-locus.c (class layout_point): Convert m_line
21421 from int to linenum_type.
21422 (line_span::comparator): Use linenum "compare" function when
21423 comparing line numbers.
21424 (test_line_span): New function.
21425 (layout_range::contains_point): Convert param "row" from int to
21427 (layout_range::intersects_line_p): Likewise.
21428 (layout::will_show_line_p): Likewise.
21429 (layout::print_source_line): Likewise.
21430 (layout::should_print_annotation_line_p): Likewise.
21431 (layout::print_annotation_line): Likewise.
21432 (layout::print_leading_fixits): Likewise.
21433 (layout::annotation_line_showed_range_p): Likewise.
21434 (struct line_corrections): Likewise for field m_row.
21435 (line_corrections::line_corrections): Likewise for param "row".
21436 (layout::print_trailing_fixits): Likewise.
21437 (layout::get_state_at_point): Likewise.
21438 (layout::get_x_bound_for_row): Likewise.
21439 (layout::print_line): Likewise.
21440 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
21441 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
21442 * input.c (selftest::test_linenum_comparisons): New function.
21443 (selftest::input_c_tests): Call it.
21444 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
21445 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
21446 * selftest.h (ASSERT_GT): New macro.
21447 (ASSERT_GT_AT): New macro.
21448 (ASSERT_LT): New macro.
21449 (ASSERT_LT_AT): New macro.
21451 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
21453 PR rtl-optimization/84780
21454 * combine.c (distribute_links): Don't make a link based on pc_rtx.
21456 2018-03-14 Martin Liska <mliska@suse.cz>
21458 * tree.c (record_node_allocation_statistics): Use
21459 get_stats_node_kind.
21460 (get_stats_node_kind): New function extracted from
21461 record_node_allocation_statistics.
21462 (free_node): Use get_stats_node_kind.
21464 2018-03-14 Richard Biener <rguenther@suse.de>
21466 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
21467 that the value-set of ANTIC_IN doesn't grow.
21470 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
21472 (BB_VISITED_WITH_VISITED_SUCCS): New define.
21473 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
21475 2018-03-14 Julia Koval <julia.koval@intel.com>
21477 * config.gcc (icelake-client, icelake-server): New.
21479 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
21480 (initial_ix86_arch_features): Ditto.
21481 (PTA_SKYLAKE): Add SGX.
21482 (PTA_ICELAKE): Remove.
21483 (PTA_ICELAKE_CLIENT): New.
21484 (PTA_ICELAKE_SERVER): New.
21485 (ix86_option_override_internal): Split up icelake on icelake client and
21487 (get_builtin_code_for_version): Ditto.
21488 (fold_builtin_cpu): Ditto.
21489 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
21490 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
21491 * config/i386/i386.h (processor_type): Ditto.
21492 * doc/invoke.texi: Ditto.
21494 2018-03-14 Jakub Jelinek <jakub@redhat.com>
21497 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
21498 INTEGER_CST offset, add it together with bitpos / 8 and
21499 sign extend based on POINTER_SIZE.
21503 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
21506 * config/i386/constraints.md (Yc): New register constraint.
21507 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
21508 Use Yc constraint for alternative 2 of operand 0. Remove
21509 preferred_for_speed attribute.
21511 2018-03-14 Richard Biener <rguenther@suse.de>
21513 PR tree-optimization/84830
21514 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
21515 with the old one to avoid oscillations.
21517 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
21520 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
21522 (assign_by_spills): Return a flag of reload assignment failure.
21523 Do not process the reload assignment failures. Do not spill other
21524 reload pseudos if they has the same reg class. Update n if
21526 (lra_assign): Add a return arg. Set up from the result of
21527 assign_by_spills call.
21528 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
21529 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
21530 usage_insns if it is not NULL.
21531 (spill_hard_reg_in_range): New function.
21532 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
21533 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
21534 function prototypes.
21535 (lra_assign): Change prototype.
21536 * lra.c (lra): Add code to deal with fails by splitting hard reg
21539 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
21541 * config/riscv/riscv.opt (mrelax): New option.
21542 * config/riscv/riscv.c (riscv_file_start): Emit ".option
21543 "norelax" when riscv_mrelax is disabled.
21544 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
21546 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
21549 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
21550 reassociation for int modes.
21552 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
21554 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
21555 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
21557 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
21558 * config/aarch64/aarch64-sve.md
21559 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
21560 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
21561 (*extend<mode><Vwide>2): Rename to...
21562 (aarch64_sve_extend<mode><Vwide>2): ...this.
21563 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
21564 renaming the old pattern to...
21565 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
21567 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
21568 define_expand, renaming the old pattern to...
21569 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
21570 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
21571 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
21572 account when deciding which SVE instruction the optab should use.
21573 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
21575 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
21577 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
21578 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
21579 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
21580 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
21581 (tlsdesc_small_<mode>): Turn a define_expand and use
21582 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
21583 (tlsdesc_small_advsimd_<mode>): ...this.
21584 (tlsdesc_small_sve_<mode>): New pattern.
21586 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
21588 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
21589 (UNSPEC_UMUL_HIGHPART): New constants.
21590 (MUL_HIGHPART): New int iteraor.
21591 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
21592 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
21594 (*<su>mul<mode>3_highpart): New define_insn.
21596 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
21599 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
21602 2018-03-13 Martin Liska <mliska@suse.cz>
21605 * (sem_item_optimizer::sem_item_optimizer): Initialize new
21607 (sem_item_optimizer::~sem_item_optimizer): Release it.
21608 (sem_item_optimizer::merge_classes): Register variable aliases.
21609 (sem_item_optimizer::fixup_pt_set): New function.
21610 (sem_item_optimizer::fixup_points_to_sets): Likewise.
21611 * ipa-icf.h: Declare new variables and functions.
21613 2018-03-13 Jakub Jelinek <jakub@redhat.com>
21615 PR middle-end/84834
21616 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
21617 integer_pow2p@2 and test integer_pow2p in condition.
21618 (A < 0 ? C : 0): Similarly for @1.
21620 PR middle-end/84831
21621 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
21622 characters starting at p contain '\0' character, don't look beyond
21626 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
21627 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
21630 * reg-stack.c (change_stack): Change update_end var from int to
21631 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
21632 also call set_block_for_insn on the newly added insns and rescan.
21635 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
21636 on the last operand.
21639 * tree.c (stabilize_reference_1): Return save_expr (e) for
21640 STATEMENT_LIST even if it doesn't have side-effects.
21642 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
21644 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
21646 2018-03-12 Renlin Li <renlin.li@arm.com>
21648 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
21649 aarch64_output_scalar_simd_mov_immediate.
21651 2018-03-12 Martin Sebor <msebor@redhat.com>
21653 PR tree-optimization/83456
21654 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
21655 for perfectly overlapping calls to memcpy.
21656 (gimple_fold_builtin_memory_chk): Same.
21657 (gimple_fold_builtin_strcpy): Handle no-warning.
21658 (gimple_fold_builtin_stxcpy_chk): Same.
21659 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
21661 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
21663 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
21664 parameter. Use it for SFmode.
21665 (rs6000_function_arg_advance_1): Adjust.
21666 (rs6000_function_arg): Adjust.
21667 (rs6000_gimplify_va_arg): Pass false for that new parameter.
21669 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
21671 PR rtl-optimization/84169
21672 PR rtl-optimization/84780
21673 * combine.c (can_combine_p): Check for a 2-insn combination whether
21674 the destination register is used between the two insns, too.
21676 2018-03-12 Richard Biener <rguenther@suse.de>
21678 PR tree-optimization/84803
21679 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
21680 for refs DR analysis didn't process.
21682 2018-03-12 Richard Biener <rguenther@suse.de>
21684 PR tree-optimization/84777
21685 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
21686 force-vectorize loops ignore whether we are optimizing for size.
21688 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
21690 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
21691 (TARGET_MD_ASM_ADJUST): Define.
21693 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
21694 Kito Cheng <kito.cheng@gmail.com>
21695 Chung-Ju Wu <jasonwucj@gmail.com>
21697 * config/nds32/nds32.c (nds32_compute_stack_frame,
21698 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
21699 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
21700 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
21701 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
21702 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
21703 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
21704 * config/nds32/nds32.md (prologue, epilogue): Use macro
21705 NDS32_V3PUSH_AVAILABLE_P to do checking.
21707 2018-03-11 Jakub Jelinek <jakub@redhat.com>
21710 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
21711 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
21712 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
21713 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
21714 addition of most attributes on !orig_type_die or the attribute not
21715 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
21717 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
21718 Chung-Ju Wu <jasonwucj@gmail.com>
21720 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
21721 __NDS32_VH__ macro.
21722 * config/nds32/nds32.opt (mvh): New option.
21724 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
21725 Chung-Ju Wu <jasonwucj@gmail.com>
21727 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
21729 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
21730 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
21733 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
21734 Chung-Ju Wu <jasonwucj@gmail.com>
21736 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
21738 * config/nds32/nds32-multiple.md (strlensi): New pattern.
21739 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
21741 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
21742 Kito Cheng <kito.cheng@gmail.com>
21743 Chung-Ju Wu <jasonwucj@gmail.com>
21745 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
21746 UNSPEC_FFMISM and UNSPEC_FLMISM.
21747 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
21748 for ffb, ffmism and flmism.
21749 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
21750 (unspec_ffmism): Ditto.
21751 (unspec_flmism): Ditto.
21752 (nds32_expand_builtin_impl): Check if string extension is available.
21753 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
21754 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
21756 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
21759 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
21762 * lra-assigns.c (assign_by_spills): Return a flag of reload
21763 assignment failure. Do not process the reload assignment
21764 failures. Do not spill other reload pseudos if they has the same
21766 (lra_assign): Add a return arg. Set up from the result of
21767 assign_by_spills call.
21768 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
21769 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
21770 usage_insns if it is not NULL.
21771 (spill_hard_reg_in_range): New function.
21772 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
21773 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
21774 function prototypes.
21775 (lra_assign): Change prototype.
21776 * lra.c (lra): Add code to deal with fails by splitting hard reg
21779 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
21782 * config/i386/i386.opt: Replace Enforcment with Enforcement.
21784 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
21787 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
21788 (dw_val_node): Add val_symbolic_view.
21789 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
21790 (symview_upper_bound): New.
21791 (new_line_info_table): Initialize symviews_since_reset.
21792 (dwarf2out_source_line): Count symviews_since_reset and set
21793 symview_upper_bound.
21794 (dw_val_equal_p): Handle symview.
21795 (add_AT_symview): New.
21796 (print_dw_val): Handle symview.
21797 (attr_checksum, attr_checksum_ordered): Likewise.
21798 (same_dw_val_p, size_of_die): Likewise.
21799 (value_format, output_die): Likewise.
21800 (add_high_low_attributes): Use add_AT_symview for entry_view.
21801 (dwarf2out_finish): Reset symview_upper_bound, clear
21804 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
21807 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
21808 Add strict argument and use it.
21809 (rs6000_split_multireg_move): Update for new strict argument.
21810 (mem_operand_gpr): Disallow all non-offsettable addresses.
21811 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
21813 2018-03-09 Jakub Jelinek <jakub@redhat.com>
21816 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
21817 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
21818 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
21821 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
21822 decl, use remap_type if we want to use the type.
21824 2018-03-09 Martin Sebor <msebor@redhat.com>
21826 PR tree-optimization/84526
21827 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
21829 (builtin_access::generic_overlap): Be prepared to handle non-array
21832 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
21834 PR rtl-optimization/84682
21835 * lra-constraints.c (process_address_1): Check is_address flag
21836 for address constraints.
21837 (process_alt_operands): Likewise.
21838 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
21839 preprocess_constraints.
21840 * recog.h (preprocess_constraints): Add oploc parameter.
21842 * recog.c (preprocess_constraints): Test address_operand for
21843 CT_ADDRESS constraints.
21845 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
21848 * lra-assigns.c (assign_by_spills): Return a flag of reload
21849 assignment failure. Do not process the reload assignment
21850 failures. Do not spill other reload pseudos if they has the same
21852 (lra_assign): Add a return arg. Set up from the result of
21853 assign_by_spills call.
21854 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
21855 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
21856 usage_insns if it is not NULL.
21857 (spill_hard_reg_in_range): New function.
21858 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
21859 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
21860 function prototypes.
21861 (lra_assign): Change prototype.
21862 * lra.c (lra): Add code to deal with fails by splitting hard reg
21865 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21868 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
21869 Accept complain bool parameter. Only emit errors if it is true.
21870 (arm_parse_cpu_option_name): Likewise.
21871 (arm_target_thumb_only): Adjust callers of the above.
21872 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
21873 prototype to take a default true bool parameter.
21874 (arm_parse_arch_option_name): Likewise.
21876 2018-03-09 David Malcolm <dmalcolm@redhat.com>
21877 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
21881 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
21882 * configure: Regenerate.
21883 * configure.ac ("linker --version-script option"): New.
21884 ("linker soname option"): New.
21886 2018-03-09 Richard Biener <rguenther@suse.de>
21888 PR tree-optimization/84775
21889 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
21890 immediate uses of predicate stmts and mark them modified.
21893 PR tree-optimization/84178
21894 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
21896 (version_loop_for_if_conversion): Delay update_ssa call.
21897 (tree_if_conversion): Delay update_ssa until after predicate
21900 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
21903 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
21904 when the function accesses prior frames.
21906 2018-03-08 Jakub Jelinek <jakub@redhat.com>
21909 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
21910 gen_llsym, otherwise call maybe_gen_llsym.
21912 PR inline-asm/84742
21913 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
21914 has ',' character inside of it.
21916 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21919 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
21920 as clobbering CC_REGNUM.
21922 2018-03-08 Richard Biener <rguenther@suse.de>
21924 PR middle-end/84552
21925 * tree-scalar-evolution.c: Include tree-into-ssa.h.
21926 (follow_copies_to_constant): Do not follow SSA names registered
21929 2018-03-08 Richard Biener <rguenther@suse.de>
21931 PR tree-optimization/84178
21932 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
21934 (version_loop_for_if_conversion): Delay update_ssa call.
21935 (tree_if_conversion): Delay update_ssa until after predicate
21938 2018-03-08 David Malcolm <dmalcolm@redhat.com>
21940 PR tree-optimization/84178
21941 * tree-if-conv.c (release_bb_predicate): Remove the
21942 the assertion that the stmts have NULL use_ops.
21943 Discard the statements, asserting that they haven't
21944 yet been added to a BB.
21946 2018-03-08 Richard Biener <rguenther@suse.de>
21948 PR tree-optimization/84746
21949 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
21950 (phi_translate): Pass in destination ANTIC_OUT set.
21951 (phi_translate_1): Likewise. For a simplified result lookup
21952 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
21953 (phi_translate_set): Adjust.
21954 (do_pre_regular_insertion): Likewise.
21955 (do_pre_partial_partial_insertion): Likewise.
21957 2018-03-08 Martin Liska <mliska@suse.cz>
21959 PR gcov-profile/84735
21960 * doc/gcov.texi: Document usage of profile files.
21961 * gcov-io.h: Document changes in the format.
21963 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
21967 * dwarf2out.c (struct dw_line_info_table): Update comments for
21969 (FORCE_RESET_NEXT_VIEW): New.
21970 (FORCE_RESETTING_VIEW_P): New.
21971 (RESETTING_VIEW_P): Check for -1 too.
21972 (ZERO_VIEW_P): Likewise.
21973 (new_line_info_table): Force-reset next view.
21974 (dwarf2out_begin_function): Likewise.
21975 (dwarf2out_source_line): Simplify zero_view_p initialization.
21976 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
21977 view directly. Omit view when omitting .loc at line 0.
21979 2018-03-08 Jakub Jelinek <jakub@redhat.com>
21981 PR tree-optimization/84740
21982 * tree-switch-conversion.c (process_switch): Call build_constructors
21983 only if info.phi_count is non-zero.
21985 PR tree-optimization/84739
21986 * tree-tailcall.c (find_tail_calls): Check call arguments against
21987 DECL_ARGUMENTS (current_function_decl) rather than
21988 DECL_ARGUMENTS (func) when checking for tail recursion.
21990 2018-03-07 Jakub Jelinek <jakub@redhat.com>
21992 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
21993 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
21994 Volker Reichelt's entry and add entries for people that perform
21995 GCC fuzzy testing and report numerous bugs.
21997 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
22000 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
22001 readonly data in sdata, if that is disabled.
22002 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
22003 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
22004 -mreadonly-in-sdata option.
22006 2018-03-07 Martin Sebor <msebor@redhat.com>
22008 PR tree-optimization/84468
22009 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
22010 basic block when looking for nul assignment.
22012 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
22015 * except.h (output_function_exception_table): Adjust prototype.
22016 * except.c (output_function_exception_table): Remove FNNAME parameter
22017 and add SECTION parameter. Ouput one part of the table at a time.
22018 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
22019 the first part of the exception table and emit unwind directives.
22020 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
22021 (i386_pe_seh_cold_init): Likewise.
22022 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
22023 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
22024 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
22025 (ix86_output_call_insn): Emit a nop in one more case for SEH.
22026 * config/i386/winnt.c: Include except.h.
22027 (struct seh_frame_state): Add reg_offset, after_prologue and
22028 in_cold_section fields.
22029 (i386_pe_seh_end_prologue): Set seh->after_prologue.
22030 (i386_pe_seh_cold_init): New function.
22031 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
22032 to seh->in_cold_section.
22033 (seh_emit_push): Record the offset of the push.
22034 (seh_emit_save): Record the offet of the save.
22035 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
22036 Test seh->after_prologue to disregard the epilogue.
22037 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
22038 (i386_pe_end_cold_function): New function.
22040 2018-03-07 Jakub Jelinek <jakub@redhat.com>
22043 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
22044 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
22047 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
22049 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
22050 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
22052 PR middle-end/84723
22053 * multiple_target.c: Include tree-inline.h and intl.h.
22054 (expand_target_clones): Diagnose and fail if node->definition and
22055 !tree_versionable_function_p (node->decl).
22057 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
22059 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
22061 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
22062 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
22063 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
22065 2018-03-06 Jakub Jelinek <jakub@redhat.com>
22068 * combine.c (try_combine): Use reg_or_subregno instead of handling
22069 just paradoxical SUBREGs and REGs.
22071 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
22073 * config/arc/arc.c (arc_finalize_pic): Remove function.
22074 (arc_must_save_register): We use single base PIC register, remove
22075 checks to save/restore the PIC register.
22076 (arc_expand_prologue): Likewise.
22077 * config/arc/arc-protos.h (arc_set_default_type_attributes):
22079 (arc_verify_short): Likewise.
22080 (arc_attr_type): Likewise.
22081 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
22082 (walk_stores): Likewise.
22083 (arc_address_cost): Make it static.
22084 (arc_verify_short): Likewise.
22085 (branch_dest): Likewise.
22086 (arc_attr_type): Likewise.
22087 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
22088 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
22089 (arc_final_prescan_insn): Remove inserting the nops due to
22090 hardware hazards. It is done in reorg step.
22091 (insn_length_variant_t): Remove.
22092 (insn_length_parameters_t): Likewise.
22093 (arc_insn_length_parameters): Likewise.
22094 (arc_get_insn_variants): Likewise.
22095 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
22097 2018-03-06 Jakub Jelinek <jakub@redhat.com>
22099 PR inline-asm/84683
22100 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
22103 PR tree-optimization/84687
22104 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
22106 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
22108 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
22110 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
22111 Rename to ppc_speculation_barrier.
22112 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
22113 __builtin_ppc_speculation_barrier.
22115 2018-03-05 Jakub Jelinek <jakub@redhat.com>
22118 * combine.c (combine_simplify_rtx): Don't try to simplify if
22119 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
22122 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
22124 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
22125 to 32 bytes when compiling for POWER9.
22127 2018-03-05 Jakub Jelinek <jakub@redhat.com>
22130 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
22131 regparm >= 3 with no arg reg available also for calls with
22132 flag_force_indirect_call. Pass decl to ix86_function_regparm.
22135 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
22137 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
22139 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
22142 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
22144 2018-03-05 Richard Biener <rguenther@suse.de>
22146 PR tree-optimization/84486
22147 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
22148 When inserting a __builtin_assume_aligned call set the LHS
22149 SSA name alignment info accordingly.
22151 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
22153 PR tree-optimization/84114
22154 * config/aarch64/aarch64.c (aarch64_reassociation_width)
22155 Avoid reassociation of FLOAT_MODE addition.
22157 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
22159 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
22160 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
22161 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
22162 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
22163 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
22164 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
22165 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
22167 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
22168 __builtin_ia32_wbinvd): New builtins.
22169 (SPECIAL_ARGS2): New.
22170 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
22171 (SPECIAL_ARGS2): New.
22172 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
22173 (ix86_valid_target_attribute_inner_p): Ditto.
22174 (ix86_init_mmx_sse_builtins): Add special_args2.
22175 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
22176 TARGET_WBNOINVD_P): New.
22177 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
22178 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
22179 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
22180 * config/i386/immintrin.h (_wbinvd): New intrinsic.
22181 * config/i386/pconfigintrin.h: New file.
22182 * config/i386/wbnoinvdintrin.h: Ditto.
22183 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
22185 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
22187 2018-03-05 Richard Biener <rguenther@suse.de>
22189 PR tree-optimization/84670
22190 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
22192 (BB_VISITED_WITH_VISITED_SUCCS): New define.
22193 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
22194 (compute_antic_aux): Only assert the number of values in ANTIC_IN
22195 doesn't grow if all successors (recursively) were visited at least
22198 2018-03-05 Richard Biener <rguenther@suse.de>
22200 PR tree-optimization/84650
22201 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
22202 if executed in the loop pipeline.
22204 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
22206 * doc/configfiles.texi (Configuration Files): Move info about
22207 conditionalizing $target-protos.h to...
22208 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
22209 differs from $target-protos.h.
22211 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
22212 Chung-Ju Wu <jasonwucj@gmail.com>
22214 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
22215 * config/nds32/nds32-multiple.md (setmemsi): Define.
22216 * config/nds32/nds32-memory-manipulation.c
22217 (nds32_gen_dup_4_byte_to_word_value): New.
22218 (emit_setmem_word_loop): New.
22219 (emit_setmem_byte_loop): New.
22220 (nds32_expand_setmem_loop): New.
22221 (nds32_expand_setmem_loop_v3m): New.
22222 (nds32_expand_setmem_unroll): New.
22223 (nds32_expand_setmem): New.
22225 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
22226 Chung-Ju Wu <jasonwucj@gmail.com>
22228 * config/nds32/nds32-memory-manipulation.c
22229 (nds32_emit_load_store): New.
22230 (nds32_emit_post_inc_load_store): New.
22231 (nds32_emit_mem_move): New.
22232 (nds32_emit_mem_move_block): New.
22233 (nds32_expand_movmemsi_loop_unknown_size): New.
22234 (nds32_expand_movmemsi_loop_known_size): New.
22235 (nds32_expand_movmemsi_loop): New.
22236 (nds32_expand_movmemsi_unroll): New.
22237 (nds32_expand_movmemqi): Rename ...
22238 (nds32_expand_movmemsi): ... to this.
22239 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
22240 (movmemsi): ... to this.
22241 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
22242 (nds32_expand_movmemsi): ... to this.
22244 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
22245 Monk Chiang <sh.chiang04@gmail.com>
22246 Chung-Ju Wu <jasonwucj@gmail.com>
22248 * config/nds32/nds32-protos.h
22249 (nds32_expand_load_multiple): New arguments.
22250 (nds32_expand_store_multiple): Ditto.
22251 (nds32_valid_multiple_load_store): Rename ...
22252 (nds32_valid_multiple_load_store_p): ... to this.
22253 * config/nds32/nds32-memory-manipulation.c
22254 (nds32_expand_load_multiple): Refine implementation.
22255 (nds32_expand_store_multiple): Ditto.
22256 * config/nds32/nds32-multiple.md
22257 (load_multiple): Update nds32_expand_load_multiple interface.
22258 (store_multiple): Update nds32_expand_store_multiple interface.
22259 * config/nds32/nds32-predicates.c
22260 (nds32_valid_multiple_load_store): Rename ...
22261 (nds32_valid_multiple_load_store_p): ... to this and refine
22263 * config/nds32/predicates.md
22264 (nds32_load_multiple_and_update_address_operation): New predicate.
22265 (nds32_store_multiple_and_update_address_operation): New predicate.
22267 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
22268 Chung-Ju Wu <jasonwucj@gmail.com>
22270 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
22271 (combo): New attribute.
22272 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
22274 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
22276 * config/nds32/nds32.opt: Change -mcmodel= default value.
22278 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
22279 Monk Chiang <sh.chiang04@gmail.com>
22280 Chung-Ju Wu <jasonwucj@gmail.com>
22282 * config/nds32/constants.md (unspec_element): New enum.
22283 * config/nds32/constraints.md (Umw): New constraint.
22284 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
22285 * config/nds32/nds32-intrinsic.md: Likewise.
22286 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
22287 (nds32_valid_smw_lwm_base_p): New.
22288 (nds32_output_smw_single_word): New.
22289 (nds32_output_lmw_single_word): New.
22290 (nds32_expand_unaligned_load): New.
22291 (nds32_expand_unaligned_store): New.
22292 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
22293 (nds32_output_smw_single_word): Declare.
22294 (nds32_output_lmw_single_word): Declare.
22295 (nds32_expand_unaligned_load): Declare.
22296 (nds32_expand_unaligned_store): Declare.
22297 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
22298 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
22299 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
22300 NDS32_BUILTIN_UASTORE_DW.
22301 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
22304 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
22305 Kito Cheng <kito.cheng@gmail.com>
22306 Chung-Ju Wu <jasonwucj@gmail.com>
22308 * config/nds32/nds32-intrinsic.c
22309 (nds32_expand_builtin_null_ftype_reg): Delete.
22310 (nds32_expand_builtin_reg_ftype_imm): Ditto.
22311 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
22312 (nds32_read_argument): New.
22313 (nds32_legitimize_target): Ditto.
22314 (nds32_legitimize_argument): Ditto.
22315 (nds32_check_constant_argument): Ditto.
22316 (nds32_expand_unop_builtin): Ditto.
22317 (nds32_expand_unopimm_builtin): Ditto.
22318 (nds32_expand_binop_builtin): Ditto.
22319 (nds32_builtin_decl_impl): Ditto.
22320 (builtin_description): Ditto.
22321 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
22322 (nds32_init_builtins_impl): Ditto.
22323 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
22324 (nds32_builtin_decl): New.
22325 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
22326 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
22328 2018-03-02 Jeff Law <law@redhat.com>
22330 * reorg.c (stop_search_p): Handle DEBUG_INSN.
22331 (redundant_insn, fill_simple_delay_slots): Likewise.
22332 (fill_slots_from_thread): Likewise.
22333 * resource.c (mark_referenced_resources): Likewise.
22334 (mark_set_resources, find_dead_or_set_registers): Likewise.
22336 2018-03-02 Jakub Jelinek <jakub@redhat.com>
22338 * substring-locations.h (format_warning_va): Formatting fix for
22339 ATTRIBUTE_GCC_DIAG.
22340 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
22342 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
22343 * substring-locations.c: Include intl.h.
22344 (format_warning_va): Turned into small wrapper around
22345 format_warning_n_va, renamed to ...
22346 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
22347 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
22349 (format_warning_at_substring_n): New function.
22350 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
22351 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
22352 format_warning_at_substring with just a shorter name instead of
22353 const function pointer.
22354 (fmtwarn_n): New function.
22355 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
22356 appropriate, get rid of all the fmtstr temporaries, move conditionals
22357 with G_() wrapped string literals directly into fmtwarn arguments,
22358 cast dir.len to (int), formatting fixes.
22360 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
22362 * doc/invoke.texi: Remove "Cilk Plus" references.
22364 2018-03-02 Jakub Jelinek <jakub@redhat.com>
22365 Richard Biener <rguenther@suse.de>
22368 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
22369 for error or warning attributes if CALL_FROM_THUNK_P is set.
22372 2018-03-02 Jakub Jelinek <jakub@redhat.com>
22375 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
22376 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
22379 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
22380 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
22382 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
22383 instead of -1U in last predictors element's probability member.
22385 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
22388 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
22389 arguments if they are comparable.
22391 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
22393 PR tree-optimization/84634
22394 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
22395 masks and masked_loop_p with a single loop_masks, making sure it's
22396 null for bb vectorization.
22398 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
22400 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
22401 (vect_analyze_data_ref_access): Use loop->safe_len rather than
22402 loop->force_vectorize to check whether there is no alias.
22404 2018-03-02 Jakub Jelinek <jakub@redhat.com>
22407 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
22409 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
22411 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
22412 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
22413 instead of a loop around prev_real_insn.
22414 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
22417 PR inline-asm/84625
22418 * config/i386/i386.c (ix86_print_operand): Use conditional
22419 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
22422 2018-03-02 Richard Biener <rguenther@suse.de>
22424 PR tree-optimization/84427
22425 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
22426 (bitmap_set_subtract_values): Rewrite to handle multiple
22429 (prune_clobbered_mems): Likewise.
22430 (phi_translate): Take edge instead of pred/phiblock.
22431 (phi_translate_1): Likewise.
22432 (phi_translate_set): Likewise. Insert all translated
22433 exprs for a value into the set, keeping possibly multiple
22434 expressions per value.
22435 (compute_antic_aux): Adjust for phi_translate changes.
22436 When intersecting union the expressions and prune those
22437 not in the final value set, keeping possibly multiple
22438 expressions per value. Do not use value-insertion
22439 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
22440 all expressions. Add verification that the value-sets
22441 only shrink during iteration.
22442 (compute_partial_antic_aux): Adjust for the phi_translate changes.
22443 (do_pre_regular_insertion): Likewise.
22444 (do_pre_partial_partial_insertion): Likewise.
22446 2018-03-02 Richard Biener <rguenther@suse.de>
22449 * config/darwin.c (saved_debug_info_level): New static global.
22450 (darwin_asm_lto_start): Disable debug info generation for LTO out.
22451 (darwin_asm_lto_end): Restore debug info generation settings.
22453 2018-03-01 Martin Liska <mliska@suse.cz>
22456 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
22457 volatile arguments.
22459 2018-03-01 Richard Biener <rguenther@suse.de>
22462 * dwarf2out.c (gen_variable_die): Properly handle late VLA
22463 type annotation with LTO when debug was disabled at compile-time.
22465 2018-03-01 Matthew Fortune <mfortune@gmail.com>
22467 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
22469 (mips_final_postscan_insn): Likewise.
22471 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
22473 PR rtl-optimization/84528
22474 * alias.c (init_alias_target): Add commentary.
22475 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
22476 a unique base value if the frame pointer is not eliminated
22477 to the stack pointer.
22479 2018-03-01 Tom de Vries <tom@codesourcery.com>
22481 PR rtl-optimization/83327
22482 * lra-int.h (hard_regs_spilled_into): Declare.
22483 * lra.c (hard_regs_spilled_into): Define.
22484 (init_reg_info): Init hard_regs_spilled_into.
22485 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
22486 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
22487 (process_bb_lives): Handle hard_regs_spilled_into.
22488 (lra_create_live_ranges_1): Before doing liveness propagation, clear
22489 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
22491 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
22493 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
22494 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
22495 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
22496 * config/rs6000/aix72.h: New file.
22498 2018-02-28 Jakub Jelinek <jakub@redhat.com>
22500 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
22501 instead of warning_at with conditional singular and plural messages
22505 * stor-layout.c (update_alignment_for_field): For
22506 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
22507 && !DECL_PACKED (field), do the alignment update, just use
22508 only desired_align instead of MAX (type_align, desired_align)
22510 (place_field): Don't do known_align < desired_align handling
22511 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
22512 is non-NULL, instead do it after rli->prev_field handling and
22513 only if not within a bitfield word. For DECL_PACKED (field)
22514 use type_align of BITS_PER_UNIT.
22516 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
22518 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
22519 superfluous parentheses and trailing spaces.
22521 2018-02-28 Richard Biener <rguenther@suse.de>
22523 PR tree-optimization/84584
22524 * graphite-scop-detection.c (scop_detection::add_scop): Discard
22525 SCoPs with fake exit edge.
22527 2018-02-28 Martin Liska <mliska@suse.cz>
22530 * timevar.c (timer::print): Fix format to properly print 100%
22533 2018-02-28 Richard Biener <rguenther@suse.de>
22535 PR middle-end/84607
22536 * genmatch.c (capture_info::walk_match): Do not mark
22537 captured expressions without operands as expr_p given
22538 they act more like predicates and should be subject to
22539 "lost tail" side-effect preserving.
22541 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
22543 PR rtl-optimization/81611
22544 * auto-inc-dec.c (attempt_change): Move dead note from
22545 mem_insn if it's the next use of regno
22546 (find_address): Take address use of reg holding
22547 non-incremented value. Add parm to limit search to the named
22549 (merge_in_block): Attempt to use a mem insn that is the next
22550 use of the original regno.
22552 2018-02-27 Martin Sebor <msebor@redhat.com>
22555 * doc/invoke.texi (-Wmissing-attributes): New option.
22556 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
22558 2018-02-27 Martin Sebor <msebor@redhat.com>
22560 PR translation/84207
22561 * diagnostic-core.h (warning_n, error_n, inform_n): Change
22562 n argument to unsigned HOST_WIDE_INT.
22563 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
22564 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
22565 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
22566 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
22568 2018-02-27 Richard Biener <rguenther@suse.de>
22570 PR tree-optimization/84512
22571 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
22572 Do not use the estimate returned from record_stmt_cost for
22573 the scalar iteration cost but sum properly using add_stmt_cost.
22575 2018-02-27 Richard Biener <rguenther@suse.de>
22577 PR tree-optimization/84466
22578 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
22579 Adjust last change to less strictly validate use operands.
22581 2018-02-27 Martin Liska <mliska@suse.cz>
22583 PR gcov-profile/84548
22584 * gcov.c (process_file): Allow partial overlap and consider it
22585 also as group functions.
22586 (output_lines): Properly calculate range of lines for a group.
22588 2018-02-27 Martin Liska <mliska@suse.cz>
22590 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
22591 'ggc' suffixes. Change first column width.
22592 (timer::print): Fix formatting of the column.
22594 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
22596 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
22597 preserve inline entry blocks for the sake of debug inline
22598 entry point markers alone.
22599 (remove_unused_locals): Suggest in comments a better place to
22600 force the preservation of inline entry blocks that are
22601 otherwise unused, but do not preserve them.
22603 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
22605 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
22607 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
22610 * config/i386/constraints.md (Bs): Replace
22611 ix86_indirect_branch_register with
22612 TARGET_INDIRECT_BRANCH_REGISTER.
22614 * config/i386/i386.md (indirect_jump): Likewise.
22615 (tablejump): Likewise.
22616 (*sibcall_memory): Likewise.
22617 (*sibcall_value_memory): Likewise.
22618 Peepholes of indirect call and jump via memory: Likewise.
22619 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
22620 (*sibcall_value_GOT_32): Likewise.
22621 * config/i386/predicates.md (indirect_branch_operand): Likewise.
22622 (GOT_memory_operand): Likewise.
22623 (call_insn_operand): Likewise.
22624 (sibcall_insn_operand): Likewise.
22625 (GOT32_symbol_operand): Likewise.
22626 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
22628 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
22630 PR rtl-optimization/83496
22631 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
22632 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
22633 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
22634 redundant insn, if any.
22635 (relax_delay_slots): Likewise.
22636 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
22638 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
22640 PR tree-optimization/83965
22641 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
22642 that grouped statements are part of a reduction chain. Return
22643 true if the statement is not marked as a reduction itself but
22644 is part of a group.
22645 (vect_recog_dot_prod_pattern): Don't check whether the statement
22646 is part of a group here.
22647 (vect_recog_sad_pattern): Likewise.
22648 (vect_recog_widen_sum_pattern): Likewise.
22650 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
22653 * final.c (rest_of_clean_state): Also look for calls inside sequences.
22655 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
22658 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
22660 (ix86_output_indirect_function_return): New prototype.
22661 (ix86_split_simple_return_pop_internal): Likewise.
22662 * config/i386/i386.c (indirect_return_via_cx): New.
22663 (indirect_return_via_cx_bnd): Likewise.
22664 (indirect_thunk_name): Handle return va CX_REG.
22665 (output_indirect_thunk_function): Create alias for
22666 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
22667 (ix86_output_indirect_jmp): Remove the bool argument.
22668 (ix86_output_indirect_function_return): New function.
22669 (ix86_split_simple_return_pop_internal): Likewise.
22670 * config/i386/i386.md (*indirect_jump): Don't pass false
22671 to ix86_output_indirect_jmp.
22672 (*tablejump_1): Likewise.
22673 (simple_return_pop_internal): Change it to define_insn_and_split.
22674 Call ix86_split_simple_return_pop_internal to split it for
22675 -mfunction-return=.
22676 (simple_return_indirect_internal): Call
22677 ix86_output_indirect_function_return instead of
22678 ix86_output_indirect_jmp.
22680 2018-02-26 Jakub Jelinek <jakub@redhat.com>
22683 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
22684 memset and value initialization afterwards.
22686 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
22688 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
22690 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
22693 * common/config/aarch64/aarch64-common.c
22694 (aarch_option_optimization_table[]): Switch
22695 off fomit-frame-pointer
22697 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
22698 Chung-Ju Wu <jasonwucj@gmail.com>
22700 * config/nds32/nds32-multiple.md (load_multiple): Disallow
22702 (store_multiple): Ditto.
22704 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
22706 * config.gcc: Add --with-cpu support for nds32 target.
22707 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
22708 * config/nds32/nds32.opt: Add -mcpu= option.
22710 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
22712 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
22713 isel=yes): Warn for these deprecated options.
22715 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
22717 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
22718 ISA_2_5_MASKS_EMBEDDED.
22720 2018-02-23 Jakub Jelinek <jakub@redhat.com>
22722 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
22723 p->max as pointers rather than using iterative_hash_expr.
22725 2018-02-23 Carl Love <cel@us.ibm.com>
22727 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
22728 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
22730 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
22731 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
22732 P8V_BUILTIN_VEC_VUNSIGNED2.
22734 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
22737 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
22738 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
22740 * lra-constraints.c (curr_insn_transform): Set up
22741 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
22743 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
22744 * lra-eliminations.c (spill_pseudos): Ditto.
22745 (process_insn_for_elimination): Ditto.
22746 * lra-lives.c (reg_early_clobber_p): Use the new macros.
22747 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
22748 LRA_NON_CLOBBERED_ALT.
22750 2018-02-22 Martin Sebor <msebor@redhat.com>
22752 PR tree-optimization/84480
22753 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
22754 to maybe_diag_stxncpy_trunc. Call it.
22755 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
22756 from gimple_fold_builtin_strcpy. Print inlining stack.
22757 (handle_builtin_stxncpy): Print inlining stack.
22758 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
22760 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
22763 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
22764 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
22765 and -fcheck-pointer-bounds are used together.
22766 (indirect_thunk_prefix): New enum.
22767 (indirect_thunk_need_prefix): New function.
22768 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
22769 "_nt" instead of "_bnd" for NOTRACK prefix.
22770 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
22771 (output_indirect_thunk_function): Likewise.
22773 (ix86_code_end): Update output_indirect_thunk_function calls.
22774 (ix86_output_indirect_branch_via_reg): Replace
22775 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
22776 (ix86_output_indirect_branch_via_push): Likewise.
22777 (ix86_output_function_return): Likewise.
22778 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
22779 incompatible with -fcf-protection=branch and
22780 -fcheck-pointer-bounds.
22782 2018-02-22 Steve Ellcey <sellcey@cavium.com>
22785 * config/aarch64/aarch64.c (aarch64_print_address_internal):
22786 Change gcc_assert call to output_operand_lossage.
22788 2018-02-22 Steve Ellcey <sellcey@cavium.com>
22790 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
22792 2018-02-22 DJ Delorie <dj@redhat.com>
22793 Sebastian Perta <sebastian.perta@renesas.com>
22794 Oleg Endo <olegendo@gcc.gnu.org>
22796 * config/rx/rx.c (rx_rtx_costs): New function.
22797 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
22799 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
22801 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
22803 2018-02-22 Martin Liska <mliska@suse.cz>
22806 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
22807 Add "native" as a possible value.
22809 2018-02-22 Martin Liska <mliska@suse.cz>
22812 * config/i386/i386.c (ix86_option_override_internal):
22813 Add "native" as a possible value for -march and -mtune.
22815 2018-02-22 Jakub Jelinek <jakub@redhat.com>
22818 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
22819 to all type variants.
22821 PR tree-optimization/84503
22822 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
22823 width as info->bitpos + info->bitsize - start.
22824 (merged_store_group::merge_overlapping): Simplify width computation.
22825 (check_no_overlap): New function.
22826 (imm_store_chain_info::try_coalesce_bswap): Compute expected
22827 start + width and last_order of the group, fail if check_no_overlap
22829 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
22830 to group if check_no_overlap fails.
22832 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
22834 * config/rs6000/altivec.md: Delete contraint arguments to
22835 define_expand, define_split, and define_peephole2, and in
22836 define_insn_and_split if always unused.
22837 * config/rs6000/darwin.md: Ditto.
22838 * config/rs6000/dfp.md: Ditto.
22839 * config/rs6000/rs6000.md: Ditto.
22840 * config/rs6000/sync.md: Ditto.
22841 * config/rs6000/vector.md: Ditto.
22842 * config/rs6000/vsx.md: Ditto.
22844 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
22846 * config/rs6000/altivec.md: Write output control strings as braced
22847 blocks instead of double-quoted strings.
22848 * config/rs6000/darwin.md: Ditto.
22849 * config/rs6000/rs6000.md: Ditto.
22850 * config/rs6000/vector.md: Ditto.
22851 * config/rs6000/vsx.md: Ditto.
22853 2018-02-21 Jason Merrill <jason@redhat.com>
22855 PR c++/84314 - ICE with templates and fastcall attribute.
22856 * attribs.c (build_type_attribute_qual_variant): Remove assert.
22858 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
22860 * ipa-cp.c (determine_versionability): Fix comment typos.
22862 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
22865 * ipa-cp.c (determine_versionability): Do not version functions caling
22868 2018-02-21 Martin Liska <mliska@suse.cz>
22871 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
22872 Add "native" as a possible value.
22873 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
22874 the macro when native cpu detection is available.
22876 2018-02-21 Martin Liska <mliska@suse.cz>
22879 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
22880 Add "native" as a possible value.
22881 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
22882 when native cpu detection is available.
22884 2018-02-21 Jakub Jelinek <jakub@redhat.com>
22885 Martin Sebor <msebor@redhat.com>
22887 PR tree-optimization/84478
22888 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
22890 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
22891 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
22892 support which is conservatively correct, for 2 only stay conservative
22893 for maxlen. Formatting and comment capitalization fixes. Add STRICT
22894 argument to the 2 argument get_range_strlen, adjust 6 arg
22895 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
22897 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
22898 (gimple_fold_builtin_strlen): Pass true as last argument to
22901 2018-02-20 Martin Sebor <msebor@redhat.com>
22903 PR middle-end/84095
22904 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
22905 (builtin_memref::set_base_and_offset): Same. Handle inner references.
22906 (builtin_memref::builtin_memref): Factor out parts into
22907 set_base_and_offset and call it.
22909 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
22911 PR middle-end/84406
22912 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
22913 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
22914 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
22915 search at the associated MODE_INT.
22917 2018-02-20 Jeff Law <law@redhat.com>
22919 PR middle-end/82123
22920 PR tree-optimization/81592
22921 PR middle-end/79257
22922 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
22923 for range data rather than using global data.
22924 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
22925 range data rather than using global data.
22926 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
22927 pass it to children as needed.
22928 (struct directive::fmtresult): Similarly.
22929 (struct directive::set_width): Similarly.
22930 (struct directive::set_precision): Similarly.
22931 (format_integer, format_directive, parse_directive): Similarly.
22932 (format_none): Accept unnamed vr_values parameter.
22933 (format_percent, format_floating, format_character): Similarly.
22934 (format_string, format_plain): Similarly.
22935 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
22936 the EVRP range analyzer for range data rather than using global data.
22937 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
22938 gimple-ssa-evrp-analyze.h
22939 (class sprintf_dom_walker): Add after_dom_children member function.
22940 Add evrp_range_analyzer member.
22941 (sprintf_dom_walker::before_dom_children): Call into the EVRP
22942 range analyzer as needed.
22943 (sprintf_dom_walker::after_dom_children): New member function.
22944 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
22946 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
22947 (evrp_range_analyzer::pop_to_marker): Likewise.
22949 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
22951 PR tree-optimization/84419
22952 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
22953 with the required type if its current type is compatible but
22956 2018-02-20 Jakub Jelinek <jakub@redhat.com>
22958 PR middle-end/82004
22959 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
22960 after vectorization.
22962 2018-02-20 Martin Liska <mliska@suse.cz>
22965 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
22966 possible values if we don't have a hint.
22968 2018-02-20 Martin Liska <mliska@suse.cz>
22972 * final.c (shorten_branches): Build align_tab array with one
22974 * opts.c (finish_options): Add alignment option limit check.
22975 (MAX_CODE_ALIGN): Likewise.
22976 (MAX_CODE_ALIGN_VALUE): Likewise.
22977 * doc/invoke.texi: Document maximum allowed option value for
22978 all -falign-* options.
22980 2018-02-19 Jakub Jelinek <jakub@redhat.com>
22983 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
22984 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
22985 * var-tracking.c (emit_note_insn_var_location): Remove all references
22986 to NOTE_INSN_CALL_ARG_LOCATION.
22987 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
22988 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
22989 Use copy_rtx_if_shared.
22990 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
22991 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
22992 (dwarf2out_var_location): Remove handling of
22993 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
22995 * final.c (final_scan_insn): Remove all references to
22996 NOTE_INSN_CALL_ARG_LOCATION.
22997 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
22998 before dumping final insns.
22999 * except.c (emit_note_eh_region_end): Remove all references to
23000 NOTE_INSN_CALL_ARG_LOCATION.
23001 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
23002 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
23003 * config/arc/arc.c (hwloop_optimize): Likewise.
23004 * config/arm/arm.c (create_fix_barrier): Likewise.
23005 * config/s390/s390.c (s390_chunkify_start): Likewise.
23006 * config/sh/sh.c (find_barrier): Likewise.
23007 * config/i386/i386.c (rest_of_insert_endbranch,
23008 ix86_seh_fixup_eh_fallthru): Likewise.
23009 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
23010 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
23011 * config/frv/frv.c (frv_function_prologue): Likewise.
23012 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
23014 (note_outside_basic_block_p): Remove all references to
23015 NOTE_INSN_CALL_ARG_LOCATION.
23016 * gengtype.c (adjust_field_rtx_def): Likewise.
23017 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
23019 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
23020 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
23023 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
23026 PR tree-optimization/84452
23027 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
23028 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
23031 2018-02-19 Martin Liska <mliska@suse.cz>
23034 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
23036 2018-02-19 Martin Liska <mliska@suse.cz>
23037 Richard Sandiford <richard.sandiford@linaro.org>
23039 PR tree-optimization/82491
23040 * gimple-fold.c (get_base_constructor): Make earlier bail out
23043 2018-02-19 Carl Love <cel@us.ibm.com>
23045 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
23046 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
23048 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
23049 P8V_BUILTIN_VEC_NEG.
23051 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
23053 * config/rl78/rl78.md (movdf): New define expand.
23055 2018-02-19 Martin Liska <mliska@suse.cz>
23058 * doc/invoke.texi: Fix typo.
23059 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
23061 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
23063 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
23064 handle rs6000_single_float and rs6000_double_float specially for
23067 2018-02-16 Jeff Law <law@redhat.com>
23069 * config/rx/rx.c (add_pop_cfi_notes): New function.;
23070 (pop_regs): Use it.
23072 2018-02-16 Jakub Jelinek <jakub@redhat.com>
23075 * ipa-inline.c (inline_small_functions): Fix a typo.
23077 2018-02-16 Nathan Sidwell <nathan@acm.org>
23079 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
23081 2018-02-16 Carl Love <cel@us.ibm.com>
23083 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
23084 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
23085 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
23086 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
23087 expansion to P8V_BUILTIN_VEC_FLOAT2.
23089 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
23091 PR rtl-optimization/70023
23092 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
23093 src_regno into account.
23095 2018-02-16 Carl Love <cel@us.ibm.com>
23097 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
23098 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
23099 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
23100 * config/rs6000/rs6000.c: Remove case statements for
23101 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
23102 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
23103 and P9V_BUILTIN_VEC_VINSERT4B.
23104 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
23105 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
23106 * config/rs6000/vsx.md:
23107 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
23110 2018-02-16 Carl Love <cel@us.ibm.com>
23112 * config/rs6000/altivec.h: Add builtin names vec_extract4b
23114 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
23116 * config/rs6000/rs6000-c.c: Add the definitions for
23117 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
23118 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
23119 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
23120 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
23121 definition for insert4b and define insn *insert3b_internal.
23122 * doc/extend.texi: Add documentation for vec_extract4b.
23124 2018-02-16 Nathan Sidwell <nathan@acm.org>
23126 * doc/extend.texi (Backwards Compatibility): Mention friend
23127 injection. Note for-scope is deprecated.
23128 * doc/invoke.texi (-ffriend-injection): Deprecate.
23130 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
23132 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
23133 that moved to I2, also allow destinations that are a paradoxical
23134 subreg (instead of a normal reg).
23136 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
23139 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
23142 2018-02-16 Richard Biener <rguenther@suse.de>
23144 PR tree-optimization/84037
23145 PR tree-optimization/84016
23147 * config/i386/i386.c (ix86_builtin_vectorization_cost):
23148 Adjust vec_construct for the fact we need additional higher latency
23149 128bit inserts for AVX256 and AVX512 vector builds.
23150 (ix86_add_stmt_cost): Scale vector construction cost for
23153 2018-02-16 Richard Biener <rguenther@suse.de>
23155 PR tree-optimization/84417
23156 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
23157 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
23158 (non_rewritable_lvalue_p): Likewise, use poly-ints.
23160 2018-02-16 Martin Liska <mliska@suse.cz>
23163 * internal-fn.def (ASAN_CHECK): Set proper flags.
23164 (ASAN_MARK): Likewise.
23166 2018-02-16 Julia Koval <julia.koval@intel.com>
23168 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
23169 from PTA_CANNONLAKE.
23171 2018-02-16 Jakub Jelinek <jakub@redhat.com>
23174 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
23175 Use ++iter rather than iter++ for std::list iterators.
23176 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
23177 defer deleting them until all nodes in the forest are processed. Do
23178 free even leaf nodes. Change to_process into auto_vec.
23181 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
23182 * vec.h (vec_default_construct): Use memset instead of placement new
23183 if BROKEN_VALUE_INITIALIZATION is defined.
23184 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
23185 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
23188 PR rtl-optimization/83723
23189 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
23190 * lra.c (lra_substitute_pseudo): Likewise. If true, use
23191 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
23193 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
23195 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
23197 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
23199 PR rtl-optimization/81443
23200 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
23201 from inner REGs to paradoxical SUBREGs.
23203 2018-02-16 Richard Biener <rguenther@suse.de>
23205 PR tree-optimization/84399
23206 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
23207 For operands we can analyze at their definition make sure we can
23208 analyze them at each use as well.
23210 2018-02-16 Richard Biener <rguenther@suse.de>
23212 PR tree-optimization/84190
23213 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
23214 volatile accesses if the decl isn't volatile.
23216 2018-02-15 Jason Merrill <jason@redhat.com>
23218 PR c++/84314 - ICE with templates and fastcall attribute.
23219 * attribs.c (build_type_attribute_qual_variant): Don't clobber
23220 TYPE_CANONICAL on an existing type.
23222 2018-02-15 Jakub Jelinek <jakub@redhat.com>
23224 PR tree-optimization/84383
23225 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
23226 dstoff nor call operand_equal_p if dstbase is NULL.
23228 PR tree-optimization/84334
23229 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
23230 also a CONSTANT_CLASS_P, punt.
23232 2018-02-14 Jim Wilson <jimw@sifive.com>
23234 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
23235 first SMALL_OPERAND check. New local min_second_step. Move assert
23236 to where locals are set. Add TARGET_RVC support.
23237 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
23239 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
23241 * doc/invoke.texi: Correct -Wformat-overflow code sample.
23243 2018-02-14 Martin Sebor <msebor@redhat.com>
23245 PR tree-optimization/83698
23246 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
23247 arrays constrain the offset range to their bounds.
23248 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
23249 (builtin_access::overlap): Avoid setting the size of overlap if it's
23251 (maybe_diag_overlap): Also consider arrays when deciding what values
23252 of offsets to include in diagnostics.
23254 2018-02-14 Martin Sebor <msebor@redhat.com>
23257 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
23258 that correspond to the kind of a declaration.
23260 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
23263 * config/pa/pa.md: Load address of PIC label using the linkage table
23264 if the label is nonlocal.
23266 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
23268 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
23269 warning message if user requests -maltivec=be.
23270 * doc/invoke.texi: Document deprecation of -maltivec=be.
23272 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
23275 * config/rs6000/rs6000-c.c: Update definitions for
23276 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
23277 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
23279 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
23282 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
23283 add _get_ssp intrinsics. Remove argument from
23284 __builtin_ia32_rdssp[d|q].
23285 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
23286 * config/i386/i386-builtin.def: Remove argument from
23287 __builtin_ia32_rdssp[d|q].
23288 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
23289 ix86_expand_special_args_builtin for _rdssp[d|q].
23290 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
23291 Clear register before usage.
23292 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
23293 Add documentation for new _get_ssp and _inc_ssp intrinsics.
23295 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
23297 PR tree-optimization/84357
23298 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
23299 operand 1 of an ARRAY_REF too.
23301 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
23304 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
23305 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
23307 (set_of_reg): New struct.
23308 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
23309 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
23310 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
23312 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
23313 Split into bitclr, bitset, bitinvert patterns if appropriate.
23314 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
23315 use rx_fuse_in_memory_bitop.
23316 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
23317 to named insn, correct maximum insn length.
23319 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
23322 * machmode.def: Define a complex mode for PARTIAL_INT.
23323 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
23325 * doc/rtl.texi: Document CSPImode.
23326 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
23328 (msp430_hard_regno_nregs_with_padding): Likewise.
23330 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
23333 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
23335 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
23337 PR rtl-optimization/84169
23338 * combine.c (try_combine): New variable split_i2i3. Set it to true if
23339 we generated a parallel as new i3 and we split that to new i2 and i3
23340 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
23341 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
23342 those to i2, not i1. Partially rewrite this scan code.
23344 2018-02-13 Jakub Jelinek <jakub@redhat.com>
23347 * stor-layout.c (place_field): For variable length fields, adjust
23348 offset_align afterwards not just based on the field's alignment,
23349 but also on the size.
23351 PR middle-end/84309
23352 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
23353 of exps and logs in the use_exp2 case.
23355 2018-02-13 Jeff Law <law@redhat.com>
23357 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
23358 entry for "vector".
23360 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
23363 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
23367 * common.opt (gas-loc-support, gas-locview-support): New.
23368 (ginline-points, ginternal-reset-location-views): New.
23369 * doc/invoke.texi: Document them. Use @itemx where intended.
23370 (gvariable-location-views): Adjust.
23371 * target.def (reset_location_view): New.
23372 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
23373 (TARGET_RESET_LOCATION_VIEW): New.
23374 * doc/tm.texi: Rebuilt.
23375 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
23376 (dwarf2out_default_as_locview_support): New.
23377 (output_asm_line_debug_info): Use option variables.
23378 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
23379 (output_loc_list): Likewise.
23380 (add_high_low_attributes): Check option variables.
23381 Don't output entry view attribute in strict mode.
23382 (gen_inlined_subroutine_die): Check option variables.
23383 (dwarf2out_inline_entry): Likewise.
23384 (init_sections_and_labels): Likewise.
23385 (dwarf2out_early_finish): Likewise.
23386 (maybe_reset_location_view): New, from...
23387 (dwarf2out_var_location): ... here. Call it.
23388 * debug.h (dwarf2out_default_as_loc_support): Declare.
23389 (dwarf2out_default_as_locview_support): Declare.
23390 * hooks.c (hook_int_rtx_insn_0): New.
23391 * hooks.h (hook_int_rtx_insn_0): Declare.
23392 * toplev.c (process_options): Take -gas-loc-support and
23393 -gas-locview-support from dwarf2out. Enable
23394 -gvariable-location-views by default only with locview
23395 assembler support. Enable -ginternal-reset-location-views by
23396 default only if the target defines the corresponding hook.
23397 Enable -ginline-points by default if location views are
23398 enabled; force it disabled if statement frontiers are
23400 * tree-inline.c (expand_call_inline): Check option variables.
23401 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
23403 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
23405 PR tree-optimization/84321
23406 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
23407 handling. Also check whether the anti-range contains any values
23408 that satisfy the mask; switch to a VR_RANGE if not.
23410 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
23413 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
23415 2018-02-13 Martin Jambor <mjambor@suse.cz>
23418 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
23419 of call statements, also set location of a load to a temporary.
23421 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
23423 * config/rl78/rl78.c (add_vector_labels): New function.
23424 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
23425 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
23426 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
23427 which checks that no arguments are passed.
23428 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
23429 * doc/extend.texi: Documentation for the new attribute.
23431 2018-02-13 Andreas Schwab <schwab@suse.de>
23433 * config/riscv/linux.h (CPP_SPEC): Define.
23435 2018-02-13 Jakub Jelinek <jakub@redhat.com>
23438 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
23439 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
23440 OPTION_MASK_ISA_AES as first argument to def_builtin_const
23441 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
23442 instead of OPTION_MASK_ISA_PCLMUL as first argument to
23443 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
23444 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
23445 temporarily for AES and PCLMUL builtins.
23447 PR tree-optimization/84339
23448 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
23449 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
23452 PR middle-end/84309
23453 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
23454 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
23455 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
23457 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
23459 * omp-simd-clone.h: New file.
23460 * omp-simd-clone.c: Include omp-simd-clone.h.
23461 (expand_simd_clones): No longer static.
23462 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
23463 cgraph.h and omp-simd-clone.h.
23464 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
23465 (vect_recog_widen_shift_pattern): Formatting fix.
23466 (vect_pattern_recog_1): Don't check optab for calls.
23469 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
23470 operands[2] into a REG before using gen_lowpart on it.
23472 2018-02-12 Jeff Law <law@redhat.com>
23475 * config/sh/sh.c (find_barrier): Consider a sibling call
23478 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
23479 successfully back substituting a reg.
23481 2018-02-12 Richard Biener <rguenther@suse.de>
23483 PR tree-optimization/84037
23484 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
23485 parameter, move visited init to caller.
23486 (vect_slp_analyze_operations): Separate cost from validity
23487 check, initialize visited once for all instances.
23488 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
23490 * tree-vect-stmts.c (vect_model_simple_cost): Make early
23492 (vect_model_promotion_demotion_cost): Likewise.
23493 (vectorizable_bswap): Guard cost modeling with !slp_node
23494 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
23496 (vectorizable_call): Likewise.
23497 (vectorizable_conversion): Likewise.
23498 (vectorizable_assignment): Likewise.
23499 (vectorizable_shift): Likewise.
23500 (vectorizable_operation): Likewise.
23501 (vectorizable_store): Likewise.
23502 (vectorizable_load): Likewise.
23503 (vectorizable_condition): Likewise.
23504 (vectorizable_comparison): Likewise.
23506 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
23509 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
23510 (ASAN_MARK): Fix fnspec to account for return value, change pointer
23511 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
23513 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
23515 PR middle-end/83665
23516 * params.def (inline-min-speedup): Increase from 8 to 15.
23517 (max-inline-insns-auto): Decrease from 40 to 30.
23518 * ipa-split.c (consider_split): Add some buffer for function to
23519 be considered inlining candidate.
23520 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
23523 2018-02-12 Richard Biener <rguenther@suse.de>
23525 PR tree-optimization/84037
23526 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
23527 matched stmts if we cannot swap the non-matched ones.
23529 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
23531 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
23532 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
23533 _mm_maskz_scalef_round_ss): New intrinsics.
23534 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
23535 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
23536 __builtin_ia32_scalefss_round): Remove.
23537 (__builtin_ia32_scalefsd_mask_round,
23538 __builtin_ia32_scalefss_mask_round): New intrinsics.
23539 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
23540 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
23541 ((match_operand:VF_128 2 "<round_nimm_predicate>"
23542 "<round_constraint>")): Changed to ...
23543 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
23544 "<round_scalar_constraint>")): ... this.
23545 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
23546 %0, %1, %2<round_op3>}"): Changed to ...
23547 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
23548 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
23549 %2<round_scalar_mask_op3>}"): ... this.
23550 * config/i386/subst.md (round_scalar_nimm_predicate): New.
23552 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
23554 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
23555 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
23556 (_mm_maskz_sqrt_round_ss): New intrinsics.
23557 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
23558 (__builtin_ia32_sqrtsd_mask_round)
23559 (__builtin_ia32_sqrtss_mask_round): New builtins.
23560 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
23561 (__builtin_ia32_sqrtss_round): Remove.
23562 (__builtin_ia32_sqrtsd_mask_round)
23563 (__builtin_ia32_sqrtss_mask_round): New builtins.
23564 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
23565 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
23566 ((match_operand:VF_128 1 "vector_operand"
23567 "xBm,<round_constraint>")): Changed to ...
23568 ((match_operand:VF_128 1 "vector_operand"
23569 "xBm,<round_scalar_constraint>")): ... this.
23570 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
23571 %0, %2, %<iptr>1<round_op3>}): Changed to ...
23572 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
23573 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
23574 %<iptr>1<round_scalar_mask_op3>}): ... this.
23575 ((set_attr "prefix" "<round_prefix>")): Changed to ...
23576 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
23578 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
23581 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
23582 Cast vec_cmpeq result to correct type.
23583 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
23584 Cast vec_cmpgt result to correct type.
23586 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
23588 * final.c (final_scan_insn_1): Renamed from...
23589 (final_scan_insn): ... this. New wrapper, to recover
23590 seen from the outermost call in recursive ones.
23591 * config/sparc/sparc.c (output_return): Drop seen from call.
23592 (output_sibcall): Likewise.
23593 * config/visium/visium.c (output_branch): Likewise.
23595 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
23597 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
23600 2018-02-10 Alan Modra <amodra@gmail.com>
23603 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
23604 Specify LR as an input.
23606 2018-02-10 Jakub Jelinek <jakub@redhat.com>
23609 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
23610 remove_member_access_dummy_vars): New functions.
23611 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
23612 lower_omp_1, execute_lower_omp): Use them.
23614 PR rtl-optimization/84308
23615 * shrink-wrap.c (spread_components): Release todo vector.
23617 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
23619 PR rtl-optimization/57193
23620 * ira-color.c (struct allocno_color_data): Add member
23621 conflict_allocno_hard_prefs.
23622 (update_conflict_allocno_hard_prefs): New.
23623 (bucket_allocno_compare_func): Add a preference based on
23624 conflict_allocno_hard_prefs.
23625 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
23626 (color_allocnos): Remove a dead code. Initiate
23627 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
23629 2018-02-09 Jakub Jelinek <jakub@redhat.com>
23632 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
23633 constraint from =wa to wa. Avoid a subreg on the output operand,
23634 instead use a pseudo and subreg it in a move.
23635 (p9_xxbrd_<mode>): Changed to ...
23636 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
23637 (p9_xxbrd_v2df): New expander.
23638 (p9_xxbrw_<mode>): Changed to ...
23639 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
23640 (p9_xxbrw_v4sf): New expander.
23642 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
23644 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
23646 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
23649 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
23650 multiply in 32-bit mode.
23651 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
23652 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
23655 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
23657 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
23658 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
23659 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
23660 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
23662 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
23665 * dwarf2out.c (is_trivial_indirect_ref): New function.
23666 (dwarf2out_late_global_decl): Do not generate a location
23667 attribute for variables that have a non-trivial DECL_VALUE_EXPR
23668 and that are not defined in the current unit.
23670 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
23672 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
23673 instead of a libcall for UNORDERED.
23675 2018-02-09 Tamar Christina <tamar.christina@arm.com>
23678 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
23679 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
23681 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
23684 * config/s390/s390.c (s390_set_current_function): Invoke
23685 s390_indirect_branch_settings also if fndecl didn't change.
23687 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
23689 * config/rs6000/rs6000.md (blockage): Set length to zero.
23691 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
23693 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
23695 2018-02-09 Jakub Jelinek <jakub@redhat.com>
23698 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
23699 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
23703 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
23704 PARALLEL incoming that failed vt_get_decl_and_offset check.
23706 PR middle-end/84237
23707 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
23708 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
23710 (get_variable_section): For decls in named .bss* sections pass true as
23711 second argument to bss_initializer_p.
23713 2018-02-09 Marek Polacek <polacek@redhat.com>
23714 Jakub Jelinek <jakub@redhat.com>
23717 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
23718 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
23719 Sync some changes from cxx_fold_indirect_ref.
23721 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
23723 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
23725 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
23726 (BLOCK_INLINE_ENTRY_LABEL): New.
23727 (dwarf2out_var_location): Disregard inline entry markers.
23728 (inline_entry_data): New struct.
23729 (inline_entry_data_hasher): New hashtable type.
23730 (inline_entry_data_hasher::hash): New.
23731 (inline_entry_data_hasher::equal): New.
23732 (inline_entry_data_table): New variable.
23733 (add_high_low_attributes): Add DW_AT_entry_pc and
23734 DW_AT_GNU_entry_view attributes if a pending entry is found
23735 in inline_entry_data_table. Add old entry_pc attribute only
23736 if debug nonbinding markers are disabled.
23737 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
23738 markers are enabled.
23739 (block_within_block_p, dwarf2out_inline_entry): New.
23740 (dwarf2out_finish): Check that no entries remained in
23741 inline_entry_data_table.
23742 * final.c (reemit_insn_block_notes): Handle inline entry notes.
23743 (final_scan_insn, notice_source_line): Likewise.
23744 (rest_of_clean_state): Skip inline entry markers.
23745 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
23747 * gimple.c (gimple_build_debug_inline_entry): New.
23748 * gimple.h (enum gimple_debug_subcode): Add
23749 GIMPLE_DEBUG_INLINE_ENTRY.
23750 (gimple_build_debug_inline_entry): Declare.
23751 (gimple_debug_inline_entry_p): New.
23752 (gimple_debug_nonbind_marker_p): Adjust.
23753 * insn-notes.def (INLINE_ENTRY): New.
23754 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
23755 inline entry marker notes.
23756 (print_insn): Likewise.
23757 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
23758 (INSN_DEBUG_MARKER_KIND): Likewise.
23759 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
23760 * tree-inline.c (expand_call_inline): Build and insert
23761 debug_inline_entry stmt.
23762 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
23763 inline entry blocks early, if nonbind markers are enabled.
23764 (dump_scope_block): Dump fragment info.
23765 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
23766 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
23767 (gimple_build_debug_inline_entry): New.
23768 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
23769 Enable/disable inline entry points too.
23770 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
23771 (DEBUG_INSN): Describe inline entry markers.
23773 * common.opt (gvariable-location-views): New.
23774 (gvariable-location-views=incompat5): New.
23775 * config.in: Rebuilt.
23776 * configure: Rebuilt.
23777 * configure.ac: Test assembler for view support.
23778 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
23779 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
23780 * dwarf2out.c (var_loc_view): New typedef.
23781 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
23782 (dwarf2out_locviews_in_attribute): New.
23783 (dwarf2out_locviews_in_loclist): New.
23784 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
23785 (enum dw_line_info_opcode): Add LI_adv_address.
23786 (struct dw_line_info_table): Add view.
23787 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
23788 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
23789 (zero_view_p): New variable.
23790 (ZERO_VIEW_P): New macro.
23791 (output_asm_line_debug_info): New.
23792 (struct var_loc_node): Add view.
23793 (add_AT_view_list, AT_loc_list): New.
23794 (add_var_loc_to_decl): Add view param. Test it against last.
23795 (new_loc_list): Add view params. Record them.
23796 (AT_loc_list_ptr): Handle loc and view lists.
23797 (view_list_to_loc_list_val_node): New.
23798 (print_dw_val): Handle dw_val_class_view_list.
23799 (size_of_die): Likewise.
23800 (value_format): Likewise.
23801 (loc_list_has_views): New.
23802 (gen_llsym): Set vl_symbol too.
23803 (maybe_gen_llsym, skip_loc_list_entry): New.
23804 (dwarf2out_maybe_output_loclist_view_pair): New.
23805 (output_loc_list): Output view list or entries too.
23806 (output_view_list_offset): New.
23807 (output_die): Handle dw_val_class_view_list.
23808 (output_dwarf_version): New.
23809 (output_compilation_unit_header): Use it.
23810 (output_skeleton_debug_sections): Likewise.
23811 (output_rnglists, output_line_info): Likewise.
23812 (output_pubnames, output_aranges): Update version comments.
23813 (output_one_line_info_table): Output view numbers in asm comments.
23814 (dw_loc_list): Determine current endview, pass it to new_loc_list.
23815 Call maybe_gen_llsym.
23816 (loc_list_from_tree_1): Adjust.
23817 (add_AT_location_description): Create view list attribute if
23818 needed, check it's absent otherwise.
23819 (convert_cfa_to_fb_loc_list): Adjust.
23820 (maybe_emit_file): Call output_asm_line_debug_info for test.
23821 (dwarf2out_var_location): Reset views as needed. Precompute
23822 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
23823 attribute. Set view.
23824 (new_line_info_table): Reset next view.
23825 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
23826 (dwarf2out_source_line): Likewise. Output view resets and labels to
23827 the assembler, or select appropriate line info opcodes.
23828 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
23829 (optimize_string_length): Catch it. Adjust.
23830 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
23831 dw_val_class_view_list, and remove it if no longer needed.
23832 (hash_loc_list): Hash view numbers.
23833 (loc_list_hasher::equal): Compare them.
23834 (optimize_location_lists): Check whether a view list symbol is
23835 needed, and whether the locview attribute is present, and
23836 whether they match. Remove the locview attribute if no longer
23838 (index_location_lists): Call skip_loc_list_entry for test.
23839 (dwarf2out_finish): Call output_asm_line_debug_info for test.
23840 Use output_dwarf_version.
23841 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
23842 (struct dw_val_node): Add val_view_list.
23843 * final.c (SEEN_NEXT_VIEW): New.
23844 (set_next_view_needed): New.
23845 (clear_next_view_needed): New.
23846 (maybe_output_next_view): New.
23847 (final_start_function): Rename to...
23848 (final_start_function_1): ... this. Take pointer to FIRST,
23849 add SEEN parameter. Emit param bindings in the initial view.
23850 (final_start_function): Reintroduce SEEN-less interface.
23851 (final): Rename to...
23852 (final_1): ... this. Take SEEN parameter. Output final pending
23853 next view at the end.
23854 (final): Reintroduce seen-less interface.
23855 (final_scan_insn): Output pending next view before switching
23856 sections or ending a block. Mark the next view as needed when
23857 outputting variable locations. Notify debug backend of section
23858 changes, and of location view changes.
23859 (rest_of_handle_final): Adjust.
23860 * toplev.c (process_options): Autodetect value for debug variable
23861 location views option. Warn on incompat5 without -gdwarf-5.
23862 * doc/invoke.texi (gvariable-location-views): New.
23863 (gvariable-location-views=incompat5): New.
23864 (gno-variable-location-views): New.
23866 2018-02-08 David Malcolm <dmalcolm@redhat.com>
23868 PR tree-optimization/84136
23869 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
23870 that the result of find_edge is non-NULL.
23872 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
23875 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
23876 storing integer register in SImode. Fix cost of 256 and 512
23877 byte aligned SSE register store.
23879 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
23881 * config/i386/i386.c (ix86_multiplication_cost): Fix
23882 multiplication cost for TARGET_AVX512DQ.
23884 2018-02-08 Marek Polacek <polacek@redhat.com>
23886 PR tree-optimization/84238
23887 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
23890 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
23892 PR tree-optimization/84265
23893 * tree-vect-stmts.c (vectorizable_store): Don't treat
23894 VMAT_CONTIGUOUS accesses as grouped.
23895 (vectorizable_load): Likewise.
23897 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
23899 PR tree-optimization/81635
23900 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
23901 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
23902 (test_round_for_mask): New functions.
23903 (wide_int_cc_tests): Call test_round_for_mask.
23904 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
23905 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
23906 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
23907 range returned by get_range_info.
23909 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
23912 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
23913 * symtab.c: Include builtins.h
23914 (symtab_node::output_to_lto_symbol_table_p): Move here
23915 from lto-streamer-out.c:output_symbol_p.
23916 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
23917 (output_symbol_p): Move all logic to symtab.c
23918 (produce_symtab): Update.
23920 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
23922 * config/s390/s390-opts.h (enum indirect_branch): Define.
23923 * config/s390/s390-protos.h (s390_return_addr_from_memory)
23924 (s390_indirect_branch_via_thunk)
23925 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
23926 (enum s390_indirect_branch_type): Define.
23927 * config/s390/s390.c (struct s390_frame_layout, struct
23928 machine_function): Remove.
23929 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
23930 (indirect_branch_table_label_no, indirect_branch_table_name):
23932 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
23933 (enum s390_indirect_branch_option): Define.
23934 (s390_return_addr_from_memory): New function.
23935 (s390_handle_string_attribute): New function.
23936 (s390_attribute_table): Add new attribute handler.
23937 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
23938 (s390_indirect_branch_via_thunk): New function.
23939 (s390_indirect_branch_via_inline_thunk): New function.
23940 (s390_function_ok_for_sibcall): When jumping via thunk disallow
23941 sibling call optimization for non z10 compiles.
23942 (s390_emit_call): Force indirect branch target to be a single
23943 register. Add r1 clobber for non-z10 compiles.
23944 (s390_emit_epilogue): Emit return jump via return_use expander.
23945 (s390_reorg): Handle JUMP_INSNs as execute targets.
23946 (s390_option_override_internal): Perform validity checks for the
23947 new command line options.
23948 (s390_indirect_branch_attrvalue): New function.
23949 (s390_indirect_branch_settings): New function.
23950 (s390_set_current_function): Invoke s390_indirect_branch_settings.
23951 (s390_output_indirect_thunk_function): New function.
23952 (s390_code_end): Implement target hook.
23953 (s390_case_values_threshold): Implement target hook.
23954 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
23956 * config/s390/s390.h (struct s390_frame_layout)
23957 (struct machine_function): Move here from s390.c.
23958 (TARGET_INDIRECT_BRANCH_NOBP_RET)
23959 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
23960 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
23961 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
23962 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
23963 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
23964 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
23965 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
23966 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
23967 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
23968 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
23969 (mnemonic attribute): Add values which aren't recognized
23971 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
23972 pattern for branch conversion. Fix mnemonic attribute.
23973 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
23974 indirect branch via thunk if requested.
23975 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
23976 ("*indirect_jump"): Disable for branch conversion using out of
23978 ("indirect_jump_via_thunk<mode>_z10")
23979 ("indirect_jump_via_thunk<mode>")
23980 ("indirect_jump_via_inlinethunk<mode>_z10")
23981 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
23982 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
23983 ("casesi_jump_via_inlinethunk<mode>_z10")
23984 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
23985 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
23986 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
23987 ("*indirect2_jump"): Disable for branch conversion.
23988 ("casesi_jump"): Turn into expander and expand patterns for branch
23990 ("return_use"): New expander.
23991 ("*return"): Emit return via thunk and rename it to ...
23992 ("*return<mode>"): ... this one.
23993 * config/s390/s390.opt: Add new options and and enum for the
23996 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
23998 * lra-constraints.c (match_reload): Unconditionally use
23999 gen_lowpart_SUBREG, rather than selecting between that
24000 and equivalent gen_rtx_SUBREG code.
24002 2018-02-08 Richard Biener <rguenther@suse.de>
24004 PR tree-optimization/84233
24005 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
24006 changed flag instead of boguously re-using phi_inserted.
24008 2018-02-08 Martin Jambor <mjambor@suse.cz>
24010 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
24011 static local variables.
24013 2018-02-08 Richard Biener <rguenther@suse.de>
24015 PR tree-optimization/84278
24016 * tree-vect-stmts.c (vectorizable_store): When looking for
24017 smaller vector types to perform grouped strided loads/stores
24018 make sure the mode is supported by the target.
24019 (vectorizable_load): Likewise.
24021 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
24023 * config/aarch64/aarch64.c (aarch64_components_for_bb):
24024 Increase LDP/STP opportunities by adding adjacent callee-saves.
24026 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
24028 PR rtl-optimization/84068
24029 PR rtl-optimization/83459
24030 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
24032 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
24034 PR tree-optimization/84224
24035 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
24036 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
24037 non-zero arguments.
24039 2018-02-07 Iain Sandoe <iain@codesourcery.com>
24042 * config/rs6000/altivec.md (*restore_world): Remove LR use.
24043 * config/rs6000/predicates.md (restore_world_operation): Adjust op
24044 count, remove one USE.
24046 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
24048 * doc/install.texi (Configuration): Document the
24049 --with-long-double-format={ibm,ieee} PowerPC configuration
24053 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
24054 Convert from define_expand to be define_insn_and_split. Rework
24055 float/double/_Float128 conversions to QI/HI/SImode to work with
24056 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
24057 conversions to QI/HImode types did a store and then a load to
24058 truncate the value. For conversions to VSX registers, don't split
24059 the insn, instead emit the code directly. Use the code iterator
24060 any_fix to combine signed and unsigned conversions.
24061 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
24062 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
24063 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
24064 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
24065 (fix_<mode>di2_hw): Likewise.
24066 (fixuns_<mode>di2_hw): Likewise.
24067 (fix_<mode>si2_hw): Likewise.
24068 (fixuns_<mode>si2_hw): Likewise.
24069 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
24070 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
24071 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
24072 fix<uns>_trunc<SFDF:mode>si2_p8.
24073 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
24074 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
24075 (fix<uns>_<mode>_mem): Likewise.
24076 (fctiw<u>z_<mode>_mem): Likewise.
24077 (fix<uns>_<mode>_mem): Likewise.
24078 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
24079 the register allocator from doing a direct move to the GPRs to do
24080 a store, and instead use the ISA 3.0 store byte/half-word from
24081 vector register instruction. For IEEE 128-bit floating point,
24082 also optimize stores of 32-bit ints.
24083 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
24085 2018-02-07 Alan Hayward <alan.hayward@arm.com>
24087 * genextract.c (push_pathstr_operand): New function to support
24089 (walk_rtx): Call push_pathstr_operand.
24090 (print_path): Support [a-zA-Z].
24092 2018-02-07 Richard Biener <rguenther@suse.de>
24094 PR tree-optimization/84037
24095 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
24096 (cse_and_gimplify_to_preheader): Declare.
24097 (vect_get_place_in_interleaving_chain): Likewise.
24098 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
24100 (_loop_vec_info::~_loop_vec_info): Delete it.
24101 (cse_and_gimplify_to_preheader): New function.
24102 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
24103 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
24104 (vectorizable_load): Likewise. For grouped stores always base
24105 the IV on the first element.
24106 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
24107 condition before gimplifying.
24109 2018-02-07 Jakub Jelinek <jakub@redhat.com>
24111 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
24112 *DIV_EXPR and *MOD_EXPR.
24114 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
24117 * config/i386/i386.c (ix86_option_override_internal): Mask out
24118 the CF_SET bit when checking -fcf-protection.
24120 2018-02-07 Tom de Vries <tom@codesourcery.com>
24123 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
24126 2018-02-07 Richard Biener <rguenther@suse.de>
24128 PR tree-optimization/84204
24129 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
24132 PR tree-optimization/84205
24133 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
24134 special-case isl_ast_op_zdiv_r.
24136 PR tree-optimization/84223
24137 * graphite-scop-detection.c (gather_bbs::before_dom_children):
24138 Only add conditions from within the region.
24139 (gather_bbs::after_dom_children): Adjust.
24141 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
24144 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
24145 * config/avr/avr.md: Only post-reload split REG-REG moves if
24146 either register is GENERAL_REG_P.
24148 2018-02-07 Jakub Jelinek <jakub@redhat.com>
24150 PR tree-optimization/84235
24151 * tree-ssa-scopedtables.c
24152 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
24153 if the subtraction is performed in floating point type where NaNs are
24154 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
24155 build 1. Formatting fix.
24157 2018-02-06 Jakub Jelinek <jakub@redhat.com>
24160 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
24161 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
24162 and skip it regardless of bb boundaries. Use CALL_P macro,
24163 don't test INSN_P (insn) together with CALL_P or JUMP_P check
24164 unnecessarily, formatting fix.
24166 2018-02-06 Michael Collison <michael.collison@arm.com>
24168 * config/arm/thumb2.md:
24169 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
24170 (*thumb_mov_notscc): Ditto.
24172 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
24175 * config/rs6000/rs6000.md (su code attribute): Use "u" for
24176 unsigned_fix, not "s".
24178 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
24180 * configure.ac (gcc_fn_eh_frame_ro): New function.
24181 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
24182 correct .eh_frame permissions.
24183 * configure: Regenerate.
24185 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
24187 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
24188 irrelevant options.
24190 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
24192 * config/rs6000/rs6000.c (rs6000_option_override_internal):
24193 Display warning message for -mno-speculate-indirect-jumps.
24195 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
24197 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
24199 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
24201 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
24203 PR tree-optimization/84225
24204 * tree-eh.c (find_trapping_overflow): Only call
24205 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
24207 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
24210 * config/i386/i386.c: Reimplement the check of possible options
24211 -mibt/-mshstk conbination. Change error messages.
24212 * doc/invoke.texi: Fix a typo: remove extra '='.
24214 2018-02-06 Marek Polacek <polacek@redhat.com>
24216 PR tree-optimization/84228
24217 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
24219 2018-02-06 Tamar Christina <tamar.christina@arm.com>
24222 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
24223 emitted arch directives.
24224 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
24225 __ARM_FEATURE_COPROC before changing architectures.
24227 2018-02-06 Richard Biener <rguenther@suse.de>
24229 * config/i386/i386.c (print_reg): Fix typo.
24230 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
24232 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
24234 * configure: Regenerate.
24236 2018-02-05 Martin Sebor <msebor@redhat.com>
24238 PR tree-optimization/83369
24239 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
24242 2018-02-05 Martin Liska <mliska@suse.cz>
24244 * doc/invoke.texi: Cherry-pick upstream r323995.
24246 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
24248 * ira.c (ira_init_register_move_cost): Adjust comment.
24250 2018-02-05 Martin Liska <mliska@suse.cz>
24252 PR gcov-profile/84137
24253 * doc/gcov.texi: Fix typo in documentation.
24255 2018-02-05 Martin Liska <mliska@suse.cz>
24257 PR gcov-profile/83879
24258 * doc/gcov.texi: Document necessity of --dynamic-list-data when
24259 using dlopen functionality.
24261 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
24263 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
24264 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
24265 _mm_maskz_range_ss, _mm_mask_range_round_ss,
24266 _mm_maskz_range_round_ss): New intrinsics.
24267 (__builtin_ia32_rangesd128_round)
24268 (__builtin_ia32_rangess128_round): Remove.
24269 (__builtin_ia32_rangesd128_mask_round,
24270 __builtin_ia32_rangess128_mask_round): New builtins.
24271 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
24272 __builtin_ia32_rangess128_round): Remove.
24273 (__builtin_ia32_rangesd128_mask_round,
24274 __builtin_ia32_rangess128_mask_round): New builtins.
24275 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
24276 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
24277 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
24278 "<round_saeonly_constraint>")): Changed to ...
24279 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
24280 "<round_saeonly_scalar_constraint>")): ... this.
24281 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
24282 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
24283 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
24284 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
24285 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
24287 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
24289 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
24291 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
24292 Remove all values except native, 8540 and 8548.
24294 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
24296 * config/i386/i386.c (ix86_output_function_return): Pass
24297 INVALID_REGNUM, instead of -1, as invalid register number to
24298 indirect_thunk_name and output_indirect_thunk.
24300 2018-02-02 Julia Koval <julia.koval@intel.com>
24302 * config.gcc: Add -march=icelake.
24303 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
24304 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
24305 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
24306 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
24307 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
24308 (processor_target_table): Add icelake.
24309 (ix86_option_override_internal): Handle new PTAs.
24310 (get_builtin_code_for_version): Handle icelake.
24311 (M_INTEL_COREI7_ICELAKE): New.
24312 (fold_builtin_cpu): Handle icelake.
24313 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
24314 * doc/invoke.texi: Add -march=icelake.
24316 2018-02-02 Julia Koval <julia.koval@intel.com>
24318 * config/i386/i386.c (ix86_option_override_internal): Change flags type
24319 to wide_int_bitmask.
24320 * wide-int-bitmask.h: New.
24322 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
24325 * config/i386/i386.md: Replace Pmode with word_mode in
24326 builtin_setjmp_setup and builtin_longjmp to support x32.
24328 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
24332 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
24334 (rs6000_supported_cpu_names): New static variable.
24335 (linux_cpu_translation_table): Likewise.
24336 (elf_platform) <cpu>: Define new static variable and use it.
24337 Translate kernel AT_PLATFORM name to canonical name if needed.
24338 Error if platform name is unknown.
24340 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
24343 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
24345 2018-02-01 Jeff Law <law@redhat.com>
24348 * config/i386/i386.c (release_scratch_register_on_entry): Add new
24349 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
24350 the scratch if RELEASE_VIA_POP is false.
24351 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
24352 If we have to save a temporary register, decrement SIZE appropriately.
24353 Pass new arguments to release_scratch_register_on_entry.
24354 (ix86_adjust_stack_and_probe): Likewise.
24355 (ix86_emit_probe_stack_range): Pass new arguments to
24356 release_scratch_register_on_entry.
24358 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
24360 PR rtl-optimization/84157
24361 * combine.c (change_zero_ext): Use REG_P predicate in
24362 front of HARD_REGISTER_P predicate.
24364 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
24366 * config/avr/avr.c (avr_option_override): Move disabling of
24367 -fdelete-null-pointer-checks to...
24368 * common/config/avr/avr-common.c (avr_option_optimization_table):
24371 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
24373 PR tree-optimization/81635
24374 * tree-data-ref.c (split_constant_offset_1): For types that
24375 wrap on overflow, try to use range info to prove that wrapping
24378 2018-02-01 Renlin Li <renlin.li@arm.com>
24381 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
24382 TAILCALL_ADDR_REGS.
24383 (aarch64_register_move_cost): Likewise.
24384 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
24385 TAILCALL_ADDR_REGS.
24386 (REG_CLASS_NAMES): Likewise.
24387 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
24388 TAILCALL_ADDR_REGS. Remove IP registers.
24389 * config/aarch64/aarch64.md (Ucs): Update register constraint.
24391 2018-02-01 Richard Biener <rguenther@suse.de>
24393 * domwalk.h (dom_walker::dom_walker): Add additional constructor
24394 for specifying RPO order and allow NULL for that.
24395 * domwalk.c (dom_walker::dom_walker): Likewise.
24396 (dom_walker::walk): Handle NULL RPO order.
24397 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
24399 (rewrite_update_dom_walker): Likewise.
24400 (mark_def_dom_walker): Likewise.
24402 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
24404 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
24405 (aarch64_maybe_expand_sve_subreg_move): Declare.
24406 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
24407 * config/aarch64/predicates.md (aarch64_any_register_operand): New
24409 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
24410 that are semantically a reverse operation.
24411 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
24412 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
24413 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
24415 (aarch64_can_change_mode_class): For big-endian, forbid changes
24416 between two SVE modes if they have different element sizes.
24418 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
24420 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
24421 the TImode handling for big-endian targets.
24423 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
24425 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
24426 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
24428 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
24429 Remove BSWAP handing for big-endian targets and use the form of
24430 LD1RQ appropariate for the mode.
24432 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
24434 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
24435 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
24436 duplicated element.
24438 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
24441 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
24442 check for operands that need to go through aarch64_sve_reload_be.
24444 2018-02-01 Jakub Jelinek <jakub@redhat.com>
24446 PR tree-optimization/81661
24447 PR tree-optimization/84117
24448 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
24449 * tree-eh.c: Include gimplify.h.
24450 (find_trapping_overflow, replace_trapping_overflow,
24451 rewrite_to_non_trapping_overflow): New functions.
24452 * tree-vect-loop.c: Include tree-eh.h.
24453 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
24454 * tree-data-ref.c: Include tree-eh.h.
24455 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
24457 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
24459 PR rtl-optimization/84123
24460 * combine.c (change_zero_ext): Check if hard register satisfies
24461 can_change_dest_mode before calling gen_lowpart_SUBREG.
24463 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
24466 * ira.c (ira_init_register_move_cost): Remove assert.
24468 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
24470 PR rtl-optimization/84071
24471 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
24472 * doc/tm.texi: Regenerate.
24474 2018-01-31 Richard Biener <rguenther@suse.de>
24476 PR tree-optimization/84132
24477 * tree-data-ref.c (analyze_miv_subscript): Properly
24478 check whether evolution_function_is_affine_multivariate_p
24479 before calling gcd_of_steps_may_divide_p.
24481 2018-01-31 Julia Koval <julia.koval@intel.com>
24484 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
24485 * config/i386/i386.md (rdpid_rex64) New.
24486 (rdpid): Make 32bit only.
24488 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
24491 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
24492 an IDENTIFIER_NODE for FUNCTION_TYPE's.
24494 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
24497 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
24499 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
24501 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
24503 PR rtl-optimization/84071
24504 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
24505 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
24507 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
24509 * config/arc/arc.c (arc_handle_aux_attribute): New function.
24510 (arc_attribute_table): Add 'aux' attribute.
24511 (arc_in_small_data_p): Consider aux like variables.
24512 (arc_is_aux_reg_p): New function.
24513 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
24514 (arc_get_aux_arg): New function.
24515 (prepare_move_operands): Handle aux-register access.
24516 (arc_handle_aux_attribute): New function.
24517 * doc/extend.texi (ARC Variable attributes): Add subsection.
24519 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
24521 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
24522 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
24523 (arc_attribute_table): Add 'uncached' attribute.
24524 (arc_print_operand): Print '.di' flag for uncached memory
24526 (arc_in_small_data_p): Do not consider for small data the uncached
24528 (arc_is_uncached_mem_p): New function.
24529 * config/arc/predicates.md (compact_store_memory_operand): Check
24530 for uncached memory accesses.
24531 (nonvol_nonimm_operand): Likewise.
24532 * doc/extend.texi (ARC Type Attribute): New subsection.
24534 2018-01-31 Jakub Jelinek <jakub@redhat.com>
24537 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
24538 falign-loops=): Add Optimization flag.
24540 2018-01-30 Jeff Law <law@redhat.com>
24543 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
24544 INT_REGISTERS_SAVED. Check it prior to calling
24545 get_scratch_register_on_entry.
24546 (ix86_adjust_stack_and_probe): Similarly.
24547 (ix86_emit_probe_stack_range): Similarly.
24548 (ix86_expand_prologue): Corresponding changes.
24550 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
24553 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
24554 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
24556 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
24559 * lra-constraints.c (curr_insn_transform): Process AND in the
24562 2018-01-30 Jakub Jelinek <jakub@redhat.com>
24564 PR rtl-optimization/83986
24565 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
24566 dependence against last_pending_memory_flush in addition to
24567 pending_jump_insns.
24569 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
24571 PR tree-optimization/81611
24572 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
24575 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
24578 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
24581 2018-01-30 Richard Biener <rguenther@suse.de>
24582 Jakub Jelinek <jakub@redhat.com>
24584 PR tree-optimization/84111
24585 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
24586 inner loops added during recursion, as they don't have up-to-date
24589 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
24592 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
24593 (can_inline_edge_by_limits_p): ... here.
24594 (can_early_inline_edge_p, check_callers,
24595 update_caller_keys, update_callee_keys, recursive_inlining,
24596 add_new_edges_to_heap, speculation_useful_p,
24597 inline_small_functions,
24598 inline_small_functions, flatten_function,
24599 inline_to_all_callers_1): Update.
24601 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
24603 * profile-count.c (profile_count::combine_with_ipa_count): Handle
24606 2018-01-30 Richard Biener <rguenther@suse.de>
24608 PR tree-optimization/83008
24609 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
24610 invariant and constant vector uses in stmts when they need
24611 more than one stmt.
24613 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
24616 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
24617 * configure: Regenerate.
24619 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
24621 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
24623 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
24624 Use gen_rtx_REG rather than gen_lowpart.
24626 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
24628 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
24629 rather than 0 when creating partial subregs.
24631 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
24633 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
24636 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
24639 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
24640 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
24641 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
24642 flags. This restores the settings used before the 2017-07-24.
24643 Turning off pre increment/decrement/modify allows IVOPTS to
24644 optimize DF/SF loops where the index is an int.
24646 2018-01-29 Richard Biener <rguenther@suse.de>
24647 Kelvin Nilsen <kelvin@gcc.gnu.org>
24650 * tree-vect-stmts.c (vectorizable_call): Don't call
24651 targetm.vectorize_builtin_md_vectorized_function if callee is
24654 2018-01-22 Carl Love <cel@us.ibm.com>
24656 * doc/extend.tex: Fix typo in second arg in
24657 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
24659 2018-01-29 Richard Biener <rguenther@suse.de>
24661 PR tree-optimization/84086
24662 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
24663 (flush_ssaname_freelist): When SSA names were released reset
24664 the SCEV hash table.
24666 2018-01-29 Richard Biener <rguenther@suse.de>
24668 PR tree-optimization/84057
24669 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
24670 removed paths when removing edges.
24672 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
24674 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
24675 -mfunction-return=@var{choice}.
24677 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
24679 PR diagnostic/84034
24680 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
24681 Handle CR like TAB.
24682 (layout::print_source_line): Likewise.
24683 (test_get_line_width_without_trailing_whitespace): Add test cases.
24685 2018-01-27 Jakub Jelinek <jakub@redhat.com>
24687 PR middle-end/84040
24688 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
24691 2018-01-26 Jim Wilson <jimw@sifive.com>
24693 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
24695 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
24698 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24700 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
24701 and CMP + SUB-immediate -> SUBS.
24703 2018-01-26 Martin Sebor <msebor@redhat.com>
24705 PR tree-optimization/83896
24706 * tree-ssa-strlen.c (get_string_len): Rename...
24707 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
24708 Avoid assuming length is constant.
24709 (handle_char_store): Use HOST_WIDE_INT for string length.
24711 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
24714 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
24715 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
24717 2018-01-26 Richard Biener <rguenther@suse.de>
24719 PR rtl-optimization/84003
24720 * dse.c (record_store): Only record redundant stores when
24721 the earlier store aliases at least all accesses the later one does.
24723 2018-01-26 Jakub Jelinek <jakub@redhat.com>
24725 PR rtl-optimization/83985
24726 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
24727 REG_CFA_RESTORE insns.
24728 (delete_unmarked_insns): Don't ignore separate shrink wrapping
24729 REG_CFA_RESTORE insns here.
24732 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
24733 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
24735 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
24737 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
24738 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
24739 (arc_init): Likewise.
24740 (arc_override_options): Likewise.
24741 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
24743 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
24745 * config/arc/arc.h (TARGET_DBNZ): Define.
24746 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
24747 properly set the tune attribute.
24748 (dbnz): Use TARGET_DBNZ guard.
24749 * config/arc/arc.opt (mtune): Add core3 option.
24751 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
24753 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
24754 recognize new pic like addresses.
24755 (arc_delegitimize_address): Clean up.
24757 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
24759 * config/arc/arc-arches.def: Option mrf16 valid for all
24761 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
24762 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
24763 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
24764 * config/arc/arc-tables.opt: Regenerate.
24765 * config/arc/arc.c (arc_conditional_register_usage): Handle
24766 reduced register file case.
24767 (arc_file_start): Set must have build attributes.
24768 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
24769 mrf16 option value.
24770 * config/arc/arc.opt (mrf16): Add new option.
24771 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
24772 * config/arc/genmultilib.awk: Handle new mrf16 option.
24773 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
24774 * config/arc/t-multilib: Regenerate.
24775 * doc/invoke.texi (ARC Options): Document mrf16 option.
24777 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
24779 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
24780 * config/arc/arc.c (arc_handle_secure_attribute): New function.
24781 (arc_attribute_table): Add 'secure_call' attribute.
24782 (arc_print_operand): Print secure call operand.
24783 (arc_function_ok_for_sibcall): Don't optimize tail calls when
24785 (arc_is_secure_call_p): New function. * config/arc/arc.md
24786 (call_i): Add support for sjli instruction.
24787 (call_value_i): Likewise.
24788 * config/arc/constraints.md (Csc): New constraint.
24790 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
24791 John Eric Martin <John.Martin@emmicro-us.com>
24793 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
24794 * config/arc/arc.c (_arc_jli_section): New struct.
24795 (arc_jli_section): New type.
24796 (rc_jli_sections): New static variable.
24797 (arc_handle_jli_attribute): New function.
24798 (arc_attribute_table): Add jli_always and jli_fixed attribute.
24799 (arc_file_end): New function.
24800 (TARGET_ASM_FILE_END): Define.
24801 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
24802 (arc_add_jli_section): New function.
24803 (jli_call_scan): Likewise.
24804 (arc_reorg): Call jli_call_scan.
24805 (arc_output_addsi): Remove 'S' from printing asm operand.
24806 (arc_is_jli_call_p): New function.
24807 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
24809 (movhi_insn): Likewise.
24810 (movsi_insn): Likewise.
24811 (movsi_set_cc_insn): Likewise.
24812 (loadqi_update): Likewise.
24813 (load_zeroextendqisi_update): Likewise.
24814 (load_signextendqisi_update): Likewise.
24815 (loadhi_update): Likewise.
24816 (load_zeroextendhisi_update): Likewise.
24817 (load_signextendhisi_update): Likewise.
24818 (loadsi_update): Likewise.
24819 (loadsf_update): Likewise.
24820 (movsicc_insn): Likewise.
24821 (bset_insn): Likewise.
24822 (bxor_insn): Likewise.
24823 (bclr_insn): Likewise.
24824 (bmsk_insn): Likewise.
24825 (bicsi3_insn): Likewise.
24826 (cmpsi_cc_c_insn): Likewise.
24827 (movsi_ne): Likewise.
24828 (movsi_cond_exec): Likewise.
24829 (clrsbsi2): Likewise.
24830 (norm_f): Likewise.
24840 (call_i): Remove 'S' asm letter, add jli instruction.
24841 (call_value_i): Likewise.
24842 * config/arc/arc.op (mjli-always): New option.
24843 * config/arc/constraints.md (Cji): New constraint.
24844 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
24846 (subsf3_fpx): Likewise.
24847 (mulsf3_fpx): Likewise.
24848 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
24850 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
24851 function attrbutes.
24852 * doc/invoke.texi (ARC): Document mjli-always option.
24854 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
24856 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
24857 avoid addition with 0 and use incw and decw where possible.
24859 2018-01-26 Richard Biener <rguenther@suse.de>
24861 PR tree-optimization/81082
24862 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
24863 association if it requires casting to unsigned.
24864 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
24865 from fold_plusminus_mult_expr to catch important cases late when
24866 range info is available.
24868 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
24870 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
24871 * configure.ac (hidden_linkonce): New test.
24872 * configure: Regenerate.
24873 * config.in: Regenerate.
24875 2018-01-26 Julia Koval <julia.koval@intel.com>
24877 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
24878 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
24879 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
24880 _mm_mask_bitshuffle_epi64_mask): Fix type.
24881 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
24882 USI_FTYPE_V4DI_V4DI_USI): Remove.
24883 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
24884 __builtin_ia32_vpshufbitqmb256_mask,
24885 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
24886 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
24887 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
24889 2018-01-26 Alan Modra <amodra@gmail.com>
24892 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
24893 UNSPEC_VBPERMQ. Sort other unspecs.
24895 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
24897 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
24899 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
24901 PR middle-end/83055
24902 * predict.c (drop_profile): Do not push/pop cfun; update also
24904 (handle_missing_profiles): Fix logic looking for zero profiles.
24906 2018-01-25 Jakub Jelinek <jakub@redhat.com>
24908 PR middle-end/83977
24909 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
24910 on functions with #pragma omp declare simd or functions with simd
24912 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
24913 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
24914 Remove trailing \n from warning_at calls.
24916 2018-01-25 Tom de Vries <tom@codesourcery.com>
24919 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
24920 for neutered workers.
24922 2018-01-24 Joseph Myers <joseph@codesourcery.com>
24925 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
24926 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
24928 2018-01-24 Jeff Law <law@redhat.com>
24931 * i386.c (get_probe_interval): Move to earlier point.
24932 (ix86_compute_frame_layout): If -fstack-clash-protection and
24933 the frame is larger than the probe interval, then use pushes
24934 to save registers rather than reg->mem moves.
24935 (ix86_expand_prologue): Remove conditional for int_registers_saved
24938 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
24941 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
24942 min/max for never referenced object.
24944 2018-01-24 Jakub Jelinek <jakub@redhat.com>
24946 PR middle-end/83977
24947 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
24949 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
24950 attributes from DECL_ATTRIBUTES (decl) without affecting
24951 DECL_ATTRIBUTES (current_function_decl).
24952 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
24953 functions with non-NULL DECL_ABSTRACT_ORIGIN.
24955 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
24957 PR tree-optimization/83979
24958 * fold-const.c (fold_comparison): Use constant_boolean_node
24959 instead of boolean_{true,false}_node.
24961 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
24963 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
24966 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
24968 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
24969 Simplify the clause that sets the length attribute.
24970 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
24971 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
24972 clause that sets the length attribute.
24973 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
24975 2018-01-24 Tom de Vries <tom@codesourcery.com>
24978 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
24979 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
24980 Add strict parameter.
24981 (prevent_branch_around_nothing): Insert dummy insn between branch to
24982 label and label with no ptx insn inbetween.
24983 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
24985 2018-01-24 Tom de Vries <tom@codesourcery.com>
24988 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
24989 for neutered threads in warp.
24990 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
24992 2018-01-24 Richard Biener <rguenther@suse.de>
24994 PR tree-optimization/83176
24995 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
24998 2018-01-24 Richard Biener <rguenther@suse.de>
25000 PR tree-optimization/82819
25001 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
25002 code generating pluses that are no-ops in the target precision.
25004 2018-01-24 Richard Biener <rguenther@suse.de>
25006 PR middle-end/84000
25007 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
25009 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
25011 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
25012 to merge probabilities.
25013 * predict.c (probably_never_executed): Also mark as cold functions
25014 with global 0 profile and guessed local profile.
25015 * profile-count.c (profile_probability::combine_with_count): New
25017 * profile-count.h (profile_probability::operator*,
25018 profile_probability::operator*=, profile_probability::operator/,
25019 profile_probability::operator/=): Reduce precision to adjusted
25020 and set value to guessed on contradictory divisions.
25021 (profile_probability::combine_with_freq): Remove.
25022 (profile_probability::combine_wiht_count): Declare.
25023 (profile_count::force_nonzero):: Set to adjusted.
25024 (profile_count::probability_in):: Set quality to adjusted.
25025 * tree-ssa-tail-merge.c (replace_block_by): Use
25026 combine_with_count.
25028 2018-01-23 Andrew Waterman <andrew@sifive.com>
25029 Jim Wilson <jimw@sifive.com>
25031 * config/riscv/riscv.c (riscv_stack_boundary): New.
25032 (riscv_option_override): Set riscv_stack_boundary. Handle
25033 riscv_preferred_stack_boundary_arg.
25034 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
25035 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
25036 (STACK_BOUNDARY): Set to riscv_stack_boundary.
25037 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
25038 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
25039 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
25041 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
25044 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
25045 of struct ix86_frame.
25046 (ix86_expand_epilogue): Likewise. Add a local variable for
25047 the reg_save_offset field in struct ix86_frame.
25049 2018-01-23 Bin Cheng <bin.cheng@arm.com>
25051 PR tree-optimization/82604
25052 * tree-loop-distribution.c (enum partition_kind): New enum item
25053 PKIND_PARTIAL_MEMSET.
25054 (partition_builtin_p): Support above new enum item.
25055 (generate_code_for_partition): Ditto.
25056 (compute_access_range): Differentiate cases that equality can be
25057 proven at all loops, the innermost loops or no loops.
25058 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
25059 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
25060 (finalize_partitions, distribute_loop): Don't fuse partition of
25061 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
25062 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
25063 parloop is enabled.
25065 2018-01-23 Martin Liska <mliska@suse.cz>
25067 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
25068 order to ignore the predictor.
25069 (PRED_POLYMORPHIC_CALL): Likewise.
25070 (PRED_RECURSIVE_CALL): Likewise.
25072 2018-01-23 Martin Liska <mliska@suse.cz>
25074 * tree-profile.c (tree_profiling): Print function header to
25075 aware reader which function we are working on.
25076 * value-prof.c (gimple_find_values_to_profile): Do not print
25077 not interesting value histograms.
25079 2018-01-23 Martin Liska <mliska@suse.cz>
25081 * profile-count.h (enum profile_quality): Add
25082 profile_uninitialized as the first value. Do not number values
25083 as they are zero based.
25084 (profile_count::verify): Update sanity check.
25085 (profile_probability::verify): Likewise.
25087 2018-01-23 Nathan Sidwell <nathan@acm.org>
25089 * doc/invoke.texi (ffor-scope): Deprecate.
25091 2018-01-23 David Malcolm <dmalcolm@redhat.com>
25093 PR tree-optimization/83510
25094 * domwalk.c (set_all_edges_as_executable): New function.
25095 (dom_walker::dom_walker): Convert bool param
25096 "skip_unreachable_blocks" to enum reachability. Move setup of
25097 edge flags to set_all_edges_as_executable and only do it when
25098 reachability is REACHABLE_BLOCKS.
25099 * domwalk.h (enum dom_walker::reachability): New enum.
25100 (dom_walker::dom_walker): Convert bool param
25101 "skip_unreachable_blocks" to enum reachability.
25102 (set_all_edges_as_executable): New decl.
25103 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
25104 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
25106 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
25107 but converting true to REACHABLE_BLOCKS.
25108 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
25110 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
25111 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
25112 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
25114 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
25115 if check_all_array_refs will be called.
25117 2018-01-23 David Malcolm <dmalcolm@redhat.com>
25119 * tree.c (selftest::test_location_wrappers): Add more test
25122 2018-01-23 David Malcolm <dmalcolm@redhat.com>
25124 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
25125 (selftest::test_bit_in_range): Likewise.
25127 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
25130 * doc/sourcebuild.texi (vect_float): Say that the selector
25131 only describes the situation when -funsafe-math-optimizations is on.
25132 (vect_float_strict): Document.
25134 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
25136 PR tree-optimization/83965
25137 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
25138 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
25139 instead of checking only for a reduction.
25140 (vect_recog_widen_sum_pattern): Likewise.
25142 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
25144 * predict.c (probably_never_executed): Only use precise profile info.
25145 (compute_function_frequency): Skip after inlining hack since we now
25146 have quality checking.
25148 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
25150 * profile-count.h (profile_probability::very_unlikely,
25151 profile_probability::unlikely, profile_probability::even): Set
25152 precision to guessed.
25154 2018-01-23 Richard Biener <rguenther@suse.de>
25156 PR tree-optimization/83963
25157 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
25158 Properly terminate dominator walk when crossing the exit edge not
25159 when visiting its source block.
25161 2018-01-23 Jakub Jelinek <jakub@redhat.com>
25164 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
25165 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
25167 2018-01-22 Jakub Jelinek <jakub@redhat.com>
25169 PR tree-optimization/83957
25170 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
25171 semicolon after for body surrounded by braces.
25173 PR tree-optimization/83081
25174 * profile-count.h (profile_probability::split): New method.
25175 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
25176 Use profile_probability::split.
25177 (do_compare_rtx_and_jump): Fix adjustment of probabilities
25178 when splitting a single conditional jump into 2.
25180 2018-01-22 David Malcolm <dmalcolm@redhat.com>
25182 PR tree-optimization/69452
25183 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
25186 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
25188 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
25189 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
25190 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
25192 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
25194 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
25196 * config/rl78/rl78.md (movdi): New define_expand.
25197 * config/rl78/rl78.c (rl78_split_movdi): New function.
25199 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
25202 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
25204 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
25205 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
25206 128-bit to produce an UNSPEC move to get the double word with the
25207 signbit and then a shift directly to do signbit.
25208 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
25209 implementation with a new version that just does either a direct
25210 move or a regular move. Move memory interface to separate insns.
25211 Move insns so they are next to the expander.
25212 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
25213 with signbit move. Split big and little endian case.
25214 (signbit<mode>2_dm_mem_le): Likewise.
25215 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
25216 (signbit<mode>2_dm2): Likewise.
25218 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
25220 * config/rl78/rl78.md (anddi3): New define_expand.
25222 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
25224 * config/rl78/rl78.md (umindi3): New define_expand.
25226 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
25228 * config/rl78/rl78.md (smindi3): New define_expand.
25230 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
25232 * config/rl78/rl78.md (smaxdi3): New define_expand.
25234 2018-01-22 Carl Love <cel@us.ibm.com>
25236 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
25237 LVX_V1TI): Add macro expansion.
25238 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
25239 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
25240 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
25241 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
25242 Change check to determine if the instruction is a byte reversing
25243 entry. Fix typo in comment.
25244 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
25245 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
25246 Add def_builtin calls for new builtins.
25247 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
25248 Add define_insn expansion.
25250 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
25252 * config/rl78/rl78.md (umaxdi3): New define_expand.
25254 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
25256 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
25257 for non-QImode registers.
25259 2018-01-22 Richard Biener <rguenther@suse.de>
25261 PR tree-optimization/83963
25262 * graphite-scop-detection.c (scop_detection::get_sese): Delay
25263 including the loop exit block.
25264 (scop_detection::merge_sese): Likewise.
25265 (scop_detection::add_scop): Do it here instead.
25267 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25269 * doc/sourcebuild.texi (arm_softfloat): Document.
25271 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
25274 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
25275 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
25276 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
25278 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
25279 David Edelsohn <dje.gcc@gmail.com>
25282 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
25283 Change "crset eq" to "crset 2".
25284 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
25285 (*call_indirect_aix<mode>_nospec): Likewise.
25286 (*call_value_indirect_aix<mode>_nospec): Likewise.
25287 (*call_indirect_elfv2<mode>_nospec): Likewise.
25288 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
25289 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
25290 change assembly output from . to $.
25291 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
25292 (indirect_jump<mode>_nospec): Change assembly output from . to $.
25293 (*tablejump<mode>_internal1_nospec): Likewise.
25295 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
25298 * config/sh/sh_optimize_sett_clrt.cc:
25299 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
25301 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
25303 PR tree-optimization/83940
25304 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
25305 offset_dt to vect_constant_def rather than vect_unknown_def_type.
25306 (vect_check_load_store_mask): Add a mask_dt_out parameter and
25307 use it to pass back the definition type.
25308 (vect_check_store_rhs): Likewise rhs_dt_out.
25309 (vect_build_gather_load_calls): Add a mask_dt argument and use
25310 it instead of a call to vect_is_simple_use.
25311 (vectorizable_store): Update calls to vect_check_load_store_mask
25312 and vect_check_store_rhs. Use the dt returned by the latter instead
25313 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
25314 instead of calls to vect_is_simple_use. Pass the scalar rather
25315 than the vector operand to vect_is_simple_use when handling
25316 second and subsequent copies of an rhs value.
25317 (vectorizable_load): Update calls to vect_check_load_store_mask
25318 and vect_build_gather_load_calls. Use the cached mask_dt and
25319 gs_info.offset_dt instead of calls to vect_is_simple_use.
25321 2018-01-20 Jakub Jelinek <jakub@redhat.com>
25323 PR middle-end/83945
25324 * tree-emutls.c: Include gimplify.h.
25325 (lower_emutls_2): New function.
25326 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
25327 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
25328 it before further processing.
25331 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
25332 UINTVAL (trueop1) instead of INTVAL (op1).
25334 2018-01-19 Jakub Jelinek <jakub@redhat.com>
25338 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
25339 INCOMING_FRAME_SP_OFFSET if not defined.
25340 (scan_trace): Add ENTRY argument. If true and
25341 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
25342 emit a note to adjust the CFA offset.
25343 (create_cfi_notes): Adjust scan_trace callers.
25344 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
25345 INCOMING_FRAME_SP_OFFSET in the CIE.
25346 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
25347 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
25349 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
25350 * doc/tm.texi: Regenerated.
25352 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
25354 PR rtl-optimization/83147
25355 * lra-constraints.c (remove_inheritance_pseudos): Use
25356 lra_substitute_pseudo_within_insn.
25358 2018-01-19 Tom de Vries <tom@codesourcery.com>
25359 Cesar Philippidis <cesar@codesourcery.com>
25362 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
25364 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
25367 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
25368 spaces for function labels.
25370 2018-01-19 Martin Liska <mliska@suse.cz>
25372 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
25373 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
25374 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
25375 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
25376 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
25377 (PRED_CONST_RETURN): Change from 69 to 65.
25378 (PRED_NULL_RETURN): Change from 91 to 71.
25379 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
25380 (PRED_LOOP_GUARD): Change from 66 to 73.
25382 2018-01-19 Martin Liska <mliska@suse.cz>
25384 * predict.c (predict_insn_def): Add new assert.
25385 (struct branch_predictor): Change type to signed integer.
25386 (test_prediction_value_range): Amend test to cover
25387 PROB_UNINITIALIZED.
25388 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
25389 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
25390 (PRED_LOOP_ITERATIONS_MAX): Likewise.
25391 (PRED_LOOP_IV_COMPARE): Likewise.
25392 * predict.h (PROB_UNINITIALIZED): Define new constant.
25394 2018-01-19 Martin Liska <mliska@suse.cz>
25396 * predict.c (dump_prediction): Add new format for
25397 analyze_brprob.py script which is enabled with -details
25399 * profile-count.h (precise_p): New function.
25401 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
25403 PR tree-optimization/83922
25404 * tree-vect-loop.c (vect_verify_full_masking): Return false if
25405 there are no statements that need masking.
25406 (vect_active_double_reduction_p): New function.
25407 (vect_analyze_loop_operations): Use it when handling phis that
25408 are not in the loop header.
25410 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
25412 PR tree-optimization/83914
25413 * tree-vect-loop.c (vectorizable_induction): Don't convert
25414 init_expr or apply the peeling adjustment for inductions
25415 that are nested within the vectorized loop.
25417 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25419 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
25422 2018-01-18 Jakub Jelinek <jakub@redhat.com>
25426 * function.h (gimplify_parameters): Add gimple_seq * argument.
25427 * function.c: Include gimple.h and options.h.
25428 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
25429 for the added local temporaries if needed.
25430 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
25431 if there are any parameter cleanups, wrap whole body into a
25432 try/finally with the cleanups.
25434 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
25437 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
25438 Use GET_MODE_CLASS for scalar floating point.
25440 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
25444 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
25445 Fix call of call_cgraph_insertion_hooks.
25447 2018-01-18 Martin Sebor <msebor@redhat.com>
25449 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
25451 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
25454 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
25457 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
25460 * common.opt: (-ffile-prefix-map): New option.
25461 * opts.c (common_handle_option): Defer it.
25462 * opts-global.c (handle_common_deferred_options): Handle it.
25463 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
25464 * file-prefix-map.h: New file.
25465 (remap_debug_filename, add_debug_prefix_map): ...here.
25466 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
25467 * final.c (debug_prefix_map, add_debug_prefix_map
25468 remap_debug_filename): Move to...
25469 * file-prefix-map.c: New file.
25470 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
25471 generalize, get rid of alloca(), use strrchr() instead of strchr().
25472 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
25473 Implement in terms of add_prefix_map().
25474 (remap_macro_filename, remap_debug_filename): Implement in term of
25476 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
25477 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
25478 * dbxout.c: Include file-prefix-map.h.
25479 * varasm.c: Likewise.
25480 * vmsdbgout.c: Likewise.
25481 * xcoffout.c: Likewise.
25482 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
25483 * doc/cppopts.texi (-fmacro-prefix-map): Document.
25484 * doc/invoke.texi (-ffile-prefix-map): Document.
25485 (-fdebug-prefix-map): Update description.
25487 2018-01-18 Martin Liska <mliska@suse.cz>
25489 * config/i386/i386.c (indirect_thunk_name): Document that also
25491 (output_indirect_thunk): Document why both instructions
25492 (pause and lfence) are generated.
25494 2018-01-18 Richard Biener <rguenther@suse.de>
25496 PR tree-optimization/83887
25497 * graphite-scop-detection.c
25498 (scop_detection::get_nearest_dom_with_single_entry): Remove.
25499 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
25500 (scop_detection::merge_sese): Re-implement with a flood-fill
25501 algorithm that properly finds a SESE region if it exists.
25503 2018-01-18 Jakub Jelinek <jakub@redhat.com>
25506 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
25507 pointer_diff optimizations use view_convert instead of convert.
25509 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
25511 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
25512 Generate different code for -mno-speculate-indirect-jumps.
25513 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
25514 (*call_indirect_aix<mode>): Disable for
25515 -mno-speculate-indirect-jumps.
25516 (*call_indirect_aix<mode>_nospec): New define_insn.
25517 (*call_value_indirect_aix<mode>): Disable for
25518 -mno-speculate-indirect-jumps.
25519 (*call_value_indirect_aix<mode>_nospec): New define_insn.
25520 (*sibcall_nonlocal_sysv<mode>): Generate different code for
25521 -mno-speculate-indirect-jumps.
25522 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
25524 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
25526 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
25527 long double type, set the flags for noting the default long double
25528 type, even if we don't pass or return a long double type.
25530 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
25533 * ipa-inline.c (flatten_function): Do not overwrite final inlining
25536 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
25538 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
25539 support for merge[hl].
25540 (fold_mergehl_helper): New helper function.
25541 (tree-vector-builder.h): New #include for tree_vector_builder usage.
25542 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
25543 (altivec_vmrglw_direct): Add xxmrglw insn.
25545 2018-01-17 Andrew Waterman <andrew@sifive.com>
25547 * config/riscv/riscv.c (riscv_conditional_register_usage): If
25548 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
25550 2018-01-17 David Malcolm <dmalcolm@redhat.com>
25553 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
25554 call the lto_location_cache before reading the
25555 DECL_SOURCE_LOCATION of the types.
25557 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
25558 Richard Sandiford <richard.sandiford@linaro.org>
25560 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
25561 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
25562 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
25563 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
25564 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
25566 * config/aarch64/constraints.md (aarch64_movti_operand):
25568 * config/aarch64/predicates.md (Uti): Add new constraint.
25570 2018-01-17 Carl Love <cel@us.ibm.com>
25572 * config/rs6000/vsx.md (define_expand xl_len_r,
25573 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
25574 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
25576 (define_expand, define_insn): Move the shift left from the
25577 define_insn to the define_expand for lxvl and stxvl instructions.
25578 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
25579 and XL_LEN_R definitions to PURE.
25581 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
25583 * config/i386/i386.c (indirect_thunk_name): Declare regno
25584 as unsigned int. Compare regno with INVALID_REGNUM.
25585 (output_indirect_thunk): Ditto.
25586 (output_indirect_thunk_function): Ditto.
25587 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
25588 in the call to output_indirect_thunk_function.
25590 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
25592 PR middle-end/83884
25593 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
25594 rather than the size of inner_type to determine the stack slot size
25595 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
25597 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
25600 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
25603 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
25605 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
25606 endian Linux systems to optionally enable multilibs for selecting
25607 the long double type if the user configured an explicit type.
25608 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
25609 have no long double multilibs if not defined.
25610 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
25611 warn if the user used -mabi={ieee,ibm}longdouble and we built
25612 multilibs for long double.
25613 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
25614 appropriate multilib option.
25615 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
25617 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
25618 for building long double multilibs.
25619 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
25621 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
25623 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
25626 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
25628 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
25631 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
25634 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
25637 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
25639 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
25640 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
25642 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
25644 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
25645 different rtl trees depending on TARGET_64BIT.
25646 (rs6000_gen_lvx): Likewise.
25648 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
25650 * config/visium/visium.md (nop): Tweak comment.
25651 (hazard_nop): Likewise.
25653 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
25655 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
25656 -mspeculate-indirect-jumps.
25657 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
25658 for -mno-speculate-indirect-jumps.
25659 (*call_indirect_elfv2<mode>_nospec): New define_insn.
25660 (*call_value_indirect_elfv2<mode>): Disable for
25661 -mno-speculate-indirect-jumps.
25662 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
25663 (indirect_jump): Emit different RTL for
25664 -mno-speculate-indirect-jumps.
25665 (*indirect_jump<mode>): Disable for
25666 -mno-speculate-indirect-jumps.
25667 (*indirect_jump<mode>_nospec): New define_insn.
25668 (tablejump): Emit different RTL for
25669 -mno-speculate-indirect-jumps.
25670 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
25671 (tablejumpsi_nospec): New define_expand.
25672 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
25673 (tablejumpdi_nospec): New define_expand.
25674 (*tablejump<mode>_internal1): Disable for
25675 -mno-speculate-indirect-jumps.
25676 (*tablejump<mode>_internal1_nospec): New define_insn.
25677 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
25680 2018-01-16 Artyom Skrobov tyomitch@gmail.com
25682 * caller-save.c (insert_save): Drop unnecessary parameter. All
25685 2018-01-16 Jakub Jelinek <jakub@redhat.com>
25686 Richard Biener <rguenth@suse.de>
25689 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
25690 return early, inline manually is_gimple_sizepos. Make sure if we
25691 call gimplify_expr we don't end up with a gimple constant.
25692 * tree.c (variably_modified_type_p): Don't return true for
25693 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
25694 * gimplify.h (is_gimple_sizepos): Remove.
25696 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
25698 PR tree-optimization/83857
25699 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
25700 vectorizable_live_operation for pure SLP statements.
25701 (vectorizable_live_operation): Handle PHIs.
25703 2018-01-16 Richard Biener <rguenther@suse.de>
25705 PR tree-optimization/83867
25706 * tree-vect-stmts.c (vect_transform_stmt): Precompute
25707 nested_in_vect_loop_p since the scalar stmt may get invalidated.
25709 2018-01-16 Jakub Jelinek <jakub@redhat.com>
25712 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
25713 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
25714 If off is not INTEGER_CST, issue a may not be aligned warning
25715 rather than isn't aligned. Use isn%'t rather than isn't.
25716 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
25718 <case MULT_EXPR>: Improve the case when bottom and one of the
25719 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
25720 operand, in that case check if the other operand is multiple of
25721 bottom divided by the INTEGER_CST operand.
25723 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
25726 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
25727 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
25728 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
25729 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
25730 * config/pa/pa.c (pa_function_arg_advance): Likewise.
25731 (pa_function_arg, pa_arg_partial_bytes): Likewise.
25732 (pa_function_arg_size): New function.
25734 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
25736 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
25737 in a separate statement.
25739 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
25741 PR tree-optimization/83847
25742 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
25743 group gathers and scatters.
25745 2018-01-16 Jakub Jelinek <jakub@redhat.com>
25747 PR rtl-optimization/86620
25748 * params.def (max-sched-ready-insns): Bump minimum value to 1.
25750 PR rtl-optimization/83213
25751 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
25752 to last if both are JUMP_INSNs.
25754 PR tree-optimization/83843
25755 * gimple-ssa-store-merging.c
25756 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
25757 store_immediate_info for bswap/nop orig_stores.
25759 2018-01-15 Andrew Waterman <andrew@sifive.com>
25761 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
25763 <UDIV>: Increase cost if !TARGET_DIV.
25765 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
25767 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
25768 (define_attr "cr_logical_3op"): New.
25769 (cceq_ior_compare): Adjust.
25770 (cceq_ior_compare_complement): Adjust.
25771 (*cceq_rev_compare): Adjust.
25772 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
25773 (is_cracked_insn): Adjust.
25774 (insn_must_be_first_in_group): Adjust.
25775 * config/rs6000/40x.md: Adjust.
25776 * config/rs6000/440.md: Adjust.
25777 * config/rs6000/476.md: Adjust.
25778 * config/rs6000/601.md: Adjust.
25779 * config/rs6000/603.md: Adjust.
25780 * config/rs6000/6xx.md: Adjust.
25781 * config/rs6000/7450.md: Adjust.
25782 * config/rs6000/7xx.md: Adjust.
25783 * config/rs6000/8540.md: Adjust.
25784 * config/rs6000/cell.md: Adjust.
25785 * config/rs6000/e300c2c3.md: Adjust.
25786 * config/rs6000/e500mc.md: Adjust.
25787 * config/rs6000/e500mc64.md: Adjust.
25788 * config/rs6000/e5500.md: Adjust.
25789 * config/rs6000/e6500.md: Adjust.
25790 * config/rs6000/mpc.md: Adjust.
25791 * config/rs6000/power4.md: Adjust.
25792 * config/rs6000/power5.md: Adjust.
25793 * config/rs6000/power6.md: Adjust.
25794 * config/rs6000/power7.md: Adjust.
25795 * config/rs6000/power8.md: Adjust.
25796 * config/rs6000/power9.md: Adjust.
25797 * config/rs6000/rs64.md: Adjust.
25798 * config/rs6000/titan.md: Adjust.
25800 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
25802 * config/i386/predicates.md (indirect_branch_operand): Rewrite
25803 ix86_indirect_branch_register logic.
25805 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
25807 * config/i386/constraints.md (Bs): Update
25808 ix86_indirect_branch_register check. Don't check
25809 ix86_indirect_branch_register with GOT_memory_operand.
25811 * config/i386/predicates.md (GOT_memory_operand): Don't check
25812 ix86_indirect_branch_register here.
25813 (GOT32_symbol_operand): Likewise.
25815 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
25817 * config/i386/predicates.md (constant_call_address_operand):
25818 Rewrite ix86_indirect_branch_register logic.
25819 (sibcall_insn_operand): Likewise.
25821 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
25823 * config/i386/constraints.md (Bs): Replace
25824 ix86_indirect_branch_thunk_register with
25825 ix86_indirect_branch_register.
25827 * config/i386/i386.md (indirect_jump): Likewise.
25828 (tablejump): Likewise.
25829 (*sibcall_memory): Likewise.
25830 (*sibcall_value_memory): Likewise.
25831 Peepholes of indirect call and jump via memory: Likewise.
25832 * config/i386/i386.opt: Likewise.
25833 * config/i386/predicates.md (indirect_branch_operand): Likewise.
25834 (GOT_memory_operand): Likewise.
25835 (call_insn_operand): Likewise.
25836 (sibcall_insn_operand): Likewise.
25837 (GOT32_symbol_operand): Likewise.
25839 2018-01-15 Jakub Jelinek <jakub@redhat.com>
25841 PR middle-end/83837
25842 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
25843 type rather than type addr's type points to.
25844 (expand_omp_atomic_mutex): Likewise.
25845 (expand_omp_atomic): Likewise.
25847 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
25850 * config/i386/i386.c (output_indirect_thunk_function): Use
25851 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
25852 for __x86_return_thunk.
25854 2018-01-15 Richard Biener <rguenther@suse.de>
25856 PR middle-end/83850
25857 * expmed.c (extract_bit_field_1): Fix typo.
25859 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25862 * config/arm/iterators.md (VF): New mode iterator.
25863 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
25864 Remove integer-related logic from pattern.
25865 (neon_vabd<mode>_3): Likewise.
25867 2018-01-15 Jakub Jelinek <jakub@redhat.com>
25869 PR middle-end/82694
25870 * common.opt (fstrict-overflow): No longer an alias.
25871 (fwrapv-pointer): New option.
25872 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
25873 also for pointer types based on flag_wrapv_pointer.
25874 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
25875 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
25876 opts->x_flag_wrapv got set.
25877 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
25878 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
25879 POINTER_TYPE_OVERFLOW_UNDEFINED.
25880 * match.pd: Likewise in address comparison pattern.
25881 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
25883 2018-01-15 Richard Biener <rguenther@suse.de>
25886 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
25887 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
25888 Reset type names to their identifier if their TYPE_DECL doesn't
25889 have linkage (and thus is used for ODR and devirt).
25890 (save_debug_info_for_decl): Remove.
25891 (save_debug_info_for_type): Likewise.
25892 (add_tree_to_fld_list): Adjust.
25893 * tree-pretty-print.c (dump_generic_node): Make dumping of
25894 type names more robust.
25896 2018-01-15 Richard Biener <rguenther@suse.de>
25898 * BASE-VER: Bump to 8.0.1.
25900 2018-01-14 Martin Sebor <msebor@redhat.com>
25903 * builtins.c (check_access): Avoid warning when the no-warning bit
25906 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
25908 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
25909 * ira-color (allocno_hard_regs_compare): Likewise.
25911 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
25914 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
25915 Use .pushsection/.popsection.
25917 2018-01-14 Martin Sebor <msebor@redhat.com>
25920 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
25922 2018-01-14 Jakub Jelinek <jakub@redhat.com>
25924 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
25925 entry from extra_headers.
25926 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
25927 extra_headers, make the list bitwise identical to the i?86-*-* one.
25929 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
25931 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
25932 -mcmodel=large with -mindirect-branch=thunk,
25933 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
25934 -mfunction-return=thunk-extern.
25935 * doc/invoke.texi: Document -mcmodel=large is incompatible with
25936 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
25937 -mfunction-return=thunk and -mfunction-return=thunk-extern.
25939 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
25941 * config/i386/i386.c (print_reg): Print the name of the full
25942 integer register without '%'.
25943 (ix86_print_operand): Handle 'V'.
25944 * doc/extend.texi: Document 'V' modifier.
25946 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
25948 * config/i386/constraints.md (Bs): Disallow memory operand for
25949 -mindirect-branch-register.
25951 * config/i386/predicates.md (indirect_branch_operand): Likewise.
25952 (GOT_memory_operand): Likewise.
25953 (call_insn_operand): Likewise.
25954 (sibcall_insn_operand): Likewise.
25955 (GOT32_symbol_operand): Likewise.
25956 * config/i386/i386.md (indirect_jump): Call convert_memory_address
25957 for -mindirect-branch-register.
25958 (tablejump): Likewise.
25959 (*sibcall_memory): Likewise.
25960 (*sibcall_value_memory): Likewise.
25961 Disallow peepholes of indirect call and jump via memory for
25962 -mindirect-branch-register.
25963 (*call_pop): Replace m with Bw.
25964 (*call_value_pop): Likewise.
25965 (*sibcall_pop_memory): Replace m with Bs.
25966 * config/i386/i386.opt (mindirect-branch-register): New option.
25967 * doc/invoke.texi: Document -mindirect-branch-register option.
25969 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
25971 * config/i386/i386-protos.h (ix86_output_function_return): New.
25972 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
25973 set function_return_type.
25974 (indirect_thunk_name): Add ret_p to indicate thunk for function
25976 (output_indirect_thunk_function): Pass false to
25977 indirect_thunk_name.
25978 (ix86_output_indirect_branch_via_reg): Likewise.
25979 (ix86_output_indirect_branch_via_push): Likewise.
25980 (output_indirect_thunk_function): Create alias for function
25981 return thunk if regno < 0.
25982 (ix86_output_function_return): New function.
25983 (ix86_handle_fndecl_attribute): Handle function_return.
25984 (ix86_attribute_table): Add function_return.
25985 * config/i386/i386.h (machine_function): Add
25986 function_return_type.
25987 * config/i386/i386.md (simple_return_internal): Use
25988 ix86_output_function_return.
25989 (simple_return_internal_long): Likewise.
25990 * config/i386/i386.opt (mfunction-return=): New option.
25991 (indirect_branch): Mention -mfunction-return=.
25992 * doc/extend.texi: Document function_return function attribute.
25993 * doc/invoke.texi: Document -mfunction-return= option.
25995 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
25997 * config/i386/i386-opts.h (indirect_branch): New.
25998 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
25999 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
26000 with local indirect jump when converting indirect call and jump.
26001 (ix86_set_indirect_branch_type): New.
26002 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
26003 (indirectlabelno): New.
26004 (indirect_thunk_needed): Likewise.
26005 (indirect_thunk_bnd_needed): Likewise.
26006 (indirect_thunks_used): Likewise.
26007 (indirect_thunks_bnd_used): Likewise.
26008 (INDIRECT_LABEL): Likewise.
26009 (indirect_thunk_name): Likewise.
26010 (output_indirect_thunk): Likewise.
26011 (output_indirect_thunk_function): Likewise.
26012 (ix86_output_indirect_branch_via_reg): Likewise.
26013 (ix86_output_indirect_branch_via_push): Likewise.
26014 (ix86_output_indirect_branch): Likewise.
26015 (ix86_output_indirect_jmp): Likewise.
26016 (ix86_code_end): Call output_indirect_thunk_function if needed.
26017 (ix86_output_call_insn): Call ix86_output_indirect_branch if
26019 (ix86_handle_fndecl_attribute): Handle indirect_branch.
26020 (ix86_attribute_table): Add indirect_branch.
26021 * config/i386/i386.h (machine_function): Add indirect_branch_type
26022 and has_local_indirect_jump.
26023 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
26025 (tablejump): Likewise.
26026 (*indirect_jump): Use ix86_output_indirect_jmp.
26027 (*tablejump_1): Likewise.
26028 (simple_return_indirect_internal): Likewise.
26029 * config/i386/i386.opt (mindirect-branch=): New option.
26030 (indirect_branch): New.
26033 (thunk-inline): Likewise.
26034 (thunk-extern): Likewise.
26035 * doc/extend.texi: Document indirect_branch function attribute.
26036 * doc/invoke.texi: Document -mindirect-branch= option.
26038 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
26041 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
26043 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
26045 * ipa-inline.c (want_inline_small_function_p): Return false if
26046 inlining has already failed with CIF_FINAL_ERROR.
26047 (update_caller_keys): Call want_inline_small_function_p before
26049 (update_callee_keys): Likewise.
26051 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
26053 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
26055 (rs6000_quadword_masked_address_p): Likewise.
26056 (quad_aligned_load_p): Likewise.
26057 (quad_aligned_store_p): Likewise.
26058 (const_load_sequence_p): Add comment to describe the outer-most loop.
26059 (mimic_memory_attributes_and_flags): New function.
26060 (rs6000_gen_stvx): Likewise.
26061 (replace_swapped_aligned_store): Likewise.
26062 (rs6000_gen_lvx): Likewise.
26063 (replace_swapped_aligned_load): Likewise.
26064 (replace_swapped_load_constant): Capitalize argument name in
26065 comment describing this function.
26066 (rs6000_analyze_swaps): Add a third pass to search for vector loads
26067 and stores that access quad-word aligned addresses and replace
26068 with stvx or lvx instructions when appropriate.
26069 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
26070 New function prototype.
26071 (rs6000_quadword_masked_address_p): Likewise.
26072 (rs6000_gen_lvx): Likewise.
26073 (rs6000_gen_stvx): Likewise.
26074 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
26075 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
26076 when memory address is aligned.
26077 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
26078 this split to select lvx instruction when memory address is aligned.
26079 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
26080 instruction when memory address is aligned.
26081 (*vsx_le_perm_load_v16qi): Likewise.
26082 (four unnamed splitters): Modify to select the stvx instruction
26083 when memory is aligned.
26085 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
26087 * predict.c (determine_unlikely_bbs): Handle correctly BBs
26088 which appears in the queue multiple times.
26090 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26091 Alan Hayward <alan.hayward@arm.com>
26092 David Sherwood <david.sherwood@arm.com>
26094 * tree-vectorizer.h (vec_lower_bound): New structure.
26095 (_loop_vec_info): Add check_nonzero and lower_bounds.
26096 (LOOP_VINFO_CHECK_NONZERO): New macro.
26097 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
26098 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
26099 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
26100 fields. Make seg_len the distance travelled, not including the
26102 (dr_direction_indicator): Declare.
26103 (dr_zero_step_indicator): Likewise.
26104 (dr_known_forward_stride_p): Likewise.
26105 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
26107 (runtime_alias_check_p): Allow runtime alias checks with
26109 (operator ==): Compare access_size and align.
26110 (prune_runtime_alias_test_list): Rework for new distinction between
26111 the access_size and seg_len.
26112 (create_intersect_range_checks_index): Likewise. Cope with polynomial
26114 (get_segment_min_max): New function.
26115 (create_intersect_range_checks): Use it.
26116 (dr_step_indicator): New function.
26117 (dr_direction_indicator): Likewise.
26118 (dr_zero_step_indicator): Likewise.
26119 (dr_known_forward_stride_p): Likewise.
26120 * tree-loop-distribution.c (data_ref_segment_size): Return
26121 DR_STEP * (niters - 1).
26122 (compute_alias_check_pairs): Update call to the dr_with_seg_len
26124 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
26125 (vect_preserves_scalar_order_p): New function, split out from...
26126 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
26127 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
26128 (vect_vfa_access_size): New function.
26129 (vect_vfa_align): Likewise.
26130 (vect_compile_time_alias): Take access_size_a and access_b arguments.
26131 (dump_lower_bound): New function.
26132 (vect_check_lower_bound): Likewise.
26133 (vect_small_gap_p): Likewise.
26134 (vectorizable_with_step_bound_p): Likewise.
26135 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
26136 depencies if the vectorization factor is 1. Convert the checks
26137 for nonzero steps into checks on the bounds of DR_STEP. Try using
26138 a bunds check for variable steps if the minimum required step is
26139 relatively small. Update calls to the dr_with_seg_len
26140 constructor and to vect_compile_time_alias.
26141 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
26143 (vect_loop_versioning): Call it.
26144 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
26146 (vect_estimate_min_profitable_iters): Account for any bounds checks.
26148 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26149 Alan Hayward <alan.hayward@arm.com>
26150 David Sherwood <david.sherwood@arm.com>
26152 * doc/sourcebuild.texi (vect_scatter_store): Document.
26153 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
26155 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
26157 * genopinit.c (main): Add supports_vec_scatter_store and
26158 supports_vec_scatter_store_cached to target_optabs.
26159 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
26160 IFN_MASK_SCATTER_STORE.
26161 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
26163 * internal-fn.h (internal_store_fn_p): Declare.
26164 (internal_fn_stored_value_index): Likewise.
26165 * internal-fn.c (scatter_store_direct): New macro.
26166 (expand_scatter_store_optab_fn): New function.
26167 (direct_scatter_store_optab_supported_p): New macro.
26168 (internal_store_fn_p): New function.
26169 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
26170 IFN_MASK_SCATTER_STORE.
26171 (internal_fn_mask_index): Likewise.
26172 (internal_fn_stored_value_index): New function.
26173 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
26174 for scatter stores.
26175 * optabs-query.h (supports_vec_scatter_store_p): Declare.
26176 * optabs-query.c (supports_vec_scatter_store_p): New function.
26177 * tree-vectorizer.h (vect_get_store_rhs): Declare.
26178 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
26179 true for scatter stores.
26180 (vect_gather_scatter_fn_p): Handle scatter stores too.
26181 (vect_check_gather_scatter): Consider using scatter stores if
26182 supports_vec_scatter_store_p.
26183 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
26184 scatter stores too.
26185 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
26186 internal_fn_stored_value_index.
26187 (check_load_store_masking): Handle scatter stores too.
26188 (vect_get_store_rhs): Make public.
26189 (vectorizable_call): Use internal_store_fn_p.
26190 (vectorizable_store): Handle scatter store internal functions.
26191 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
26192 when deciding whether the end of the group has been reached.
26193 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
26194 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
26195 (mask_scatter_store<mode>): New insns.
26197 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26198 Alan Hayward <alan.hayward@arm.com>
26199 David Sherwood <david.sherwood@arm.com>
26201 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
26202 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
26203 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
26205 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
26206 Use vect_truncate_gather_scatter_offset if we can't treat the
26207 operation as a normal gather load or scatter store.
26208 (get_group_load_store_type): Take the gather_scatter_info
26209 as argument. Try using a gather load or scatter store for
26210 single-element groups.
26211 (get_load_store_type): Update calls to get_group_load_store_type
26212 and vect_use_strided_gather_scatters_p.
26214 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26215 Alan Hayward <alan.hayward@arm.com>
26216 David Sherwood <david.sherwood@arm.com>
26218 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
26219 optional tree argument.
26220 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
26222 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
26223 but continue to use the current value as a fallback.
26224 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
26225 to compare the updates.
26226 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
26227 (get_load_store_type): Use it when handling a strided access.
26228 (vect_get_strided_load_store_ops): New function.
26229 (vect_get_data_ptr_increment): Likewise.
26230 (vectorizable_load): Handle strided gather loads. Always pass
26231 a step to vect_create_data_ref_ptr and bump_vector_ptr.
26233 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26234 Alan Hayward <alan.hayward@arm.com>
26235 David Sherwood <david.sherwood@arm.com>
26237 * doc/md.texi (gather_load@var{m}): Document.
26238 (mask_gather_load@var{m}): Likewise.
26239 * genopinit.c (main): Add supports_vec_gather_load and
26240 supports_vec_gather_load_cached to target_optabs.
26241 * optabs-tree.c (init_tree_optimization_optabs): Use
26242 ggc_cleared_alloc to allocate target_optabs.
26243 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
26244 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
26246 * internal-fn.h (internal_load_fn_p): Declare.
26247 (internal_gather_scatter_fn_p): Likewise.
26248 (internal_fn_mask_index): Likewise.
26249 (internal_gather_scatter_fn_supported_p): Likewise.
26250 * internal-fn.c (gather_load_direct): New macro.
26251 (expand_gather_load_optab_fn): New function.
26252 (direct_gather_load_optab_supported_p): New macro.
26253 (direct_internal_fn_optab): New function.
26254 (internal_load_fn_p): Likewise.
26255 (internal_gather_scatter_fn_p): Likewise.
26256 (internal_fn_mask_index): Likewise.
26257 (internal_gather_scatter_fn_supported_p): Likewise.
26258 * optabs-query.c (supports_at_least_one_mode_p): New function.
26259 (supports_vec_gather_load_p): Likewise.
26260 * optabs-query.h (supports_vec_gather_load_p): Declare.
26261 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
26262 and memory_type field.
26263 (NUM_PATTERNS): Bump to 15.
26264 * tree-vect-data-refs.c: Include internal-fn.h.
26265 (vect_gather_scatter_fn_p): New function.
26266 (vect_describe_gather_scatter_call): Likewise.
26267 (vect_check_gather_scatter): Try using internal functions for
26268 gather loads. Recognize existing calls to a gather load function.
26269 (vect_analyze_data_refs): Consider using gather loads if
26270 supports_vec_gather_load_p.
26271 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
26272 (vect_get_gather_scatter_offset_type): Likewise.
26273 (vect_convert_mask_for_vectype): Likewise.
26274 (vect_add_conversion_to_patterm): Likewise.
26275 (vect_try_gather_scatter_pattern): Likewise.
26276 (vect_recog_gather_scatter_pattern): New pattern recognizer.
26277 (vect_vect_recog_func_ptrs): Add it.
26278 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
26279 internal_fn_mask_index and internal_gather_scatter_fn_p.
26280 (check_load_store_masking): Take the gather_scatter_info as an
26281 argument and handle gather loads.
26282 (vect_get_gather_scatter_ops): New function.
26283 (vectorizable_call): Check internal_load_fn_p.
26284 (vectorizable_load): Likewise. Handle gather load internal
26286 (vectorizable_store): Update call to check_load_store_masking.
26287 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
26288 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
26289 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
26290 (aarch64_gather_scale_operand_d): New predicates.
26291 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
26292 (mask_gather_load<mode>): New insns.
26294 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26295 Alan Hayward <alan.hayward@arm.com>
26296 David Sherwood <david.sherwood@arm.com>
26298 * optabs.def (fold_left_plus_optab): New optab.
26299 * doc/md.texi (fold_left_plus_@var{m}): Document.
26300 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
26301 * internal-fn.c (fold_left_direct): Define.
26302 (expand_fold_left_optab_fn): Likewise.
26303 (direct_fold_left_optab_supported_p): Likewise.
26304 * fold-const-call.c (fold_const_fold_left): New function.
26305 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
26306 * tree-parloops.c (valid_reduction_p): New function.
26307 (gather_scalar_reductions): Use it.
26308 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
26309 (vect_finish_replace_stmt): Declare.
26310 * tree-vect-loop.c (fold_left_reduction_fn): New function.
26311 (needs_fold_left_reduction_p): New function, split out from...
26312 (vect_is_simple_reduction): ...here. Accept reductions that
26313 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
26314 (vect_force_simple_reduction): Also store the reduction type in
26315 the assignment's STMT_VINFO_REDUC_TYPE.
26316 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
26317 (merge_with_identity): New function.
26318 (vect_expand_fold_left): Likewise.
26319 (vectorize_fold_left_reduction): Likewise.
26320 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
26321 scalar phi in place for it. Check for target support and reject
26322 cases that would reassociate the operation. Defer the transform
26323 phase to vectorize_fold_left_reduction.
26324 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
26325 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
26326 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
26328 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26330 * tree-if-conv.c (predicate_mem_writes): Remove redundant
26331 call to ifc_temp_var.
26333 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26334 Alan Hayward <alan.hayward@arm.com>
26335 David Sherwood <david.sherwood@arm.com>
26337 * target.def (legitimize_address_displacement): Take the original
26338 offset as a poly_int.
26339 * targhooks.h (default_legitimize_address_displacement): Update
26341 * targhooks.c (default_legitimize_address_displacement): Likewise.
26342 * doc/tm.texi: Regenerate.
26343 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
26344 as an argument, moving assert of ad->disp == ad->disp_term to...
26345 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
26346 Try calling targetm.legitimize_address_displacement before expanding
26347 the address rather than afterwards, and adjust for the new interface.
26348 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
26349 Match the new hook interface. Handle SVE addresses.
26350 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
26351 new hook interface.
26353 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26355 * Makefile.in (OBJS): Add early-remat.o.
26356 * target.def (select_early_remat_modes): New hook.
26357 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
26358 * doc/tm.texi: Regenerate.
26359 * targhooks.h (default_select_early_remat_modes): Declare.
26360 * targhooks.c (default_select_early_remat_modes): New function.
26361 * timevar.def (TV_EARLY_REMAT): New timevar.
26362 * passes.def (pass_early_remat): New pass.
26363 * tree-pass.h (make_pass_early_remat): Declare.
26364 * early-remat.c: New file.
26365 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
26367 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
26369 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26370 Alan Hayward <alan.hayward@arm.com>
26371 David Sherwood <david.sherwood@arm.com>
26373 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
26374 vfm1 with a bound_epilog parameter.
26375 (vect_do_peeling): Update calls accordingly, and move the prologue
26376 call earlier in the function. Treat the base bound_epilog as 0 for
26377 fully-masked loops and retain vf - 1 for other loops. Add 1 to
26378 this base when peeling for gaps.
26379 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
26380 with fully-masked loops.
26381 (vect_estimate_min_profitable_iters): Handle the single peeled
26382 iteration in that case.
26384 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26385 Alan Hayward <alan.hayward@arm.com>
26386 David Sherwood <david.sherwood@arm.com>
26388 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
26389 single-element interleaving even if the size is not a power of 2.
26390 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
26391 accesses for single-element interleaving if the group size is
26394 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26395 Alan Hayward <alan.hayward@arm.com>
26396 David Sherwood <david.sherwood@arm.com>
26398 * doc/md.texi (fold_extract_last_@var{m}): Document.
26399 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
26400 * optabs.def (fold_extract_last_optab): New optab.
26401 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
26402 * internal-fn.c (fold_extract_direct): New macro.
26403 (expand_fold_extract_optab_fn): Likewise.
26404 (direct_fold_extract_optab_supported_p): Likewise.
26405 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
26406 * tree-vect-loop.c (vect_model_reduction_cost): Handle
26407 EXTRACT_LAST_REDUCTION.
26408 (get_initial_def_for_reduction): Do not create an initial vector
26409 for EXTRACT_LAST_REDUCTION reductions.
26410 (vectorizable_reduction): Leave the scalar phi in place for
26411 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
26412 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
26413 epilogue code for EXTRACT_LAST_REDUCTION and defer the
26414 transform phase to vectorizable_condition.
26415 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
26417 (vect_finish_stmt_generation): ...here.
26418 (vect_finish_replace_stmt): New function.
26419 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
26420 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
26422 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
26424 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26425 Alan Hayward <alan.hayward@arm.com>
26426 David Sherwood <david.sherwood@arm.com>
26428 * doc/md.texi (extract_last_@var{m}): Document.
26429 * optabs.def (extract_last_optab): New optab.
26430 * internal-fn.def (EXTRACT_LAST): New internal function.
26431 * internal-fn.c (cond_unary_direct): New macro.
26432 (expand_cond_unary_optab_fn): Likewise.
26433 (direct_cond_unary_optab_supported_p): Likewise.
26434 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
26435 loops using EXTRACT_LAST.
26436 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
26437 (extract_last_<mode>): ...this optab.
26438 (vec_extract<mode><Vel>): Update accordingly.
26440 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26441 Alan Hayward <alan.hayward@arm.com>
26442 David Sherwood <david.sherwood@arm.com>
26444 * target.def (empty_mask_is_expensive): New hook.
26445 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
26446 * doc/tm.texi: Regenerate.
26447 * targhooks.h (default_empty_mask_is_expensive): Declare.
26448 * targhooks.c (default_empty_mask_is_expensive): New function.
26449 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
26450 if the target says that empty masks are expensive.
26451 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
26453 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
26455 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26456 Alan Hayward <alan.hayward@arm.com>
26457 David Sherwood <david.sherwood@arm.com>
26459 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
26460 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
26461 (vect_use_loop_mask_for_alignment_p): New function.
26462 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
26463 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
26464 niters_skip argument. Make sure that the first niters_skip elements
26465 of the first iteration are inactive.
26466 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
26467 Update call to vect_set_loop_masks_directly.
26468 (get_misalign_in_elems): New function, split out from...
26469 (vect_gen_prolog_loop_niters): ...here.
26470 (vect_update_init_of_dr): Take a code argument that specifies whether
26471 the adjustment should be added or subtracted.
26472 (vect_update_init_of_drs): Likewise.
26473 (vect_prepare_for_masked_peels): New function.
26474 (vect_do_peeling): Skip prologue peeling if we're using a mask
26475 instead. Update call to vect_update_inits_of_drs.
26476 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
26478 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
26479 alignment. Do not include the number of peeled iterations in
26480 the minimum threshold in that case.
26481 (vectorizable_induction): Adjust the start value down by
26482 LOOP_VINFO_MASK_SKIP_NITERS iterations.
26483 (vect_transform_loop): Call vect_prepare_for_masked_peels.
26484 Take the number of skipped iterations into account when calculating
26486 * tree-vect-stmts.c (vect_gen_while_not): New function.
26488 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26489 Alan Hayward <alan.hayward@arm.com>
26490 David Sherwood <david.sherwood@arm.com>
26492 * doc/sourcebuild.texi (vect_fully_masked): Document.
26493 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
26494 default value to 0.
26495 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
26497 (vect_analyze_loop_2): ...here. Don't check the vectorization
26498 factor against the number of loop iterations if the loop is
26501 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26502 Alan Hayward <alan.hayward@arm.com>
26503 David Sherwood <david.sherwood@arm.com>
26505 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
26506 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
26507 (dump_groups): Update accordingly.
26508 (iv_use::mem_type): New member variable.
26509 (address_p): New function.
26510 (record_use): Add a mem_type argument and initialize the new
26512 (record_group_use): Add a mem_type argument. Use address_p.
26513 Remove obsolete null checks of base_object. Update call to record_use.
26514 (find_interesting_uses_op): Update call to record_group_use.
26515 (find_interesting_uses_cond): Likewise.
26516 (find_interesting_uses_address): Likewise.
26517 (get_mem_type_for_internal_fn): New function.
26518 (find_address_like_use): Likewise.
26519 (find_interesting_uses_stmt): Try find_address_like_use before
26520 calling find_interesting_uses_op.
26521 (addr_offset_valid_p): Use the iv mem_type field as the type
26522 of the addressed memory.
26523 (add_autoinc_candidates): Likewise.
26524 (get_address_cost): Likewise.
26525 (split_small_address_groups_p): Use address_p.
26526 (split_address_groups): Likewise.
26527 (add_iv_candidate_for_use): Likewise.
26528 (autoinc_possible_for_pair): Likewise.
26529 (rewrite_groups): Likewise.
26530 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
26531 (determine_group_iv_cost): Update after split of USE_ADDRESS.
26532 (get_alias_ptr_type_for_ptr_address): New function.
26533 (rewrite_use_address): Rewrite address uses in calls that were
26534 identified by find_address_like_use.
26536 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26537 Alan Hayward <alan.hayward@arm.com>
26538 David Sherwood <david.sherwood@arm.com>
26540 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
26542 * gimple-expr.h (is_gimple_addressable: Likewise.
26543 * gimple-expr.c (is_gimple_address): Likewise.
26544 * internal-fn.c (expand_call_mem_ref): New function.
26545 (expand_mask_load_optab_fn): Use it.
26546 (expand_mask_store_optab_fn): Likewise.
26548 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26549 Alan Hayward <alan.hayward@arm.com>
26550 David Sherwood <david.sherwood@arm.com>
26552 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
26553 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
26554 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
26555 (cond_umax@var{mode}): Document.
26556 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
26557 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
26558 (cond_umin_optab, cond_umax_optab): New optabs.
26559 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
26560 (COND_IOR, COND_XOR): New internal functions.
26561 * internal-fn.h (get_conditional_internal_fn): Declare.
26562 * internal-fn.c (cond_binary_direct): New macro.
26563 (expand_cond_binary_optab_fn): Likewise.
26564 (direct_cond_binary_optab_supported_p): Likewise.
26565 (get_conditional_internal_fn): New function.
26566 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
26567 Cope with reduction statements that are vectorized as calls rather
26569 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
26570 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
26571 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
26572 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
26573 (UNSPEC_COND_EOR): New unspecs.
26574 (optab): Add mappings for them.
26575 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
26576 (sve_int_op, sve_fp_op): New int attributes.
26578 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26579 Alan Hayward <alan.hayward@arm.com>
26580 David Sherwood <david.sherwood@arm.com>
26582 * optabs.def (while_ult_optab): New optab.
26583 * doc/md.texi (while_ult@var{m}@var{n}): Document.
26584 * internal-fn.def (WHILE_ULT): New internal function.
26585 * internal-fn.h (direct_internal_fn_supported_p): New override
26586 that takes two types as argument.
26587 * internal-fn.c (while_direct): New macro.
26588 (expand_while_optab_fn): New function.
26589 (convert_optab_supported_p): Likewise.
26590 (direct_while_optab_supported_p): New macro.
26591 * wide-int.h (wi::udiv_ceil): New function.
26592 * tree-vectorizer.h (rgroup_masks): New structure.
26593 (vec_loop_masks): New typedef.
26594 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
26595 and fully_masked_p.
26596 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
26597 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
26598 (vect_max_vf): New function.
26599 (slpeel_make_loop_iterate_ntimes): Delete.
26600 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
26601 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
26602 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
26603 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
26604 internal-fn.h, stor-layout.h and optabs-query.h.
26605 (vect_set_loop_mask): New function.
26606 (add_preheader_seq): Likewise.
26607 (add_header_seq): Likewise.
26608 (interleave_supported_p): Likewise.
26609 (vect_maybe_permute_loop_masks): Likewise.
26610 (vect_set_loop_masks_directly): Likewise.
26611 (vect_set_loop_condition_masked): Likewise.
26612 (vect_set_loop_condition_unmasked): New function, split out from
26613 slpeel_make_loop_iterate_ntimes.
26614 (slpeel_make_loop_iterate_ntimes): Rename to..
26615 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
26616 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
26617 (vect_do_peeling): Update call accordingly.
26618 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
26620 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
26621 mask_compare_type, can_fully_mask_p and fully_masked_p.
26622 (release_vec_loop_masks): New function.
26623 (_loop_vec_info): Use it to free the loop masks.
26624 (can_produce_all_loop_masks_p): New function.
26625 (vect_get_max_nscalars_per_iter): Likewise.
26626 (vect_verify_full_masking): Likewise.
26627 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
26628 retries, and free the mask rgroups before retrying. Check loop-wide
26629 reasons for disallowing fully-masked loops. Make the final decision
26630 about whether use a fully-masked loop or not.
26631 (vect_estimate_min_profitable_iters): Do not assume that peeling
26632 for the number of iterations will be needed for fully-masked loops.
26633 (vectorizable_reduction): Disable fully-masked loops.
26634 (vectorizable_live_operation): Likewise.
26635 (vect_halve_mask_nunits): New function.
26636 (vect_double_mask_nunits): Likewise.
26637 (vect_record_loop_mask): Likewise.
26638 (vect_get_loop_mask): Likewise.
26639 (vect_transform_loop): Handle the case in which the final loop
26640 iteration might handle a partial vector. Call vect_set_loop_condition
26641 instead of slpeel_make_loop_iterate_ntimes.
26642 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
26643 (check_load_store_masking): New function.
26644 (prepare_load_store_mask): Likewise.
26645 (vectorizable_store): Handle fully-masked loops.
26646 (vectorizable_load): Likewise.
26647 (supportable_widening_operation): Use vect_halve_mask_nunits for
26649 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
26650 (vect_gen_while): New function.
26651 * config/aarch64/aarch64.md (umax<mode>3): New expander.
26652 (aarch64_uqdec<mode>): New insn.
26654 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26655 Alan Hayward <alan.hayward@arm.com>
26656 David Sherwood <david.sherwood@arm.com>
26658 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
26659 (reduc_xor_scal_optab): New optabs.
26660 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
26661 (reduc_xor_scal_@var{m}): Document.
26662 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
26663 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
26664 internal functions.
26665 * fold-const-call.c (fold_const_call): Handle them.
26666 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
26667 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
26668 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
26669 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
26670 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
26671 (UNSPEC_XORV): New unspecs.
26672 (optab): Add entries for them.
26673 (BITWISEV): New int iterator.
26674 (bit_reduc_op): New int attributes.
26676 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26677 Alan Hayward <alan.hayward@arm.com>
26678 David Sherwood <david.sherwood@arm.com>
26680 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
26681 * internal-fn.def (VEC_SHL_INSERT): New internal function.
26682 * optabs.def (vec_shl_insert_optab): New optab.
26683 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
26684 (duplicate_and_interleave): Likewise.
26685 * tree-vect-loop.c: Include internal-fn.h.
26686 (neutral_op_for_slp_reduction): New function, split out from
26687 get_initial_defs_for_reduction.
26688 (get_initial_def_for_reduction): Handle option 2 for variable-length
26689 vectors by loading the neutral value into a vector and then shifting
26690 the initial value into element 0.
26691 (get_initial_defs_for_reduction): Replace the code argument with
26692 the neutral value calculated by neutral_op_for_slp_reduction.
26693 Use gimple_build_vector for constant-length vectors.
26694 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
26695 but the first group_size elements have a neutral value.
26696 Use duplicate_and_interleave otherwise.
26697 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
26698 Update call to get_initial_defs_for_reduction. Handle SLP
26699 reductions for variable-length vectors by creating one vector
26700 result for each scalar result, with the elements associated
26701 with other scalar results stubbed out with the neutral value.
26702 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
26703 Require IFN_VEC_SHL_INSERT for double reductions on
26704 variable-length vectors, or SLP reductions that have
26705 a neutral value. Require can_duplicate_and_interleave_p
26706 support for variable-length unchained SLP reductions if there
26707 is no neutral value, such as for MIN/MAX reductions. Also require
26708 the number of vector elements to be a multiple of the number of
26709 SLP statements when doing variable-length unchained SLP reductions.
26710 Update call to vect_create_epilog_for_reduction.
26711 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
26712 and remove initial values.
26713 (duplicate_and_interleave): Make public.
26714 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
26715 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
26717 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26718 Alan Hayward <alan.hayward@arm.com>
26719 David Sherwood <david.sherwood@arm.com>
26721 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
26722 (can_duplicate_and_interleave_p): New function.
26723 (vect_get_and_check_slp_defs): Take the vector of statements
26724 rather than just the current one. Remove excess parentheses.
26725 Restriction rejectinon of vect_constant_def and vect_external_def
26726 for variable-length vectors to boolean types, or types for which
26727 can_duplicate_and_interleave_p is false.
26728 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
26729 (duplicate_and_interleave): New function.
26730 (vect_get_constant_vectors): Use gimple_build_vector for
26731 constant-length vectors and suitable variable-length constant
26732 vectors. Use duplicate_and_interleave for other variable-length
26733 vectors. Don't defer the update when inserting new statements.
26735 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26736 Alan Hayward <alan.hayward@arm.com>
26737 David Sherwood <david.sherwood@arm.com>
26739 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
26740 min_profitable_iters doesn't go negative.
26742 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26743 Alan Hayward <alan.hayward@arm.com>
26744 David Sherwood <david.sherwood@arm.com>
26746 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
26747 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
26748 * optabs.def (vec_mask_load_lanes_optab): New optab.
26749 (vec_mask_store_lanes_optab): Likewise.
26750 * internal-fn.def (MASK_LOAD_LANES): New internal function.
26751 (MASK_STORE_LANES): Likewise.
26752 * internal-fn.c (mask_load_lanes_direct): New macro.
26753 (mask_store_lanes_direct): Likewise.
26754 (expand_mask_load_optab_fn): Handle masked operations.
26755 (expand_mask_load_lanes_optab_fn): New macro.
26756 (expand_mask_store_optab_fn): Handle masked operations.
26757 (expand_mask_store_lanes_optab_fn): New macro.
26758 (direct_mask_load_lanes_optab_supported_p): Likewise.
26759 (direct_mask_store_lanes_optab_supported_p): Likewise.
26760 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
26762 (vect_load_lanes_supported): Likewise.
26763 * tree-vect-data-refs.c (strip_conversion): New function.
26764 (can_group_stmts_p): Likewise.
26765 (vect_analyze_data_ref_accesses): Use it instead of checking
26766 for a pair of assignments.
26767 (vect_store_lanes_supported): Take a masked_p parameter.
26768 (vect_load_lanes_supported): Likewise.
26769 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
26770 vect_store_lanes_supported and vect_load_lanes_supported.
26771 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
26772 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
26773 parameter. Don't allow gaps for masked accesses.
26774 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
26775 and vect_load_lanes_supported.
26776 (get_load_store_type): Take a masked_p parameter and update
26777 call to get_group_load_store_type.
26778 (vectorizable_store): Update call to get_load_store_type.
26779 Handle IFN_MASK_STORE_LANES.
26780 (vectorizable_load): Update call to get_load_store_type.
26781 Handle IFN_MASK_LOAD_LANES.
26783 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26784 Alan Hayward <alan.hayward@arm.com>
26785 David Sherwood <david.sherwood@arm.com>
26787 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
26789 * config/aarch64/aarch64-protos.h
26790 (aarch64_sve_struct_memory_operand_p): Declare.
26791 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
26792 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
26793 (VPRED, vpred): Handle SVE structure modes.
26794 * config/aarch64/constraints.md (Utx): New constraint.
26795 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
26796 (aarch64_sve_struct_nonimmediate_operand): New predicates.
26797 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
26798 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
26799 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
26800 structure modes. Split into pieces after RA.
26801 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
26802 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
26804 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
26805 SVE structure modes.
26806 (aarch64_classify_address): Likewise.
26807 (sizetochar): Move earlier in file.
26808 (aarch64_print_operand): Handle SVE register lists.
26809 (aarch64_array_mode): New function.
26810 (aarch64_sve_struct_memory_operand_p): Likewise.
26811 (TARGET_ARRAY_MODE): Redefine.
26813 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26814 Alan Hayward <alan.hayward@arm.com>
26815 David Sherwood <david.sherwood@arm.com>
26817 * target.def (array_mode): New target hook.
26818 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
26819 * doc/tm.texi: Regenerate.
26820 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
26821 * hooks.c (hook_optmode_mode_uhwi_none): New function.
26822 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
26823 targetm.array_mode.
26824 * stor-layout.c (mode_for_array): Likewise. Support polynomial
26827 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26828 Alan Hayward <alan.hayward@arm.com>
26829 David Sherwood <david.sherwood@arm.com>
26831 * fold-const.c (fold_binary_loc): Check the argument types
26832 rather than the result type when testing for a vector operation.
26834 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26836 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
26837 * doc/tm.texi: Regenerate.
26839 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26840 Alan Hayward <alan.hayward@arm.com>
26841 David Sherwood <david.sherwood@arm.com>
26843 * doc/invoke.texi (-msve-vector-bits=): Document new option.
26844 (sve): Document new AArch64 extension.
26845 * doc/md.texi (w): Extend the description of the AArch64
26846 constraint to include SVE vectors.
26847 (Upl, Upa): Document new AArch64 predicate constraints.
26848 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
26850 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
26851 (msve-vector-bits=): New option.
26852 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
26853 SVE when these are disabled.
26854 (sve): New extension.
26855 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
26856 modes. Adjust their number of units based on aarch64_sve_vg.
26857 (MAX_BITSIZE_MODE_ANY_MODE): Define.
26858 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
26859 aarch64_addr_query_type.
26860 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
26861 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
26862 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
26863 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
26864 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
26865 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
26866 (aarch64_simd_imm_zero_p): Delete.
26867 (aarch64_check_zero_based_sve_index_immediate): Declare.
26868 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
26869 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
26870 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
26871 (aarch64_sve_float_mul_immediate_p): Likewise.
26872 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
26873 rather than an rtx.
26874 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
26875 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
26876 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
26877 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
26878 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
26879 (aarch64_regmode_natural_size): Likewise.
26880 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
26881 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
26883 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
26884 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
26885 for VG and the SVE predicate registers.
26886 (V_ALIASES): Add a "z"-prefixed alias.
26887 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
26888 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
26889 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
26890 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
26891 (REG_CLASS_NAMES): Add entries for them.
26892 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
26893 and the predicate registers.
26894 (aarch64_sve_vg): Declare.
26895 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
26896 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
26897 (REGMODE_NATURAL_SIZE): Define.
26898 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
26900 * config/aarch64/aarch64.c: Include cfgrtl.h.
26901 (simd_immediate_info): Add a constructor for series vectors,
26902 and an associated step field.
26903 (aarch64_sve_vg): New variable.
26904 (aarch64_dbx_register_number): Handle VG and the predicate registers.
26905 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
26906 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
26907 (VEC_ANY_DATA, VEC_STRUCT): New constants.
26908 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
26909 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
26910 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
26911 (aarch64_get_mask_mode): New functions.
26912 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
26913 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
26914 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
26915 predicate modes and predicate registers. Explicitly restrict
26916 GPRs to modes of 16 bytes or smaller. Only allow FP registers
26917 to store a vector mode if it is recognized by
26918 aarch64_classify_vector_mode.
26919 (aarch64_regmode_natural_size): New function.
26920 (aarch64_hard_regno_caller_save_mode): Return the original mode
26922 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
26923 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
26924 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
26925 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
26927 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
26928 does not overlap dest if the function is frame-related. Handle
26930 (aarch64_split_add_offset): New function.
26931 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
26932 them aarch64_add_offset.
26933 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
26934 and update call to aarch64_sub_sp.
26935 (aarch64_add_cfa_expression): New function.
26936 (aarch64_expand_prologue): Pass extra temporary registers to the
26937 functions above. Handle the case in which we need to emit new
26938 DW_CFA_expressions for registers that were originally saved
26939 relative to the stack pointer, but now have to be expressed
26940 relative to the frame pointer.
26941 (aarch64_output_mi_thunk): Pass extra temporary registers to the
26943 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
26944 IP0 and IP1 values for SVE frames.
26945 (aarch64_expand_vec_series): New function.
26946 (aarch64_expand_sve_widened_duplicate): Likewise.
26947 (aarch64_expand_sve_const_vector): Likewise.
26948 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
26949 Handle SVE constants. Use emit_move_insn to move a force_const_mem
26950 into the register, rather than emitting a SET directly.
26951 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
26952 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
26953 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
26954 (offset_9bit_signed_scaled_p): New functions.
26955 (aarch64_replicate_bitmask_imm): New function.
26956 (aarch64_bitmask_imm): Use it.
26957 (aarch64_cannot_force_const_mem): Reject expressions involving
26958 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
26959 (aarch64_classify_index): Handle SVE indices, by requiring
26960 a plain register index with a scale that matches the element size.
26961 (aarch64_classify_address): Handle SVE addresses. Assert that
26962 the mode of the address is VOIDmode or an integer mode.
26963 Update call to aarch64_classify_symbol.
26964 (aarch64_classify_symbolic_expression): Update call to
26965 aarch64_classify_symbol.
26966 (aarch64_const_vec_all_in_range_p): New function.
26967 (aarch64_print_vector_float_operand): Likewise.
26968 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
26969 "vN" for FP registers with SVE modes. Handle (const ...) vectors
26970 and the FP immediates 1.0 and 0.5.
26971 (aarch64_print_address_internal): Handle SVE addresses.
26972 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
26973 (aarch64_regno_regclass): Handle predicate registers.
26974 (aarch64_secondary_reload): Handle big-endian reloads of SVE
26976 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
26977 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
26978 (aarch64_convert_sve_vector_bits): New function.
26979 (aarch64_override_options): Use it to handle -msve-vector-bits=.
26980 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
26981 rather than an rtx.
26982 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
26983 Handle SVE vector and predicate modes. Accept VL-based constants
26984 that need only one temporary register, and VL offsets that require
26985 no temporary registers.
26986 (aarch64_conditional_register_usage): Mark the predicate registers
26987 as fixed if SVE isn't available.
26988 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
26989 Return true for SVE vector and predicate modes.
26990 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
26991 rather than an unsigned int. Handle SVE modes.
26992 (aarch64_preferred_simd_mode): Update call accordingly. Handle
26994 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
26996 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
26997 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
26998 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
26999 (aarch64_sve_float_mul_immediate_p): New functions.
27000 (aarch64_sve_valid_immediate): New function.
27001 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
27002 Explicitly reject structure modes. Check for INDEX constants.
27003 Handle PTRUE and PFALSE constants.
27004 (aarch64_check_zero_based_sve_index_immediate): New function.
27005 (aarch64_simd_imm_zero_p): Delete.
27006 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
27007 vector modes. Accept constants in the range of CNT[BHWD].
27008 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
27009 ask for an Advanced SIMD mode.
27010 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
27011 (aarch64_simd_vector_alignment): Handle SVE predicates.
27012 (aarch64_vectorize_preferred_vector_alignment): New function.
27013 (aarch64_simd_vector_alignment_reachable): Use it instead of
27015 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
27016 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
27018 (MAX_VECT_LEN): Delete.
27019 (expand_vec_perm_d): Add a vec_flags field.
27020 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
27021 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
27022 (aarch64_evpc_ext): Don't apply a big-endian lane correction
27024 (aarch64_evpc_rev): Rename to...
27025 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
27026 (aarch64_evpc_rev_global): New function.
27027 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
27028 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
27030 (aarch64_evpc_sve_tbl): New function.
27031 (aarch64_expand_vec_perm_const_1): Update after rename of
27032 aarch64_evpc_rev. Handle SVE permutes too, trying
27033 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
27034 than aarch64_evpc_tbl.
27035 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
27036 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
27037 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
27038 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
27039 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
27040 (aarch64_expand_sve_vcond): New functions.
27041 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
27042 of aarch64_vector_mode_p.
27043 (aarch64_dwarf_poly_indeterminate_value): New function.
27044 (aarch64_compute_pressure_classes): Likewise.
27045 (aarch64_can_change_mode_class): Likewise.
27046 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
27047 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
27048 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
27049 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
27050 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
27051 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
27052 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
27053 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
27055 (Dn, Dl, Dr): Accept const as well as const_vector.
27056 (Dz): Likewise. Compare against CONST0_RTX.
27057 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
27058 of "vector" where appropriate.
27059 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
27060 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
27061 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
27062 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
27063 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
27064 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
27065 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
27066 (v_int_equiv): Extend to SVE modes.
27067 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
27069 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
27070 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
27071 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
27072 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
27073 (SVE_COND_FP_CMP): New int iterators.
27074 (perm_hilo): Handle the new unpack unspecs.
27075 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
27077 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
27078 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
27079 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
27080 (aarch64_equality_operator, aarch64_constant_vector_operand)
27081 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
27082 (aarch64_sve_nonimmediate_operand): Likewise.
27083 (aarch64_sve_general_operand): Likewise.
27084 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
27085 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
27086 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
27087 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
27088 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
27089 (aarch64_sve_float_arith_immediate): Likewise.
27090 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
27091 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
27092 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
27093 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
27094 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
27095 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
27096 (aarch64_sve_float_arith_operand): Likewise.
27097 (aarch64_sve_float_arith_with_sub_operand): Likewise.
27098 (aarch64_sve_float_mul_operand): Likewise.
27099 (aarch64_sve_vec_perm_operand): Likewise.
27100 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
27101 (aarch64_mov_operand): Accept const_poly_int and const_vector.
27102 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
27103 as well as const_vector.
27104 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
27105 in file. Use CONST0_RTX and CONSTM1_RTX.
27106 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
27107 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
27108 Use aarch64_simd_imm_zero.
27109 * config/aarch64/aarch64-sve.md: New file.
27110 * config/aarch64/aarch64.md: Include it.
27111 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
27112 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
27113 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
27114 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
27115 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
27116 (sve): New attribute.
27117 (enabled): Disable instructions with the sve attribute unless
27119 (movqi, movhi): Pass CONST_POLY_INT operaneds through
27120 aarch64_expand_mov_immediate.
27121 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
27122 CNT[BHSD] immediates.
27123 (movti): Split CONST_POLY_INT moves into two halves.
27124 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
27125 Split additions that need a temporary here if the destination
27126 is the stack pointer.
27127 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
27128 (*add<mode>3_poly_1): New instruction.
27129 (set_clobber_cc): New expander.
27131 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
27133 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
27134 parameter and use it instead of GET_MODE_SIZE (innermode). Use
27135 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
27136 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
27137 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
27138 Change innermode from fixed_mode_size to machine_mode.
27139 (simplify_subreg): Update call accordingly. Handle a constant-sized
27140 subreg of a variable-length CONST_VECTOR.
27142 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
27143 Alan Hayward <alan.hayward@arm.com>
27144 David Sherwood <david.sherwood@arm.com>
27146 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
27147 (add_offset_to_base): New function, split out from...
27148 (create_mem_ref): ...here. When handling a scale other than 1,
27149 check first whether the address is valid without the offset.
27150 Add it into the base if so, leaving the index and scale as-is.
27152 2018-01-12 Jakub Jelinek <jakub@redhat.com>
27155 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
27156 fold_for_warn before checking if arg2 is INTEGER_CST.
27158 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
27160 * config/rs6000/predicates.md (load_multiple_operation): Delete.
27161 (store_multiple_operation): Delete.
27162 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
27163 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
27164 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
27165 guarded by TARGET_STRING.
27166 (rs6000_output_load_multiple): Delete.
27167 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
27168 OPTION_MASK_STRING / TARGET_STRING handling.
27169 (print_operand) <'N', 'O'>: Add comment that these are unused now.
27170 (const rs6000_opt_masks) <"string">: Change mask to 0.
27171 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
27172 (MASK_STRING): Delete.
27173 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
27175 (load_multiple): Delete.
27182 (store_multiple): Delete.
27189 (movmemsi_8reg): Delete.
27190 (corresponding unnamed define_insn): Delete.
27191 (movmemsi_6reg): Delete.
27192 (corresponding unnamed define_insn): Delete.
27193 (movmemsi_4reg): Delete.
27194 (corresponding unnamed define_insn): Delete.
27195 (movmemsi_2reg): Delete.
27196 (corresponding unnamed define_insn): Delete.
27197 (movmemsi_1reg): Delete.
27198 (corresponding unnamed define_insn): Delete.
27199 * config/rs6000/rs6000.opt (mno-string): New.
27200 (mstring): Replace by deprecation warning stub.
27201 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
27203 2018-01-12 Jakub Jelinek <jakub@redhat.com>
27205 * regrename.c (regrename_do_replace): If replacing the same
27206 reg multiple times, try to reuse last created gen_raw_REG.
27209 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
27210 main to workaround a bug in GDB.
27212 2018-01-12 Tom de Vries <tom@codesourcery.com>
27215 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
27217 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
27219 PR rtl-optimization/80481
27220 * ira-color.c (get_cap_member): New function.
27221 (allocnos_conflict_by_live_ranges_p): Use it.
27222 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
27223 (setup_slot_coalesced_allocno_live_ranges): Ditto.
27225 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
27228 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
27229 (*saddl_se_1): Ditto.
27230 (*ssubsi_1): Ditto.
27231 (*ssubl_se_1): Ditto.
27233 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
27235 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
27236 rather than wi::to_widest for DR_INITs.
27237 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
27238 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
27239 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
27241 (vect_analyze_group_access_1): Note that here.
27243 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
27245 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
27246 polynomial type sizes.
27248 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
27250 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
27251 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
27252 (gimple_add_tmp_var): Likewise.
27254 2018-01-12 Martin Liska <mliska@suse.cz>
27256 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
27257 (gimple_alloc_sizes): Likewise.
27258 (dump_gimple_statistics): Use PRIu64 in printf format.
27259 * gimple.h: Change uint64_t to int.
27261 2018-01-12 Martin Liska <mliska@suse.cz>
27263 * tree-core.h: Use uint64_t instead of int.
27264 * tree.c (tree_node_counts): Likewise.
27265 (tree_node_sizes): Likewise.
27266 (dump_tree_statistics): Use PRIu64 in printf format.
27268 2018-01-12 Martin Liska <mliska@suse.cz>
27270 * Makefile.in: As qsort_chk is implemented in vec.c, add
27271 vec.o to linkage of gencfn-macros.
27272 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
27273 passing the info to record_node_allocation_statistics.
27274 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
27276 * ggc-common.c (struct ggc_usage): Add operator== and use
27277 it in operator< and compare function.
27278 * mem-stats.h (struct mem_usage): Likewise.
27279 * vec.c (struct vec_usage): Remove operator< and compare
27280 function. Can be simply inherited.
27282 2018-01-12 Martin Jambor <mjambor@suse.cz>
27285 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
27286 * tree-ssa-math-opts.c: Include domwalk.h.
27287 (convert_mult_to_fma_1): New function.
27288 (fma_transformation_info): New type.
27289 (fma_deferring_state): Likewise.
27290 (cancel_fma_deferring): New function.
27291 (result_of_phi): Likewise.
27292 (last_fma_candidate_feeds_initial_phi): Likewise.
27293 (convert_mult_to_fma): Added deferring logic, split actual
27294 transformation to convert_mult_to_fma_1.
27295 (math_opts_dom_walker): New type.
27296 (math_opts_dom_walker::after_dom_children): New method, body moved
27297 here from pass_optimize_widening_mul::execute, added deferring logic
27299 (pass_optimize_widening_mul::execute): Moved most of code to
27300 math_opts_dom_walker::after_dom_children.
27301 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
27302 * config/i386/i386.c (ix86_option_override_internal): Added
27303 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
27305 2018-01-12 Richard Biener <rguenther@suse.de>
27308 * dwarf2out.c (gen_variable_die): Do not reset old_die for
27309 inline instance vars.
27311 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
27314 * config/rx/rx.c (rx_is_restricted_memory_address):
27315 Handle SUBREG case.
27317 2018-01-12 Richard Biener <rguenther@suse.de>
27319 PR tree-optimization/80846
27320 * target.def (split_reduction): New target hook.
27321 * targhooks.c (default_split_reduction): New function.
27322 * targhooks.h (default_split_reduction): Declare.
27323 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
27324 target requests first reduce vectors by combining low and high
27326 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
27327 (get_vectype_for_scalar_type_and_size): Export.
27328 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
27329 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
27330 * doc/tm.texi: Regenerate.
27331 * config/i386/i386.c (ix86_split_reduction): Implement
27332 TARGET_VECTORIZE_SPLIT_REDUCTION.
27334 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
27337 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
27338 in PIC mode except for TARGET_VXWORKS_RTP.
27339 * config/sparc/sparc.c: Include cfgrtl.h.
27340 (TARGET_INIT_PIC_REG): Define.
27341 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
27342 (sparc_pic_register_p): New predicate.
27343 (sparc_legitimate_address_p): Use it.
27344 (sparc_legitimize_pic_address): Likewise.
27345 (sparc_delegitimize_address): Likewise.
27346 (sparc_mode_dependent_address_p): Likewise.
27347 (gen_load_pcrel_sym): Remove 4th parameter.
27348 (load_got_register): Adjust call to above. Remove obsolete stuff.
27349 (sparc_expand_prologue): Do not call load_got_register here.
27350 (sparc_flat_expand_prologue): Likewise.
27351 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
27352 (sparc_use_pseudo_pic_reg): New function.
27353 (sparc_init_pic_reg): Likewise.
27354 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
27355 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
27357 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
27359 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
27360 Add item for branch_cost.
27362 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
27364 PR rtl-optimization/83565
27365 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
27366 not extend the result to a larger mode for rotate operations.
27367 (num_sign_bit_copies1): Likewise.
27369 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
27372 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
27374 Use values-Xc.o for -pedantic.
27375 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
27377 2018-01-12 Martin Liska <mliska@suse.cz>
27380 * ipa-devirt.c (final_warning_record::grow_type_warnings):
27382 (possible_polymorphic_call_targets): Use it.
27383 (ipa_devirt): Likewise.
27385 2018-01-12 Martin Liska <mliska@suse.cz>
27387 * profile-count.h (enum profile_quality): Use 0 as invalid
27388 enum value of profile_quality.
27390 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
27392 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
27393 -mext-string options.
27395 2018-01-12 Richard Biener <rguenther@suse.de>
27397 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
27398 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
27399 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
27401 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
27403 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
27405 * configure.ac (--with-long-double-format): Add support for the
27406 configuration option to change the default long double format on
27408 * config.gcc (powerpc*-linux*-*): Likewise.
27409 * configure: Regenerate.
27410 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
27411 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
27412 used without modification.
27414 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
27416 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
27417 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
27418 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
27419 MISC_BUILTIN_SPEC_BARRIER.
27420 (rs6000_init_builtins): Likewise.
27421 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
27423 (speculation_barrier): New define_insn.
27424 * doc/extend.texi: Document __builtin_speculation_barrier.
27426 2018-01-11 Jakub Jelinek <jakub@redhat.com>
27429 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
27430 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
27431 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
27433 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
27434 integral modes instead of "ss" and "sd".
27435 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
27436 vectors with 32-bit and 64-bit elements.
27437 (vecdupssescalarmodesuffix): New mode attribute.
27438 (vec_dup<mode>): Use it.
27440 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
27443 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
27444 frame if argument is passed on stack.
27446 2018-01-11 Jakub Jelinek <jakub@redhat.com>
27449 * ree.c (combine_reaching_defs): Optimize also
27450 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
27451 reg2=any_extend(exp); reg1=reg2;, formatting fix.
27453 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
27455 PR middle-end/83189
27456 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
27458 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
27460 PR middle-end/83718
27461 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
27462 after they are computed.
27464 2018-01-11 Bin Cheng <bin.cheng@arm.com>
27466 PR tree-optimization/83695
27467 * gimple-loop-linterchange.cc
27468 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
27469 reset cached scev information after interchange.
27470 (pass_linterchange::execute): Remove call to scev_reset_htab.
27472 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27474 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
27475 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
27476 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
27477 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
27478 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
27479 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
27480 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
27481 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
27482 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
27483 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
27484 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
27485 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
27486 (V_lane_reg): Likewise.
27487 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
27489 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
27490 (vfmal_lane_low<mode>_intrinsic,
27491 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
27492 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
27493 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
27494 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
27495 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
27496 vfmsl_lane_high<mode>_intrinsic): New define_insns.
27498 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27500 * config/arm/arm-cpus.in (fp16fml): New feature.
27501 (ALL_SIMD): Add fp16fml.
27502 (armv8.2-a): Add fp16fml as an option.
27503 (armv8.3-a): Likewise.
27504 (armv8.4-a): Add fp16fml as part of fp16.
27505 * config/arm/arm.h (TARGET_FP16FML): Define.
27506 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
27508 * config/arm/arm-modes.def (V2HF): Define.
27509 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
27510 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
27511 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
27512 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
27513 vfmsl_low, vfmsl_high): New set of builtins.
27514 * config/arm/iterators.md (PLUSMINUS): New code iterator.
27515 (vfml_op): New code attribute.
27516 (VFMLHALVES): New int iterator.
27517 (VFML, VFMLSEL): New mode attributes.
27518 (V_reg): Define mapping for V2HF.
27519 (V_hi, V_lo): New mode attributes.
27520 (VF_constraint): Likewise.
27521 (vfml_half, vfml_half_selector): New int attributes.
27522 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
27524 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
27525 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
27527 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
27528 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
27529 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
27530 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
27532 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
27533 Document new effective target and option set.
27535 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27537 * config/arm/arm-cpus.in (armv8_4): New feature.
27538 (ARMv8_4a): New fgroup.
27539 (armv8.4-a): New arch.
27540 * config/arm/arm-tables.opt: Regenerate.
27541 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
27542 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
27543 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
27544 Add matching rules for -march=armv8.4-a and extensions.
27545 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
27547 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
27550 * config/rx/rx.md (BW): New mode attribute.
27551 (sync_lock_test_and_setsi): Add mode suffix to insn output.
27553 2018-01-11 Richard Biener <rguenther@suse.de>
27555 PR tree-optimization/83435
27556 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
27557 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
27558 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
27560 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
27561 Alan Hayward <alan.hayward@arm.com>
27562 David Sherwood <david.sherwood@arm.com>
27564 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
27566 (aarch64_classify_address): Initialize it. Track polynomial offsets.
27567 (aarch64_print_address_internal): Use it to check for a zero offset.
27569 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
27570 Alan Hayward <alan.hayward@arm.com>
27571 David Sherwood <david.sherwood@arm.com>
27573 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
27574 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
27575 Return a poly_int64 rather than a HOST_WIDE_INT.
27576 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
27577 rather than a HOST_WIDE_INT.
27578 * config/aarch64/aarch64.h (aarch64_frame): Protect with
27579 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
27580 hard_fp_offset, frame_size, initial_adjust, callee_offset and
27581 final_offset from HOST_WIDE_INT to poly_int64.
27582 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
27583 to_constant when getting the number of units in an Advanced SIMD
27585 (aarch64_builtin_vectorized_function): Check for a constant number
27587 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
27589 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
27590 attribute instead of GET_MODE_NUNITS.
27591 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
27592 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
27593 GET_MODE_SIZE for fixed-size registers.
27594 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
27595 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
27596 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
27597 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
27598 (aarch64_print_operand, aarch64_print_address_internal)
27599 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
27600 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
27601 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
27602 Handle polynomial GET_MODE_SIZE.
27603 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
27604 wider than SImode without modification.
27605 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
27606 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
27607 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
27608 passing and returning SVE modes.
27609 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
27610 rather than GEN_INT.
27611 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
27612 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
27613 (aarch64_allocate_and_probe_stack_space): Likewise.
27614 (aarch64_layout_frame): Cope with polynomial offsets.
27615 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
27616 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
27617 polynomial offsets.
27618 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
27619 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
27620 poly_int64 rather than a HOST_WIDE_INT.
27621 (aarch64_get_separate_components, aarch64_process_components)
27622 (aarch64_expand_prologue, aarch64_expand_epilogue)
27623 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
27624 (aarch64_anchor_offset): New function, split out from...
27625 (aarch64_legitimize_address): ...here.
27626 (aarch64_builtin_vectorization_cost): Handle polynomial
27627 TYPE_VECTOR_SUBPARTS.
27628 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
27630 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
27631 number of elements from the PARALLEL rather than the mode.
27632 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
27633 rather than GET_MODE_BITSIZE.
27634 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
27635 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
27636 (aarch64_expand_vec_perm_const_1): Handle polynomial
27637 d->perm.length () and d->perm elements.
27638 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
27639 Apply to_constant to d->perm elements.
27640 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
27641 polynomial CONST_VECTOR_NUNITS.
27642 (aarch64_move_pointer): Take amount as a poly_int64 rather
27644 (aarch64_progress_pointer): Avoid temporary variable.
27645 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
27646 the mode attribute instead of GET_MODE.
27648 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
27649 Alan Hayward <alan.hayward@arm.com>
27650 David Sherwood <david.sherwood@arm.com>
27652 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
27653 x exists before using it.
27654 (aarch64_add_constant_internal): Rename to...
27655 (aarch64_add_offset_1): ...this. Replace regnum with separate
27656 src and dest rtxes. Handle the case in which they're different,
27657 including when the offset is zero. Replace scratchreg with an rtx.
27658 Use 2 additions if there is no spare register into which we can
27659 move a 16-bit constant.
27660 (aarch64_add_constant): Delete.
27661 (aarch64_add_offset): Replace reg with separate src and dest
27662 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
27663 Use aarch64_add_offset_1.
27664 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
27665 an rtx rather than an int. Take the delta as a poly_int64
27666 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
27667 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
27668 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
27669 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
27670 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
27671 and aarch64_add_sp.
27672 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
27673 aarch64_add_constant.
27675 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
27677 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
27678 Use scalar_float_mode.
27680 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
27682 * config/aarch64/aarch64-simd.md
27683 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
27684 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
27685 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
27686 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
27687 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
27688 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
27689 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
27690 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
27691 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
27692 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
27694 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
27697 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
27698 targ_options->x_arm_arch_string is non NULL.
27700 2018-01-11 Tamar Christina <tamar.christina@arm.com>
27702 * config/aarch64/aarch64.h
27703 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
27705 2018-01-11 Sudakshina Das <sudi.das@arm.com>
27708 * expmed.c (emit_store_flag_force): Swap if const op0
27709 and change VOIDmode to mode of op0.
27711 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
27713 PR rtl-optimization/83761
27714 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
27715 than bytes to mode_for_size.
27717 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
27719 PR middle-end/83189
27720 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
27721 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
27724 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
27726 PR middle-end/83575
27727 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
27728 when in layout mode.
27729 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
27730 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
27733 2018-01-10 Michael Collison <michael.collison@arm.com>
27735 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
27736 * config/aarch64/aarch64-option-extension.def: Add
27737 AARCH64_OPT_EXTENSION of 'fp16fml'.
27738 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
27739 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
27740 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
27741 * config/aarch64/constraints.md (Ui7): New constraint.
27742 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
27743 (VFMLA_SEL_W): Ditto.
27746 (VFMLA16_LOW): New int iterator.
27747 (VFMLA16_HIGH): Ditto.
27748 (UNSPEC_FMLAL): New unspec.
27749 (UNSPEC_FMLSL): Ditto.
27750 (UNSPEC_FMLAL2): Ditto.
27751 (UNSPEC_FMLSL2): Ditto.
27752 (f16mac): New code attribute.
27753 * config/aarch64/aarch64-simd-builtins.def
27754 (aarch64_fmlal_lowv2sf): Ditto.
27755 (aarch64_fmlsl_lowv2sf): Ditto.
27756 (aarch64_fmlalq_lowv4sf): Ditto.
27757 (aarch64_fmlslq_lowv4sf): Ditto.
27758 (aarch64_fmlal_highv2sf): Ditto.
27759 (aarch64_fmlsl_highv2sf): Ditto.
27760 (aarch64_fmlalq_highv4sf): Ditto.
27761 (aarch64_fmlslq_highv4sf): Ditto.
27762 (aarch64_fmlal_lane_lowv2sf): Ditto.
27763 (aarch64_fmlsl_lane_lowv2sf): Ditto.
27764 (aarch64_fmlal_laneq_lowv2sf): Ditto.
27765 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
27766 (aarch64_fmlalq_lane_lowv4sf): Ditto.
27767 (aarch64_fmlsl_lane_lowv4sf): Ditto.
27768 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
27769 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
27770 (aarch64_fmlal_lane_highv2sf): Ditto.
27771 (aarch64_fmlsl_lane_highv2sf): Ditto.
27772 (aarch64_fmlal_laneq_highv2sf): Ditto.
27773 (aarch64_fmlsl_laneq_highv2sf): Ditto.
27774 (aarch64_fmlalq_lane_highv4sf): Ditto.
27775 (aarch64_fmlsl_lane_highv4sf): Ditto.
27776 (aarch64_fmlalq_laneq_highv4sf): Ditto.
27777 (aarch64_fmlsl_laneq_highv4sf): Ditto.
27778 * config/aarch64/aarch64-simd.md:
27779 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
27780 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
27781 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
27782 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
27783 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
27784 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
27785 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
27786 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
27787 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
27788 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
27789 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
27790 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
27791 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
27792 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
27793 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
27794 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
27795 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
27796 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
27797 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
27798 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
27799 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
27800 (vfmlsl_low_u32): Ditto.
27801 (vfmlalq_low_u32): Ditto.
27802 (vfmlslq_low_u32): Ditto.
27803 (vfmlal_high_u32): Ditto.
27804 (vfmlsl_high_u32): Ditto.
27805 (vfmlalq_high_u32): Ditto.
27806 (vfmlslq_high_u32): Ditto.
27807 (vfmlal_lane_low_u32): Ditto.
27808 (vfmlsl_lane_low_u32): Ditto.
27809 (vfmlal_laneq_low_u32): Ditto.
27810 (vfmlsl_laneq_low_u32): Ditto.
27811 (vfmlalq_lane_low_u32): Ditto.
27812 (vfmlslq_lane_low_u32): Ditto.
27813 (vfmlalq_laneq_low_u32): Ditto.
27814 (vfmlslq_laneq_low_u32): Ditto.
27815 (vfmlal_lane_high_u32): Ditto.
27816 (vfmlsl_lane_high_u32): Ditto.
27817 (vfmlal_laneq_high_u32): Ditto.
27818 (vfmlsl_laneq_high_u32): Ditto.
27819 (vfmlalq_lane_high_u32): Ditto.
27820 (vfmlslq_lane_high_u32): Ditto.
27821 (vfmlalq_laneq_high_u32): Ditto.
27822 (vfmlslq_laneq_high_u32): Ditto.
27823 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
27824 (AARCH64_FL_FOR_ARCH8_4): New.
27825 (AARCH64_ISA_F16FML): New ISA flag.
27826 (TARGET_F16FML): New feature flag for fp16fml.
27827 (doc/invoke.texi): Document new fp16fml option.
27829 2018-01-10 Michael Collison <michael.collison@arm.com>
27831 * config/aarch64/aarch64-builtins.c:
27832 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
27833 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
27834 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
27835 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
27836 (AARCH64_ISA_SHA3): New ISA flag.
27837 (TARGET_SHA3): New feature flag for sha3.
27838 * config/aarch64/iterators.md (sha512_op): New int attribute.
27839 (CRYPTO_SHA512): New int iterator.
27840 (UNSPEC_SHA512H): New unspec.
27841 (UNSPEC_SHA512H2): Ditto.
27842 (UNSPEC_SHA512SU0): Ditto.
27843 (UNSPEC_SHA512SU1): Ditto.
27844 * config/aarch64/aarch64-simd-builtins.def
27845 (aarch64_crypto_sha512hqv2di): New builtin.
27846 (aarch64_crypto_sha512h2qv2di): Ditto.
27847 (aarch64_crypto_sha512su0qv2di): Ditto.
27848 (aarch64_crypto_sha512su1qv2di): Ditto.
27849 (aarch64_eor3qv8hi): Ditto.
27850 (aarch64_rax1qv2di): Ditto.
27851 (aarch64_xarqv2di): Ditto.
27852 (aarch64_bcaxqv8hi): Ditto.
27853 * config/aarch64/aarch64-simd.md:
27854 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
27855 (aarch64_crypto_sha512su0qv2di): Ditto.
27856 (aarch64_crypto_sha512su1qv2di): Ditto.
27857 (aarch64_eor3qv8hi): Ditto.
27858 (aarch64_rax1qv2di): Ditto.
27859 (aarch64_xarqv2di): Ditto.
27860 (aarch64_bcaxqv8hi): Ditto.
27861 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
27862 (vsha512h2q_u64): Ditto.
27863 (vsha512su0q_u64): Ditto.
27864 (vsha512su1q_u64): Ditto.
27865 (veor3q_u16): Ditto.
27866 (vrax1q_u64): Ditto.
27867 (vxarq_u64): Ditto.
27868 (vbcaxq_u16): Ditto.
27869 * config/arm/types.md (crypto_sha512): New type attribute.
27870 (crypto_sha3): Ditto.
27871 (doc/invoke.texi): Document new sha3 option.
27873 2018-01-10 Michael Collison <michael.collison@arm.com>
27875 * config/aarch64/aarch64-builtins.c:
27876 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
27877 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
27878 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
27879 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
27880 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
27881 (AARCH64_ISA_SM4): New ISA flag.
27882 (TARGET_SM4): New feature flag for sm4.
27883 * config/aarch64/aarch64-simd-builtins.def
27884 (aarch64_sm3ss1qv4si): Ditto.
27885 (aarch64_sm3tt1aq4si): Ditto.
27886 (aarch64_sm3tt1bq4si): Ditto.
27887 (aarch64_sm3tt2aq4si): Ditto.
27888 (aarch64_sm3tt2bq4si): Ditto.
27889 (aarch64_sm3partw1qv4si): Ditto.
27890 (aarch64_sm3partw2qv4si): Ditto.
27891 (aarch64_sm4eqv4si): Ditto.
27892 (aarch64_sm4ekeyqv4si): Ditto.
27893 * config/aarch64/aarch64-simd.md:
27894 (aarch64_sm3ss1qv4si): Ditto.
27895 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
27896 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
27897 (aarch64_sm4eqv4si): Ditto.
27898 (aarch64_sm4ekeyqv4si): Ditto.
27899 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
27900 (sm3part_op): Ditto.
27901 (CRYPTO_SM3TT): Ditto.
27902 (CRYPTO_SM3PART): Ditto.
27903 (UNSPEC_SM3SS1): New unspec.
27904 (UNSPEC_SM3TT1A): Ditto.
27905 (UNSPEC_SM3TT1B): Ditto.
27906 (UNSPEC_SM3TT2A): Ditto.
27907 (UNSPEC_SM3TT2B): Ditto.
27908 (UNSPEC_SM3PARTW1): Ditto.
27909 (UNSPEC_SM3PARTW2): Ditto.
27910 (UNSPEC_SM4E): Ditto.
27911 (UNSPEC_SM4EKEY): Ditto.
27912 * config/aarch64/constraints.md (Ui2): New constraint.
27913 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
27914 * config/arm/types.md (crypto_sm3): New type attribute.
27915 (crypto_sm4): Ditto.
27916 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
27917 (vsm3tt1aq_u32): Ditto.
27918 (vsm3tt1bq_u32): Ditto.
27919 (vsm3tt2aq_u32): Ditto.
27920 (vsm3tt2bq_u32): Ditto.
27921 (vsm3partw1q_u32): Ditto.
27922 (vsm3partw2q_u32): Ditto.
27923 (vsm4eq_u32): Ditto.
27924 (vsm4ekeyq_u32): Ditto.
27925 (doc/invoke.texi): Document new sm4 option.
27927 2018-01-10 Michael Collison <michael.collison@arm.com>
27929 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
27930 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
27931 (AARCH64_FL_FOR_ARCH8_4): New.
27932 (AARCH64_FL_V8_4): New flag.
27933 (doc/invoke.texi): Document new armv8.4-a option.
27935 2018-01-10 Michael Collison <michael.collison@arm.com>
27937 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
27938 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
27939 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
27940 * config/aarch64/aarch64-option-extension.def: Add
27941 AARCH64_OPT_EXTENSION of 'sha2'.
27942 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
27943 (crypto): Disable sha2 and aes if crypto disabled.
27944 (crypto): Enable aes and sha2 if enabled.
27945 (simd): Disable sha2 and aes if simd disabled.
27946 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
27948 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
27949 (TARGET_SHA2): New feature flag for sha2.
27950 (TARGET_AES): New feature flag for aes.
27951 * config/aarch64/aarch64-simd.md:
27952 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
27953 conditional on TARGET_AES.
27954 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
27955 (aarch64_crypto_sha1hsi): Make pattern conditional
27957 (aarch64_crypto_sha1hv4si): Ditto.
27958 (aarch64_be_crypto_sha1hv4si): Ditto.
27959 (aarch64_crypto_sha1su1v4si): Ditto.
27960 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
27961 (aarch64_crypto_sha1su0v4si): Ditto.
27962 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
27963 (aarch64_crypto_sha256su0v4si): Ditto.
27964 (aarch64_crypto_sha256su1v4si): Ditto.
27965 (doc/invoke.texi): Document new aes and sha2 options.
27967 2018-01-10 Martin Sebor <msebor@redhat.com>
27969 PR tree-optimization/83781
27970 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
27973 2018-01-11 Martin Sebor <msebor@gmail.com>
27974 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
27976 PR tree-optimization/83501
27977 PR tree-optimization/81703
27979 * tree-ssa-strlen.c (get_string_cst): Rename...
27980 (get_string_len): ...to this. Handle global constants.
27981 (handle_char_store): Adjust.
27983 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
27984 Jim Wilson <jimw@sifive.com>
27986 * config/riscv/riscv-protos.h (riscv_output_return): New.
27987 * config/riscv/riscv.c (struct machine_function): New naked_p field.
27988 (riscv_attribute_table, riscv_output_return),
27989 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
27990 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
27991 (riscv_compute_frame_info): Only compute frame->mask if not a naked
27993 (riscv_expand_prologue): Add early return for naked function.
27994 (riscv_expand_epilogue): Likewise.
27995 (riscv_function_ok_for_sibcall): Return false for naked function.
27996 (riscv_set_current_function): New.
27997 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
27998 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
27999 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
28000 * doc/extend.texi (RISC-V Function Attributes): New.
28002 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
28004 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
28005 check for 128-bit long double before checking TCmode.
28006 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
28007 128-bit long doubles before checking TFmode or TCmode.
28008 (FLOAT128_IBM_P): Likewise.
28010 2018-01-10 Martin Sebor <msebor@redhat.com>
28012 PR tree-optimization/83671
28013 * builtins.c (c_strlen): Unconditionally return zero for the empty
28015 Use -Warray-bounds for warnings.
28016 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
28017 for non-constant array indices with COMPONENT_REF, arrays of
28018 arrays, and pointers to arrays.
28019 (gimple_fold_builtin_strlen): Determine and set length range for
28020 non-constant character arrays.
28022 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
28024 PR middle-end/81897
28025 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
28028 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
28030 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
28032 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
28035 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
28036 VECTOR_MEM_ALTIVEC_OR_VSX_P.
28037 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
28038 indexed_or_indirect_operand predicate.
28039 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
28040 (*vsx_le_perm_load_v8hi): Likewise.
28041 (*vsx_le_perm_load_v16qi): Likewise.
28042 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
28043 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
28044 (*vsx_le_perm_store_v8hi): Likewise.
28045 (*vsx_le_perm_store_v16qi): Likewise.
28046 (eight unnamed splitters): Likewise.
28048 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
28050 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
28051 * config/rs6000/emmintrin.h: Likewise.
28052 * config/rs6000/mmintrin.h: Likewise.
28053 * config/rs6000/xmmintrin.h: Likewise.
28055 2018-01-10 David Malcolm <dmalcolm@redhat.com>
28058 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
28060 * tree.c (tree_nop_conversion): Return true for location wrapper
28062 (maybe_wrap_with_location): New function.
28063 (selftest::check_strip_nops): New function.
28064 (selftest::test_location_wrappers): New function.
28065 (selftest::tree_c_tests): Call it.
28066 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
28067 (maybe_wrap_with_location): New decl.
28068 (EXPR_LOCATION_WRAPPER_P): New macro.
28069 (location_wrapper_p): New inline function.
28070 (tree_strip_any_location_wrapper): New inline function.
28072 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
28075 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
28076 stack_realign_offset for the largest alignment of stack slot
28078 (ix86_find_max_used_stack_alignment): New function.
28079 (ix86_finalize_stack_frame_flags): Use it. Set
28080 max_used_stack_alignment if we don't realign stack.
28081 * config/i386/i386.h (machine_function): Add
28082 max_used_stack_alignment.
28084 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
28086 * config/arm/arm.opt (-mbranch-cost): New option.
28087 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
28090 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
28093 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
28094 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
28096 2018-01-10 Richard Biener <rguenther@suse.de>
28099 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
28100 early out so it also covers the case where we have a non-NULL
28103 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
28105 PR tree-optimization/83753
28106 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
28107 for non-strided grouped accesses if the number of elements is 1.
28109 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
28112 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
28113 * i386.h (TARGET_USE_GATHER): Define.
28114 * x86-tune.def (X86_TUNE_USE_GATHER): New.
28116 2018-01-10 Martin Liska <mliska@suse.cz>
28119 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
28120 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
28122 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
28123 CLEANUP_NO_PARTITIONING is not set.
28125 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
28127 * doc/rtl.texi: Remove documentation of (const ...) wrappers
28128 for vectors, as a partial revert of r254296.
28129 * rtl.h (const_vec_p): Delete.
28130 (const_vec_duplicate_p): Don't test for vector CONSTs.
28131 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
28132 * expmed.c (make_tree): Likewise.
28135 * common.md (E, F): Use CONSTANT_P instead of checking for
28137 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
28138 checking for CONST_VECTOR.
28140 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
28142 PR middle-end/83575
28143 * predict.c (force_edge_cold): Handle in more sane way edges
28144 with no prediction.
28146 2018-01-09 Carl Love <cel@us.ibm.com>
28148 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
28150 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
28151 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
28152 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
28153 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
28154 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
28155 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
28156 * config/rs6000/rs6000-protos.h: Add extern defition for
28157 rs6000_generate_float2_double_code.
28158 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
28160 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
28161 (float2_v2df): Add define_expand.
28163 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
28166 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
28167 op_mode in the force_to_mode call.
28169 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
28171 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
28172 instead of checking each element individually.
28173 (aarch64_evpc_uzp): Likewise.
28174 (aarch64_evpc_zip): Likewise.
28175 (aarch64_evpc_ext): Likewise.
28176 (aarch64_evpc_rev): Likewise.
28177 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
28178 instead of checking each element individually. Return true without
28180 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
28181 whether all selected elements come from the same input, instead of
28182 checking each element individually. Remove calls to gen_rtx_REG,
28183 start_sequence and end_sequence and instead assert that no rtl is
28186 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
28188 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
28189 order of HIGH and CONST checks.
28191 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
28193 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
28194 if the destination isn't an SSA_NAME.
28196 2018-01-09 Richard Biener <rguenther@suse.de>
28198 PR tree-optimization/83668
28199 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
28201 (canonicalize_loop_form): ... here, renamed from ...
28202 (canonicalize_loop_closed_ssa_form): ... this and amended to
28203 swap successor edges for loop exit blocks to make us use
28204 the RPO order we need for initial schedule generation.
28206 2018-01-09 Joseph Myers <joseph@codesourcery.com>
28208 PR tree-optimization/64811
28209 * match.pd: When optimizing comparisons with Inf, avoid
28210 introducing or losing exceptions from comparisons with NaN.
28212 2018-01-09 Martin Liska <mliska@suse.cz>
28215 * asan.c (shadow_mem_size): Add gcc_assert.
28217 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
28219 Don't save registers in main().
28222 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
28223 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
28224 * config/avr/avr.c (avr_set_current_function): Don't error if
28225 naked, OS_task or OS_main are specified at the same time.
28226 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
28228 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
28230 * common/config/avr/avr-common.c (avr_option_optimization_table):
28231 Switch on -mmain-is-OS_task for optimizing compilations.
28233 2018-01-09 Richard Biener <rguenther@suse.de>
28235 PR tree-optimization/83572
28236 * graphite.c: Include cfganal.h.
28237 (graphite_transform_loops): Connect infinite loops to exit
28238 and remove fake edges at the end.
28240 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
28242 * ipa-inline.c (edge_badness): Revert accidental checkin.
28244 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
28247 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
28248 symbols; not inline clones.
28250 2018-01-09 Jakub Jelinek <jakub@redhat.com>
28253 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
28254 hard registers. Formatting fixes.
28256 PR preprocessor/83722
28257 * gcc.c (try_generate_repro): Pass
28258 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
28259 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
28262 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
28263 Kito Cheng <kito.cheng@gmail.com>
28265 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
28266 (riscv_leaf_function_p): Delete.
28267 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
28269 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
28271 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
28273 (do_ifelse): New function.
28274 (do_isel): New function.
28275 (do_sub3): New function.
28276 (do_add3): New function.
28277 (do_load_mask_compare): New function.
28278 (do_overlap_load_compare): New function.
28279 (expand_compare_loop): New function.
28280 (expand_block_compare): Call expand_compare_loop() when appropriate.
28281 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
28282 option description.
28283 (-mblock-compare-inline-loop-limit): New option.
28285 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
28288 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
28289 Reverse order of second and third operands in first alternative.
28290 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
28291 of first and second elements in UNSPEC_VPERMR vector.
28292 (altivec_expand_vec_perm_le): Likewise.
28294 2018-01-08 Jeff Law <law@redhat.com>
28296 PR rtl-optimizatin/81308
28297 * tree-switch-conversion.c (cfg_altered): New file scoped static.
28298 (process_switch): If group_case_labels makes a change, then set
28300 (pass_convert_switch::execute): If a switch is converted, then
28301 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
28303 PR rtl-optimization/81308
28304 * recog.c (split_all_insns): Conditionally cleanup the CFG after
28307 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
28309 PR target/83663 - Revert r255946
28310 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
28311 generation for cases where splatting a value is not useful.
28312 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
28313 across a vec_duplicate and a paradoxical subreg forming a vector
28314 mode to a vec_concat.
28316 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28318 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
28319 -march=armv8.3-a variants.
28320 * config/arm/t-multilib: Likewise.
28321 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
28323 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
28325 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
28327 (cceq_ior_compare_complement): Give it a name so I can use it, and
28328 change boolean_or_operator predicate to boolean_operator so it can
28329 be used to generate a crand.
28330 (eqne): New code iterator.
28331 (bd/bd_neg): New code_attrs.
28332 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
28333 a single define_insn.
28334 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
28335 decrement (bdnzt/bdnzf/bdzt/bdzf).
28336 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
28337 with the new names of the branch decrement patterns, and added the
28338 names of the branch decrement conditional patterns.
28340 2018-01-08 Richard Biener <rguenther@suse.de>
28342 PR tree-optimization/83563
28343 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
28346 2018-01-08 Richard Biener <rguenther@suse.de>
28348 PR middle-end/83713
28349 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
28351 2018-01-08 Richard Biener <rguenther@suse.de>
28353 PR tree-optimization/83685
28354 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
28355 references to abnormals.
28357 2018-01-08 Richard Biener <rguenther@suse.de>
28360 * dwarf2out.c (output_indirect_strings): Handle empty
28361 skeleton_debug_str_hash.
28362 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
28364 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
28366 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
28367 (emit_store_direct): Likewise.
28368 (arc_trampoline_adjust_address): Likewise.
28369 (arc_asm_trampoline_template): New function.
28370 (arc_initialize_trampoline): Use asm_trampoline_template.
28371 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
28372 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
28373 * config/arc/arc.md (flush_icache): Delete pattern.
28375 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
28377 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
28378 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
28381 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
28384 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
28385 by not USED_FOR_TARGET.
28386 (make_pass_resolve_sw_modes): Likewise.
28388 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
28390 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
28393 2018-01-08 Richard Biener <rguenther@suse.de>
28395 PR middle-end/83580
28396 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
28398 2018-01-08 Richard Biener <rguenther@suse.de>
28400 PR middle-end/83517
28401 * match.pd ((t * 2) / 2) -> t): Add missing :c.
28403 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
28405 PR middle-end/81897
28406 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
28407 basic blocks with a small number of successors.
28408 (convert_control_dep_chain_into_preds): Improve handling of
28410 (dump_predicates): Split apart into...
28411 (dump_pred_chain): ...here...
28412 (dump_pred_info): ...and here.
28413 (can_one_predicate_be_invalidated_p): Add debugging printfs.
28414 (can_chain_union_be_invalidated_p): Improve check for invalidation
28416 (uninit_uses_cannot_happen): Avoid unnecessary if
28417 convert_control_dep_chain_into_preds yielded nothing.
28419 2018-01-06 Martin Sebor <msebor@redhat.com>
28421 PR tree-optimization/83640
28422 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
28423 subtracting negative offset from size.
28424 (builtin_access::overlap): Adjust offset bounds of the access to fall
28425 within the size of the object if possible.
28427 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
28429 PR rtl-optimization/83699
28430 * expmed.c (extract_bit_field_1): Restrict the vector usage of
28431 extract_bit_field_as_subreg to cases in which the extracted
28432 value is also a vector.
28434 * lra-constraints.c (process_alt_operands): Test for the equivalence
28435 substitutions when detecting a possible reload cycle.
28437 2018-01-06 Jakub Jelinek <jakub@redhat.com>
28440 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
28441 by default if flag_selective_schedling{,2}. Formatting fixes.
28443 PR rtl-optimization/83682
28444 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
28445 if it has non-VECTOR_MODE element mode.
28446 (vec_duplicate_p): Likewise.
28448 PR middle-end/83694
28449 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
28450 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
28452 2018-01-05 Jakub Jelinek <jakub@redhat.com>
28455 * config/i386/i386-builtin.def
28456 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
28457 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
28458 Require also OPTION_MASK_ISA_AVX512F in addition to
28459 OPTION_MASK_ISA_GFNI.
28460 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
28461 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
28462 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
28463 to OPTION_MASK_ISA_GFNI.
28464 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
28465 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
28466 OPTION_MASK_ISA_AVX512BW.
28467 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
28468 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
28469 addition to OPTION_MASK_ISA_GFNI.
28470 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
28471 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
28472 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
28473 to OPTION_MASK_ISA_GFNI.
28474 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
28475 a requirement for all ISAs rather than any of them with a few
28477 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
28479 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
28480 bitmasks to be enabled with 3 exceptions, instead of requiring any
28481 enabled ISA with lots of exceptions.
28482 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
28483 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
28484 Change avx512bw in isa attribute to avx512f.
28485 * config/i386/sgxintrin.h: Add license boilerplate.
28486 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
28487 to __AVX512F__ and __AVX512VL to __AVX512VL__.
28488 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
28489 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
28491 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
28492 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
28493 temporarily sse2 rather than sse if not enabled already.
28496 * config/i386/sse.md (VI248_VLBW): Rename to ...
28497 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
28498 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
28499 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
28500 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
28501 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
28502 mode iterator instead of VI248_VLBW.
28504 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
28506 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
28507 (record_modified): Skip clobbers; add debug output.
28508 (param_change_prob): Use sreal frequencies.
28510 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
28512 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
28513 punt for user-aligned variables.
28515 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
28517 * tree-chrec.c (chrec_contains_symbols): Return true for
28520 2018-01-05 Sudakshina Das <sudi.das@arm.com>
28523 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
28524 of (x|y) == x for BICS pattern.
28526 2018-01-05 Jakub Jelinek <jakub@redhat.com>
28528 PR tree-optimization/83605
28529 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
28530 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
28533 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
28535 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
28536 * config/epiphany/rtems.h: New file.
28538 2018-01-04 Jakub Jelinek <jakub@redhat.com>
28539 Uros Bizjak <ubizjak@gmail.com>
28542 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
28543 QIreg_operand instead of register_operand predicate.
28544 * config/i386/i386.c (ix86_rop_should_change_byte_p,
28545 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
28546 comments instead of -fmitigate[-_]rop.
28548 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
28551 * cgraphunit.c (symbol_table::compile): Switch to text_section
28552 before calling assembly_start debug hook.
28553 * run-rtl-passes.c (run_rtl_passes): Likewise.
28556 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
28558 * tree-vrp.c (extract_range_from_binary_expr_1): Check
28559 range_int_cst_p rather than !symbolic_range_p before calling
28560 extract_range_from_multiplicative_op_1.
28562 2018-01-04 Jeff Law <law@redhat.com>
28564 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
28565 redundant test in assertion.
28567 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
28569 * doc/rtl.texi: Document machine_mode wrapper classes.
28571 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
28573 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
28574 using tree_to_uhwi.
28576 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
28578 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
28579 the VEC_PERM_EXPR fold to fail.
28581 2018-01-04 Jakub Jelinek <jakub@redhat.com>
28584 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
28585 to switched_sections.
28587 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
28590 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
28591 test for d.testing.
28593 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
28596 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
28597 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
28599 2018-01-04 Jakub Jelinek <jakub@redhat.com>
28602 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
28603 is BLKmode and bitpos not zero or mode change is needed.
28605 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
28608 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
28611 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
28614 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
28615 instead of MULT rtx. Update all corresponding splitters.
28616 (*saddl_se): Ditto.
28617 (*ssub<modesuffix>): Ditto.
28618 (*ssubl_se): Ditto.
28619 (*cmp_sadd_di): Update split patterns.
28620 (*cmp_sadd_si): Ditto.
28621 (*cmp_sadd_sidi): Ditto.
28622 (*cmp_ssub_di): Ditto.
28623 (*cmp_ssub_si): Ditto.
28624 (*cmp_ssub_sidi): Ditto.
28625 * config/alpha/predicates.md (const23_operand): New predicate.
28626 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
28627 Look for ASHIFT, not MULT inner operand.
28628 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
28630 2018-01-04 Martin Liska <mliska@suse.cz>
28632 PR gcov-profile/83669
28633 * gcov.c (output_intermediate_file): Add version to intermediate
28635 * doc/gcov.texi: Document new field 'version' in intermediate
28636 file format. Fix location of '-k' option of gcov command.
28638 2018-01-04 Martin Liska <mliska@suse.cz>
28641 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
28643 2018-01-04 Jakub Jelinek <jakub@redhat.com>
28645 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
28647 2018-01-03 Martin Sebor <msebor@redhat.com>
28649 PR tree-optimization/83655
28650 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
28651 checking calls with invalid arguments.
28653 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28655 * tree-vect-stmts.c (vect_get_store_rhs): New function.
28656 (vectorizable_mask_load_store): Delete.
28657 (vectorizable_call): Return false for masked loads and stores.
28658 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
28659 instead of gimple_assign_rhs1.
28660 (vectorizable_load): Handle IFN_MASK_LOAD.
28661 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
28663 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28665 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
28667 (vectorizable_mask_load_store): ...here.
28668 (vectorizable_load): ...and here.
28670 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28672 * tree-vect-stmts.c (vect_build_all_ones_mask)
28673 (vect_build_zero_merge_argument): New functions, split out from...
28674 (vectorizable_load): ...here.
28676 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28678 * tree-vect-stmts.c (vect_check_store_rhs): New function,
28680 (vectorizable_mask_load_store): ...here.
28681 (vectorizable_store): ...and here.
28683 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28685 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
28687 (vectorizable_mask_load_store): ...here.
28689 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28691 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
28692 (vect_model_store_cost): Take a vec_load_store_type instead of a
28694 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
28695 (vect_model_store_cost): Take a vec_load_store_type instead of a
28697 (vectorizable_mask_load_store): Update accordingly.
28698 (vectorizable_store): Likewise.
28699 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
28701 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28703 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
28704 IFN_MASK_LOAD calls here rather than...
28705 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
28707 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28708 Alan Hayward <alan.hayward@arm.com>
28709 David Sherwood <david.sherwood@arm.com>
28711 * expmed.c (extract_bit_field_1): For vector extracts,
28712 fall back to extract_bit_field_as_subreg if vec_extract
28715 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28716 Alan Hayward <alan.hayward@arm.com>
28717 David Sherwood <david.sherwood@arm.com>
28719 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
28720 they are variable or constant sized.
28721 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
28722 slots for constant-sized data.
28724 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28725 Alan Hayward <alan.hayward@arm.com>
28726 David Sherwood <david.sherwood@arm.com>
28728 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
28729 handling COND_EXPRs with boolean comparisons, try to find a better
28730 basis for the mask type than the boolean itself.
28732 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28734 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
28735 is calculated and how it can be overridden.
28736 * genmodes.c (max_bitsize_mode_any_mode): New variable.
28737 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
28739 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
28742 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28743 Alan Hayward <alan.hayward@arm.com>
28744 David Sherwood <david.sherwood@arm.com>
28746 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
28747 Remove the mode argument.
28748 (aarch64_simd_valid_immediate): Remove the mode and inverse
28750 * config/aarch64/iterators.md (bitsize): New iterator.
28751 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
28752 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
28753 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
28754 aarch64_simd_valid_immediate.
28755 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
28756 (aarch64_reg_or_bic_imm): Likewise.
28757 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
28758 with an insn_type enum and msl with a modifier_type enum.
28759 Replace element_width with a scalar_mode. Change the shift
28760 to unsigned int. Add constructors for scalar_float_mode and
28761 scalar_int_mode elements.
28762 (aarch64_vect_float_const_representable_p): Delete.
28763 (aarch64_can_const_movi_rtx_p)
28764 (aarch64_simd_scalar_immediate_valid_for_move)
28765 (aarch64_simd_make_constant): Update call to
28766 aarch64_simd_valid_immediate.
28767 (aarch64_advsimd_valid_immediate_hs): New function.
28768 (aarch64_advsimd_valid_immediate): Likewise.
28769 (aarch64_simd_valid_immediate): Remove mode and inverse
28770 arguments. Rewrite to use the above. Use const_vec_duplicate_p
28771 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
28772 and aarch64_float_const_representable_p on the result.
28773 (aarch64_output_simd_mov_immediate): Remove mode argument.
28774 Update call to aarch64_simd_valid_immediate and use of
28775 simd_immediate_info.
28776 (aarch64_output_scalar_simd_mov_immediate): Update call
28779 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28780 Alan Hayward <alan.hayward@arm.com>
28781 David Sherwood <david.sherwood@arm.com>
28783 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
28784 (mode_nunits): Likewise CONST_MODE_NUNITS.
28785 * machmode.def (ADJUST_NUNITS): Document.
28786 * genmodes.c (mode_data::need_nunits_adj): New field.
28787 (blank_mode): Update accordingly.
28788 (adj_nunits): New variable.
28789 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
28791 (emit_mode_size_inline): Set need_bytesize_adj for all modes
28792 listed in adj_nunits.
28793 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
28794 listed in adj_nunits. Don't emit case statements for such modes.
28795 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
28796 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
28797 nothing if adj_nunits is nonnull.
28798 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
28799 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
28800 (emit_mode_fbit): Update use of print_maybe_const_decl.
28801 (emit_move_size): Likewise. Treat the array as non-const
28803 (emit_mode_adjustments): Handle adj_nunits.
28805 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28807 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
28808 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
28809 (VECTOR_MODES): Use it.
28810 (make_vector_modes): Take the prefix as an argument.
28812 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28813 Alan Hayward <alan.hayward@arm.com>
28814 David Sherwood <david.sherwood@arm.com>
28816 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
28817 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
28818 for MODE_VECTOR_BOOL.
28819 * machmode.def (VECTOR_BOOL_MODE): Document.
28820 * genmodes.c (VECTOR_BOOL_MODE): New macro.
28821 (make_vector_bool_mode): New function.
28822 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
28824 * lto-streamer-in.c (lto_input_mode_table): Likewise.
28825 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
28827 * stor-layout.c (int_mode_for_mode): Likewise.
28828 * tree.c (build_vector_type_for_mode): Likewise.
28829 * varasm.c (output_constant_pool_2): Likewise.
28830 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
28831 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
28832 for MODE_VECTOR_BOOL.
28833 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
28834 of mode class checks.
28835 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
28836 instead of a list of mode class checks.
28837 (expand_vector_scalar_condition): Likewise.
28838 (type_for_widest_vector_mode): Handle BImode as an inner mode.
28840 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28841 Alan Hayward <alan.hayward@arm.com>
28842 David Sherwood <david.sherwood@arm.com>
28844 * machmode.h (mode_size): Change from unsigned short to
28846 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
28847 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
28848 or if measurement_type is not polynomial.
28849 (fixed_size_mode::includes_p): Check for constant-sized modes.
28850 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
28851 return a poly_uint16 rather than an unsigned short.
28852 (emit_mode_size): Change the type of mode_size from unsigned short
28853 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
28854 (emit_mode_adjustments): Cope with polynomial vector sizes.
28855 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
28857 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
28859 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
28860 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
28861 * caller-save.c (setup_save_areas): Likewise.
28862 (replace_reg_with_saved_mem): Likewise.
28863 * calls.c (emit_library_call_value_1): Likewise.
28864 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
28865 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
28866 (gen_lowpart_for_combine): Likewise.
28867 * convert.c (convert_to_integer_1): Likewise.
28868 * cse.c (equiv_constant, cse_insn): Likewise.
28869 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
28870 (cselib_subst_to_values): Likewise.
28871 * dce.c (word_dce_process_block): Likewise.
28872 * df-problems.c (df_word_lr_mark_ref): Likewise.
28873 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
28874 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
28875 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
28876 (rtl_for_decl_location): Likewise.
28877 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
28878 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
28879 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
28880 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
28881 (expand_expr_real_1): Likewise.
28882 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
28883 (pad_below): Likewise.
28884 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
28885 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
28886 * ira.c (get_subreg_tracking_sizes): Likewise.
28887 * ira-build.c (ira_create_allocno_objects): Likewise.
28888 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
28889 (ira_sort_regnos_for_alter_reg): Likewise.
28890 * ira-costs.c (record_operand_costs): Likewise.
28891 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
28892 (resolve_simple_move): Likewise.
28893 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
28894 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
28895 (lra_constraints): Likewise.
28896 (CONST_POOL_OK_P): Reject variable-sized modes.
28897 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
28898 (add_pseudo_to_slot, lra_spill): Likewise.
28899 * omp-low.c (omp_clause_aligned_alignment): Likewise.
28900 * optabs-query.c (get_best_extraction_insn): Likewise.
28901 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
28902 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
28903 (expand_mult_highpart, valid_multiword_target_p): Likewise.
28904 * recog.c (offsettable_address_addr_space_p): Likewise.
28905 * regcprop.c (maybe_mode_change): Likewise.
28906 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
28907 * regrename.c (build_def_use): Likewise.
28908 * regstat.c (dump_reg_info): Likewise.
28909 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
28910 (find_reloads, find_reloads_subreg_address): Likewise.
28911 * reload1.c (eliminate_regs_1): Likewise.
28912 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
28913 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
28914 (simplify_binary_operation_1, simplify_subreg): Likewise.
28915 * targhooks.c (default_function_arg_padding): Likewise.
28916 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
28917 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
28918 (verify_gimple_assign_ternary): Likewise.
28919 * tree-inline.c (estimate_move_cost): Likewise.
28920 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
28921 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
28922 (get_address_cost_ainc): Likewise.
28923 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
28924 (vect_supportable_dr_alignment): Likewise.
28925 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
28926 (vectorizable_reduction): Likewise.
28927 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
28928 (vectorizable_operation, vectorizable_load): Likewise.
28929 * tree.c (build_same_sized_truth_vector_type): Likewise.
28930 * valtrack.c (cleanup_auto_inc_dec): Likewise.
28931 * var-tracking.c (emit_note_insn_var_location): Likewise.
28932 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
28933 (ADDR_VEC_ALIGN): Likewise.
28935 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28936 Alan Hayward <alan.hayward@arm.com>
28937 David Sherwood <david.sherwood@arm.com>
28939 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
28941 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
28942 or if measurement_type is polynomial.
28943 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
28944 * combine.c (make_extraction): Likewise.
28945 * dse.c (find_shift_sequence): Likewise.
28946 * dwarf2out.c (mem_loc_descriptor): Likewise.
28947 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
28948 (extract_bit_field, extract_low_bits): Likewise.
28949 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
28950 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
28951 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
28952 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
28953 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
28954 * reload.c (find_reloads): Likewise.
28955 * reload1.c (alter_reg): Likewise.
28956 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
28957 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
28958 * tree-if-conv.c (predicate_mem_writes): Likewise.
28959 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
28960 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
28961 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
28962 * valtrack.c (dead_debug_insert_temp): Likewise.
28963 * varasm.c (mergeable_constant_section): Likewise.
28964 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
28966 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28967 Alan Hayward <alan.hayward@arm.com>
28968 David Sherwood <david.sherwood@arm.com>
28970 * expr.c (expand_assignment): Cope with polynomial mode sizes
28971 when assigning to a CONCAT.
28973 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28974 Alan Hayward <alan.hayward@arm.com>
28975 David Sherwood <david.sherwood@arm.com>
28977 * machmode.h (mode_precision): Change from unsigned short to
28979 (mode_to_precision): Return a poly_uint16 rather than an unsigned
28981 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
28982 or if measurement_type is not polynomial.
28983 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
28984 in which the mode is already known to be a scalar_int_mode.
28985 * genmodes.c (emit_mode_precision): Change the type of mode_precision
28986 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
28988 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
28989 for GET_MODE_PRECISION.
28990 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
28991 for GET_MODE_PRECISION.
28992 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
28994 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
28995 (expand_field_assignment, make_extraction): Likewise.
28996 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
28997 (get_last_value): Likewise.
28998 * convert.c (convert_to_integer_1): Likewise.
28999 * cse.c (cse_insn): Likewise.
29000 * expr.c (expand_expr_real_1): Likewise.
29001 * lra-constraints.c (simplify_operand_subreg): Likewise.
29002 * optabs-query.c (can_atomic_load_p): Likewise.
29003 * optabs.c (expand_atomic_load): Likewise.
29004 (expand_atomic_store): Likewise.
29005 * ree.c (combine_reaching_defs): Likewise.
29006 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
29007 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
29008 * tree.h (type_has_mode_precision_p): Likewise.
29009 * ubsan.c (instrument_si_overflow): Likewise.
29011 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29012 Alan Hayward <alan.hayward@arm.com>
29013 David Sherwood <david.sherwood@arm.com>
29015 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
29016 polynomial numbers of units.
29017 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
29018 (valid_vector_subparts_p): New function.
29019 (build_vector_type): Remove temporary shim and take the number
29020 of units as a poly_uint64 rather than an int.
29021 (build_opaque_vector_type): Take the number of units as a
29022 poly_uint64 rather than an int.
29023 * tree.c (build_vector_from_ctor): Handle polynomial
29024 TYPE_VECTOR_SUBPARTS.
29025 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
29026 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
29027 (build_vector_from_val): If the number of units is variable,
29028 use build_vec_duplicate_cst for constant operands and
29029 VEC_DUPLICATE_EXPR otherwise.
29030 (make_vector_type): Remove temporary is_constant ().
29031 (build_vector_type, build_opaque_vector_type): Take the number of
29032 units as a poly_uint64 rather than an int.
29033 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
29035 * cfgexpand.c (expand_debug_expr): Likewise.
29036 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
29037 (store_constructor, expand_expr_real_1): Likewise.
29038 (const_scalar_mask_from_tree): Likewise.
29039 * fold-const-call.c (fold_const_reduction): Likewise.
29040 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
29041 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
29042 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
29043 (fold_relational_const): Likewise.
29044 (native_interpret_vector): Likewise. Change the size from an
29045 int to an unsigned int.
29046 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
29047 TYPE_VECTOR_SUBPARTS.
29048 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
29049 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
29050 duplicating a non-constant operand into a variable-length vector.
29051 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
29052 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
29053 * ipa-icf.c (sem_variable::equals): Likewise.
29054 * match.pd: Likewise.
29055 * omp-simd-clone.c (simd_clone_subparts): Likewise.
29056 * print-tree.c (print_node): Likewise.
29057 * stor-layout.c (layout_type): Likewise.
29058 * targhooks.c (default_builtin_vectorization_cost): Likewise.
29059 * tree-cfg.c (verify_gimple_comparison): Likewise.
29060 (verify_gimple_assign_binary): Likewise.
29061 (verify_gimple_assign_ternary): Likewise.
29062 (verify_gimple_assign_single): Likewise.
29063 * tree-pretty-print.c (dump_generic_node): Likewise.
29064 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
29065 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
29066 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
29067 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
29068 (vect_shift_permute_load_chain): Likewise.
29069 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
29070 (expand_vector_condition, optimize_vector_constructor): Likewise.
29071 (lower_vec_perm, get_compute_type): Likewise.
29072 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
29073 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
29074 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
29075 (vect_recog_mask_conversion_pattern): Likewise.
29076 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
29077 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
29078 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
29079 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
29080 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
29081 (vectorizable_shift, vectorizable_operation, vectorizable_store)
29082 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
29083 (supportable_widening_operation): Likewise.
29084 (supportable_narrowing_operation): Likewise.
29085 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
29087 * varasm.c (output_constant): Likewise.
29089 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29090 Alan Hayward <alan.hayward@arm.com>
29091 David Sherwood <david.sherwood@arm.com>
29093 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
29094 so that both the length == 3 and length != 3 cases set up their
29095 own permute vectors. Add comments explaining why we know the
29096 number of elements is constant.
29097 (vect_permute_load_chain): Likewise.
29099 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29100 Alan Hayward <alan.hayward@arm.com>
29101 David Sherwood <david.sherwood@arm.com>
29103 * machmode.h (mode_nunits): Change from unsigned char to
29105 (ONLY_FIXED_SIZE_MODES): New macro.
29106 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
29107 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
29108 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
29110 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
29111 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
29112 or if measurement_type is not polynomial.
29113 * genmodes.c (ZERO_COEFFS): New macro.
29114 (emit_mode_nunits_inline): Make mode_nunits_inline return a
29116 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
29117 Use ZERO_COEFFS when emitting initializers.
29118 * data-streamer.h (bp_pack_poly_value): New function.
29119 (bp_unpack_poly_value): Likewise.
29120 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
29121 for GET_MODE_NUNITS.
29122 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
29123 for GET_MODE_NUNITS.
29124 * tree.c (make_vector_type): Remove temporary shim and make
29125 the real function take the number of units as a poly_uint64
29126 rather than an int.
29127 (build_vector_type_for_mode): Handle polynomial nunits.
29128 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
29129 * emit-rtl.c (const_vec_series_p_1): Likewise.
29130 (gen_rtx_CONST_VECTOR): Likewise.
29131 * fold-const.c (test_vec_duplicate_folding): Likewise.
29132 * genrecog.c (validate_pattern): Likewise.
29133 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
29134 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
29135 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
29136 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
29137 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
29138 * rtlanal.c (subreg_get_info): Likewise.
29139 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
29140 (vect_grouped_load_supported): Likewise.
29141 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
29142 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
29143 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
29144 (simplify_const_unary_operation, simplify_binary_operation_1)
29145 (simplify_const_binary_operation, simplify_ternary_operation)
29146 (test_vector_ops_duplicate, test_vector_ops): Likewise.
29147 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
29148 instead of CONST_VECTOR_NUNITS.
29149 * varasm.c (output_constant_pool_2): Likewise.
29150 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
29151 explicit-encoded elements in the XVEC for variable-length vectors.
29153 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29155 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
29157 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29158 Alan Hayward <alan.hayward@arm.com>
29159 David Sherwood <david.sherwood@arm.com>
29161 * coretypes.h (fixed_size_mode): Declare.
29162 (fixed_size_mode_pod): New typedef.
29163 * builtins.h (target_builtins::x_apply_args_mode)
29164 (target_builtins::x_apply_result_mode): Change type to
29165 fixed_size_mode_pod.
29166 * builtins.c (apply_args_size, apply_result_size, result_vector)
29167 (expand_builtin_apply_args_1, expand_builtin_apply)
29168 (expand_builtin_return): Update accordingly.
29170 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29172 * cse.c (hash_rtx_cb): Hash only the encoded elements.
29173 * cselib.c (cselib_hash_rtx): Likewise.
29174 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
29175 CONST_VECTOR encoding.
29177 2018-01-03 Jakub Jelinek <jakub@redhat.com>
29178 Jeff Law <law@redhat.com>
29181 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
29182 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
29183 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
29184 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
29187 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
29188 explicitly probe *sp in a noreturn function if there were any callee
29189 register saves or frame pointer is needed.
29191 2018-01-03 Jakub Jelinek <jakub@redhat.com>
29194 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
29195 BLKmode for ternary, binary or unary expressions.
29198 * var-tracking.c (delete_vta_debug_insn): New inline function.
29199 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
29200 insns from get_insns () to NULL instead of each bb separately.
29201 Use delete_vta_debug_insn. No longer static.
29202 (vt_debug_insns_local, variable_tracking_main_1): Adjust
29203 delete_vta_debug_insns callers.
29204 * rtl.h (delete_vta_debug_insns): Declare.
29205 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
29206 instead of variable_tracking_main.
29208 2018-01-03 Martin Sebor <msebor@redhat.com>
29210 PR tree-optimization/83603
29211 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
29212 arguments past the endof the argument list in functions declared
29213 without a prototype.
29214 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
29215 Avoid checking when arguments are null.
29217 2018-01-03 Martin Sebor <msebor@redhat.com>
29220 * doc/extend.texi (attribute const): Fix a typo.
29221 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
29222 issuing -Wsuggest-attribute for void functions.
29224 2018-01-03 Martin Sebor <msebor@redhat.com>
29226 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
29227 offset_int::from instead of wide_int::to_shwi.
29228 (maybe_diag_overlap): Remove assertion.
29229 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
29230 * gimple-ssa-sprintf.c (format_directive): Same.
29231 (parse_directive): Same.
29232 (sprintf_dom_walker::compute_format_length): Same.
29233 (try_substitute_return_value): Same.
29235 2018-01-03 Jeff Law <law@redhat.com>
29237 PR middle-end/83654
29238 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
29239 non-constant residual for zero at runtime and avoid probing in
29240 that case. Reorganize code for trailing problem to mirror handling
29243 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
29245 PR tree-optimization/83501
29246 * tree-ssa-strlen.c (get_string_cst): New.
29247 (handle_char_store): Call get_string_cst.
29249 2018-01-03 Martin Liska <mliska@suse.cz>
29251 PR tree-optimization/83593
29252 * tree-ssa-strlen.c: Include tree-cfg.h.
29253 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
29254 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
29255 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
29257 (strlen_dom_walker::before_dom_children): Call
29258 gimple_purge_dead_eh_edges. Dump tranformation with details
29260 (strlen_dom_walker::before_dom_children): Update call by adding
29261 new argument cleanup_eh.
29262 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
29264 2018-01-03 Martin Liska <mliska@suse.cz>
29267 * cif-code.def (VARIADIC_THUNK): New enum value.
29268 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
29271 2018-01-03 Jan Beulich <jbeulich@suse.com>
29273 * sse.md (mov<mode>_internal): Tighten condition for when to use
29274 vmovdqu<ssescalarsize> for TI and OI modes.
29276 2018-01-03 Jakub Jelinek <jakub@redhat.com>
29278 Update copyright years.
29280 2018-01-03 Martin Liska <mliska@suse.cz>
29283 * ipa-visibility.c (function_and_variable_visibility): Skip
29284 functions with noipa attribure.
29286 2018-01-03 Jakub Jelinek <jakub@redhat.com>
29288 * gcc.c (process_command): Update copyright notice dates.
29289 * gcov-dump.c (print_version): Ditto.
29290 * gcov.c (print_version): Ditto.
29291 * gcov-tool.c (print_version): Ditto.
29292 * gengtype.c (create_file): Ditto.
29293 * doc/cpp.texi: Bump @copying's copyright year.
29294 * doc/cppinternals.texi: Ditto.
29295 * doc/gcc.texi: Ditto.
29296 * doc/gccint.texi: Ditto.
29297 * doc/gcov.texi: Ditto.
29298 * doc/install.texi: Ditto.
29299 * doc/invoke.texi: Ditto.
29301 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29303 * vector-builder.h (vector_builder::m_full_nelts): Change from
29304 unsigned int to poly_uint64.
29305 (vector_builder::full_nelts): Update prototype accordingly.
29306 (vector_builder::new_vector): Likewise.
29307 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
29308 (vector_builder::operator ==): Likewise.
29309 (vector_builder::finalize): Likewise.
29310 * int-vector-builder.h (int_vector_builder::int_vector_builder):
29311 Take the number of elements as a poly_uint64 rather than an
29313 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
29314 from unsigned int to poly_uint64.
29315 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
29316 (vec_perm_indices::new_vector): Likewise.
29317 (vec_perm_indices::length): Likewise.
29318 (vec_perm_indices::nelts_per_input): Likewise.
29319 (vec_perm_indices::input_nelts): Likewise.
29320 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
29321 number of elements per input as a poly_uint64 rather than an
29322 unsigned int. Use the original encoding for variable-length
29323 vectors, rather than clamping each individual element.
29324 For the second and subsequent elements in each pattern,
29325 clamp the step and base before clamping their sum.
29326 (vec_perm_indices::series_p): Handle polynomial element counts.
29327 (vec_perm_indices::all_in_range_p): Likewise.
29328 (vec_perm_indices_to_tree): Likewise.
29329 (vec_perm_indices_to_rtx): Likewise.
29330 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
29331 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
29332 (tree_vector_builder::new_binary_operation): Handle polynomial
29333 element counts. Return false if we need to know the number
29334 of elements at compile time.
29335 * fold-const.c (fold_vec_perm): Punt if the number of elements
29336 isn't known at compile time.
29338 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29340 * vec-perm-indices.h (vec_perm_builder): Change element type
29341 from HOST_WIDE_INT to poly_int64.
29342 (vec_perm_indices::element_type): Update accordingly.
29343 (vec_perm_indices::clamp): Handle polynomial element_types.
29344 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
29345 (vec_perm_indices::all_in_range_p): Likewise.
29346 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
29348 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
29349 polynomial vec_perm_indices element types.
29350 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
29351 * fold-const.c (fold_vec_perm): Likewise.
29352 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
29353 * tree-vect-generic.c (lower_vec_perm): Likewise.
29354 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
29355 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
29356 element type to HOST_WIDE_INT.
29358 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29359 Alan Hayward <alan.hayward@arm.com>
29360 David Sherwood <david.sherwood@arm.com>
29362 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
29363 rather than an int. Use plus_constant.
29364 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
29365 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
29367 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29368 Alan Hayward <alan.hayward@arm.com>
29369 David Sherwood <david.sherwood@arm.com>
29371 * calls.c (emit_call_1, expand_call): Change struct_value_size from
29372 a HOST_WIDE_INT to a poly_int64.
29374 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29375 Alan Hayward <alan.hayward@arm.com>
29376 David Sherwood <david.sherwood@arm.com>
29378 * calls.c (load_register_parameters): Cope with polynomial
29379 mode sizes. Require a constant size for BLKmode parameters
29380 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
29381 forces a parameter to be padded at the lsb end in order to
29382 fill a complete number of words, require the parameter size
29383 to be ordered wrt UNITS_PER_WORD.
29385 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29386 Alan Hayward <alan.hayward@arm.com>
29387 David Sherwood <david.sherwood@arm.com>
29389 * reload1.c (spill_stack_slot_width): Change element type
29390 from unsigned int to poly_uint64_pod.
29391 (alter_reg): Treat mode sizes as polynomial.
29393 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29394 Alan Hayward <alan.hayward@arm.com>
29395 David Sherwood <david.sherwood@arm.com>
29397 * reload.c (complex_word_subreg_p): New function.
29398 (reload_inner_reg_of_subreg, push_reload): Use it.
29400 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29401 Alan Hayward <alan.hayward@arm.com>
29402 David Sherwood <david.sherwood@arm.com>
29404 * lra-constraints.c (process_alt_operands): Reject matched
29405 operands whose sizes aren't ordered.
29406 (match_reload): Refer to this check here.
29408 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29409 Alan Hayward <alan.hayward@arm.com>
29410 David Sherwood <david.sherwood@arm.com>
29412 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
29413 that the mode size is in the set {1, 2, 4, 8, 16}.
29415 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29416 Alan Hayward <alan.hayward@arm.com>
29417 David Sherwood <david.sherwood@arm.com>
29419 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
29420 Use plus_constant instead of gen_rtx_PLUS.
29422 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29423 Alan Hayward <alan.hayward@arm.com>
29424 David Sherwood <david.sherwood@arm.com>
29426 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
29427 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
29428 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
29429 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
29430 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
29431 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
29432 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
29433 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
29434 * config/i386/i386.c (ix86_push_rounding): ...this new function.
29435 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
29437 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
29438 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
29439 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
29440 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
29441 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
29442 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
29443 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
29444 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
29445 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
29446 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
29448 * expr.c (emit_move_resolve_push): Treat the input and result
29449 of PUSH_ROUNDING as a poly_int64.
29450 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
29451 (emit_push_insn): Likewise.
29452 * lra-eliminations.c (mark_not_eliminable): Likewise.
29453 * recog.c (push_operand): Likewise.
29454 * reload1.c (elimination_effects): Likewise.
29455 * rtlanal.c (nonzero_bits1): Likewise.
29456 * calls.c (store_one_arg): Likewise. Require the padding to be
29457 known at compile time.
29459 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29460 Alan Hayward <alan.hayward@arm.com>
29461 David Sherwood <david.sherwood@arm.com>
29463 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
29464 Use plus_constant instead of gen_rtx_PLUS.
29466 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29467 Alan Hayward <alan.hayward@arm.com>
29468 David Sherwood <david.sherwood@arm.com>
29470 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
29471 rather than an int.
29473 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29474 Alan Hayward <alan.hayward@arm.com>
29475 David Sherwood <david.sherwood@arm.com>
29477 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
29478 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
29479 via stack temporaries. Treat the mode size as polynomial too.
29481 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29482 Alan Hayward <alan.hayward@arm.com>
29483 David Sherwood <david.sherwood@arm.com>
29485 * expr.c (expand_expr_real_2): When handling conversions involving
29486 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
29487 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
29488 as a poly_uint64 too.
29490 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29491 Alan Hayward <alan.hayward@arm.com>
29492 David Sherwood <david.sherwood@arm.com>
29494 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
29496 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29497 Alan Hayward <alan.hayward@arm.com>
29498 David Sherwood <david.sherwood@arm.com>
29500 * combine.c (can_change_dest_mode): Handle polynomial
29501 REGMODE_NATURAL_SIZE.
29502 * expmed.c (store_bit_field_1): Likewise.
29503 * expr.c (store_constructor): Likewise.
29504 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
29505 and polynomial REGMODE_NATURAL_SIZE.
29506 (gen_lowpart_common): Likewise.
29507 * reginfo.c (record_subregs_of_mode): Likewise.
29508 * rtlanal.c (read_modify_subreg_p): Likewise.
29510 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29511 Alan Hayward <alan.hayward@arm.com>
29512 David Sherwood <david.sherwood@arm.com>
29514 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
29515 numbers of elements.
29517 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29518 Alan Hayward <alan.hayward@arm.com>
29519 David Sherwood <david.sherwood@arm.com>
29521 * match.pd: Cope with polynomial numbers of vector elements.
29523 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29524 Alan Hayward <alan.hayward@arm.com>
29525 David Sherwood <david.sherwood@arm.com>
29527 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
29528 in a POINTER_PLUS_EXPR.
29530 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29531 Alan Hayward <alan.hayward@arm.com>
29532 David Sherwood <david.sherwood@arm.com>
29534 * omp-simd-clone.c (simd_clone_subparts): New function.
29535 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
29536 (ipa_simd_modify_function_body): Likewise.
29538 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29539 Alan Hayward <alan.hayward@arm.com>
29540 David Sherwood <david.sherwood@arm.com>
29542 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
29543 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
29544 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
29545 (expand_vector_condition, vector_element): Likewise.
29546 (subparts_gt): New function.
29547 (get_compute_type): Use subparts_gt.
29548 (count_type_subparts): Delete.
29549 (expand_vector_operations_1): Use subparts_gt instead of
29550 count_type_subparts.
29552 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29553 Alan Hayward <alan.hayward@arm.com>
29554 David Sherwood <david.sherwood@arm.com>
29556 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
29557 (vect_compile_time_alias): ...this new function. Do the calculation
29558 on poly_ints rather than trees.
29559 (vect_prune_runtime_alias_test_list): Update call accordingly.
29561 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29562 Alan Hayward <alan.hayward@arm.com>
29563 David Sherwood <david.sherwood@arm.com>
29565 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
29567 (vect_schedule_slp_instance): Likewise.
29569 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29570 Alan Hayward <alan.hayward@arm.com>
29571 David Sherwood <david.sherwood@arm.com>
29573 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
29574 constant and extern definitions for variable-length vectors.
29575 (vect_get_constant_vectors): Note that the number of units
29576 is known to be constant.
29578 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29579 Alan Hayward <alan.hayward@arm.com>
29580 David Sherwood <david.sherwood@arm.com>
29582 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
29583 of units as polynomial. Choose between WIDE and NARROW based
29586 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29587 Alan Hayward <alan.hayward@arm.com>
29588 David Sherwood <david.sherwood@arm.com>
29590 * tree-vect-stmts.c (simd_clone_subparts): New function.
29591 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
29593 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29594 Alan Hayward <alan.hayward@arm.com>
29595 David Sherwood <david.sherwood@arm.com>
29597 * tree-vect-stmts.c (vectorizable_call): Treat the number of
29598 vectors as polynomial. Use build_index_vector for
29599 IFN_GOMP_SIMD_LANE.
29601 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29602 Alan Hayward <alan.hayward@arm.com>
29603 David Sherwood <david.sherwood@arm.com>
29605 * tree-vect-stmts.c (get_load_store_type): Treat the number of
29606 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
29607 for variable-length vectors.
29608 (vectorizable_mask_load_store): Treat the number of units as
29609 polynomial, asserting that it is constant if the condition has
29610 already been enforced.
29611 (vectorizable_store, vectorizable_load): Likewise.
29613 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29614 Alan Hayward <alan.hayward@arm.com>
29615 David Sherwood <david.sherwood@arm.com>
29617 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
29618 of units as polynomial. Punt if we can't tell at compile time
29619 which vector contains the final result.
29621 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29622 Alan Hayward <alan.hayward@arm.com>
29623 David Sherwood <david.sherwood@arm.com>
29625 * tree-vect-loop.c (vectorizable_induction): Treat the number
29626 of units as polynomial. Punt on SLP inductions. Use an integer
29627 VEC_SERIES_EXPR for variable-length integer reductions. Use a
29628 cast of such a series for variable-length floating-point
29631 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29632 Alan Hayward <alan.hayward@arm.com>
29633 David Sherwood <david.sherwood@arm.com>
29635 * tree.h (build_index_vector): Declare.
29636 * tree.c (build_index_vector): New function.
29637 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
29638 of units as polynomial, forcibly converting it to a constant if
29639 vectorizable_reduction has already enforced the condition.
29640 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
29641 to create a {1,2,3,...} vector.
29642 (vectorizable_reduction): Treat the number of units as polynomial.
29643 Choose vectype_in based on the largest scalar element size rather
29644 than the smallest number of units. Enforce the restrictions
29647 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29648 Alan Hayward <alan.hayward@arm.com>
29649 David Sherwood <david.sherwood@arm.com>
29651 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
29652 number of units as polynomial.
29654 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29655 Alan Hayward <alan.hayward@arm.com>
29656 David Sherwood <david.sherwood@arm.com>
29658 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
29659 * target.def (autovectorize_vector_sizes): Return the vector sizes
29660 by pointer, using vector_sizes rather than a bitmask.
29661 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
29662 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
29663 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
29665 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
29666 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
29667 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
29668 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
29669 * omp-general.c (omp_max_vf): Likewise.
29670 * omp-low.c (omp_clause_aligned_alignment): Likewise.
29671 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
29672 * tree-vect-loop.c (vect_analyze_loop): Likewise.
29673 * tree-vect-slp.c (vect_slp_bb): Likewise.
29674 * doc/tm.texi: Regenerate.
29675 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
29677 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
29678 the vector size as a poly_uint64 rather than an unsigned int.
29679 (current_vector_size): Change from an unsigned int to a poly_uint64.
29680 (get_vectype_for_scalar_type): Update accordingly.
29681 * tree.h (build_truth_vector_type): Take the size and number of
29682 units as a poly_uint64 rather than an unsigned int.
29683 (build_vector_type): Add a temporary overload that takes
29684 the number of units as a poly_uint64 rather than an unsigned int.
29685 * tree.c (make_vector_type): Likewise.
29686 (build_truth_vector_type): Take the number of units as a poly_uint64
29687 rather than an unsigned int.
29689 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29690 Alan Hayward <alan.hayward@arm.com>
29691 David Sherwood <david.sherwood@arm.com>
29693 * target.def (get_mask_mode): Take the number of units and length
29694 as poly_uint64s rather than unsigned ints.
29695 * targhooks.h (default_get_mask_mode): Update accordingly.
29696 * targhooks.c (default_get_mask_mode): Likewise.
29697 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
29698 * doc/tm.texi: Regenerate.
29700 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29701 Alan Hayward <alan.hayward@arm.com>
29702 David Sherwood <david.sherwood@arm.com>
29704 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
29705 * omp-general.c (omp_max_vf): Likewise.
29706 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
29707 (expand_omp_simd): Handle polynomial safelen.
29708 * omp-low.c (omplow_simd_context): Add a default constructor.
29709 (omplow_simd_context::max_vf): Change from int to poly_uint64.
29710 (lower_rec_simd_input_clauses): Update accordingly.
29711 (lower_rec_input_clauses): Likewise.
29713 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29714 Alan Hayward <alan.hayward@arm.com>
29715 David Sherwood <david.sherwood@arm.com>
29717 * tree-vectorizer.h (vect_nunits_for_cost): New function.
29718 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
29719 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
29720 (vect_analyze_slp_cost): Likewise.
29721 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
29722 (vect_model_load_cost): Likewise.
29724 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29725 Alan Hayward <alan.hayward@arm.com>
29726 David Sherwood <david.sherwood@arm.com>
29728 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
29729 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
29730 from an unsigned int * to a poly_uint64_pod *.
29731 (calculate_unrolling_factor): New function.
29732 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
29734 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29735 Alan Hayward <alan.hayward@arm.com>
29736 David Sherwood <david.sherwood@arm.com>
29738 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
29739 from an unsigned int to a poly_uint64.
29740 (_loop_vec_info::slp_unrolling_factor): Likewise.
29741 (_loop_vec_info::vectorization_factor): Change from an int
29743 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
29744 (vect_get_num_vectors): New function.
29745 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
29746 (vect_get_num_copies): Use vect_get_num_vectors.
29747 (vect_analyze_data_ref_dependences): Change max_vf from an int *
29748 to an unsigned int *.
29749 (vect_analyze_data_refs): Change min_vf from an int * to a
29751 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
29752 than an unsigned HOST_WIDE_INT.
29753 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
29754 (vect_analyze_data_ref_dependence): Change max_vf from an int *
29755 to an unsigned int *.
29756 (vect_analyze_data_ref_dependences): Likewise.
29757 (vect_compute_data_ref_alignment): Handle polynomial vf.
29758 (vect_enhance_data_refs_alignment): Likewise.
29759 (vect_prune_runtime_alias_test_list): Likewise.
29760 (vect_shift_permute_load_chain): Likewise.
29761 (vect_supportable_dr_alignment): Likewise.
29762 (dependence_distance_ge_vf): Take the vectorization factor as a
29763 poly_uint64 rather than an unsigned HOST_WIDE_INT.
29764 (vect_analyze_data_refs): Change min_vf from an int * to a
29766 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
29767 vfm1 as a poly_uint64 rather than an int. Make the same change
29768 for the returned bound_scalar.
29769 (vect_gen_vector_loop_niters): Handle polynomial vf.
29770 (vect_do_peeling): Likewise. Update call to
29771 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
29772 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
29774 * tree-vect-loop.c (vect_determine_vectorization_factor)
29775 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
29776 (vect_get_known_peeling_cost): Likewise.
29777 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
29778 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
29779 (vect_transform_loop): Likewise. Use the lowest possible VF when
29780 updating the upper bounds of the loop.
29781 (vect_min_worthwhile_factor): Make static. Return an unsigned int
29782 rather than an int.
29783 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
29784 polynomial unroll factors.
29785 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
29786 (vect_make_slp_decision): Likewise.
29787 (vect_supported_load_permutation_p): Likewise, and polynomial
29789 (vect_analyze_slp_cost): Handle polynomial vf.
29790 (vect_slp_analyze_node_operations): Likewise.
29791 (vect_slp_analyze_bb_1): Likewise.
29792 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
29793 than an unsigned HOST_WIDE_INT.
29794 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
29795 (vectorizable_load): Handle polynomial vf.
29796 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
29798 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
29800 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29801 Alan Hayward <alan.hayward@arm.com>
29802 David Sherwood <david.sherwood@arm.com>
29804 * match.pd: Handle bit operations involving three constants
29805 and try to fold one pair.
29807 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29809 * tree-vect-loop-manip.c: Include gimple-fold.h.
29810 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
29811 niters_maybe_zero parameters. Handle other cases besides a step of 1.
29812 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
29813 Add a path that uses a step of VF instead of 1, but disable it
29815 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
29816 and niters_no_overflow parameters. Update calls to
29817 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
29818 Create a new SSA name if the latter choses to use a ste other
29819 than zero, and return it via niters_vector_mult_vf_var.
29820 * tree-vect-loop.c (vect_transform_loop): Update calls to
29821 vect_do_peeling, vect_gen_vector_loop_niters and
29822 slpeel_make_loop_iterate_ntimes.
29823 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
29824 (vect_gen_vector_loop_niters): Update declarations after above changes.
29826 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
29828 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
29829 128-bit round to integer instructions.
29830 (ceil<mode>2): Likewise.
29831 (btrunc<mode>2): Likewise.
29832 (round<mode>2): Likewise.
29834 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
29836 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
29837 unaligned VSX load/store on P8/P9.
29838 (expand_block_clear): Allow the use of unaligned VSX
29839 load/store on P8/P9.
29841 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
29843 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
29845 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
29846 swap associated with both a load and a store.
29848 2018-01-02 Andrew Waterman <andrew@sifive.com>
29850 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
29851 * config/riscv/riscv.md (clear_cache): Use it.
29853 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
29855 * web.c: Remove out-of-date comment.
29857 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
29859 * expr.c (fixup_args_size_notes): Check that any existing
29860 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
29861 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
29862 (emit_single_push_insn): ...here.
29864 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
29866 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
29867 (const_vector_encoded_nelts): New function.
29868 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
29869 (const_vector_int_elt, const_vector_elt): Declare.
29870 * emit-rtl.c (const_vector_int_elt_1): New function.
29871 (const_vector_elt): Likewise.
29872 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
29873 of CONST_VECTOR_ELT.
29875 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
29877 * expr.c: Include rtx-vector-builder.h.
29878 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
29879 directly on the tree encoding.
29880 (const_vector_from_tree): Likewise.
29881 * optabs.c: Include rtx-vector-builder.h.
29882 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
29883 sequence of "u" values.
29884 * vec-perm-indices.c: Include rtx-vector-builder.h.
29885 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
29886 directly on the vec_perm_indices encoding.
29888 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
29890 * doc/rtl.texi (const_vector): Describe new encoding scheme.
29891 * Makefile.in (OBJS): Add rtx-vector-builder.o.
29892 * rtx-vector-builder.h: New file.
29893 * rtx-vector-builder.c: Likewise.
29894 * rtl.h (rtx_def::u2): Add a const_vector field.
29895 (CONST_VECTOR_NPATTERNS): New macro.
29896 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
29897 (CONST_VECTOR_DUPLICATE_P): Likewise.
29898 (CONST_VECTOR_STEPPED_P): Likewise.
29899 (CONST_VECTOR_ENCODED_ELT): Likewise.
29900 (const_vec_duplicate_p): Check for a duplicated vector encoding.
29901 (unwrap_const_vec_duplicate): Likewise.
29902 (const_vec_series_p): Check for a non-duplicated vector encoding.
29903 Say that the function only returns true for integer vectors.
29904 * emit-rtl.c: Include rtx-vector-builder.h.
29905 (gen_const_vec_duplicate_1): Delete.
29906 (gen_const_vector): Call gen_const_vec_duplicate instead of
29907 gen_const_vec_duplicate_1.
29908 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
29909 (gen_const_vec_duplicate): Use rtx_vector_builder.
29910 (gen_const_vec_series): Likewise.
29911 (gen_rtx_CONST_VECTOR): Likewise.
29912 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
29913 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
29914 Build a new vector rather than modifying a CONST_VECTOR in-place.
29915 (handle_special_swappables): Update call accordingly.
29916 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
29917 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
29918 Build a new vector rather than modifying a CONST_VECTOR in-place.
29919 (handle_special_swappables): Update call accordingly.
29921 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
29923 * simplify-rtx.c (simplify_const_binary_operation): Use
29924 CONST_VECTOR_ELT instead of XVECEXP.
29926 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
29928 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
29929 the selector elements to be different from the data elements
29930 if the selector is a VECTOR_CST.
29931 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
29932 ssizetype for the selector.
29934 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
29936 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
29937 before testing each element individually.
29938 * tree-vect-generic.c (lower_vec_perm): Likewise.
29940 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
29942 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
29943 * selftest-run-tests.c (selftest::run_tests): Call it.
29944 * vector-builder.h (vector_builder::operator ==): New function.
29945 (vector_builder::operator !=): Likewise.
29946 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
29947 (vec_perm_indices::all_from_input_p): New function.
29948 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
29949 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
29950 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
29951 instead of reading the VECTOR_CST directly. Detect whether both
29952 vector inputs are the same before constructing the vec_perm_indices,
29953 and update the number of inputs argument accordingly. Use the
29954 utility functions added above. Only construct sel2 if we need to.
29956 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
29958 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
29959 the broadcast of the low byte.
29960 (expand_mult_highpart): Use an explicit encoding for the permutes.
29961 * optabs-query.c (can_mult_highpart_p): Likewise.
29962 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
29963 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
29964 (vectorizable_bswap): Likewise.
29965 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
29966 explicit encoding for the power-of-2 permutes.
29967 (vect_permute_store_chain): Likewise.
29968 (vect_grouped_load_supported): Likewise.
29969 (vect_permute_load_chain): Likewise.
29971 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
29973 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
29974 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
29975 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
29976 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
29977 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
29978 (vect_gen_perm_mask_any): Likewise.
29980 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
29982 * int-vector-builder.h: New file.
29983 * vec-perm-indices.h: Include int-vector-builder.h.
29984 (vec_perm_indices): Redefine as an int_vector_builder.
29985 (auto_vec_perm_indices): Delete.
29986 (vec_perm_builder): Redefine as a stand-alone class.
29987 (vec_perm_indices::vec_perm_indices): New function.
29988 (vec_perm_indices::clamp): Likewise.
29989 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
29990 (vec_perm_indices::new_vector): New function.
29991 (vec_perm_indices::new_expanded_vector): Update for new
29992 vec_perm_indices class.
29993 (vec_perm_indices::rotate_inputs): New function.
29994 (vec_perm_indices::all_in_range_p): Operate directly on the
29995 encoded form, without computing elided elements.
29996 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
29997 encoding. Update for new vec_perm_indices class.
29998 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
29999 the given vec_perm_builder.
30000 (expand_vec_perm_var): Update vec_perm_builder constructor.
30001 (expand_mult_highpart): Use vec_perm_builder instead of
30002 auto_vec_perm_indices.
30003 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
30004 vec_perm_indices instead of auto_vec_perm_indices. Use a single
30005 or double series encoding as appropriate.
30006 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
30007 vec_perm_indices instead of auto_vec_perm_indices.
30008 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
30009 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
30010 (vect_permute_store_chain): Likewise.
30011 (vect_grouped_load_supported): Likewise.
30012 (vect_permute_load_chain): Likewise.
30013 (vect_shift_permute_load_chain): Likewise.
30014 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
30015 (vect_transform_slp_perm_load): Likewise.
30016 (vect_schedule_slp_instance): Likewise.
30017 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
30018 (vectorizable_mask_load_store): Likewise.
30019 (vectorizable_bswap): Likewise.
30020 (vectorizable_store): Likewise.
30021 (vectorizable_load): Likewise.
30022 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
30023 vec_perm_indices instead of auto_vec_perm_indices. Use
30024 tree_to_vec_perm_builder to read the vector from a tree.
30025 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
30026 vec_perm_builder instead of a vec_perm_indices.
30027 (have_whole_vector_shift): Use vec_perm_builder and
30028 vec_perm_indices instead of auto_vec_perm_indices. Leave the
30029 truncation to calc_vec_perm_mask_for_shift.
30030 (vect_create_epilog_for_reduction): Likewise.
30031 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
30032 from auto_vec_perm_indices to vec_perm_indices.
30033 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
30034 instead of changing individual elements.
30035 (aarch64_vectorize_vec_perm_const): Use new_vector to install
30036 the vector in d.perm.
30037 * config/arm/arm.c (expand_vec_perm_d::perm): Change
30038 from auto_vec_perm_indices to vec_perm_indices.
30039 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
30040 instead of changing individual elements.
30041 (arm_vectorize_vec_perm_const): Use new_vector to install
30042 the vector in d.perm.
30043 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
30044 Update vec_perm_builder constructor.
30045 (rs6000_expand_interleave): Likewise.
30046 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
30047 (rs6000_expand_interleave): Likewise.
30049 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
30051 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
30052 to qimode could truncate the indices.
30053 * optabs.c (expand_vec_perm_var): Likewise.
30055 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
30057 * Makefile.in (OBJS): Add vec-perm-indices.o.
30058 * vec-perm-indices.h: New file.
30059 * vec-perm-indices.c: Likewise.
30060 * target.h (vec_perm_indices): Replace with a forward class
30062 (auto_vec_perm_indices): Move to vec-perm-indices.h.
30063 * optabs.h: Include vec-perm-indices.h.
30064 (expand_vec_perm): Delete.
30065 (selector_fits_mode_p, expand_vec_perm_var): Declare.
30066 (expand_vec_perm_const): Declare.
30067 * target.def (vec_perm_const_ok): Replace with...
30068 (vec_perm_const): ...this new hook.
30069 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
30070 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
30071 * doc/tm.texi: Regenerate.
30072 * optabs.def (vec_perm_const): Delete.
30073 * doc/md.texi (vec_perm_const): Likewise.
30074 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
30075 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
30076 expand_vec_perm for constant permutation vectors. Assert that
30077 the mode of variable permutation vectors is the integer equivalent
30078 of the mode that is being permuted.
30079 * optabs-query.h (selector_fits_mode_p): Declare.
30080 * optabs-query.c: Include vec-perm-indices.h.
30081 (selector_fits_mode_p): New function.
30082 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
30083 is defined, instead of checking whether the vec_perm_const_optab
30084 exists. Use targetm.vectorize.vec_perm_const instead of
30085 targetm.vectorize.vec_perm_const_ok. Check whether the indices
30086 fit in the vector mode before using a variable permute.
30087 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
30088 vec_perm_indices instead of an rtx.
30089 (expand_vec_perm): Replace with...
30090 (expand_vec_perm_const): ...this new function. Take the selector
30091 as a vec_perm_indices rather than an rtx. Also take the mode of
30092 the selector. Update call to shift_amt_for_vec_perm_mask.
30093 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
30094 Use vec_perm_indices::new_expanded_vector to expand the original
30095 selector into bytes. Check whether the indices fit in the vector
30096 mode before using a variable permute.
30097 (expand_vec_perm_var): Make global.
30098 (expand_mult_highpart): Use expand_vec_perm_const.
30099 * fold-const.c: Includes vec-perm-indices.h.
30100 * tree-ssa-forwprop.c: Likewise.
30101 * tree-vect-data-refs.c: Likewise.
30102 * tree-vect-generic.c: Likewise.
30103 * tree-vect-loop.c: Likewise.
30104 * tree-vect-slp.c: Likewise.
30105 * tree-vect-stmts.c: Likewise.
30106 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
30108 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
30109 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
30110 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
30111 (aarch64_vectorize_vec_perm_const): ...this new function.
30112 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
30113 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
30114 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
30115 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
30116 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
30117 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
30118 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
30120 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
30121 check for NEON modes.
30122 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
30123 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
30124 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
30125 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
30127 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
30128 the old VEC_PERM_CONST conditions.
30129 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
30130 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
30131 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
30132 (ia64_vectorize_vec_perm_const_ok): Merge into...
30133 (ia64_vectorize_vec_perm_const): ...this new function.
30134 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
30135 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
30136 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
30137 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
30138 * config/mips/mips.c (mips_expand_vec_perm_const)
30139 (mips_vectorize_vec_perm_const_ok): Merge into...
30140 (mips_vectorize_vec_perm_const): ...this new function.
30141 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
30142 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
30143 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
30144 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
30145 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
30146 (rs6000_expand_vec_perm_const): Delete.
30147 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
30149 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
30150 (altivec_expand_vec_perm_const_le): Take each operand individually.
30151 Operate on constant selectors rather than rtxes.
30152 (altivec_expand_vec_perm_const): Likewise. Update call to
30153 altivec_expand_vec_perm_const_le.
30154 (rs6000_expand_vec_perm_const): Delete.
30155 (rs6000_vectorize_vec_perm_const_ok): Delete.
30156 (rs6000_vectorize_vec_perm_const): New function.
30157 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
30158 an element count and rtx array.
30159 (rs6000_expand_extract_even): Update call accordingly.
30160 (rs6000_expand_interleave): Likewise.
30161 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
30162 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
30163 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
30164 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
30165 (rs6000_expand_vec_perm_const): Delete.
30166 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
30167 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
30168 (altivec_expand_vec_perm_const_le): Take each operand individually.
30169 Operate on constant selectors rather than rtxes.
30170 (altivec_expand_vec_perm_const): Likewise. Update call to
30171 altivec_expand_vec_perm_const_le.
30172 (rs6000_expand_vec_perm_const): Delete.
30173 (rs6000_vectorize_vec_perm_const_ok): Delete.
30174 (rs6000_vectorize_vec_perm_const): New function. Remove stray
30175 reference to the SPE evmerge intructions.
30176 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
30177 an element count and rtx array.
30178 (rs6000_expand_extract_even): Update call accordingly.
30179 (rs6000_expand_interleave): Likewise.
30180 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
30181 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
30183 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
30185 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
30187 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
30188 vector mode and that that mode matches the mode of the data
30190 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
30191 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
30192 directly using expand_vec_perm_1 when forcing selectors into
30194 (expand_vec_perm_var): New function, split out from expand_vec_perm.
30196 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
30198 * optabs-query.h (can_vec_perm_p): Delete.
30199 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
30200 * optabs-query.c (can_vec_perm_p): Split into...
30201 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
30202 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
30203 particular selector is valid.
30204 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
30205 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
30206 (vect_grouped_load_supported): Likewise.
30207 (vect_shift_permute_load_chain): Likewise.
30208 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
30209 (vect_transform_slp_perm_load): Likewise.
30210 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
30211 (vectorizable_bswap): Likewise.
30212 (vect_gen_perm_mask_checked): Likewise.
30213 * fold-const.c (fold_ternary_loc): Likewise. Don't take
30214 implementations of variable permutation vectors into account
30215 when deciding which selector to use.
30216 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
30217 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
30218 with a false third argument.
30219 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
30220 to test whether the constant selector is valid and can_vec_perm_var_p
30221 to test whether a variable selector is valid.
30223 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
30225 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
30226 * optabs-query.c (can_vec_perm_p): Likewise.
30227 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
30228 instead of vec_perm_indices.
30229 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
30230 (vect_gen_perm_mask_checked): Likewise,
30231 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
30232 (vect_gen_perm_mask_checked): Likewise,
30234 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
30236 * optabs-query.h (qimode_for_vec_perm): Declare.
30237 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
30238 (qimode_for_vec_perm): ...this new function.
30239 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
30241 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
30243 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
30244 does not have a conditional at the top.
30246 2018-01-02 Richard Biener <rguenther@suse.de>
30248 * ipa-inline.c (big_speedup_p): Fix expression.
30250 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
30253 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
30256 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
30260 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
30261 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
30262 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
30263 cond_taken_branch_cost 3->4.
30265 2018-01-01 Jakub Jelinek <jakub@redhat.com>
30267 PR tree-optimization/83581
30268 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
30269 TODO_cleanup_cfg if any changes have been made.
30271 PR middle-end/83608
30272 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
30273 convert_modes if target mode has the right side, but different mode
30276 PR middle-end/83609
30277 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
30278 last argument when extracting from CONCAT. If either from_real or
30279 from_imag is NULL, use expansion through memory. If result is not
30280 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
30281 the parts directly to inner mode, if even that fails, use expansion
30284 PR middle-end/83623
30285 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
30286 check for bswap in mode rather than HImode and use that in expand_unop
30289 Copyright (C) 2018 Free Software Foundation, Inc.
30291 Copying and distribution of this file, with or without modification,
30292 are permitted in any medium without royalty provided the copyright
30293 notice and this notice are preserved.