1 2018-11-05 Richard Biener <rguenther@suse.de>
3 * tree-scalar-evolution.h (final_value_replacement_loop): Update
5 * tree-scalar-evolution.c (final_value_replacement_loop): Return
6 whether anything was done.
7 (scev_const_prop): Remove constant propagation part, fold
9 * tree-ssa-loop.c (pass_scev_cprop::execute): ... here.
10 (pass_data_scev_cprop): TODO_cleanup_cfg is now done
13 2018-11-05 Jakub Jelinek <jakub@redhat.com>
15 PR tree-optimization/87859
16 * gimple-ssa-store-merging.c (struct merged_store_group): Add
17 only_constants and first_nonmergeable_order members.
18 (merged_store_group::merged_store_group): Initialize them.
19 (merged_store_group::do_merge): Clear only_constants member if
20 adding something other than INTEGER_CST store.
21 (imm_store_chain_info::coalesce_immediate_stores): Don't merge
22 stores with order >= first_nonmergeable_order. Use
23 merged_store->only_constants instead of always recomputing it.
24 Set merged_store->first_nonmergeable_order if we've skipped any
25 stores. Attempt to merge overlapping INTEGER_CST stores that
26 we would otherwise skip.
29 * match.pd (X + Y < X): Don't optimize if TYPE_OVERFLOW_SANITIZED.
31 2018-11-05 Xuepeng Guo <xuepeng.guo@intel.com>
34 * config/i386/emmintrin.h (__v16qs): New to cope with option
36 (_mm_cmpeq_epi8): Replace __v16qi with __v16qs.
37 (_mm_cmplt_epi8): Likewise.
38 (_mm_cmpgt_epi8): Likewise.
40 2018-11-05 Richard Biener <rguenther@suse.de>
42 PR rtl-optimization/87852
43 * fwprop.c (use_killed_between): Only consider single-defs of the
44 use whose definition statement dominates the use.
46 2018-11-05 Martin Liska <mliska@suse.cz>
49 * doc/invoke.texi: Remove options that are
50 not disabled with -Os.
52 2018-11-05 Martin Liska <mliska@suse.cz>
55 * doc/extend.texi: Update constrain about the last argument
56 of __builtin_expect_with_probability.
58 2018-11-05 Martin Liska <mliska@suse.cz>
61 * predict.c (expr_expected_value_1): Verify
62 that last argument is a real constants and emit
65 2018-11-05 Martin Liska <mliska@suse.cz>
68 * ipa-profile.c (ipa_profile): Adjust hotness threshold
71 2018-11-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
73 PR tree-optimization/86572
74 * builtins.c (c_strlen): Handle negative offsets in a safe way.
76 2018-11-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
78 PR tree-optimization/87672
79 * gimple-fold.c (gimple_fold_builtin_stxcpy_chk): Gimplify.
80 * tree-ssa-strlen.c (handle_builtin_strcat): Adjust object size.
82 2018-11-04 Uros Bizjak <ubizjak@gmail.com>
85 * cfgexpand.c (pass_expand::execute): Move the call to
86 finish_eh_generation in front of the call to expand_stack_alignment.
88 2018-11-04 Venkataramanan Kumar <venkataramanan.kumar@amd.com>
90 * common/config/i386/i386-common.c (processor_alias_table): Add
92 * config.gcc (i[34567]86-*-linux* | ...): Add znver2.
93 (case ${target}): Add znver2.
94 * config/i386/driver-i386.c: (host_detect_local_cpu): Let
95 -march=native recognize znver2 processors.
96 * config/i386/i386-c.c (ix86_target_macros_internal): Add znver2.
97 * config/i386/i386.c (m_znver2): New definition.
98 (m_ZNVER): New definition.
99 (m_AMD_MULTIPLE): Includes m_znver2.
100 (processor_cost_table): Add znver2 entry.
101 (processor_target_table): Add znver2 entry.
102 (get_builtin_code_for_version): Set priority for
104 (processor_model): Add M_AMDFAM17H_ZNVER2.
105 (arch_names_table): Ditto.
106 (ix86_reassociation_width): Include znver2.
107 * config/i386/i386.h (TARGET_znver2): New definition.
108 (struct ix86_size_cost): Add TARGET_ZNVER2.
109 (enum processor_type): Add PROCESSOR_ZNVER2.
110 * config/i386/i386.md (define_attr "cpu"): Add znver2.
111 * config/i386/x86-tune-costs.h: (processor_costs) Add znver2 costs.
112 * config/i386/x86-tune-sched.c: (ix86_issue_rate): Add znver2.
113 (ix86_adjust_cost): Add znver2.
114 * config/i386/x86-tune.def: Replace m_ZNVER1 by m_ZNVER.
115 * gcc/doc/extend.texi: Add details about znver2.
116 * gcc/doc/invoke.texi: Add details about znver2.
118 2018-11-03 Sandra Loosemore <sandra@codesourcery.com>
122 * config/nios2/nios2.c (nios2_rtx_costs): Recognize <mul>sidi3
125 2018-11-02 Aaron Sawdey <acsawdey@linux.ibm.com>
127 * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Pay
128 attention to TARGET_AVOID_XFORM and BYTES_BIG_ENDIAN.
130 2018-11-02 Richard Earnshaw <rearnsha@arm.com>
132 * config/aarch64/aarch64.c ((aarch64_override_options): Disable
133 shrink-wrapping when -mtrack-speculation.
135 2018-11-02 Richard Biener <rguenther@suse.de>
137 * tree-ssa-coalesce.c (struct coalesce_list): Add obstack member.
138 (pop_cost_one_pair): Do not free pair.
139 (pop_best_coalesce): Likewise.
140 (create_coalesce_list): Initialize obstack.
141 (delete_coalesce_list): Free obstack.
142 (find_coalesce_pair): Obstack-allocate coalesce pairs.
143 (add_cost_one_coalesce): Likewise.
144 (struct live_track): Remove bitmap pointer indirections.
145 (new_live_track): Adjust.
146 (delete_live_track): Likewise.
147 (live_track_remove_partition): Likewise.
148 (live_track_add_partition): Likewise.
149 (live_track_live_p): Likewise.
150 (live_track_process_def): Likewise.
151 (live_track_clear_base_vars): Likewise.
153 2018-11-02 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
155 * configure.ac (gcc_cv_as_sparc_register_op): Remove.
156 * configure: Regenerate.
157 * config.in: Regenerate.
158 * config/sparc/sparc.c (sparc_output_scratch_registers): Remove
159 HAVE_AS_REGISTER_PSEUDO_OP guard.
160 * config/sparc/sparc.h (ASM_ARCH64_SPEC): Remove
161 !HAVE_AS_REGISTER_PSEUDO_OP support.
162 (ASM_DECLARE_REGISTER_GLOBAL): Remove HAVE_AS_REGISTER_PSEUDO_OP
165 2018-11-02 Richard Biener <rguenther@suse.de>
167 * dwarf2out.c (build_abbrev_table): Guard lookup_external_ref call
170 2018-11-02 Richard Biener <rguenther@suse.de>
172 PR tree-optimization/87776
173 * tree-ssa-sccvn.c (do_rpo_vn): Do not mark backedges
174 executable when iterating but running into rpo-vn-max-loop-depth
175 and not eliding the iteration.
177 2018-11-30 Jan Hubicka <jh@suse.cz>
179 * tree.c (free_lang_data_in_decl): Free TREE_TYPE and alignment of
182 2018-11-01 Aaron Sawdey <acsawdey@linux.ibm.com>
184 * config/rs6000/rs6000-protos.h (rs6000_address_for_fpconvert): Remove
186 * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem):
187 Combine with rs6000_address_for_fpconvert.
188 (rs6000_address_for_fpconvert): Combine with
189 rs6000_force_indexed_or_indirect_mem.
190 (rs6000_expand_vector_init): Change function call from
191 rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
192 * config/rs6000/rs6000.md (floatsi<mode>2_lfiwax): Change call from
193 rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
194 (floatsi<mode>2_lfiwax_mem): Ditto.
195 (floatunssi<mode>2_lfiwzx): Ditto.
196 (floatunssi<mode>2_lfiwzx_mem): Ditto.
197 (float<QHI:mode><FP_ISA3:mode>2): Ditto.
198 (floatuns<QHI:mode><FP_ISA3:mode>2): Ditto.
199 (fix_trunc<mode>si2_stfiwx): Ditto.
200 (fixuns_trunc<mode>si2_stfiwx): Ditto.
201 (float_<mode>si2_hw): Ditto.
202 (floatuns_<mode>si2_hw): Ditto.
203 * config/rs6000/vsx.md (*vsx_extract_si): Ditto.
204 (vsx_splat_<mode>): Ditto.
206 2018-11-01 Joseph Myers <joseph@codesourcery.com>
208 * configure.ac (gcc_cv_as_ia64_ltoffx_ldxmov_relocs): Add newline
209 at end of assembler input text.
210 * configure: Regenerate.
212 2018-11-01 Jakub Jelinek <jakub@redhat.com>
214 PR tree-optimization/87826
215 * gimple-ssa-store-merging.c (do_shift_rotate): Punt if count is
216 negative or larger or equal to type's precision.
218 2018-10-31 Alexandre Oliva <aoliva@redhat.com>
220 * opts.c (default_options_table): Do not enable
221 OPT_fdelayed_branch at -Og.
222 * doc/invoke.texi (-fdelayed-branch): Document it.
224 2018-10-31 Richard Henderson <richard.henderson@linaro.org>
226 * optabs-libfuncs.c (build_libfunc_function_visibility):
227 New, split out from...
228 (build_libfunc_function): ... here.
229 (init_one_libfunc_visibility): New, split out from ...
230 (init_one_libfunc): ... here.
232 * config/aarch64/atomics.md (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse):
233 scratch register need not be early-clobber. Document the reason
234 why we cannot use ST<OP>.
236 2018-10-31 Joseph Myers <joseph@codesourcery.com>
239 * configure.ac: Remove AC_PREREQ. Use AC_LANG_SOURCE. Use single
240 line for second argument of AC_DEFINE_UNQUOTED.
241 * doc/install.texi (Tools/packages necessary for modifying GCC):
242 Update to autoconf 2.69 and automake 1.15.1.
243 * aclocal.m4, config.in, configure: Regenerate.
245 2018-10-31 Pat Haugen <pthaugen@us.ibm.com>
247 * tree-ssa-loop-ivcanon.c (try_peel_loop): Fix typo and remove dead
250 2018-10-31 Martin Liska <mliska@suse.cz>
253 * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
254 Add new argument invalid_extension.
255 (aarch64_get_all_extension_candidates): New function.
256 (aarch64_rewrite_selected_cpu): Add NULL to function call.
257 * config/aarch64/aarch64-protos.h (aarch64_parse_extension): Add
259 (aarch64_get_all_extension_candidates): New function.
260 * config/aarch64/aarch64.c (aarch64_parse_arch): Add new
261 argument invalid_extension.
262 (aarch64_parse_cpu): Likewise.
263 (aarch64_print_hint_for_extensions): New function.
264 (aarch64_validate_mcpu): Provide hint about invalid extension.
265 (aarch64_validate_march): Likewise.
266 (aarch64_handle_attr_arch): Pass new argument.
267 (aarch64_handle_attr_cpu): Provide hint about invalid extension.
268 (aarch64_handle_attr_isa_flags): Likewise.
270 2018-10-31 Richard Biener <rguenther@suse.de>
274 * tree-outof-ssa.c (insert_backedge_copies): Restrict
275 copy generation to useful cases. Place the copy before
276 the definition of the backedge value when possible.
278 2018-10-31 Claudiu Zissulescu <claziss@synopsys.com>
280 * config/arc/arc-arch.h (ARC_TUNE_ARC7XX): New tune value.
281 * config/arc/arc.c (arc_active_insn): New function.
282 (check_store_cacheline_hazard): Likewise.
283 (workaround_arc_anomaly): Use check_store_cacheline_hazard.
284 (arc_override_options): Disable delay slot scheduler for older
286 (arc_store_addr_hazard_p): New implementation, old one renamed to
288 (arc_store_addr_hazard_internal_p): Renamed.
289 (arc_reorg): Don't combine into brcc instructions which are part
290 of hardware hazard solution.
291 * config/arc/arc.md (attr tune): Consider new arc7xx tune value.
292 (tune_arc700): Likewise.
293 * config/arc/arc.opt (arc7xx): New tune value.
294 * config/arc/arc700.md: Improve A7 scheduler.
296 2018-10-31 Claudiu Zissulescu <claziss@synopsys.com>
298 * config/arc/arc.c (arc_override_options): Remove
299 TARGET_COMPACT_CASESI.
300 * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Update.
301 (CASE_VECTOR_MODE): Likewise.
302 (CASE_VECTOR_PC_RELATIVE): Likewise.
303 (CASE_VECTOR_SHORTEN_MODE): Likewise.
304 (CASE_VECTOR_SHORTEN_MODE1): Delete.
305 (ADDR_VEC_ALIGN): Update.
306 (ASM_OUTPUT_CASE_LABEL): Undefine.
307 (ASM_OUTPUT_BEFORE_CASE_LABEL): Undefine.
308 (TARGET_BI_BIH): Define.
309 (DEFAULT_BRANCH_INDEX): Likewise.
310 * config/arc/arc.md (casesi): Rework to accept BI/BIH
311 instructions, remove compact_casesi use case.
312 (casesi_compact_jump): Remove.
313 (casesi_dispatch): New pattern.
314 * config/arc/arc.opt: Add mbranch-index option. Deprecate
315 compact_casesi option.
316 * doc/invoke.texi: Document mbranch-index option.
318 2018-10-31 Claudiu Zissulescu <claziss@synopsys.com>
320 * config/arc/arc.c (arc_get_tp): Remove function.
321 (arc_emit_call_tls_get_addr): Likewise.
322 (arc_call_tls_get_addr): New function.
323 (arc_legitimize_tls_address): Make use of arc_call_tls_get_addr.
324 * config/arc/arc.md (tls_load_tp_soft): Remove.
325 (tls_gd_get_addr): Likewise.
327 2018-10-31 Claudiu Zissulescu <claziss@synopsys.com>
329 * config/arc/arc.md (mulsi3): Remove call to mulsi_600_lib.
330 (mulsi3_600_lib): Remove pattern.
331 (umulsi3_highpart_600_lib_le): Likewise.
332 (umulsi3_highpart): Remove call to umulsi3_highpart_600_lib_le.
333 (umulsidi3): Remove call to umulsidi3_600_lib.
334 (umulsidi3_600_lib): Remove pattern.
335 (peephole2): Remove peephole using the above deprecated patterns.
337 2018-10-31 Thomas Preud'homme <thomas.preudhomme@linaro.org>
340 * config/arm/arm.c (arm_option_check_internal): Disable the combined
341 use of -mslow-flash-data and -mword-relocations.
342 (arm_option_override): Enable -mword-relocations if -fpic or -fPIC.
343 * config/arm/arm.md (SYMBOL_REF MOVT splitter): Stop checking for
345 * doc/invoke.texi (-mword-relocations): Mention conflict with
347 (-mslow-flash-data): Reciprocally.
349 2018-10-31 Richard Henderson <richard.henderson@linaro.org>
351 * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Force
352 16-byte modes held in GP registers to use an even regno.
354 * config/aarch64/aarch64.c (aarch64_emit_bic): Remove.
355 (aarch64_atomic_ldop_supported_p): Remove.
356 (aarch64_gen_atomic_ldop): Remove.
357 * config/aarch64/atomic.md (atomic_<atomic_optab><ALLI>):
358 Fully expand LSE operations here.
359 (atomic_fetch_<atomic_optab><ALLI>): Likewise.
360 (atomic_<atomic_optab>_fetch<ALLI>): Likewise.
361 (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse): Drop atomic_op iterator
362 and use ATOMIC_LDOP instead; use register_operand for the input;
363 drop the split and emit insns directly.
364 (aarch64_atomic_fetch_<ATOMIC_LDOP><ALLI>_lse): Likewise.
365 (aarch64_atomic_<atomic_op>_fetch<ALLI>_lse): Remove.
366 (@aarch64_atomic_load<ATOMIC_LDOP><ALLI>): Remove.
368 * config/aarch64/aarch64.c (aarch64_emit_atomic_swap): Remove.
369 (aarch64_gen_atomic_ldop): Don't call it.
370 * config/aarch64/atomics.md (atomic_exchange<ALLI>):
371 Use aarch64_reg_or_zero.
372 (aarch64_atomic_exchange<ALLI>): Likewise.
373 (aarch64_atomic_exchange<ALLI>_lse): Remove split; remove & from
374 operand 0; use aarch64_reg_or_zero for input; merge ...
375 (@aarch64_atomic_swp<ALLI>): ... this and remove.
377 * config/aarch64/aarch64.c (aarch64_gen_compare_reg_maybe_ze): New.
378 (aarch64_split_compare_and_swap): Use it.
379 (aarch64_expand_compare_and_swap): Likewise. Remove convert_modes;
380 test oldval against the proper predicate.
381 * config/aarch64/atomics.md (@atomic_compare_and_swap<ALLI>):
382 Use nonmemory_operand for expected.
383 (cas_short_expected_pred): New.
384 (@aarch64_compare_and_swap<SHORT>): Use it; use "rn" not "rI" to match.
385 (@aarch64_compare_and_swap<GPI>): Use "rn" not "rI" for expected.
386 * config/aarch64/predicates.md (aarch64_plushi_immediate): New.
387 (aarch64_plushi_operand): New.
389 * config/aarch64/aarch64.c (aarch64_expand_compare_and_swap):
390 Force oldval into the rval register for TARGET_LSE; emit the compare
391 during initial expansion so that it may be deleted if unused.
392 (aarch64_gen_atomic_cas): Remove.
393 * config/aarch64/atomics.md (@aarch64_compare_and_swap<SHORT>_lse):
394 Change =&r to +r for operand 0; use match_dup for operand 2;
395 remove is_weak and mod_f operands as unused. Drop the split
397 (@aarch64_atomic_cas<SHORT>): ... this pattern's output; remove.
398 (@aarch64_compare_and_swap<GPI>_lse): Similarly.
399 (@aarch64_atomic_cas<GPI>): Similarly.
401 2018-10-31 Richard Biener <rguenther@suse.de>
403 * tree-eh.c (replace_trapping_overflow): Simplify ABS_EXPR case
406 2018-10-31 Sameera Deshpande <sameera.deshpande@linaro.org>
408 * config/aarch64/aarch64-cores.def (saphira): Use saphira pipeline.
409 * config/aarch64/aarch64.md: Include saphira.md
410 * config/aarch64/saphira.md: New file for pipeline description.
412 2018-10-30 Martin Sebor <msebor@redhat.com>
415 * gimple-ssa-sprintf.c (format_directive): Use %G to include
417 (sprintf_dom_walker::compute_format_length):
418 Avoid setting POSUNDER4K here.
419 (get_destination_size): Handle null argument values.
420 (get_user_idx_format): New function.
421 (sprintf_dom_walker::handle_gimple_call): Handle all printf-like
422 functions, including user-defined with attribute format printf.
423 Use %G to include inlining context.
426 2018-10-30 Jan Hubicka <jh@suse.cz>
428 * params.def (lto-partitions): Bump from 32 to 128.
430 2018-10-30 Jan Hubicka <jh@suse.cz>
433 (free_lang_data_d, add_tree_to_fld_list, fld_worklist_push): Move
435 (free_lang_data_in_type): Forward declare.
436 (fld_type_variant_equal_p): New function.
437 (fld_type_variant): New function
438 (fld_incomplete_types): New hash.
439 (fld_incomplete_type_of): New function
440 (fld_simplfied-type): New function.
441 (free_lang_data_in_decl): Add fld parameter; simplify type of FIELD_DECL
442 (free_lang_data): Allocate and free fld_incomplete_type; update call
443 of free_lang_data_in_decl.
445 2018-10-30 Eric Botcazou <ebotcazou@adacore.com>
447 * gcov.c (output_lines): Remove duplicate line.
449 2018-10-30 Aaron Sawdey <acsawdey@linux.ibm.com>
451 * config/rs6000/rs6000.md (bswapdi2): Force address into register
452 if not in indexed or indirect form.
453 (bswapdi2_load): Change predicate to indexed_or_indirect_operand.
454 (bswapdi2_store): Ditto.
455 * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem): New
457 * config/rs6000/rs6000-protos.h (rs6000_force_indexed_or_indirect_mem):
458 Prototype for helper function.
460 2018-10-30 Martin Sebor <msebor@redhat.com>
462 * doc/extend.texi (optimize): Clarify/expand attribute documentation.
463 (target, pragma GCC optimize, pragma GCC target): Ditto.
465 2018-10-30 Jonathan Wakely <jwakely@redhat.com>
467 * doc/extend.texi: Fix prototype and description of
468 __builtin_expect_with_probability.
470 2018-10-30 Michael Ploujnikov <michael.ploujnikov@oracle.com>
472 * cgraph.h (clone_function_name_1): Replaced by new
473 clone_function_name_numbered that takes name as string; for
474 privatize_symbol_name_1 use only.
475 (clone_function_name): Renamed to
476 clone_function_name_numbered to be explicit about numbering.
477 (clone_function_name): New two-argument function that does
478 not number its output.
479 (clone_function_name): New three-argument function that
480 takes a number to append to its output.
481 * cgraphclones.c (duplicate_thunk_for_node):
482 (clone_function_name_1): Renamed.
483 (clone_function_name_numbered): Two new functions.
484 (clone_function_name): Improved documentation.
485 (cgraph_node::create_virtual_clone): Use clone_function_name_numbered.
486 * config/rs6000/rs6000.c (make_resolver_func): Ditto.
487 * final.c (final_scan_insn_1): Use the new clone_function_name
489 * multiple_target.c (create_dispatcher_calls): Ditto.
490 (create_target_clone): Ditto.
491 * omp-expand.c (grid_expand_target_grid_body): Ditto.
492 * omp-low.c (create_omp_child_function_name): Ditto.
493 * omp-simd-clone.c (simd_clone_create): Ditto.
494 * symtab.c (simd_symtab_node::noninterposable_alias): Use the
495 new clone_function_name without numbering.
497 2018-10-30 Richard Earnshaw <rearnsha@arm.com>
499 * alloc-pool.h (base_pool_allocator <TBlockAllocator>::initialize):
500 Assert that the allocation size is not zero.
502 2018-10-30 Richard Biener <rguenther@suse.de>
504 PR tree-optimization/87800
505 * tree-vect-slp.c (vect_build_slp_tree_2): Reject any
506 non-induction or reduction PHIs.
508 2018-10-30 Sameera Deshpande <sameera.deshpande@linaro.org
510 * config/aarch64/falkor-tag-collision-avoidance.c
511 (execute_tag_collision_avoidance): Call df_note_add_problem.
513 2018-10-30 Martin Liska <mliska@suse.cz>
515 * doc/extend.texi: Fix typo in documentation
516 of __builtin_expect_with_probability.
518 2018-10-29 David Malcolm <dmalcolm@redhat.com>
521 * input.c (get_substring_ranges_for_loc): Detect if
522 linemap_resolve_location gives us a NULL map, and reject
525 2018-10-29 Iain Buclaw <ibuclaw@gdcproject.org>
527 * config.gcc (xstormy16-*-elf): Set tm_d_file.
529 2018-10-29 David Malcolm <dmalcolm@redhat.com>
530 Martin Sebor <msebor@redhat.com>
531 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
533 * doc/ux.texi (Quoting): New subsection, adapted from material at
534 https://gcc.gnu.org/wiki/DiagnosticsGuidelines written by
535 MartinSebor and ManuelLopezIbanez.
536 (Fix-it hints): Note that fix-it hints shouldn't be marked for
539 2018-10-29 Kugan Vivekanandarajah <kuganv@linaro.org>
542 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Fix niter
545 2018-10-29 Paul A. Clarke <pc@us.ibm.com>
547 * gcc.target/powerpc/ssse3-check.h: Remove inaccurate comment.
549 2018-10-29 Paul Koning <ni1d@arrl.net>
551 * doc/rtl.texi (CONST_WIDE_INT_ELT): Give correct macro name.
553 2018-10-29 Paul A. Clarke <pc@us.ibm.com>
555 * gcc/config/rs6000/mmintrin.h (_mm_packs_pi16, _mm_packs_pi32,
556 _mm_packs_pu16, _mm_unpackhi_pi8, _mm_unpacklo_pi8, _mm_add_pi8,
557 _mm_add_pi16, _mm_add_pi32, _mm_sub_pi8, _mm_sub_pi16, _mm_sub_pi32,
558 _mm_cmpgt_pi8, _mm_cmpeq_pi16, _mm_cmpgt_pi16, _mm_cmpeq_pi32,
559 _mm_cmpgt_pi32, _mm_adds_pi8, _mm_adds_pi16, _mm_adds_pu8,
560 _mm_adds_pu16, _mm_subs_pi8, _mm_subs_pi16, _mm_subs_pu8,
561 _mm_subs_pu16, _mm_madd_pi16, _mm_mulhi_pi16, _mm_mullo_pi16,
562 _mm_sll_pi16, _mm_sra_pi16, _mm_srl_pi16, _mm_set1_pi16, _mm_set1_pi8):
563 Change 'vector' to '__vector'.
564 * gcc/config/rs6000/xmmintrin.h (_mm_cvtps_pi32, _mm_cvttps_pi32,
565 _mm_cvtps_pi16, _mm_cvtps_pi8, _mm_max_pi16, _mm_max_pu8, _mm_min_pi16,
566 _mm_min_pu8, _mm_mulhi_pu16, _mm_shuffle_pi16, _mm_avg_pu8,
567 _mm_avg_pu16): Likewise. And, whitespace corrections.
569 2018-10-29 Richard Biener <rguenther@suse.de>
571 PR tree-optimization/87785
572 * tree-vect-slp.c (vect_gather_slp_loads): Only gather
575 2018-10-29 Olivier Hainque <hainque@adacore.com>
577 * VXWORKS_NET_LIBS_RTP: New macro, network part of VXWORKS_LIBS_RTP.
579 2018-10-29 Olivier Hainque <hainque@adacore.com>
581 * config/vxworks.h (VXWORKS_PERSONALITY): New VxWorks
582 ports configuration macro, defaults to "gnu".
583 (VXWORKS_OS_CPP_BUILTINS): Use it. Feed TOOL and
584 TOOL_FAMILY instead of _VX_TOOL/_VXTOOL_FAMILY.
586 2018-10-29 Olivier Hainque <hainque@adacore.com>
588 * config/vxworks.h (CLEAR_INSN_CACHE): #undef before #define.
590 2018-10-29 Richard Biener <rguenther@suse.de>
592 PR tree-optimization/87790
593 * tree-vect-slp.c (vect_mark_slp_stmts): Simplify.
594 (vect_make_slp_decision): Adjust.
595 (vect_slp_analyze_bb_1): Likewise.
596 (vect_detect_hybrid_slp_stmts): Properly union SLP type over
599 2018-10-29 Richard Biener <rguenther@suse.de>
601 PR tree-optimization/87785
602 * tree-vect-slp.c (vect_build_slp_tree_2): Remove loads argument
604 (vect_build_slp_tree): Likewise.
605 (vect_gather_slp_loads): New function.
606 (vect_analyze_slp_instance): Gather loads separately from the
609 2018-10-29 Martin Liska <mliska@suse.cz>
611 * Makefile.in: Make dependency to json.o.
612 * doc/gcov.texi: Document new JSON format, remove
613 old intermediate format documentation.
614 * gcov.c (struct function_info): Come up with m_name and
616 (function_info::function_info): Initialize it.
617 (function_info::~function_info): Release it.
618 (main): Rename flag_intermediate_format to flag_json_format.
619 (print_usage): Describe --json-format.
620 (process_args): Set flag_json_format.
621 (output_intermediate_line): Remove.
622 (output_intermediate_json_line): Likewise.
623 (get_gcov_intermediate_filename): Return new extension
625 (output_intermediate_file): Implement JSON emission.
626 (output_json_intermediate_file): Implement JSON emission.
627 (generate_results): Use ::get_name for function name.
628 Handle JSON output file.
629 (read_graph_file): Use ::get_name instead of cplus_demangle.
630 (read_count_file): Likewise.
631 (solve_flow_graph): Likewise.
632 (add_line_counts): Likewise.
633 (accumulate_line_counts): Use new flag_json_format.
634 (output_function_details): Use ::get_name instead of cplus_demangle.
635 (output_lines): Likewise.
636 * json.cc (test_writing_literals): Add new tests.
637 * json.h (class literal): Add new boolean constructor.
639 2018-10-29 Segher Boessenkool <segher@kernel.crashing.org>
641 PR rtl-optimization/87701
642 PR rtl-optimization/87780
643 * combine.c (make_more_copies): Rewrite.
645 2018-10-28 Kugan Vivekanandarajah <kuganv@linaro.org>
647 * doc/generic.texi (ABSU_EXPR): Document.
648 * match.pd (absu(x)*absu(x) -> x*x): Handle.
649 (absu(absu(X)) -> absu(X)): Likewise.
650 (absu(-X) -> absu(X)): Likewise.
651 (absu(X) where X is nonnegative -> X): Likewise.
653 2018-10-28 Iain Buclaw <ibuclaw@gdcproject.org>
655 * Makefile.in (tm_d_file_list, tm_d_include_list): New variables.
656 (TM_D_H, D_TARGET_DEF, D_TARGET_H, D_TARGET_OBJS): New variables.
657 (tm_d.h, cs-tm_d.h, default-d.o): New rules.
658 (d/d-target-hooks-def.h, s-d-target-hooks-def-h): New rules.
659 (s-tm-texi): Also check timestamp on d-target.def.
660 (generated_files): Add TM_D_H and d-target-hooks-def.h.
661 (build/genhooks.o): Also depend on D_TARGET_DEF.
662 * config.gcc (tm_d_file, d_target_objs, target_has_targetdm): New
664 * config/aarch64/aarch64-d.c: New file.
665 * config/aarch64/aarch64-linux.h (GNU_USER_TARGET_D_CRITSEC_SIZE):
667 * config/aarch64/aarch64-protos.h (aarch64_d_target_versions): New
669 * config/aarch64/aarch64.h (TARGET_D_CPU_VERSIONS): Define.
670 * config/aarch64/t-aarch64 (aarch64-d.o): New rule.
671 * config/arm/arm-d.c: New file.
672 * config/arm/arm-protos.h (arm_d_target_versions): New prototype.
673 * config/arm/arm.h (TARGET_D_CPU_VERSIONS): Define.
674 * config/arm/linux-eabi.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
675 * config/arm/t-arm (arm-d.o): New rule.
676 * config/default-d.c: New file.
677 * config/glibc-d.c: New file.
678 * config/gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
679 * config/i386/i386-d.c: New file.
680 * config/i386/i386-protos.h (ix86_d_target_versions): New prototype.
681 * config/i386/i386.h (TARGET_D_CPU_VERSIONS): Define.
682 * config/i386/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
683 (GNU_USER_TARGET_D_CRITSEC_SIZE): Define.
684 * config/i386/t-i386 (i386-d.o): New rule.
685 * config/kfreebsd-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
686 * config/kopensolaris-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
687 * config/linux-android.h (ANDROID_TARGET_D_OS_VERSIONS): Define.
688 * config/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
689 * config/mips/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
690 * config/mips/mips-d.c: New file.
691 * config/mips/mips-protos.h (mips_d_target_versions): New prototype.
692 * config/mips/mips.h (TARGET_D_CPU_VERSIONS): Define.
693 * config/mips/t-mips (mips-d.o): New rule.
694 * config/powerpcspe/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
695 * config/powerpcspe/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
696 * config/powerpcspe/powerpcspe-d.c: New file.
697 * config/powerpcspe/powerpcspe-protos.h (rs6000_d_target_versions):
699 * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
700 Support GNU D by using 0 as the language type.
701 * config/powerpcspe/powerpcspe.h (TARGET_D_CPU_VERSIONS): Define.
702 * config/powerpcspe/t-powerpcspe (powerpcspe-d.o): New rule.
703 * config/riscv/riscv-d.c: New file.
704 * config/riscv/riscv-protos.h (riscv_d_target_versions): New
706 * config/riscv/riscv.h (TARGET_D_CPU_VERSIONS): Define.
707 * config/riscv/t-riscv (riscv-d.o): New rule.
708 * config/rs6000/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
709 * config/rs6000/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
710 * config/rs6000/rs6000-d.c: New file.
711 * config/rs6000/rs6000-protos.h (rs6000_d_target_versions): New
713 * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
714 Support GNU D by using 0 as the language type.
715 * config/rs6000/rs6000.h (TARGET_D_CPU_VERSIONS): Define.
716 * config/rs6000/t-rs6000 (rs6000-d.o): New rule.
717 * config/s390/s390-d.c: New file.
718 * config/s390/s390-protos.h (s390_d_target_versions): New prototype.
719 * config/s390/s390.h (TARGET_D_CPU_VERSIONS): Define.
720 * config/s390/t-s390 (s390-d.o): New rule.
721 * config/sparc/sparc-d.c: New file.
722 * config/sparc/sparc-protos.h (sparc_d_target_versions): New
724 * config/sparc/sparc.h (TARGET_D_CPU_VERSIONS): Define.
725 * config/sparc/t-sparc (sparc-d.o): New rule.
726 * config/t-glibc (glibc-d.o): New rule.
727 * configure: Regenerated.
728 * configure.ac (tm_d_file): New variable.
729 (tm_d_file_list, tm_d_include_list, d_target_objs): Add substitutes.
730 * doc/contrib.texi (Contributors): Add self for the D frontend.
731 * doc/frontends.texi (G++ and GCC): Mention D as a supported language.
732 * doc/install.texi (Configuration): Mention libphobos as an option for
733 --enable-shared. Mention d as an option for --enable-languages.
734 (Testing): Mention check-d as a target.
735 * doc/invoke.texi (Overall Options): Mention .d, .dd, and .di as file
736 name suffixes. Mention d as a -x option.
737 * doc/sourcebuild.texi (Top Level): Mention libphobos.
738 * doc/standards.texi (Standards): Add section on D language.
739 * doc/tm.texi: Regenerated.
740 * doc/tm.texi.in: Add @node for D language and ABI, and @hook for
741 TARGET_CPU_VERSIONS, TARGET_D_OS_VERSIONS, and TARGET_D_CRITSEC_SIZE.
742 * dwarf2out.c (is_dlang): New function.
743 (gen_compile_unit_die): Use DW_LANG_D for D.
744 (declare_in_namespace): Return module die for D, instead of adding
745 extra declarations into the namespace.
746 (gen_namespace_die): Generate DW_TAG_module for D.
747 (gen_decl_die): Handle CONST_DECLSs for D.
748 (dwarf2out_decl): Likewise.
749 (prune_unused_types_walk_local_classes): Handle DW_tag_interface_type.
750 (prune_unused_types_walk): Handle DW_tag_interface_type same as other
752 * gcc.c (default_compilers): Add entries for .d, .dd and .di.
753 * genhooks.c: Include d/d-target.def.
755 2018-10-28 Iain Sandoe <iain@sandoe.co.uk>
758 * config/rs6000/darwin.h (STACK_BOUNDARY): New.
759 (RS6000_STARTING_FRAME_OFFSET): Adjust to preserve 16byte alignment.
760 (STACK_DYNAMIC_OFFSET): Likewise.
762 2018-10-27 Sandra Loosemore <sandra@codesourcery.com>
765 * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Fix
768 2018-10-26 Tulio Magno Quites Machado Filho <tuliom@linux.ibm.com>
770 * doc/extend.texi (PowerPC builtins): Fix __builtin_unpack_ibm128
771 return type and other typos.
773 2018-10-26 Aaron Sawdey <acsawdey@linux.ibm.com>
775 * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Change to
776 a shorter sequence with fewer branches.
777 (emit_final_str_compare_gpr): Ditto.
779 2018-10-26 Paul A. Clarke <pc@us.ibm.com>
781 * config/rs6000/tmmintrin.h: New file.
782 * config.gcc (powerpc*-*-*): Add tmmintrin.h to extra_headers.
784 2018-10-26 Paul A. Clarke <pc@us.ibm.com>
786 * config/rs6000/mmintrin.h: Enable 32bit compilation.
787 * config/rs6000/xmmintrin.h: Likewise.
789 2018-10-26 Paul A. Clarke <pc@us.ibm.com>
791 * config/rs6000/xmmintrin.h (_mm_extract_pi16): Fix for big-endian.
793 2018-10-26 Richard Biener <rguenther@suse.de>
795 * tree-vect-slp.c (vect_mark_slp_stmts): Add visited hash_set
797 (vect_mark_slp_stmts_relevant): Likewise.
798 (vect_detect_hybrid_slp_stmts): Likewise.
799 (vect_bb_slp_scalar_cost): Likewise.
800 (vect_remove_slp_scalar_calls): Likewise.
802 2018-10-26 Jan Hubicka <jh@suse.cz>
804 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix recursion.
805 (warn_types_mismatch): Fix walk of DECL_NAME.
806 (odr_types_equivalent_p): Fix overactive assert.
808 2018-10-26 Richard Biener <rguenther@suse.de>
810 PR tree-optimization/87105
811 * tree-vectorizer.h (_slp_tree::refcnt): New member.
812 * tree-vect-slp.c (vect_free_slp_tree): Decrement and honor
814 (vect_create_new_slp_node): Initialize refcnt to one.
816 (scalar_stmts_set_t, bst_fail): Remove.
817 (vect_build_slp_tree_2): Add bst_map argument and adjust calls.
818 (vect_build_slp_tree): Add bst_map argument and lookup
819 already created SLP nodes.
820 (vect_print_slp_tree): Handle a SLP graph, print SLP node
822 (vect_slp_rearrange_stmts): Handle a SLP graph.
823 (vect_analyze_slp_instance): Adjust and free SLP nodes from
824 the CSE map. Fix indenting.
825 (vect_schedule_slp_instance): Add short-cut.
827 2018-10-26 Martin Liska <mliska@suse.cz>
830 * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Hash with
831 addr_expr and not with pointers.
833 2018-10-26 Jan Hubicka <jh@suse.cz>
835 * tree.c (free_lang_data_in_type): Only check main variants.
836 * ipa-devirt.c (warn_odr): Make static.
837 (types_same_for_odr): Drop strict variant.
838 (types_odr_comparable): Likewise.
839 (odr_or_derived_type_p): Look for main variants.
840 (odr_name_hasher::equal): Cleanup comment.
841 (odr_subtypes_equivalent): Add warn and warned arguments; check main
843 (type_variants_equivalent_p): break out from ...
844 (odr_types_equivalent): ... here; go for main variants where needed.
845 (warn_odr): ... here; turn static.
846 (warn_types_mismatch): Compare mangled names of main variants.
847 * ipa-utils.h (types_odr_comparable): Drop strict parameter.
848 (type_with_linkage_p): Sanity check that we look at main variant.
849 * lto.c (lto_read_decls): Only consider main variant to be ODR type.
850 * tree.h (types_same_for_odr): Drop strict argument.
852 2018-10-26 Richard Biener <rguenther@suse.de>
854 PR tree-optimization/87746
855 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
856 Simplify and fix WRT strided store groups with size not
857 equal to step in element count.
858 (vect_analyze_group_access_1): Dump the whole group.
860 2018-10-25 Carl Love <cel@us.ibm.com>
862 * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_VSCEDPGT,
863 P9V_BUILTIN_VEC_VSCEDPLT, P9V_BUILTIN_VEC_VSCEDPEQ,
864 P9V_BUILTIN_VEC_VSCEDPUO): Rename base overloaded name. Add quad
865 precicion entry for each overloaded builtin.
866 * config/rs6000/rs6000-builtin.def (VSCEDPGT, VSCEDPLT, VSCEDPEQ,
867 VSCEDPUO): Rename overloaded name.
868 (VSCEDPGT, VSCEQPGT, VSCEDPLT, VSCEQPLT, VSCEDPEQ, VSCEQPEQ,
869 VSCEDPUO, VSCEQPUO): Add defitions for overloaded builtins.
870 * config/rs6000/vsx.md (xscmpexpqp_<code>_<mode>): Add
871 define_expand for xscmpexqp instruction.
872 (*xscmpexpqp): Add define_insn for the xscmpexqp instruction.
874 2018-10-25 Bill Schmidt <wschmidt@linux.ibm.com>
875 Jinsong Ji <jji@us.ibm.com>
877 * config/rs6000/emmintrin.h (_mm_slli_epi16): Replace deprecated
878 function with vec_sl.
879 (_mm_slli_epi32): Likewise.
880 (_mm_slli_epi64): Likewise.
881 (_mm_srai_epi16): Replace deprecated function with vec_sra.
882 (_mm_srai_epi32): Likewise.
883 (_mm_srli_epi16): Replace deprecated function with vec_sr.
884 (_mm_srli_epi32): Likewise.
885 (_mm_srli_epi64): Likewise.
886 (_mm_sll_epi16): Replace deprecated function with vec_sl.
887 (_mm_sll_epi32): Likewise.
888 (_mm_sll_epi64): Likewise.
889 (_mm_sra_epi16): Replace deprecated function with vec_sra.
890 (_mm_sra_epi32): Likewise.
891 (_mm_srl_epi16): Replace deprecated function with vec_sr.
892 (_mm_srl_epi32): Likewise.
893 (_mm_srl_epi64): Likewise.
895 2018-10-25 Bill Schmidt <wschmidt@linux.ibm.com>
896 Jinsong Ji <jji@us.ibm.com>
898 * gcc/config/rs6000/emmintrin.h (_mm_sll_epi16): Replace
899 comparison operators with vec_cmp* for compatibility due to
900 unfortunate history; clean up formatting and use types more
902 (_mm_sll_epi32): Likewise.
903 (_mm_sll_epi64): Likewise.
904 (_mm_srl_epi16): Likewise.
905 (_mm_srl_epi32): Likewise.
906 (_mm_srl_epi64): Likewise.
908 2018-10-25 Bill Schmidt <wschmidt@linux.ibm.com>
909 Jinsong Ji <jji@us.ibm.com>
911 * config/rs6000/emmintrin.h (_mm_sll_epi64): Remove wrong cast.
912 * config/rs6000/xmmintrin.h (_mm_min_ps): Change m's type to
913 __vector __bool int. Use vec_cmpgt in preference to deprecated
914 function vec_vcmpgtfp.
915 (_mm_max_ps): Likewise.
917 2018-10-25 Jeff Law <law@redhat.com>
919 * config/rl78/rl78.c (insn_ok_now): Always re-recognize the insn
922 2018-10-25 Martin Sebor <msebor@redhat.com>
924 * doc/extend.texi (aligned): Expand attribute description.
925 (Alignment): Rename section. Discuss function arguments.
927 2018-10-25 Jan Hubicka <jh@suse.cz>
929 * ipa-devirt.c (main_odr_variant): Remove.
930 (hash_odr_name, types_same_for_odr, types_odr_comparable,
931 odr_name_hasher::equal, odr_subtypes_equivalent_p):
932 Drop use of main_odr_variant.
933 (add_type_duplicate): Silence confused warnings on integer types.
934 (get_odr_type): Always look for main variant.
935 (register_odr_type): Simplify.
937 2018-10-25 Richard Biener <rguenther@suse.de>
939 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses):
940 Initialize ng to silence error with release checking bootstrap.
942 2018-10-25 Richard Biener <rguenther@suse.de>
944 * tree-if-conv.c: Include tree-ssa-sccvn.h.
945 (tree_if_conversion): Run CSE on the if-converted loop body.
947 2018-10-25 Ilya Leoshkevich <iii@linux.ibm.com>
949 * config/s390/constraints.md (ZL): New constraint.
950 * config/s390/s390.c (legitimate_pic_operand_p): Accept LARL
952 * config/s390/s390.md (movdi_larl): Remove.
953 (movdi_64): Add the LARL alternative.
955 2018-10-25 Ilya Leoshkevich <iii@linux.ibm.com>
958 * rtl.c (RTX_CODE_HWINT_P_1): New helper macro.
959 (RTX_CODE_HWINT_P): New macro.
960 (rtx_code_size): Use RTX_CODE_HWINT_P ().
962 2018-10-25 Jan Hubicka <jh@suse.cz>
964 * ipa-devirt.c (odr_types_equivalent_p): Do not ICE if one of types
967 2018-10-25 Richard Biener <rguenther@suse.de>
969 PR tree-optimization/87665
970 PR tree-optimization/87745
971 * tree-vectorizer.h (get_earlier_stmt): Remove.
972 (get_later_stmt): Pick up UID from the original non-pattern stmt.
974 2018-10-25 Sam Tebbs <sam.tebbs@arm.com>
976 * options.texi (Deprecated): Move list to Var section.
978 2018-10-24 Bill Schmidt <wschmidt@linux.ibm.com>
979 Jinsong Ji <jji@us.ibm.com>
981 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Change deprecated
982 __vector long to __vector long long.
983 (_mm_cvtpd_ps): Likewise.
984 (_mm_cvttpd_epi32): Likewise.
985 (_mm_cvtpi32_pd): Likewise.
986 (_mm_unpackhi_epi64): Likewise.
987 (_mm_unpacklo_epi64): Likewise.
989 2018-10-24 Segher Boessenkool <segher@kernel.crashing.org>
991 PR rtl-optimization/87720
992 * combine.c (make_more_copies): Skip if the dest is pc_rtx.
994 2018-10-24 Alexandre Oliva <aoliva@redhat.com>
996 * gimple-ssa-isolate-paths.c
997 (find_implicit_erroneous_behavior): Do not change code if the
998 pass is running for warnings only.
999 (find_explicit_erroneous_behavior): Likewise.
1001 2018-10-24 Michael Meissner <meissner@linux.ibm.com>
1003 * config/rs6000/rs6000.c (TARGET_MANGLE_DECL_ASSEMBLER_NAME):
1004 Define as rs6000_mangle_decl_assembler_name.
1005 (rs6000_mangle_decl_assembler_name): If the user switched from IBM
1006 long double to IEEE long double, switch the names of the long
1007 double built-in functions to be <func>f128 instead of <func>l.
1009 2018-10-24 Martin Sebor <msebor@redhat.com>
1011 * doc/extend.texi (nonnull): List no-argument form. Reference
1012 -fno-delete-null-pointer-checks and -fisolate-erroneous-paths-attribute.
1014 2018-10-24 Richard Biener <rguenther@suse.de>
1016 * tree-ssa-sccvn.c (do_rpo_vn): Free rpo_state.
1018 2018-10-24 Martin Liska <mliska@suse.cz>
1020 PR tree-optimization/84436
1021 * tree-switch-conversion.c (switch_conversion::contains_same_values_p):
1023 (switch_conversion::contains_linear_function_p): New.
1024 (switch_conversion::build_one_array): Support linear
1025 transformation on input.
1026 * tree-switch-conversion.h (struct switch_conversion): Add
1027 contains_linear_function_p declaration.
1029 2018-10-24 Richard Biener <rguenther@suse.de>
1031 * varasm.c (const_hash_1): Return hash of ADDR_EXPR
1032 if its argument is CONSTANT_CLASS_P.
1034 2018-10-11 Jan Hubicka <hubicka@ucw.cz>
1036 * ipa-utils.h (type_with_linkage_p): No longer check for TYPE_STUB_DECL;
1037 it is wrong for forward declarations.
1039 2018-10-24 Ilya Leoshkevich <iii@linux.ibm.com>
1041 * config/s390/s390.c (s390_check_qrst_address): Add the missing
1042 SYMBOL_REF_P () check.
1044 2018-10-24 Richard Biener <rguenther@suse.de>
1046 PR tree-optimization/87105
1047 * tree-vect-data-refs.c (vect_analyze_group_access_1): Adjust
1048 dump classification.
1049 (vect_analyze_data_ref_accesses): Handle duplicate loads and
1050 stores by splitting the affected group after the fact.
1051 * tree-vect-slp.c (vect_build_slp_tree_2): Dump when we
1052 fail the SLP build because of size constraints.
1054 2018-10-24 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1056 * configure.ac (gcc_cv_ld_aligned_shf_merge): New test.
1057 * configure: Regenerate.
1058 * config.in: Regenerate.
1059 * varasm.c (mergeable_string_section): Use readonly_data_section
1060 if linker doesn't support SHF_MERGE with alignment > 8.
1061 (mergeable_constant_section): Likewise.
1063 2018-10-24 Richard Biener <rguenther@suse.de>
1065 PR tree-optimization/84013
1066 * tree-ssa-structalias.c (struct msdi_data): New struct for
1067 marshalling data to walk_stmt_load_store_ops.
1068 (maybe_set_dependence_info): Refactor as callback for
1069 walk_stmt_load_store_ops.
1070 (compute_dependence_clique): Set restrict info on all stmt kinds.
1072 2018-10-24 Martin Liska <mliska@suse.cz>
1074 * cgraph.c (cgraph_node::dump):
1075 Remove reduntant dumps and make tp_first_run dump more compact.
1077 2018-10-24 Richard Biener <rguenther@suse.de>
1079 PR tree-optimization/87665
1080 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Adjust
1083 2018-10-12 Jeff Law <law@redhat.com>
1085 * config/h8300/h8300.c (h8300_expand_prologue): Fix stm generation
1088 2018-10-23 Richard Biener <rguenther@suse.de>
1090 * tree-vrp.c (add_assert_info): Guard dump_printf with
1092 * gimple-ssa-evrp-analyze.c
1093 (evrp_range_analyzer::record_ranges_from_incoming_edge):
1094 Use value_range::ignore_equivs_equal_p.
1096 2018-10-23 Richard Biener <rguenther@suse.de>
1098 PR tree-optimization/87105
1099 PR tree-optimization/87608
1100 * passes.def (pass_all_early_optimizations): Add early phi-opt
1102 * tree-ssa-phiopt.c (value_replacement): Ignore NOPs and predicts in
1103 addition to debug stmts.
1104 (tree_ssa_phiopt_worker): Add early_p argument, do only min/max
1105 and abs replacement early.
1106 * tree-cfg.c (gimple_empty_block_p): Likewise.
1108 2018-10-23 Richard Earnshaw <rearnsha@arm.com>
1111 * config.gcc (arm*-*-netbsdelf*): Default to StrongARM if no CPU
1112 specified to configure.
1113 (arm*-*-*): Use ARM7TDMI as the target CPU if no default provided.
1115 2018-10-23 Richard Biener <rguenther@suse.de>
1117 PR tree-optimization/87700
1118 * tree-ssa-copy.c (set_copy_of_val): Fix change detection logic.
1120 2018-10-23 Jakub Jelinek <jakub@redhat.com>
1123 * config/i386/avx512vlintrin.h (_mm_mask_mullo_epi32): Change type of
1124 second argument from __mmask16 to __mmask8.
1125 * config/i386/avx512vlbwintrin.h (_mm_mask_packus_epi32,
1126 _mm_mask_packs_epi32): Likewise.
1127 * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i64scatter_ps):
1129 (_mm512_mask_prefetch_i64scatter_pd): Likewise. Formatting fix.
1131 2018-10-23 Richard Biener <rguenther@suse.de>
1133 * tree-vect-stmts.c (vect_analyze_stmt): Fix typo in comment.
1135 2018-10-23 Richard Biener <rguenther@suse.de>
1137 PR tree-optimization/86144
1138 * tree-vect-stmts.c (vect_analyze_stmt): Prefer -mveclibabi
1139 over simd attribute.
1141 2018-10-23 Richard Biener <rguenther@suse.de>
1143 PR tree-optimization/87693
1144 * tree-ssa-threadedge.c (thread_around_empty_blocks): Handle
1145 the case we do not find the taken edge.
1147 2018-10-22 Bill Schmidt <wschmidt@linux.ibm.com>
1148 Jinsong Ji <jji@us.ibm.com>
1150 * config/rs6000/emmintrin.h (_MM_SHUFFLE2): Comment cleanup.
1151 (_mm_store_pd): Use unaligned vector type for pointer cast.
1152 (_mm_maskmoveu_si128): Likewise.
1153 * config/rs6000/xmmintrin.h (__m128_u): New typedef.
1154 (_mm_store_ps): Use unaligned vector type for pointer cast.
1156 2018-10-22 Paul Koning <ni1d@arrl.net>
1158 * symtab.c (symtab_node::increase_alignment): Correct max
1161 2018-10-22 Yury Gribov <tetra2005@gmail.com>
1163 PR tree-optimization/87633
1164 * match.pd: Do not generate unordered integer comparisons.
1166 2018-10-22 Segher Boessenkool <segher@kernel.crashing.org>
1168 PR rtl-optimization/87600
1169 * combine.c: Add include of expr.h.
1170 (cant_combine_insn_p): Do not combine moves from any hard non-fixed
1171 register to a pseudo.
1172 (make_more_copies): New function, add a copy to a new pseudo after
1173 the moves from hard registers into pseudos.
1174 (rest_of_handle_combine): Declare rebuild_jump_labels_after_combine
1175 later. Call make_more_copies.
1177 2018-10-22 Andrew Stubbs <ams@codesourcery.com>
1179 * lra-constraints.c (process_alt_operands): New local array,
1180 matching_early_clobber. Check matching_early_clobber before
1181 decrementing reject, and set matching_early_clobber after.
1183 2018-10-22 Segher Boessenkool <segher@kernel.crashing.org>
1186 * config/rs6000/rs6000.c (print_operand_address): For unexpected RTL
1187 call output_addr_const and hope for the best.
1189 2018-10-22 Richard Biener <rguenther@suse.de>
1191 * gimple-ssa-evrp-analyze.c
1192 (evrp_range_analyzer::record_ranges_from_incoming_edge): Be
1193 smarter about what ranges to use.
1194 * tree-vrp.c (add_assert_info): Dump here.
1195 (register_edge_assert_for_2): Instead of here at multiple but
1198 * gcc.dg/tree-ssa/evrp12.c: New testcase.
1199 * gcc.dg/predict-6.c: Adjust.
1200 * gcc.dg/tree-ssa/vrp33.c: Disable EVRP.
1201 * gcc.dg/tree-ssa/vrp02.c: Likewise.
1202 * gcc.dg/tree-ssa/cunroll-9.c: Likewise.
1204 2018-10-22 Steven Bosscher <steven@gcc.gnu.org>
1205 Richard Biener <rguenther@suse.de>
1207 * bitmap.h: Update data structure documentation, including a
1208 description of bitmap views as either linked-lists or splay trees.
1209 (struct bitmap_element_def): Update comments for splay tree bitmaps.
1210 (struct bitmap_head_def): Likewise.
1211 (bitmap_list_view, bitmap_tree_view): New prototypes.
1212 (bitmap_initialize_stat): Initialize a bitmap_head's indx and
1214 (bmp_iter_set_init): Assert the iterated bitmaps are in list form.
1215 (bmp_iter_and_init, bmp_iter_and_compl_init): Likewise.
1216 * bitmap.c (bitmap_elem_to_freelist): Unregister overhead of a
1217 released bitmap element here.
1218 (bitmap_element_free): Remove.
1219 (bitmap_elt_clear_from): Work on splay tree bitmaps.
1220 (bitmap_list_link_element): Renamed from bitmap_element_link. Move
1221 this function similar ones such that linked-list bitmap implementation
1222 functions are grouped.
1223 (bitmap_list_unlink_element): Renamed from bitmap_element_unlink,
1224 and moved for grouping.
1225 (bitmap_list_insert_element_after): Renamed from
1226 bitmap_elt_insert_after, and moved for grouping.
1227 (bitmap_list_find_element): New function spliced from bitmap_find_bit.
1228 (bitmap_tree_link_left, bitmap_tree_link_right,
1229 bitmap_tree_rotate_left, bitmap_tree_rotate_right, bitmap_tree_splay,
1230 bitmap_tree_link_element, bitmap_tree_unlink_element,
1231 bitmap_tree_find_element): New functions for splay-tree bitmap
1233 (bitmap_element_link, bitmap_element_unlink, bitmap_elt_insert_after):
1234 Renamed and moved, see above entries.
1235 (bitmap_tree_listify_from): New function to convert part of a splay
1236 tree bitmap to a linked-list bitmap.
1237 (bitmap_list_view): Convert a splay tree bitmap to linked-list form.
1238 (bitmap_tree_view): Convert a linked-list bitmap to splay tree form.
1239 (bitmap_find_bit): Remove.
1240 (bitmap_clear, bitmap_clear_bit, bitmap_set_bit,
1241 bitmap_single_bit_set_p, bitmap_first_set_bit, bitmap_last_set_bit):
1242 Handle splay tree bitmaps.
1243 (bitmap_copy, bitmap_count_bits, bitmap_and, bitmap_and_into,
1244 bitmap_elt_copy, bitmap_and_compl, bitmap_and_compl_into,
1245 bitmap_compl_and_into, bitmap_elt_ior, bitmap_ior, bitmap_ior_into,
1246 bitmap_xor, bitmap_xor_into, bitmap_equal_p, bitmap_intersect_p,
1247 bitmap_intersect_compl_p, bitmap_ior_and_compl,
1248 bitmap_ior_and_compl_into, bitmap_set_range, bitmap_clear_range,
1249 bitmap_hash): Reject trying to act on splay tree bitmaps. Make
1250 corresponding changes to use linked-list specific bitmap_element
1251 manipulation functions as applicable for efficiency.
1252 (bitmap_tree_to_vec): New function.
1253 (debug_bitmap_elt_file): New function split out from ...
1254 (debug_bitmap_file): ... here. Handle splay tree bitmaps.
1255 (bitmap_print): Likewise.
1257 PR tree-optimization/63155
1258 * tree-ssa-propagate.c (ssa_prop_init): Use tree-view for the
1260 * tree-ssa-coalesce.c (coalesce_ssa_name): Populate used_in_copies
1263 2018-10-22 Martin Liska <mliska@suse.cz>
1265 PR tree-optimization/87686
1267 2018-08-29 Martin Liska <mliska@suse.cz>
1269 * tree-switch-conversion.c (switch_conversion::expand):
1270 Strenghten assumption about gswitch statements.
1272 2018-10-22 Martin Liska <mliska@suse.cz>
1274 * ipa-icf.c (sem_item::compare_attributes): Remove.
1275 (sem_item::compare_referenced_symbol_properties): Use
1276 attribute_list_equal instead.
1277 (sem_function::equals_wpa): Likewise.
1278 * ipa-icf.h: Remove compare_attributes.
1280 2018-10-22 Richard Biener <rguenther@suse.de>
1283 * mem-stats.h (mem_usage::operator==): Fix pasto.
1285 2018-10-22 Richard Biener <rguenther@suse.de>
1287 PR tree-optimization/87640
1288 * tree-vrp.c (set_value_range_with_overflow): Decompose
1290 (extract_range_from_binary_expr_1): Adjust.
1292 2018-10-22 Martin Jambor <mjambor@suse.cz>
1294 * tree-eh.h (stmt_could_throw_p): Add function parameter.
1295 (stmt_can_throw_external): Likewise.
1296 (stmt_can_throw_internal): Likewise.
1297 * tree-eh.c (lower_eh_constructs_2): Pass cfun to stmt_could_throw_p.
1298 (lower_eh_constructs_2): Likewise.
1299 (stmt_could_throw_p): Add fun parameter, use it instead of cfun.
1300 (stmt_can_throw_external): Likewise.
1301 (stmt_can_throw_internal): Likewise.
1302 (maybe_clean_eh_stmt_fn): Pass cfun to stmt_could_throw_p.
1303 (maybe_clean_or_replace_eh_stmt): Pass cfun to stmt_could_throw_p.
1304 (maybe_duplicate_eh_stmt_fn): Pass new_fun to stmt_could_throw_p.
1305 (maybe_duplicate_eh_stmt): Pass cfun to stmt_could_throw_p.
1306 (pass_lower_eh_dispatch::execute): Pass cfun to
1307 stmt_can_throw_external.
1308 (cleanup_empty_eh): Likewise.
1309 (verify_eh_edges): Pass cfun to stmt_could_throw_p.
1310 * cgraph.c (cgraph_edge::set_call_stmt): Pass a function to
1311 stmt_can_throw_external instead of pushing it to cfun.
1312 (symbol_table::create_edge): Likewise.
1313 * gimple-fold.c (fold_builtin_atomic_compare_exchange): Pass cfun to
1314 stmt_can_throw_internal.
1315 * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Pass cfun
1316 to stmt_could_throw_p.
1317 * gimple-ssa-store-merging.c (handled_load): Pass cfun to
1318 stmt_can_throw_internal.
1319 (pass_store_merging::execute): Likewise.
1320 * gimple-ssa-strength-reduction.c
1321 (find_candidates_dom_walker::before_dom_children): Pass cfun to
1323 * gimplify-me.c (gimple_regimplify_operands): Pass cfun to
1324 stmt_can_throw_internal.
1325 * ipa-pure-const.c (check_call): Pass cfun to stmt_could_throw_p and
1326 to stmt_can_throw_external.
1327 (check_stmt): Pass cfun to stmt_could_throw_p.
1328 (check_stmt): Pass cfun to stmt_can_throw_external.
1329 (pass_nothrow::execute): Likewise.
1330 * trans-mem.c (expand_call_tm): Pass cfun to stmt_can_throw_internal.
1331 * tree-cfg.c (is_ctrl_altering_stmt): Pass cfun to
1332 stmt_can_throw_internal.
1333 (verify_gimple_in_cfg): Pass cfun to stmt_could_throw_p.
1334 (stmt_can_terminate_bb_p): Pass cfun to stmt_can_throw_external.
1335 (gimple_purge_dead_eh_edges): Pass cfun to stmt_can_throw_internal.
1336 * tree-complex.c (expand_complex_libcall): Pass cfun to
1337 stmt_could_throw_p and to stmt_can_throw_internal.
1338 (expand_complex_multiplication): Pass cfun to stmt_can_throw_internal.
1339 * tree-inline.c (copy_edges_for_bb): Likewise.
1340 (maybe_move_debug_stmts_to_successors): Likewise.
1341 * tree-outof-ssa.c (ssa_is_replaceable_p): Pass cfun to
1343 * tree-parloops.c (oacc_entry_exit_ok_1): Likewise.
1344 * tree-sra.c (scan_function): Pass cfun to stmt_can_throw_external.
1345 * tree-ssa-alias.c (stmt_kills_ref_p): Pass cfun to
1346 stmt_can_throw_internal.
1347 * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
1348 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Pass cfun to
1350 (mark_aliased_reaching_defs_necessary_1): Pass cfun to
1351 stmt_can_throw_internal.
1352 * tree-ssa-forwprop.c (pass_forwprop::execute): Likewise.
1353 * tree-ssa-loop-im.c (movement_possibility): Pass cfun to
1355 * tree-ssa-loop-ivopts.c (find_givs_in_stmt_scev): Likewise.
1356 (add_autoinc_candidates): Pass cfun to stmt_can_throw_internal.
1357 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
1358 (convert_mult_to_fma_1): Likewise.
1359 (convert_to_divmod): Likewise.
1360 * tree-ssa-phiprop.c (propagate_with_phi): Likewise.
1361 * tree-ssa-pre.c (compute_avail): Pass cfun to stmt_could_throw_p.
1362 * tree-ssa-propagate.c
1363 (substitute_and_fold_dom_walker::before_dom_children): Likewise.
1364 * tree-ssa-reassoc.c (suitable_cond_bb): Likewise.
1365 (maybe_optimize_range_tests): Likewise.
1366 (linearize_expr_tree): Likewise.
1367 (reassociate_bb): Likewise.
1368 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
1369 * tree-ssa-scopedtables.c (hashable_expr_equal_p): Likewise.
1370 * tree-ssa-strlen.c (adjust_last_stmt): Likewise.
1371 (handle_char_store): Likewise.
1372 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Pass cfun to
1373 stmt_can_throw_internal.
1374 * tree-vect-patterns.c (check_bool_pattern): Pass cfun to
1376 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
1377 (vectorizable_call): Pass cfun to stmt_can_throw_internal.
1378 (vectorizable_simd_clone_call): Likewise.
1379 * value-prof.c (gimple_ic): Pass cfun to stmt_could_throw_p.
1380 (gimple_stringop_fixed_value): Likewise.
1382 2018-10-22 Ilya Leoshkevich <iii@linux.ibm.com>
1384 * config/s390/s390.c (s390_loadrelative_operand_p): Accept
1385 literal pool references.
1386 (s390_check_qrst_address): Adapt to the new behavior of
1387 s390_loadrelative_operand_p ().
1389 2018-10-22 H.J. Lu <hongjiu.lu@intel.com>
1392 * config/i386/sse.md (*andnot<mode>3_bcst): New.
1394 2018-10-22 H.J. Lu <hongjiu.lu@intel.com>
1397 * config/i386/sse.md (*<code><mode>3_bcst): New.
1399 2018-10-22 H.J. Lu <hongjiu.lu@intel.com>
1402 * config/i386/sse.md (avx512bcst): Updated for V4SI, V2DI, V8SI,
1403 V4DI, V16SI and V8DI.
1404 (*sub<mode>3<mask_name>_bcst): New.
1405 (*add<mode>3<mask_name>_bcst): Likewise.
1407 2018-10-21 Bill Schmidt <wschmidt@linux.ibm.com>
1408 Jinsong Ji <jji@us.ibm.com>
1410 * config/rs6000/emmintrin.h (_mm_movemask_pd): Replace __vector
1411 __m64 with __vector unsigned long long for compatibility.
1412 (_mm_movemask_epi8): Likewise.
1413 * config/rs6000/xmmintrin.h (_mm_cvtps_pi32): Likewise.
1414 (_mm_cvttps_pi32): Likewise.
1415 (_mm_cvtpi32_ps): Likewise.
1416 (_mm_cvtps_pi16): Likewise.
1417 (_mm_loadh_pi): Likewise.
1418 (_mm_storeh_pi): Likewise.
1419 (_mm_movehl_ps): Likewise.
1420 (_mm_movelh_ps): Likewise.
1421 (_mm_loadl_pi): Likewise.
1422 (_mm_storel_pi): Likewise.
1423 (_mm_movemask_ps): Likewise.
1424 (_mm_shuffle_pi16): Likewise.
1426 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
1429 * config/i386/avx512fintrin.h (_mm512_fnmsub_round_pd): Use
1430 __builtin_ia32_vfnmsubpd512_mask.
1431 (_mm512_mask_fnmsub_round_pd): Likewise.
1432 (_mm512_fnmsub_pd): Likewise.
1433 (_mm512_mask_fnmsub_pd): Likewise.
1434 (_mm512_maskz_fnmsub_round_pd): Use
1435 __builtin_ia32_vfnmsubpd512_maskz.
1436 (_mm512_maskz_fnmsub_pd): Likewise.
1437 (_mm512_fnmsub_round_ps): Use __builtin_ia32_vfnmsubps512_mask.
1438 (_mm512_mask_fnmsub_round_ps): Likewise.
1439 (_mm512_fnmsub_ps): Likewise.
1440 (_mm512_mask_fnmsub_ps): Likewise.
1441 (_mm512_maskz_fnmsub_round_ps): Use
1442 __builtin_ia32_vfnmsubps512_maskz.
1443 (_mm512_maskz_fnmsub_ps): Likewise.
1444 * config/i386/avx512vlintrin.h (_mm256_mask_fnmsub_pd): Use
1445 __builtin_ia32_vfnmsubpd256_mask.
1446 (_mm256_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256_maskz.
1447 (_mm_mask_fnmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
1448 (_mm_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd128_maskz.
1449 (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
1450 (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
1451 (_mm256_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_maskz.
1452 (_mm_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_mask.
1453 (_mm_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_maskz.
1454 * config/i386/fmaintrin.h (_mm_fnmsub_pd): Use
1455 __builtin_ia32_vfnmsubpd.
1456 (_mm256_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256.
1457 (_mm_fnmsub_ps): Use __builtin_ia32_vfnmsubps.
1458 (_mm256_fnmsub_ps): Use __builtin_ia32_vfnmsubps256.
1459 (_mm_fnmsub_sd): Use __builtin_ia32_vfnmsubsd3.
1460 (_mm_fnmsub_ss): Use __builtin_ia32_vfnmsubss3.
1461 * config/i386/i386-builtin.def: Add
1462 __builtin_ia32_vfnmsubpd256_mask,
1463 __builtin_ia32_vfnmsubpd256_maskz,
1464 __builtin_ia32_vfnmsubpd128_mask,
1465 __builtin_ia32_vfnmsubpd128_maskz,
1466 __builtin_ia32_vfnmsubps256_mask,
1467 __builtin_ia32_vfnmsubps256_maskz,
1468 __builtin_ia32_vfnmsubps128_mask,
1469 __builtin_ia32_vfnmsubps128_maskz,
1470 __builtin_ia32_vfnmsubpd512_mask,
1471 __builtin_ia32_vfnmsubpd512_maskz,
1472 __builtin_ia32_vfnmsubps512_mask,
1473 __builtin_ia32_vfnmsubps512_maskz, __builtin_ia32_vfnmsubss3,
1474 __builtin_ia32_vfnmsubsd3, __builtin_ia32_vfnmsubps,
1475 __builtin_ia32_vfnmsubpd, __builtin_ia32_vfnmsubps256 and.
1476 __builtin_ia32_vfnmsubpd256.
1477 * config/i386/sse.md (fma4i_fnmsub_<mode>): New.
1478 (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Likewise.
1479 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1):
1481 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2):
1483 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3):
1485 (fmai_vmfnmsub_<mode><round_name>): Likewise.
1487 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
1490 * config/i386/avx512fintrin.h (_mm512_fnmadd_round_pd): Use
1491 __builtin_ia32_vfnmaddpd512_mask.
1492 (_mm512_mask_fnmadd_round_pd): Likewise.
1493 (_mm512_fnmadd_pd): Likewise.
1494 (_mm512_mask_fnmadd_pd): Likewise.
1495 (_mm512_maskz_fnmadd_round_pd): Use
1496 __builtin_ia32_vfnmaddpd512_maskz.
1497 (_mm512_maskz_fnmadd_pd): Likewise.
1498 (_mm512_fnmadd_round_ps): Use __builtin_ia32_vfnmaddps512_mask.
1499 (_mm512_mask_fnmadd_round_ps): Likewise.
1500 (_mm512_fnmadd_ps): Likewise.
1501 (_mm512_mask_fnmadd_ps): Likewise.
1502 (_mm512_maskz_fnmadd_round_ps): Use
1503 __builtin_ia32_vfnmaddps512_maskz.
1504 (_mm512_maskz_fnmadd_ps): Likewise.
1505 * config/i386/avx512vlintrin.h (_mm256_mask_fnmadd_pd): Use
1506 __builtin_ia32_vfnmaddpd256_mask.
1507 (_mm256_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256_maskz.
1508 (_mm_mask_fnmadd_pd): Use __builtin_ia32_vfmaddpd128_mask
1509 (_mm_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd128_maskz.
1510 (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
1511 (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
1512 (_mm256_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_maskz.
1513 (_mm_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_mask.
1514 (_mm_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_maskz.
1515 * config/i386/fmaintrin.h (_mm_fnmadd_pd): Use
1516 __builtin_ia32_vfnmaddpd.
1517 (_mm256_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256.
1518 (_mm_fnmadd_ps): Use __builtin_ia32_vfnmaddps.
1519 (_mm256_fnmadd_ps): Use __builtin_ia32_vfnmaddps256.
1520 (_mm_fnmadd_sd): Use __builtin_ia32_vfnmaddsd3.
1521 (_mm_fnmadd_ss): Use __builtin_ia32_vfnmaddss3.
1522 * config/i386/i386-builtin.def: Add
1523 __builtin_ia32_vfnmaddpd256_mask,
1524 __builtin_ia32_vfnmaddpd256_maskz,
1525 __builtin_ia32_vfnmaddpd128_mask,
1526 __builtin_ia32_vfnmaddpd128_maskz,
1527 __builtin_ia32_vfnmaddps256_mask,
1528 __builtin_ia32_vfnmaddps256_maskz,
1529 __builtin_ia32_vfnmaddps128_mask,
1530 __builtin_ia32_vfnmaddps128_maskz,
1531 __builtin_ia32_vfnmaddpd512_mask,
1532 __builtin_ia32_vfnmaddpd512_maskz,
1533 __builtin_ia32_vfnmaddps512_mask,
1534 __builtin_ia32_vfnmaddps512_maskz, __builtin_ia32_vfnmaddss3,
1535 __builtin_ia32_vfnmaddsd3, __builtin_ia32_vfnmaddps,
1536 __builtin_ia32_vfnmaddpd, __builtin_ia32_vfnmaddps256 and.
1537 __builtin_ia32_vfnmaddpd256.
1538 * config/i386/sse.md (fma4i_fnmadd_<mode>): New.
1539 (<avx512>_fnmadd_<mode>_maskz<round_expand_name>): Likewise.
1540 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1):
1542 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2):
1544 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3):
1546 (fmai_vmfnmadd_<mode><round_name>): Likewise.
1548 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
1551 * config/i386/avx512fintrin.h (_mm512_fmsub_round_pd): Use
1552 __builtin_ia32_vfmsubpd512_mask.
1553 (_mm512_mask_fmsub_round_pd): Likewise.
1554 (_mm512_fmsub_pd): Likewise.
1555 (_mm512_mask_fmsub_pd): Likewise.
1556 (_mm512_maskz_fmsub_round_pd): Use
1557 __builtin_ia32_vfmsubpd512_maskz.
1558 (_mm512_maskz_fmsub_pd): Likewise.
1559 (_mm512_fmsub_round_ps): Use __builtin_ia32_vfmsubps512_mask.
1560 (_mm512_mask_fmsub_round_ps): Likewise.
1561 (_mm512_fmsub_ps): Likewise.
1562 (_mm512_mask_fmsub_ps): Likewise.
1563 (_mm512_maskz_fmsub_round_ps): Use
1564 __builtin_ia32_vfmsubps512_maskz.
1565 (_mm512_maskz_fmsub_ps): Likewise.
1566 * config/i386/avx512vlintrin.h (_mm256_mask_fmsub_pd): Use
1567 __builtin_ia32_vfmsubpd256_mask.
1568 (_mm256_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd256_maskz.
1569 (_mm_mask_fmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
1570 (_mm_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd128_maskz.
1571 (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
1572 (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
1573 (_mm256_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps256_maskz.
1574 (_mm_mask_fmsub_ps): Use __builtin_ia32_vfmsubps128_mask.
1575 (_mm_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps128_maskz.
1576 * config/i386/fmaintrin.h (_mm_fmsub_pd): Use
1577 __builtin_ia32_vfmsubpd.
1578 (_mm256_fmsub_pd): Use __builtin_ia32_vfmsubpd256.
1579 (_mm_fmsub_ps): Use __builtin_ia32_vfmsubps.
1580 (_mm256_fmsub_ps): Use __builtin_ia32_vfmsubps256.
1581 (_mm_fmsub_sd): Use __builtin_ia32_vfmsubsd3.
1582 (_mm_fmsub_ss): Use __builtin_ia32_vfmsubss3.
1583 * config/i386/i386-builtin.def: Add
1584 __builtin_ia32_vfmsubpd256_mask,
1585 __builtin_ia32_vfmsubpd256_maskz,
1586 __builtin_ia32_vfmsubpd128_mask,
1587 __builtin_ia32_vfmsubpd128_maskz,
1588 __builtin_ia32_vfmsubps256_mask,
1589 __builtin_ia32_vfmsubps256_maskz,
1590 __builtin_ia32_vfmsubps128_mask,
1591 __builtin_ia32_vfmsubps128_maskz,
1592 __builtin_ia32_vfmsubpd512_mask,
1593 __builtin_ia32_vfmsubpd512_maskz,
1594 __builtin_ia32_vfmsubps512_mask,
1595 __builtin_ia32_vfmsubps512_maskz, __builtin_ia32_vfmsubss3,
1596 __builtin_ia32_vfmsubsd3, __builtin_ia32_vfmsubps,
1597 __builtin_ia32_vfmsubpd, __builtin_ia32_vfmsubps256 and.
1598 __builtin_ia32_vfmsubpd256.
1599 * config/i386/sse.md (fma4i_fmsub_<mode>): New.
1600 (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Likewise.
1601 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1):
1603 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2):
1605 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3):
1607 (fmai_vmfmsub_<mode><round_name>): Likewise.
1609 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
1611 * config/i386/sse.md (*<plusminus_insn><mode>3<mask_name>_bcst_1):
1612 Remove plus. Renamed to ...
1613 (*sub<mode>3<mask_name>_bcst): This.
1614 (*add<mode>3<mask_name>_bcst_2): Renamede to ...
1615 (*add<mode>3<mask_name>_bcst): This.
1617 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
1620 * config/i386/sse.md (*mul<mode>3<mask_name>_bcst): New.
1622 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
1625 * i386/avx512vlintrin.h (_mm256_or_epi32): New.
1626 (_mm_or_epi32): Likewise.
1627 (_mm256_xor_epi32): Likewise.
1628 (_mm_xor_epi32): Likewise.
1629 (_mm256_or_epi64): Likewise.
1630 (_mm_or_epi64): Likewise.
1631 (_mm256_xor_epi64): Likewise.
1632 (_mm_xor_epi64): Likewise.
1634 2018-10-20 H.J. Lu <hongjiu.lu@intel.com>
1637 * config/i386/sse.md (*<avx512>_div<mode>3<mask_name>_bcst): New.
1639 2018-10-20 Jakub Jelinek <jakub@redhat.com>
1642 * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR.
1644 2018-10-20 Andreas Schwab <schwab@linux-m68k.org>
1646 * doc/ux.texi: Move @section directly after @node.
1648 2018-10-19 Jakub Jelinek <jakub@redhat.com>
1652 * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
1653 depend closely nested inside of loop with ordered clause with
1656 2018-10-19 David Malcolm <dmalcolm@redhat.com>
1658 * Makefile.in (TEXI_GCCINT_FILES): Add ux.texi.
1659 * doc/gccint.texi: Include ux.texi and use it in top-level menu.
1660 * doc/ux.texi: New file.
1662 2018-10-19 Segher Boessenkool <segher@kernel.crashing.org>
1664 * config/rs6000/rs6000.h (REG_ALLOC_ORDER): Move 68 (that is, CR0) to
1665 be the first CR field allocated.
1667 2018-10-19 Richard Biener <rguenther@suse.de>
1670 * config/i386/i386.c (ix86_builtin_vectorization_cost): Use
1671 TYPE_VECTOR_SUBPARTS and avoid relying on vector mode.
1673 2018-10-19 H.J. Lu <hongjiu.lu@intel.com>
1676 * config/i386/sse.md
1677 (*<plusminus_insn><mode>3<mask_name>_bcst_1): New.
1678 (*add<mode>3<mask_name>_bcst_2): Likewise.
1680 2018-10-19 H.J. Lu <hongjiu.lu@intel.com>
1682 * config/i386/sse.md
1683 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
1684 Replace nonimmediate_operand with register_operand.
1685 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
1687 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
1690 2018-10-19 Ilya Leoshkevich <iii@linux.ibm.com>
1692 PR rtl-optimization/87596
1693 * lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () +
1694 lra_get_insn_recog_data () instead of lra_insn_recog_data[]
1695 for instructions in FROM..TO range.
1697 2018-10-19 Eric Botcazou <ebotcazou@adacore.com>
1699 * cfgexpand.c (expand_one_var): Use specific wording in error message
1700 for non-local frame variables.
1701 * stor-layout.c (layout_decl): Do not issue a warning for them.
1703 2018-10-19 Robin Dapp <rdapp@linux.ibm.com>
1705 * haifa-sched.c (priority): Add force_recompute parameter.
1706 (apply_replacement): Call priority () with force_recompute = true.
1707 (restore_pattern): Likewise.
1709 2018-10-18 H.J. Lu <hongjiu.lu@intel.com>
1711 * simplify-rtx.c (simplify_subreg): Limit mask of vec_merge to
1712 HOST_BITS_PER_WIDE_INT.
1713 (test_vector_ops_duplicate): Likewise.
1715 2018-10-18 H.J. Lu <hongjiu.lu@intel.com>
1718 * config/i386/sse.md (VF_AVX512): New.
1719 (avx512bcst): Likewise.
1720 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
1722 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
1724 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
1727 2018-10-18 Jonathan Wakely <jwakely@redhat.com>
1729 * doc/invoke.texi (-dumpversion): Improve grammar.
1730 (-dumpfullversion): Make more consistent with -dumpversion.
1732 2018-10-18 Uros Bizjak <ubizjak@gmail.com>
1734 * config/i386/i386.c (ix86_emit_fp_unordered_jump):
1735 Set JUMP_LABEL to the jump insn.
1736 (ix86_emit_i387_log1p): Use ix86_expand_branch to expand branch.
1737 Predict emitted jump and add label to jump insn.
1739 2018-10-18 David Malcolm <dmalcolm@redhat.com>
1741 PR tree-optimization/87562
1742 * input.c (get_substring_ranges_for_loc): Use
1743 LRK_SPELLING_LOCATION rather than LRK_MACRO_EXPANSION_POINT when
1744 getting the linemap for the endpoint. Verify that it's either
1745 in the same linemap as the start point's spelling location, or
1746 at least in the same file.
1748 2018-10-18 Richard Biener <rguenther@suse.de>
1750 * config/i386/i386.c (ix86_builtin_vectorization_cost): Do not
1751 feed width-specific load/store costs through ix86_vec_cost.
1752 * config/i386/x86-tune-costs.h (athlon_cost): Adjust.
1753 (k8_cost): Likewise.
1754 (bdver_cost): Likewise.
1755 (znver1_cost): Likewise.
1756 (btver1_cost): Likewise.
1757 (btver2_cost): Likewise.
1759 2018-10-18 H.J. Lu <hongjiu.lu@intel.com>
1761 * simplify-rtx.c (simplify_subreg): Call simplify_gen_subreg
1762 to simplify subreg of vec_merge.
1764 2018-10-18 Richard Biener <rguenther@suse.de>
1766 * config/i386/i386.c: Fix costing of vector FMA.
1768 2018-10-18 Richard Biener <rguenther@suse.de>
1770 * config/i386/i386.c (ix86_vec_cost): Remove !parallel path
1772 (ix86_builtin_vectorization_cost): For vec_construct properly
1773 cost insertion into SSE regs.
1774 (...): Adjust calls to ix86_vec_cost.
1776 2018-10-18 Richard Biener <rguenther@suse.de>
1780 2018-02-07 Richard Biener <rguenther@suse.de>
1782 PR tree-optimization/84204
1783 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
1786 2018-10-18 H.J. Lu <hongjiu.lu@intel.com>
1789 * simplify-rtx.c (simplify_subreg): Simplify subreg of vec_merge
1791 (test_vector_ops_duplicate): Add test for a scalar subreg of a
1792 VEC_MERGE of a VEC_DUPLICATE.
1794 2018-10-17 Joseph Myers <joseph@codesourcery.com>
1796 * doc/cpp.texi (__STDC_VERSION__): Document C2X handling.
1797 * doc/invoke.texi (-std=c2x, -std=gnu2x): Document new options.
1798 * doc/standards.texi (C Language): Document C2X.
1799 * dwarf2out.c (highest_c_language), config/rl78/rl78.c
1800 (rl78_option_override): Handle "GNU C2X" language name.
1802 2018-10-17 Joseph Myers <joseph@codesourcery.com>
1804 * doc/invoke.texi (-std=c17), doc/standards.texi (C Language):
1805 Document C17 as published in 2018.
1807 2018-10-17 Eric Botcazou <ebotcazou@adacore.com>
1810 * fold-const.c (fold_truth_andor_1): If the right side is not constant,
1811 bail out if both sides do not have the same storage order.
1813 2018-10-17 Aldy Hernandez <aldyh@redhat.com>
1815 * bitmap.c (bitmap_head::dump): New.
1816 * bitmap.h (bitmap_head): Add dump().
1817 * gimple-ssa-evrp-analyze.c
1818 (evrp_range_analyzer::try_find_new_range): Adjust for value_range API.
1819 (evrp_range_analyzer::set_ssa_range_info): Same.
1820 (evrp_range_analyzer::record_ranges_from_phis): Same.
1821 (evrp_range_analyzer::record_ranges_from_stmt): Same.
1822 * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Same.
1823 * gimple-ssa-sprintf.c (get_int_range): Same.
1824 (format_integer): Same.
1825 (sprintf_dom_walker::handle_gimple_call): Same.
1826 * ipa-cp.c (ipcp_vr_lattice::meet_with_1): Same.
1827 (ipcp_vr_lattice::top_p): Same.
1828 (ipcp_vr_lattice::bottom_p): Same.
1829 (ipcp_vr_lattice::set_to_bottom): Same.
1830 (ipa_vr_operation_and_type_effects): Same.
1831 (propagate_vr_across_jump_function): Same.
1832 (ipcp_store_vr_results): Same.
1833 * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Same.
1834 (ipa_print_node_jump_functions_for_edge): Same.
1835 (ipa_get_value_range): Same.
1836 (ipa_compute_jump_functions_for_edge): Same.
1837 (ipa_write_jump_function): Same.
1838 * tree-ssa-dom.c (simplify_stmt_for_jump_threading): Same.
1839 * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
1841 * vr-values.c (set_value_range_to_nonnegative): Same.
1842 (set_value_range_to_truthvalue): Same.
1843 (vr_values::get_value_range): Same.
1844 (vr_values::set_defs_to_varying): Same.
1845 (vr_values::update_value_range): Same.
1846 (symbolic_range_based_on_p): Same.
1847 (vr_values::op_with_boolean_value_range_p): Same.
1848 (vr_values::extract_range_for_var_from_comparison_expr): Same.
1849 (vr_values::extract_range_from_ssa_name): Same.
1850 (vr_values::extract_range_from_binary_expr): Same.
1851 (vr_values::extract_range_from_unary_expr): Same.
1852 (vr_values::extract_range_from_cond_expr): Same.
1853 (vr_values::extract_range_from_comparison): Same.
1854 (vr_values::check_for_binary_op_overflow): Same.
1855 (vr_values::extract_range_basic): Same.
1856 (vr_values::extract_range_from_assignment): Same.
1857 (compare_ranges): Same.
1858 (compare_range_with_value): Same.
1859 (vr_values::adjust_range_with_scev): Same.
1860 (vrp_valueize): Same.
1861 (vrp_valueize_1): Same.
1862 (vr_values::get_vr_for_comparison): Same.
1863 (vr_values::compare_name_with_value): Same.
1864 (vr_values::compare_names): Same.
1865 (vr_values::vrp_evaluate_conditional): Same.
1866 (find_case_label_ranges): Same.
1867 (vr_values::vrp_visit_switch_stmt): Same.
1868 (vr_values::extract_range_from_phi_node): Same.
1869 (vr_values::simplify_div_or_mod_using_ranges): Same.
1870 (vr_values::simplify_bit_ops_using_ranges): Same.
1871 (test_for_singularity): Same.
1872 (range_fits_type_p): Same.
1873 (vr_values::simplify_cond_using_ranges_1): Same.
1874 (vr_values::simplify_switch_using_ranges): Same.
1875 (vr_values::simplify_float_conversion_using_ranges): Same.
1876 (vr_values::two_valued_val_range_p): Same.
1877 (vr_values::add_equivalence): Move to value_range::equiv_add.
1878 * vr-values.h (vr_values::add_equivalence): Remove.
1879 (VR_INITIALIZER): Remove.
1880 * tree-vrp.c (value_range::set): New.
1881 (value_range::equiv_add): New.
1882 (value_range::value_range): New.
1883 (value_range::deep_copy): New.
1884 (value_range::check): New.
1885 (value_range::equal_p): New.
1886 (value_range::ignore_equivs_equal_p): New.
1887 (value_range::operator==): New.
1888 (value_range::operator!=): New.
1889 (value_range::symbolic_p): New.
1890 (value_range::numeric_p): New.
1891 (value_range::set_undefined): New.
1892 (value_range::set_varying): New.
1893 (value_range::may_contain_p): New.
1894 (value_range::equiv_clear): New.
1895 (value_range::singleton_p): New.
1896 (value_range::intersect): New.
1897 (value_range::dump): New.
1898 (value_range::set_and_canonicalize): New.
1899 (set_value_range): Adjust for value_range API.
1900 (set_value_range_to_undefined): Same.
1901 (set_value_range_to_varying): Same.
1902 (set_and_canonicalize_value_range): Same.
1903 (set_value_range_to_nonnull): Same.
1904 (set_value_range_to_null): Same.
1905 (range_is_null): Same.
1906 (range_is_nonnull): Same.
1907 (range_int_cst_p): Same.
1908 (range_int_cst_singleton_p): Same.
1909 (symbolic_range_p): Same.
1910 (range_includes_zero_p): Same.
1911 (value_range_constant_singleton): Same.
1912 (vrp_set_zero_nonzero_bits): Same.
1913 (ranges_from_anti_range): Same.
1914 (extract_range_into_wide_ints): Same.
1915 (extract_range_from_multiplicative_op): Same.
1916 (set_value_range_with_overflow): Same.
1917 (extract_range_from_binary_expr_1): Same.
1918 (extract_range_from_unary_expr): Same.
1919 (dump_value_range): Same.
1920 (debug_value_range): Same.
1921 (vrp_prop::check_array_ref): Same.
1922 (vrp_prop::check_mem_ref): Same.
1923 (vrp_prop::vrp_initialize): Same.
1924 (vrp_prop::visit_stmt): Same.
1925 (intersect_ranges): Same.
1926 (vrp_prop::visit_phi): Same.
1927 (vrp_prop::vrp_finalize): Same.
1928 (determine_value_range_1): Same.
1929 (determine_value_range): Same.
1930 (vrp_intersect_ranges_1): Rename to...
1931 (vrp_intersect_1): this.
1932 (vrp_intersect_ranges): Rename to...
1933 (value_range::intersect_helper): ...this.
1934 (vrp_meet_1): Rename to...
1935 (value_range::union_helper): ...this.
1936 (vrp_meet): Rename to...
1937 (value_range::union_): ...this.
1938 (copy_value_range): Remove.
1939 * tree-vrp.h (struct value_range): Rewrite into a proper class.
1940 (value_range::vrtype): New.
1941 (value_range::type): New.
1942 (value_range::equiv): New.
1943 (value_range::min): New.
1944 (value_range::max): New.
1945 (value_range::varying_p): New.
1946 (value_range::undefined_p): New.
1947 (value_range::null_p): New.
1948 (value_range::equiv_add): New.
1949 (copy_value_range): Remove.
1951 2018-10-17 David Malcolm <dmalcolm@redhat.com>
1953 * Makefile.in (SELFTEST_TARGETS): New.
1954 (selftest) Change from s-selftest-c to $(SELFTEST_TARGETS).
1955 (C_SELFTEST_FLAGS, C_SELFTEST_DEPS, s-selftest-c, selftest-c-gdb)
1956 (selftest-gdb, selftest-c-valgrind, selftest-valgrind): Move to
1958 (CPP_SELFTEST_FLAGS, CPP_SELFTEST_DEPS, s-selftest-c++)
1959 (selftest-c++-gdb, selftest-c++-valgrind): Move to
1961 * configure: Regenerate.
1962 * configure.ac (selftest_languages): New.
1964 2018-10-17 Aldy Hernandez <aldyh@redhat.com>
1966 * tree-vrp.c (extract_range_from_multiplicative_op): Remove
1967 overflow wraps argument.
1968 (extract_range_from_binary_expr_1): Do not pass overflow wraps to
1969 wide_int_range_multiplicative_op.
1970 * wide-int-range.cc (wide_int_range_mult_wrapping): Remove
1971 overflow wraps argument.
1972 (wide_int_range_multiplicative_op): Same.
1973 (wide_int_range_lshift): Same.
1974 (wide_int_range_div): Same.
1975 * wide-int-range.h (wide_int_range_multiplicative_op): Same.
1976 (wide_int_range_lshift): Same.
1977 (wide_int_range_div): Same.
1979 2018-10-17 Aldy Hernandez <aldyh@redhat.com>
1981 * wide-int-range.h (wide_int_range_shift_undefined_p): Adjust to
1982 use sign as argument.
1983 * tree-vrp.c (extract_range_from_binary_expr_1): Pass sign to
1984 wide_int_range_shift_undefined_p.
1986 2018-10-16 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1988 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>):
1990 (@despeculate_copy<ALLI_TI:mode>): ... This.
1991 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): Remove
1994 2018-10-16 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1996 * config.gcc: Obsolete *-*-solaris2.10*.
1997 * doc/install.texi (Specific, *-*-solaris2*): Document it.
1999 2018-10-12 Jeff Law <law@redhat.com>
2001 * config/ft32/ft32.md (ft32_general_movsrc_operand): Disable
2002 reg + sym +- const_int addressing modes.
2004 2018-10-15 David Malcolm <dmalcolm@redhat.com>
2006 * common.opt (fdiagnostics-minimum-margin-width=): New option.
2007 * diagnostic-show-locus.c (layout::layout): Apply the minimum
2009 (layout::start_annotation_line): Only print up to 3 of the
2010 margin character, to avoid touching the left-hand side.
2011 (selftest::test_diagnostic_show_locus_fixit_lines): Update for
2012 minimum margin width, as set by test_diagnostic_context's ctor.
2013 (selftest::test_fixit_insert_containing_newline): Likewise.
2014 (selftest::test_fixit_insert_containing_newline_2): Likewise.
2015 (selftest::test_line_numbers_multiline_range): Clear
2016 dc.min_margin_width.
2017 * diagnostic.c (diagnostic_initialize): Initialize
2019 * diagnostic.h (struct diagnostic_context): Add field
2021 * doc/invoke.texi: Add -fdiagnostics-minimum-margin-width=.
2022 * opts.c (common_handle_option): Handle
2023 OPT_fdiagnostics_minimum_margin_width_.
2024 * selftest-diagnostic.c
2025 (selftest::test_diagnostic_context::test_diagnostic_context):
2026 Initialize min_margin_width to 6.
2027 * toplev.c (general_init): Initialize global_dc->min_margin_width.
2029 2018-10-15 David Malcolm <dmalcolm@redhat.com>
2031 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
2032 Fix usage of "error_at_rich_loc" in the comment.
2034 2018-10-15 Renlin Li <renlin.li@arm.com>
2037 * tree-vectorizer.c (try_vectorize_loop_1): Don't use
2038 if-conversioned loop when it contains ifn with types not
2039 supported by backend.
2040 * internal-fn.c (expand_direct_optab_fn): Add an assert.
2041 (direct_internal_fn_supported_p): New helper function.
2042 * internal-fn.h (direct_internal_fn_supported_p): Declare.
2044 2018-10-15 Jakub Jelinek <jakub@redhat.com>
2047 * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
2048 Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
2049 OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
2051 2018-10-15 Bin Cheng <bin.cheng@linux.alibaba.com>
2053 PR tree-optimization/87022
2054 * tree-loop-distribution.c (pg_add_dependence_edges): Check all
2055 bits in dist vector rather than the first one.
2057 2018-10-15 Richard Biener <rguenther@suse.de>
2060 * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
2061 (visit_loadstore): When a used restrict tag escaped verify that
2062 the points-to solution of "other" pointers do not include
2064 (compute_dependence_clique): If a used restrict tag escaped
2065 communicated that down to visit_loadstore.
2067 2018-10-15 Andreas Krebbel <krebbel@linux.ibm.com>
2069 * config/s390/s390.c (s390_expand_vec_init): Force vector element
2070 into reg if it isn't a general operand.
2072 2018-10-14 H.J. Lu <hongjiu.lu@intel.com>
2075 * config/i386/sse.md (*vec_dupv2di): Add register source to
2078 2018-10-14 H.J. Lu <hongjiu.lu@intel.com>
2081 * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
2082 Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
2083 OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
2084 OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
2086 2018-10-13 Eric Botcazou <ebotcazou@adacore.com>
2088 * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
2089 (notice_args_size): Set it in the current trace if no insn that can
2090 throw internally has been seen yet.
2091 (connect_traces): When connecting args_size between traces, allow the
2092 incoming values not to match if there is an insn setting it before the
2093 first insn that can throw internally; in that case, force the creation
2094 of a CFI note on this latter insn.
2096 2018-10-13 Jonathan Wakely <jwakely@redhat.com>
2098 * opt-problem.h (opt_wrapper): Use template-argument-list when naming
2099 the base class, because using the injected-class-name was not clearly
2100 specified until DR 176.
2102 2018-10-12 Paul Koning <ni1d@arrl.net>
2104 * config/pdp11/pdp11.md (doloop_end): New expander.
2105 (doloop_end_insn): renamed from "doloop_end".
2106 (addqi3): New pattern.
2107 (subqi3): New pattern.
2108 * config/pdp11/predicates.md (incdec_operand): New predicate.
2110 2018-10-12 Yury Gribov <tetra2005@gmail.com>
2113 * real.c (format_helper::can_represent_integral_type_p): New function
2114 * real.h (format_helper::can_represent_integral_type_p): Ditto.
2115 * match.pd: New pattern.
2117 2018-10-12 Alexandre Oliva <oliva@adacore.com>
2119 * configure.ac: Introduce --enable-large-address-aware
2120 to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
2121 * doc/install.texi: Document it.
2122 * configure, config.in: Rebuilt.
2123 * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
2124 based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
2125 (LINK_SPEC): Insert it.
2126 * config/i386/mingw-w64.h: Likewise.
2128 * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
2130 2018-10-12 Peter Bergner <bergner@linux.ibm.com>
2132 PR rtl-optimization/87600
2133 * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
2135 2018-10-12 Paul Koning <ni1d@arrl.net>
2137 * doc/md.texi (doloop_end): Document that the pattern code may
2138 need to check operand mode.
2140 2018-10-12 Wilco Dijkstra <wdijkstr@arm.com>
2142 * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
2143 to zero-extend between int and floating-point registers.
2144 (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
2145 ldp into floating-point registers. Add type and arch attributes.
2146 (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
2147 Use f_loads for type attribute.
2149 2018-10-11 Martin Sebor <msebor@redhat.com>
2151 * doc/extend.texi (attribute packed): Correct typos.
2153 2018-10-11 Martin Sebor <msebor@redhat.com>
2155 * doc/extend.texi (attribute flatten): Mention interaction with
2158 2018-10-11 Jan Hubicka <hubicka@ucw.cz>
2161 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
2162 Set new_decl virtual flag to zero.
2164 2018-10-11 Martin Sebor <msebor@redhat.com>
2167 * doc/extend.texi (attribute format_arg): Discuss using multiple
2168 attributes on a single function.
2170 2018-10-11 Giuliano Belinassi <giuliano.belinassi@usp.br>
2172 PR tree-optimization/86829
2173 * match.pd (sin (atan (x))): New simplification rules.
2174 (cos (atan (x))): Likewise.
2175 * real.c (build_sinatan_real): New function.
2176 * real.h (build_sinatan_real): Prototype.
2178 2018-10-11 Will Schmidt <will_schmidt@vnet.ibm.com>
2180 * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
2182 (fold_mergeeo_helper): New helper function.
2183 (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
2184 intrinsics. Correct some whitespace indentation issues.
2186 2018-10-11 Wilco Dijkstra <wdijkstr@arm.com>
2189 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
2190 Use HOST_WIDE_INT_1U for shift.
2192 2018-10-11 Doug Rupp <rupp@adacore.com>
2193 Olivier Hainque <hainque@adacore.com>
2195 * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
2196 Pass --relax to the linker for RTPs.
2197 (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
2199 2018-10-11 Andrew Stubbs <ams@codesourcery.com>
2200 Jan Hubicka <jh@suse.cz>
2201 Martin Jambor <mjambor@suse.cz>
2203 * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
2204 the same elements are repeated rather than printing all of them.
2205 * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
2206 "repeated" elements.
2207 * read-rtl-function.c (test_loading_repeat): New function.
2208 (read_rtl_function_c_tests): Call test_loading_repeat.
2209 * rtl-tests.c (test_dumping_repeat): New function.
2210 (rtl_tests_c_tests): Call test_dumping_repeat.
2212 2018-10-11 Richard Biener <rguenther@suse.de>
2214 * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
2215 bdver?_cost): Unify to ...
2216 (bdver_memcpy, bdver_memset, bdver_cost): ... this.
2217 * config/i386/i386.c (processor_cost_table): Adjust.
2219 2018-10-10 Eric Botcazou <ebotcazou@adacore.com>
2222 * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
2223 the thunk when expanding to GIMPLE.
2225 2018-10-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
2227 * varasm.c (mergeable_string_section): Don't try to move zero-length
2228 strings to the merge section.
2230 2018-10-10 Uros Bizjak <ubizjak@gmail.com>
2233 * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
2235 2018-10-10 Jakub Jelinek <jakub@redhat.com>
2238 * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
2239 to special_args set.
2241 2018-10-10 Richard Biener <rguenther@suse.de>
2243 * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
2244 reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
2245 reduc_plus_scal_v4sf): Merge into pattern reducing to half width
2246 and recursing and pattern terminating the recursion on SSE
2247 vector width using ix86_expand_reduc.
2248 (reduc_sminmax_scal_<mode>): Split into part reducing to half
2249 width and recursing and SSE2 vector variant doing the final
2250 reduction with ix86_expand_reduc.
2251 (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
2252 with terminating the recursion at AVX level, splitting that
2255 2018-10-09 David Malcolm <dmalcolm@redhat.com>
2257 * genmatch.c (error_cb): Rename to...
2258 (diagnostic_cb): ...this, converting int params to enums.
2259 (fatal_at): Update for renaming.
2260 (warning_at): Likewise.
2262 * input.c (selftest::ebcdic_execution_charset::apply):
2263 Update for renaming of...
2264 (selftest::ebcdic_execution_charset::on_error): ...this, renaming
2266 (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
2267 converting level and reason to enums.
2268 (class selftest::lexer_error_sink): Rename to...
2269 (class selftest::lexer_test_options): ...this, renaming field
2270 "m_errors" to "m_diagnostics".
2271 (selftest::lexer_test_options::apply): Update for renaming of...
2272 (selftest::lexer_test_options::on_error): ...this, renaming to...
2273 (selftest::lexer_test_options::on_diagnostic): ...this
2274 converting level and reason to enums.
2275 (selftest::test_lexer_string_locations_raw_string_unterminated):
2276 Update for renamings.
2277 * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
2280 2018-10-09 Paul A. Clarke <pc@us.ibm.com>
2282 * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
2283 * config/rs6000/pmmintrin.h: New file.
2285 2018-10-09 Eric Botcazou <ebotcazou@adacore.com>
2287 PR tree-optimization/86659
2288 * gimple-match.h (gimple_match_op constructors): Initialize reverse.
2290 2018-10-09 Richard Biener <rguenther@suse.de>
2292 PR tree-optimization/63155
2293 * tree-ssa-structalias.c: Include tree-ssa.h.
2294 (get_constraint_for_ssa_var): For undefs return nothing_id.
2295 (find_func_aliases): Cleanup PHI handling.
2297 2018-10-09 Richard Biener <rguenther@suse.de>
2299 * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
2302 2018-10-09 Martin Liska <mliska@suse.cz>
2304 * asan.c (asan_emit_stack_protection): If a stack variable
2305 is located in a same file as current function, then emit
2306 line info into variable definition string.
2308 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
2310 * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
2313 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
2315 * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
2318 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
2320 PR tree-optimization/86659
2321 * gimple-match.h (struct gimple_match_op): Add reverse field.
2322 (gimple_match_op::set_op): New overloaded method.
2323 * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
2324 the REF_REVERSE_STORAGE_ORDER flag on the value.
2325 (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
2326 REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
2328 2018-10-08 Richard Sandiford <richard.sandiford@arm.com>
2331 * gimple-ssa-backprop.c (backprop::intersect_uses): Use
2332 FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
2334 2018-10-08 H.J. Lu <hongjiu.lu@intel.com>
2337 * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
2338 Defined with __builtin_ia32_vfmaddsubpd512_mask.
2340 2018-10-08 Richard Biener <rguenther@suse.de>
2342 * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
2343 cost the same as AVX128 ones.
2345 2018-10-08 Paul Koning <ni1d@arrl.net>
2347 * config/pdp11/pdp11-protos.h (output_block_move): Remove.
2348 (expand_block_move): New function.
2349 * config/pdp11/pdp11.c (output_block_move): Remove.
2350 (expand_block_move): New function.
2351 * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
2352 * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
2353 (*movmemhi1): Remove.
2355 2018-10-08 Robin Dapp <rdapp@linux.ibm.com>
2357 * config/s390/2827.md: Increase latencies for some FP instructions.
2359 2018-10-08 Richard Biener <rguenther@suse.de>
2361 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
2363 * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
2364 * tree-vectorizer.h (dump_stmt_cost): Adjust.
2365 (add_stmt_cost): Dump return value of the hook.
2367 2018-10-08 Richard Biener <rguenther@suse.de>
2369 PR tree-optimization/63155
2370 * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
2371 (ssa_propagation_engine::ssa_propagate): Remove redundant
2372 bitmap bit clearing.
2374 2018-10-05 Peter Bergner <bergner@linux.ibm.com>
2376 PR rtl-optimization/86939
2377 PR rtl-optimization/87479
2378 * ira.h (non_conflicting_reg_copy_p): New prototype.
2379 * ira-lives.c (ignore_reg_for_conflicts): New static variable.
2380 (make_hard_regno_dead): Don't add conflicts for register
2381 ignore_reg_for_conflicts.
2382 (make_object_dead): Likewise.
2383 (non_conflicting_reg_copy_p): New function.
2384 (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
2385 Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
2386 * lra-lives.c (ignore_reg_for_conflicts): New static variable.
2387 (make_hard_regno_dead): Don't add conflicts for register
2388 ignore_reg_for_conflicts. Remove special conflict handling of
2389 REAL_PIC_OFFSET_TABLE_REGNUM. Remove now unused argument
2390 check_pic_pseudo_p and update callers.
2391 (mark_pseudo_dead): Don't add conflicts for register
2392 ignore_reg_for_conflicts.
2393 (process_bb_lives): Set ignore_reg_for_conflicts for copies.
2395 2018-10-05 Andrew Waterman <andrew@sifive.com>
2396 Jim Wilson <jimw@sifive.com>
2398 * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
2399 Add define_expand. Add ! HONOR_SNANS check to current pattern. Add
2400 new pattern using HONOR_SNANS that emits one extra instruction.
2402 2018-10-05 Segher Boessenkool <segher@kernel.crashing.org>
2404 * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
2405 patterns): Merge SI and DI patterns to a GPR pattern.
2406 (unnamed define_insn and define_split for record form of that): Merge
2407 to a single define_insn_and_split pattern.
2409 2018-10-05 David Malcolm <dmalcolm@redhat.com>
2412 * input.c (expand_location_to_spelling_point): Add param "aspect"
2413 and use rather than hardcoding LOCATION_ASPECT_CARET.
2414 (get_substring_ranges_for_loc): Handle the case of a single token
2415 within a macro expansion.
2416 * input.h (expand_location_to_spelling_point): Add "aspect" param,
2417 defaulting to LOCATION_ASPECT_CARET.
2419 2018-10-05 Paul Koning <ni1d@arrl.net>
2421 * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
2422 (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
2423 (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
2424 (pdp11_guard_type): New function.
2426 2018-10-05 Paul Koning <ni1d@arrl.net>
2428 * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
2429 * config/pdp11/pdp11.opt (mfloat32): Remove.
2431 * doc/invoke.texi (pdp11 -mfloat32): Remove:
2432 (pdp11 -mfloat64): Remove.
2434 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
2436 * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
2437 (*cmp<mode>_cc_i387): Ditto.
2438 (*cmpu<mode>_cc_i387): Ditto.
2439 (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
2440 * config/i386/i386.c (ix86_expand_fp_compare): Remove
2442 <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
2443 Emit x86_sahf_1 pattern.
2444 (ix86_expand_compare): Update call to ix86_expand_fp_compare.
2445 (ix86_expand_carry_flag_compare): Ditto.
2447 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
2449 * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
2450 to reg_or_0_operand. Add "C" constraint.
2451 (*cmpxf_cc_i387): Ditto.
2452 (*cmp<mode>_i387): Change operand 2 predicate
2453 to nonimm_or_0_operand. Add "C" constraint.
2454 (*cmp<mode>_cc_i387): Ditto.
2455 (*cmp<mode>_0_i387): Remove insn pattern.
2456 (*cmp<mode>_0_cc_i387): Ditto.
2458 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
2460 * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
2461 * config/i386/predicates.md (nonimm_or_0_operand): Rename
2462 from vector_move_operand. Update all uses.
2464 2018-10-05 Martin Sebor <msebor@redhat.com>
2466 PR tree-optimization/87490
2467 * builtins.c (expand_builtin_strnlen): Handle a null data.decl
2470 2018-10-05 Richard Biener <rguenther@suse.de>
2472 PR tree-optimization/63155
2473 * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
2474 vertical space in dumpfiles.
2475 * tree-ssa-propagate.h
2476 (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
2477 * tree-ssa-propagate.c (cfg_blocks_back): New global.
2478 (ssa_edge_worklist_back): Likewise.
2479 (curr_order): Likewise.
2480 (cfg_blocks_get): Remove abstraction.
2481 (cfg_blocks_add): Likewise.
2482 (cfg_blocks_empty_p): Likewise.
2483 (add_ssa_edge): Add to current or next worklist based on
2485 (add_control_edge): Likewise.
2486 (ssa_propagation_engine::process_ssa_edge_worklist): Fold
2488 (ssa_propagation_engine::ssa_propagate): ... here. Unify
2489 iteration from CFG and SSA edge worklist so we process
2490 everything in RPO order, prioritizing forward progress
2492 (ssa_prop_init): Allocate new worklists, do not dump
2494 (ssa_prop_fini): Free new worklists.
2496 2018-10-05 Richard Biener <rguenther@suse.de>
2498 * tree-core.h (tree_block::abstract_flag): Remove.
2499 (tree_block::block_num): Make full 32bits.
2500 * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
2501 * tree.h (BLOCK_ABSTRACT): Remove.
2502 * dwarf2out.c (gen_lexical_block_die): Remove dead code
2503 resulting from BLOCK_ABSTRACT being always false.
2504 (gen_inlined_subroutine_die): Likewise.
2505 (gen_block_die): Likewise.
2506 * tree.c (block_ultimate_origin): Likewise.
2507 * tree-pretty-print.c (dump_block_node): Remove code dealing
2508 with BLOCK_ABSTRACT.
2509 * tree-ssa-live.c (dump_scope_block): Likewise.
2510 * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
2511 * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
2513 2018-10-05 Richard Biener <rguenther@suse.de>
2515 * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
2516 is asked for initialize mode to the component mode of the
2519 2018-10-05 H.J. Lu <hongjiu.lu@intel.com>
2522 * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
2523 assembler for -mavx.
2524 * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
2526 2018-10-05 Segher Boessenkool <segher@kernel.crashing.org>
2529 * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
2531 * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
2532 to be DImode. When using mffscrn, force the operand to a register.
2534 2018-10-04 Uros Bizjak <ubizjak@gmail.com>
2536 * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
2537 from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
2538 X87MODEF mode iterator.
2539 (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
2540 *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
2541 X87MODEF mode iterator.
2543 2018-10-04 Vinay Kumar <vinay.kumar@blackfigtech.com>
2545 * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
2546 -Wno-prio-ctor-dtor.
2548 2018-10-04 David Malcolm <dmalcolm@redhat.com>
2550 * Makefile.in (OBJS): Add opt-problem.o.
2551 * dump-context.h: Include "selftest.h.
2552 (selftest::temp_dump_context): New forward decl.
2553 (class dump_context): Make friend of class
2554 selftest::temp_dump_context.
2555 (dump_context::dump_loc_immediate): New decl.
2556 (class dump_pretty_printer): Move here from dumpfile.c.
2557 (class temp_dump_context): Move to namespace selftest.
2558 (temp_dump_context::temp_dump_context): Add param
2559 "forcibly_enable_dumping".
2560 (selftest::verify_dumped_text):
2561 (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
2562 (selftest::verify_item):
2563 (ASSERT_IS_TEXT): Move here from dumpfile.c.
2564 (ASSERT_IS_TREE): Likewise.
2565 (ASSERT_IS_GIMPLE): Likewise.
2566 * dumpfile.c (dump_context::dump_loc): Move immediate dumping
2568 (dump_context::dump_loc_immediate): ...this new function.
2569 (class dump_pretty_printer): Move to dump-context.h.
2570 (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
2571 (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
2572 (temp_dump_context::temp_dump_context): Move to "selftest"
2573 namespace. Add param "forcibly_enable_dumping", and use it to
2574 conditionalize the use of m_pp;
2575 (selftest::verify_dumped_text): Make non-static.
2576 (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
2577 (selftest::verify_item): Make non-static.
2578 (ASSERT_IS_TEXT): Move to dump-context.h.
2579 (ASSERT_IS_TREE): Likewise.
2580 (ASSERT_IS_GIMPLE): Likewise.
2581 (selftest::test_capture_of_dump_calls): Pass "true" for new
2582 param of temp_dump_context.
2583 * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
2584 it to MSG_ALL_PRIORITIES. Update values of TDF_COMPARE_DEBUG and
2586 * opt-problem.cc: New file.
2587 * opt-problem.h: New file.
2588 * optinfo-emit-json.cc
2589 (selftest::test_building_json_from_dump_calls): Pass "true" for
2590 new param of temp_dump_context.
2591 * optinfo.cc (optinfo_kind_to_dump_flag): New function.
2592 (optinfo::emit_for_opt_problem): New function.
2593 (optinfo::emit): Clarity which emit_item is used.
2594 * optinfo.h (optinfo::get_dump_location): New accessor.
2595 (optinfo::emit_for_opt_problem): New decl.
2596 (optinfo::emit): Make const.
2597 * selftest-run-tests.c (selftest::run_tests): Call
2598 selftest::opt_problem_cc_tests.
2599 * selftest.h (selftest::opt_problem_cc_tests): New decl.
2600 * tree-data-ref.c (dr_analyze_innermost): Convert return type from
2601 bool to opt_result, converting fprintf messages to
2602 opt_result::failure_at calls. Add "stmt" param for use by the
2604 (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
2605 (runtime_alias_check_p): Convert return type from bool to
2606 opt_result, converting dump_printf calls to
2607 opt_result::failure_at, using the statement DDR_A for their
2609 (find_data_references_in_stmt): Convert return type from bool to
2610 opt_result, converting "return false" to opt_result::failure_at
2612 * tree-data-ref.h: Include "opt-problem.h".
2613 (dr_analyze_innermost): Convert return type from bool to opt_result,
2614 and add a const gimple * param.
2615 (find_data_references_in_stmt): Convert return type from bool to
2617 (runtime_alias_check_p): Likewise.
2618 * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
2619 dr_analyze_innermost.
2620 * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
2621 Convert return type from bool to opt_result, adding a message for
2622 the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
2623 (vect_analyze_data_ref_dependence): Convert return type from bool
2624 to opt_result. Change sense of return type from "false"
2625 effectively meaning "no problems" to "false" meaning a problem,
2626 so that "return false" becomes "return opt_result::success".
2627 Convert "return true" calls to opt_result::failure_at, using
2628 the location of statement A rather than vect_location.
2629 (vect_analyze_data_ref_dependences): Convert return type from bool
2631 (verify_data_ref_alignment): Likewise, converting dump_printf_loc
2632 calls to opt_result::failure_at, using the stmt location rather
2634 (vect_verify_datarefs_alignment): Convert return type from bool
2636 (vect_enhance_data_refs_alignment): Likewise. Split local "stat"
2637 into multiple more-tightly-scoped copies.
2638 (vect_analyze_data_refs_alignment): Convert return type from bool
2640 (vect_analyze_data_ref_accesses): Likewise, converting a
2641 "return false" to a "return opt_result::failure_at", adding a
2643 (vect_prune_runtime_alias_test_list): Convert return type from
2644 bool to opt_result, converting dump_printf_loc to
2645 opt_result::failure_at. Add a %G to show the pertinent statement,
2646 and use the stmt's location rather than vect_location.
2647 (vect_find_stmt_data_reference): Convert return type from
2648 bool to opt_result, converting dump_printf_loc to
2649 opt_result::failure_at, using stmt's location.
2650 (vect_analyze_data_refs): Convert return type from bool to
2651 opt_result. Convert "return false" to "return
2652 opt_result::failure_at", adding messages as needed.
2653 * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
2654 type from bool to opt_result.
2655 (vect_determine_vf_for_stmt): Likewise.
2656 (vect_determine_vectorization_factor): Likewise, converting
2657 dump_printf_loc to opt_result::failure_at, using location of phi
2658 rather than vect_location.
2659 (vect_analyze_loop_form_1): Convert return type from bool to
2660 opt_result, converting dump_printf_loc calls, retaining the use of
2662 (vect_analyze_loop_form): Convert return type from loop_vec_info
2663 to opt_loop_vec_info.
2664 (vect_analyze_loop_operations): Convert return type from bool to
2665 opt_result, converting dump_printf_loc calls, using the location
2666 of phi/stmt rather than vect_location where available. Convert
2667 various "return false" to "return opt_result::failure_at" with
2668 "unsupported phi" messages.
2669 (vect_get_datarefs_in_loop): Convert return type from bool to
2670 opt_result. Add a message for the
2671 PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
2672 (vect_analyze_loop_2): Convert return type from bool to
2673 opt_result. Ensure "ok" is set to a opt_result::failure_at before
2674 each "goto again;", adding new messages where needed.
2675 Add "unsupported grouped {store|load}" messages.
2676 (vect_analyze_loop): Convert return type from loop_vec_info to
2678 * tree-vect-slp.c (vect_analyze_slp): Convert return type from
2680 * tree-vect-stmts.c (process_use): Likewise, converting
2681 dump_printf_loc call and using stmt location, rather than
2683 (vect_mark_stmts_to_be_vectorized): Likeise.
2684 (vect_analyze_stmt): Likewise, adding a %G.
2685 (vect_get_vector_types_for_stmt): Convert return type from bool to
2686 opt_result, converting dump_printf_loc calls and using stmt
2687 location, rather than vect_location.
2688 (vect_get_mask_type_for_stmt): Convert return type from tree to
2689 opt_tree, converting dump_printf_loc calls and using stmt location.
2690 * tree-vectorizer.c: Include "opt-problem.h.
2691 (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
2692 MSG_PRIORITY_INTERNALS. Convert local "loop_vinfo" from
2693 loop_vec_info to opt_loop_vec_info. If if fails, and dumping is
2694 enabled, use it to report at the top level "couldn't vectorize
2695 loop" followed by the problem.
2696 * tree-vectorizer.h (opt_loop_vec_info): New typedef.
2697 (vect_mark_stmts_to_be_vectorized): Convert return type from bool
2699 (vect_analyze_stmt): Likewise.
2700 (vect_get_vector_types_for_stmt): Likewise.
2701 (tree vect_get_mask_type_for_stmt): Likewise.
2702 (vect_analyze_data_ref_dependences): Likewise.
2703 (vect_enhance_data_refs_alignment): Likewise.
2704 (vect_analyze_data_refs_alignment): Likewise.
2705 (vect_verify_datarefs_alignment): Likewise.
2706 (vect_analyze_data_ref_accesses): Likewise.
2707 (vect_prune_runtime_alias_test_list): Likewise.
2708 (vect_find_stmt_data_reference): Likewise.
2709 (vect_analyze_data_refs): Likewise.
2710 (vect_analyze_loop): Convert return type from loop_vec_info to
2712 (vect_analyze_loop_form): Likewise.
2713 (vect_analyze_slp): Convert return type from bool to opt_result.
2715 2018-10-04 David Malcolm <dmalcolm@redhat.com>
2717 * doc/invoke.texi (-fopt-info): Document new "internals"
2719 * dump-context.h (dump_context::apply_dump_filter_p): New decl.
2720 * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
2722 (optinfo_verbosity_options): Add "internals".
2723 (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
2724 (dump_context::apply_dump_filter_p): New member function.
2725 (dump_context::dump_loc): Use apply_dump_filter_p rather than
2726 explicitly masking the dump_kind.
2727 (dump_context::begin_scope): Increment the scope depth first. Use
2728 apply_dump_filter_p rather than explicitly masking the dump_kind.
2729 (dump_context::emit_item): Use apply_dump_filter_p rather than
2730 explicitly masking the dump_kind.
2731 (dump_dec): Likewise.
2732 (dump_hex): Likewise.
2733 (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
2734 (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
2735 (opt_info_switch_p): Update handling of default
2736 MSG_OPTIMIZED_LOCATIONS to cope with default of
2737 MSG_PRIORITY_USER_FACING.
2738 (dump_basic_block): Use apply_dump_filter_p rather than explicitly
2739 masking the dump_kind.
2740 (selftest::test_capture_of_dump_calls): Update test_dump_context
2741 instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
2742 than MSG_ALL. Generalize scope test to be run at all four
2743 combinations of with/without MSG_PRIORITY_USER_FACING and
2744 MSG_PRIORITY_INTERNALS, adding examples of explicit priority
2745 for each of the two values.
2746 * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
2747 Rename MSG_ALL to MSG_ALL_KINDS. Add MSG_PRIORITY_USER_FACING,
2748 MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
2749 values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
2750 (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
2751 with MSG_PRIORITY_*.
2752 * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
2753 dump messages as MSG_PRIORITY_USER_FACING.
2754 * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
2755 about the interaction with MSG_PRIORITY_*.
2757 2018-10-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
2759 * varasm.c (output_constant): Add new parameter merge_strings.
2760 Make strings properly zero terminated in merge string sections.
2761 (mergeable_string_section): Don't fail if the last char is non-zero.
2762 (assemble_variable_contents): Handle merge string sections.
2763 (assemble_variable): Likewise.
2764 (assemble_constant_contents): Likewise.
2765 (output_constant_def_contents): Likewise.
2766 (output_constructor_array_range,
2767 output_constructor_regular_field): Adjust call to output_constant.
2768 (output_object_block): Adjust call to assemble_constant_contents
2769 and assemble_variable_contents.
2771 2018-10-04 Martin Liska <mliska@suse.cz>
2774 * cgraphunit.c (process_function_and_variable_attributes):
2775 Warn about a function with alias attribute and a body.
2777 2018-10-04 Martin Liska <mliska@suse.cz>
2780 * multiple_target.c (redirect_to_specific_clone): New function.
2781 (ipa_target_clone): Use it.
2782 * tree-inline.c: Fix comment.
2784 2018-10-04 David Malcolm <dmalcolm@redhat.com>
2786 * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
2788 (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
2789 (gcc::dump_manager::register_pass): New member function, adapted
2790 from loop body in gcc::pass_manager::register_pass, adding a
2791 call to update_dfi_for_opt_info.
2792 (gcc::dump_manager::opt_info_enable_passes): Store the
2793 -fopt-info options into the new fields. Move the loop
2795 (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
2797 * dumpfile.h (struct opt_pass): New forward decl.
2798 (gcc::dump_manager::register_pass): New decl.
2799 (gcc::dump_manager::update_dfi_for_opt_info): New decl.
2800 (class gcc::dump_manager): Add fields "m_optgroup_flags",
2801 "m_optinfo_flags", and "m_optinfo_filename".
2802 * passes.c (gcc::pass_manager::register_pass): Move all of the
2803 dump-handling code to gcc::dump_manager::register_pass.
2805 2018-10-04 Peter Bergner <bergner@linux.ibm.com>
2807 PR rtl-optimization/87466
2808 * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
2809 * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
2810 * doc/tm.texi: Regenerate.
2811 * ira-lives.c (process_bb_node_lives): Use the new target hook.
2812 * lra-lives.c (process_bb_lives): Likewise.
2813 * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
2816 2018-10-04 Tamar Christina <tamar.christina@arm.com>
2818 * params.c (add_params): Fix initialization.
2820 2018-10-04 Martin Liska <mliska@suse.cz>
2822 PR gcov-profile/84107
2823 * tree-profile.c (init_ic_make_global_vars):
2824 Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
2825 Come up with new ic_tuple* variables. Emit
2826 __gcov_indirect_call{,_topn} variables.
2827 (gimple_gen_ic_profiler): Access the variable
2829 (gimple_gen_ic_func_profiler): Access
2830 __gcov_indirect_call.callee field.
2831 (gimple_init_gcov_profiler): Use ptr_type_node.
2832 * value-prof.c (gimple_ic): Use ptr_type_node.
2834 2018-10-04 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2836 PR tree-optimization/85787
2837 * ipa-pure-const.c (malloc_candidate_p_1): Move most of malloc_candidate_p
2838 into this function and add support for detecting multiple phis.
2839 (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
2841 2018-10-04 Martin Liska <mliska@suse.cz>
2844 * ipa-inline.c (inline_to_all_callers_1):
2845 Call ultimate_alias_target for node being inlined.
2847 2018-10-03 Jeff Law <law@redhat.com>
2849 * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
2851 * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
2852 * tree.h (get_typenode_from_name): Prototype.
2854 2018-10-03 Uros Bizjak <ubizjak@gmail.com>
2856 * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
2857 Change operand 2 predicate to nonimmediate_operand.
2858 (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
2860 2018-10-03 Martin Sebor <msebor@redhat.com>
2861 Jeff Law <law@redhat.com>
2863 * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
2865 (get_string_length): Detect unterminated arrays.
2866 (format_string): Same.
2867 (format_directive): Warn about unterminated arrays.
2868 (handle_gimple_call): Mark statements with no_warning as needed.
2870 2018-10-03 Jim Wilson <jimw@sifive.com>
2872 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
2873 also define __riscv_abi_rve. Delete trailing white space.
2875 2018-10-03 Paul Koning <ni1d@arrl.net>
2877 Enable LRA register allocator for PDP11.
2878 * config/pdp11/constraints.md (Q): Use define_memory_constraint.
2881 * config/pdp11/pdp11.c (pdp11_lra_p): New function.
2882 * config/pdp11/pdp11.opt (-mlra): New option.
2883 * doc/invoke.texi (PDP-11 Options): Document -mlra.
2885 2018-10-03 Uros Bizjak <ubizjak@gmail.com>
2887 * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
2888 (*<absneg:code>extend<mode>xf2): Ditto.
2890 2018-10-03 Aldy Hernandez <aldyh@redhat.com>
2892 PR tree-optimization/87415
2893 * tree-vrp.c (set_value_range_with_overflow): Special case one bit
2896 2018-10-02 Jeff Law <law@redhat.com>
2898 * gimple-fold.c (get_range_strlen): Only set *nonstr when
2899 an unterminated string is discovered. Bubble up range
2900 even for unterminated strings.
2901 (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
2902 indicates the string was not terminated via NONSTR.
2904 2018-10-03 Aldy Hernandez <aldyh@redhat.com>
2906 * tree-vrp.c (extract_range_from_unary_expr): Special case all
2907 pointer conversions.
2908 Do not do anything special for anti-ranges.
2910 2018-10-03 Jérôme Lambourg <lambourg@adacore.com>
2912 * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
2913 DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
2915 2018-10-03 Martin Liska <mliska@suse.cz>
2917 PR gcov-profile/86109
2918 * coverage.c (coverage_begin_function): Do not
2919 mark lambdas as artificial.
2920 * tree-core.h (struct GTY): Remove tm_clone_flag
2921 and introduce new lambda_function.
2922 * tree.h (DECL_LAMBDA_FUNCTION): New macro.
2924 2018-10-02 Aaron Sawdey <acsawdey@linux.ibm.com>
2927 * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
2928 P8_VECTOR and VSX are enabled.
2930 2018-10-02 Andreas Krebbel <krebbel@linux.ibm.com>
2932 * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
2933 0x3907 as CPU model number.
2935 2018-10-02 Andreas Krebbel <krebbel@linux.ibm.com>
2937 * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
2938 * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
2939 PF_Z14. Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
2940 TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
2941 TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
2942 * config/s390/s390.md: Likewise. Rename also the cpu attribute
2943 value from arch12 to z14.
2945 2018-10-02 Uros Bizjak <ubizjak@gmail.com>
2947 * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
2949 (isinf<mode>2): Ditto.
2951 2018-10-02 Uros Bizjak <ubizjak@gmail.com>
2953 * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
2954 before emitting fxam. Perform calculations in XFmode.
2956 2018-10-02 Marc Glisse <marc.glisse@inria.fr>
2958 * match.pd (((X /[ex] A) +- B) * A): New transformation.
2960 2018-10-02 Marc Glisse <marc.glisse@inria.fr>
2963 * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
2964 * tree.c (signed_or_unsigned_type_for): Handle complex.
2966 2018-10-02 Jeff Law <law@redhat.com>
2968 * gimple-fold.c (get_range_strlen): Remove dead code.
2970 2018-10-02 Martin Sebor <msebor@redhat.com>
2971 Jeff Law <law@redhat.com>
2973 * builtins.c (unterminated_array): Add new arguments.
2974 If argument is not terminated, bubble up size and exact
2976 (expand_builtin_strnlen): Detect, avoid expanding
2977 and diagnose unterminated arrays.
2978 (c_strlen): Fill in offset of start of unterminated strings.
2979 * builtins.h (unterminated_array): Update prototype.
2981 2018-10-02 Richard Biener <rguenther@suse.de>
2983 * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
2984 of haddv4df, first reduce to SSE width and exploit the fact
2985 that we only need element zero with the reduction result.
2986 (reduc_plus_scal_v2df): Likewise.
2988 2018-10-02 Eric Botcazou <ebotcazou@adacore.com>
2990 * dojump.h (do_jump): Delete.
2991 (do_jump_1): Likewise.
2992 (split_comparison): Move around.
2993 * dojump.c (do_jump): Make static.
2994 (do_jump_1): Likewise.
2995 (jumpifnot): Move around.
2996 (jumpifnot_1): Likewise.
2998 (jumpif_1): Likewise.
2999 * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
3001 2018-10-02 Eric Botcazou <ebotcazou@adacore.com>
3003 * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
3004 insns in the delay slot and add_insn_after for the jump insn.
3006 2018-10-02 Richard Biener <rguenther@suse.de>
3008 * tree-inline.c (expand_call_inline): Use the location of
3009 the callee declaration for the inline-entry marker.
3010 * final.c (notice_source_line): Remove special-casing of
3011 NOTE_INSN_INLINE_ENTRY.
3013 2018-10-01 Carl Love <cel@us.ibm.com>
3016 * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
3017 (__builtin_mtfsb0): New.
3018 (__builtin_mtfsb1): New.
3019 ( __builtin_set_fpscr_rn): New.
3020 (__builtin_set_fpscr_drn): New.
3021 * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
3022 (rs6000_expand_set_fpscr_rn_builtin): Add.
3023 (rs6000_expand_set_fpscr_drn_builtin): Add.
3024 (rs6000_expand_builtin): Add case statement entries for
3025 RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
3026 RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
3027 RS6000_BUILTIN_MFFSL.
3028 (rs6000_init_builtins): Add ftype initialization and def_builtin
3029 calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
3030 __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
3031 * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
3032 rs6000_mffscdrn): Add define_insn.
3033 (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
3034 * doc/extend.texi: Add documentation for the builtins.
3036 2018-10-01 Richard Biener <rguenther@suse.de>
3038 PR tree-optimization/87465
3039 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
3040 causing branch miscounts.
3042 2018-10-01 Tamar Christina <tamar.christina@arm.com>
3044 * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
3045 aarch64_option_default_param): New.
3046 (params.h): Include.
3047 (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
3048 * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
3049 stack-clash protection validation code.
3051 2018-10-01 Tamar Christina <tamar.christina@arm.com>
3053 * params.c (validate_param): New.
3054 (add_params): Use it.
3055 (set_param_value): Refactor param validation into validate_param.
3056 (diagnostic.h): Include.
3057 * diagnostic.h (diagnostic_ready_p): New.
3059 2018-10-01 Tamar Christina <tamar.christina@arm.com>
3061 * params.c (set_param_value):
3062 Add index of parameter being validated.
3063 * common/common-target.def (option_validate_param): New.
3064 * common/common-targhooks.h (default_option_validate_param): New.
3065 * common/common-targhooks.c (default_option_validate_param): New.
3066 * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
3067 * doc/tm.texi: Regenerate.
3069 2018-10-01 Tamar Christina <tamar.christina@arm.com>
3072 * config/aarch64/aarch64.c (aarch64_override_options_internal):
3073 Add validation for stack-clash parameters and set defaults.
3075 2018-10-01 Tamar Christina <tamar.christina@arm.com>
3078 * configure.ac: Add stack-clash-protection-guard-size.
3079 * doc/install.texi: Document it.
3080 * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
3081 * params.def: Update comment for guard-size.
3082 (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
3083 PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
3084 * configure: Regenerate.
3086 2018-10-01 Tamar Christina <tamar.christina@arm.com>
3089 * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
3090 STACK_DYNAMIC_OFFSET): New.
3091 * config/aarch64/aarch64.c (aarch64_layout_frame):
3092 Update outgoing args size.
3093 (aarch64_stack_clash_protection_alloca_probe_range,
3094 TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
3096 2018-10-01 Tamar Christina <tamar.christina@arm.com>
3099 * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
3101 * target.def (stack_clash_protection_alloca_probe_range): New.
3102 (stack_clash_protection_final_dynamic_probe): Remove.
3103 * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
3104 (default_stack_clash_protection_final_dynamic_probe): Remove.
3105 * targhooks.c: Likewise.
3106 * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
3107 (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
3108 * doc/tm.texi: Regenerate.
3110 2018-10-01 Tamar Christina <tamar.christina@arm.com>
3113 * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
3114 * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
3115 aarch64_clamp_to_uimm12_shift): New.
3116 (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
3117 * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
3119 2018-10-01 Tamar Christina <tamar.christina@arm.com>
3122 * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
3124 2018-10-01 Jeff Law <law@redhat.com>
3125 Richard Sandiford <richard.sandiford@linaro.org>
3126 Tamar Christina <tamar.christina@arm.com>
3129 * config/aarch64/aarch64.md
3130 (probe_stack_range): Add k (SP) constraint.
3131 * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
3132 STACK_CLASH_MAX_UNROLL_PAGES): New.
3133 * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
3134 stack probes for stack clash.
3135 (aarch64_allocate_and_probe_stack_space): New.
3136 (aarch64_expand_prologue): Use it.
3137 (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
3138 (aarch64_sub_sp): Add emit_move_imm optional param.
3140 2018-10-01 MCC CS <deswurstes@users.noreply.github.com>
3142 PR tree-optimization/87261
3143 * match.pd: Remove trailing whitespace.
3144 Add (x & y) | ~(x | y) -> ~(x ^ y),
3145 (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
3147 2018-10-01 Claudiu Zissulescu <claziss@synopsys.com>
3149 * config/arc/arc.md (*add_n): Clean up pattern, update instruction
3151 (ashlsi3_insn): Update instruction constraints.
3152 (ashrsi3_insn): Likewise.
3153 (rotrsi3): Likewise.
3154 (add_shift): Likewise.
3155 * config/arc/constraints.md (Csz): New 32 bit constraint. It
3156 avoids placing in the limm field small constants which, otherwise,
3157 could end into a small instruction.
3159 2018-10-01 Claudiu Zissulescu <claziss@synopsys.com>
3161 * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
3162 destination register is not odd-even.
3163 (umaddsidi4_split): Likewise.
3165 2018-10-01 Richard Biener <rguenther@suse.de>
3167 * tree-inline.c (expand_call_inline): Store origin of fn
3168 in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
3169 * tree.c (block_ultimate_origin): Simplify and do some
3172 2018-09-30 Uros Bizjak <ubizjak@gmail.com>
3174 * config/i386/mmx.md (EMMS): New int iterator.
3175 (emms): New int attribute.
3176 (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
3177 EMMS int iterator. Explicitly declare clobbers.
3178 (mmx_emms): Remove expander.
3180 * config/i386/predicates.md (emms_operation): Remove predicate.
3181 (vzeroall_pattern): New predicate.
3182 (vzeroupper_pattern): Rename from vzeroupper_operation.
3183 * config/i386/i386.c (ix86_avx_u128_mode_after): Use
3184 vzeroupper_pattern and vzeroall_pattern predicates.
3186 2018-09-30 Peter Bergner <bergner@linux.ibm.com>
3188 PR rtl-optimization/86939
3189 * ira-lives.c (make_hard_regno_born): Rename from this...
3190 (make_hard_regno_live): ... to this. Remove update to conflict
3191 information. Update function comment.
3192 (make_hard_regno_dead): Add conflict information update. Update
3194 (make_object_born): Rename from this...
3195 (make_object_live): ... to this. Remove update to conflict information.
3196 Update function comment.
3197 (make_object_dead): Add conflict information update. Update function
3199 (mark_pseudo_regno_live): Call make_object_live.
3200 (mark_pseudo_regno_subword_live): Likewise.
3201 (mark_hard_reg_dead): Update function comment.
3202 (mark_hard_reg_live): Call make_hard_regno_live.
3203 (process_bb_node_lives): Likewise.
3204 * lra-lives.c (make_hard_regno_born): Rename from this...
3205 (make_hard_regno_live): ... to this. Remove update to conflict
3206 information. Remove now uneeded check_pic_pseudo_p argument.
3207 Update function comment.
3208 (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
3209 to conflict information. Update function comment.
3210 (mark_pseudo_live): Remove update to conflict information. Update
3212 (mark_pseudo_dead): Add conflict information update.
3213 (mark_regno_live): Call make_hard_regno_live.
3214 (mark_regno_dead): Call make_hard_regno_dead with new arguement.
3215 (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
3217 2018-09-29 H.J. Lu <hongjiu.lu@intel.com>
3220 * config/i386/i386.c (construct_container): Use TImode for
3221 BLKmode values in 2 integer registers.
3223 2018-09-29 Jeff Law <law@redhat.com>
3225 * builtins.c (unterminated_array): Pass in c_strlen_data * to
3226 c_strlen rather than just a tree *.
3227 (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
3228 Update recursive calls appropriately. If caller did not provide a
3229 suitable data pointer, create a local one. When a non-terminated
3230 string is discovered, bubble up information about the string via the
3231 c_strlen_data object.
3232 * builtins.h (c_strlen): Update prototype.
3233 (c_strlen_data): New structure.
3234 * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
3235 For a type 2 call, if c_strlen indicates a non-terminated string
3236 use the length of the non-terminated string.
3237 (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
3239 2018-09-29 Jakub Jelinek <jakub@redhat.com>
3242 * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
3243 __m512d type for __A argument rather than __m512.
3245 2018-09-28 John David Anglin <danglin@gcc.gnu.org>
3247 * match.pd (simple_comparison): Don't optimize if either operand is
3248 a function pointer when target needs function pointer canonicalization.
3250 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
3252 * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
3253 power5 .. power9 to remove indirection.
3254 * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
3255 ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
3256 ASM_CPU_476_SPEC): Delete.
3257 (ASM_CPU_SPEC): Adjust.
3258 (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
3259 asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
3261 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
3263 * config.in: Delete HAVE_AS_DCI.
3264 * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
3265 * config/rs6000/rs6000.h: Ditto.
3266 * configure.ac: Delete HAVE_AS_DCI.
3267 * configure: Regenerate.
3269 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
3271 * config.in (HAVE_AS_LWSYNC): Delete.
3272 * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
3273 * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
3275 * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
3276 * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
3278 * configure.ac: Delete HAVE_AS_LWSYNC.
3279 * configure: Regenerate.
3281 2018-09-28 Eric Botcazou <ebotcazou@adacore.com>
3282 Pierre-Marie de Rodat <derodat@adacore.com>
3284 * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
3285 * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
3286 (cgraph_node::create_thunk): Add indirect_offset parameter.
3287 (thunk_adjust): Likewise.
3288 * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
3289 and initialize the corresponding field with it.
3290 (cgraph_node::dump): Dump indirect_offset field.
3291 * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
3292 * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
3293 (thunk_adjust): Add indirect_offset parameter and deal with it.
3294 (cgraph_node::expand_thunk): Deal with the indirect_offset field and
3295 pass it to thunk_adjust. Do not call the target hook if it's non-zero
3296 or if the thunk is external or local. Fix formatting. Do not chain
3297 the RESULT_DECL to BLOCK_VARS. Pass the static chain to the target,
3298 if any, in the GIMPLE representation.
3299 * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
3300 * lto-cgraph.c (lto_output_node): Write indirect_offset field.
3301 (input_node): Read indirect_offset field.
3302 * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
3303 call to thunk_adjust.
3304 * tree-nested.c (struct nesting_info): Add thunk_p field.
3305 (create_nesting_tree): Set it.
3306 (convert_all_function_calls): Copy static chain from targets to thunks.
3307 (finalize_nesting_tree_1): Return early for thunks.
3308 (unnest_nesting_tree_1): Do not finalize thunks.
3309 (gimplify_all_functions): Do not gimplify thunks.
3311 2018-09-28 David Malcolm <dmalcolm@redhat.com>
3313 * opt-suggestions.c (option_proposer::build_option_suggestions):
3314 Release "option_values".
3316 2018-09-28 David Malcolm <dmalcolm@redhat.com>
3318 * coverage.c (get_coverage_counts): Convert problem-reporting dump
3319 messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
3320 * dumpfile.c (kind_as_string): New function.
3321 (dump_loc): Rather than a hardcoded prefix of "note: ", use
3322 kind_as_string to vary the prefix based on dump_kind.
3323 (selftest::test_capture_of_dump_calls): Update for above.
3325 2018-09-28 Uros Bizjak <ubizjak@gmail.com>
3327 * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
3328 (GET_SSE_REGNO): Rename from SSE_REGNO. Update all uses for rename.
3330 2018-09-28 Uros Bizjak <ubizjak@gmail.com>
3332 * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
3333 * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
3334 INVALID_REGNUM instead of FPSR_REG.
3335 (ix86_md_asm_adjust): Do not clobber FPSR_REG.
3336 * config/i386/i386.md: Update comment of FP compares.
3337 (fldenv): Do not clobber FPSR_REG.
3339 2018-09-28 Richard Biener <rguenther@suse.de>
3341 * tree.h (BLOCK_ORIGIN): New.
3342 * omp-expand.c (grid_expand_target_grid_body): Assign
3343 BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
3344 * tree-inline.c (remap_block): Likewise.
3345 * auto-profile.c (get_function_decl_from_block): Simplify
3346 by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
3347 * langhooks.c (lhd_print_error_function): Likewise.
3348 * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
3350 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
3351 * tree.c (block_nonartificial_location): Likewise.
3352 (block_ultimate_origin): Likewise.
3353 * tree-pretty-print.c (percent_K_format): Likewise. Remove
3354 no longer needed LTO case.
3356 2018-09-28 Andrew Stubbs <ams@codesourcery.com>
3357 Jan Hubicka <jh@suse.cz>
3358 Martin Jambor <mjambor@suse.cz>
3360 * simplify-rtx.c (simplify_merge_mask): New function.
3361 (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
3362 same masks are used in op1 or op2.
3363 (test_vec_merge): New function.
3364 (test_vector_ops): Call test_vec_merge.
3366 2018-09-28 Eric Botcazou <ebotcazou@adacore.com>
3368 * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
3369 * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
3370 * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
3371 (cypress_costs): Set it.
3372 (supersparc_costs): Likewise.
3373 (hypersparc_costs): Likewise.
3374 (leon_cost): Likewise.
3375 (leon3_costs): Likewise.
3376 (sparclet_costs): Likewise.
3377 (ultrasparc_costs): Likewise.
3378 (ultrasparc_costs): Likewise.
3379 (niagara_costs): Likewise.
3380 (niagara2_costs): Likewise.
3381 (niagara3_costs): Likewise.
3382 (niagara4_costs): Likewise.
3383 (niagara7_costs): Likewise.
3384 (m8_costs): Likewise.
3385 (TARGET_CAN_FOLLOW_JUMP): Define.
3386 (pass_work_around_errata::gate): Minor tweak.
3387 (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
3388 Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
3389 Automaitcally clear MASK_FSMULD mask for V7 processors.
3390 (sparc_can_follow_jump): New static function.
3391 (output_ubranch): Deal with CROSSING_JUMP_P.
3392 (sparc_use_sched_lookahead): Rewrite using switch statement.
3393 (sparc_issue_rate): Reorder.
3394 (sparc_branch_cost): New function.
3396 2018-09-27 Martin Sebor <msebor@redhat.com>
3398 * tree.h (tree_to_shwi): Add attribute nonnull and pure.
3399 (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
3400 (int_fits_type_p): Same.
3402 2018-09-27 Uros Bizjak <ubizjak@gmail.com>
3404 * config/i386/i386.md (FPCR_REG): Remove.
3405 (UNSPEC_FLDCW): Remove.
3406 (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
3407 (x86_fldcw_1): Remove insn pattern.
3408 (fnstenv): Do not clobber FPCR_REG.
3410 * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
3411 (CALL_USED_REGISTERS): Ditto.
3412 (REG_ALLOC_ORDER): Ditto.
3413 (REG_CLASS_CONTENTS): Ditto.
3414 (HI_REGISTER_NAMES): Ditto.
3415 (ADDITIONAL_REGISTER_NAMES): Use defines instead
3416 of numerical constants.
3417 * config/i386/i386.c (regclass_map): Remove fpsr register.
3418 (dbx_register_map): Ditto.
3419 (dbx64_register_map): Ditto.
3420 (svr4_dbx_register_map): Ditto.
3421 (print_reg): Do not handle FPCR_REG.
3423 2018-09-27 Segher Boessenkool <segher@kernel.crashing.org>
3426 * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
3427 HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
3428 Delete, always treat as true.
3429 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
3430 Ditto. Simplify remaining code.
3431 * config/powerpcspe/powerpcspe.h: Ditto.
3432 * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
3433 Simplify remaining code.
3434 (rs6000_expand_builtin): Ditto.
3435 * config/rs6000/rs6000.h: Ditto.
3436 * configure.ac: Ditto.
3437 * configure: Regenerate.
3439 2018-09-27 Martin Liska <mliska@suse.cz>
3441 * coverage.c (get_coverage_counts): Revert the formatting
3442 of missing profile opt info.
3444 2018-09-27 Richard Biener <rguenther@suse.de>
3448 * dwarf2out.c (set_block_origin_self): Do not mark outermost
3449 block as we do not output that.
3450 (gen_inlined_subroutine_die): Elide the originally outermost
3451 block, matching what we do for concrete instances.
3452 (decls_for_scope): Add parameter specifying whether to recurse
3455 2018-09-27 Andrew Stubbs <ams@codesourcery.com>
3456 Tom de Vries <tom@codesourcery.com>
3460 * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
3461 STORE_FLAG_VALUE == 1.
3463 2018-09-27 Andreas Krebbel <krebbel@linux.ibm.com>
3465 * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
3466 constant definitions.
3467 ("tx_assist"): Replace magic number with PPA_TX_ABORT.
3468 ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
3469 ("speculation_barrier"): New expander definition.
3471 2018-09-26 Indu Bhagat <indu.bhagat@oracle.com>
3473 PR gcov-profile/86957
3474 * common.opt: New warning option -Wmissing-profile.
3475 * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
3476 * doc/invoke.texi: Document -Wmissing-profile.
3478 2018-09-26 Jim Wilson <jimw@sifive.com>
3480 * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
3481 (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
3483 2018-09-26 Martin Sebor <msebor@redhat.com>
3485 * tree.c (zerop): Change return type to bool.
3486 (integer_zerop, integer_onep, integer_each_onep): Same.
3487 (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
3488 (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
3489 (real_onep, real_minus_onep, chain_index): Same.
3490 (print_type_hash_statistics, type_list_equal): Same.
3491 * tree.h (zerop): Same.
3492 (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
3493 (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
3494 (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
3495 (real_onep, real_minus_onep, chain_index): Same.
3496 (print_type_hash_statistics, type_list_equal): Same.
3498 2018-09-26 Jim Wilson <jimw@sifive.com>
3500 * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
3502 2018-09-26 Jakub Jelinek <jakub@redhat.com>
3505 * config/i386/i386.c: Include debug.h and dwarf2out.h.
3506 (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
3509 2018-09-25 Andrew Stubbs <ams@codesourcery.com>
3511 * builtins.c (get_builtin_sync_mem): Force address mode conversion.
3513 2018-09-26 Uros Bizjak <ubizjak@gmail.com>
3515 * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
3516 and FP_SECOND_SSE_REGS.
3517 (REG_CLASS_NAMES): Ditto.
3518 (REG_CLASS_CONTENTS): Ditto.
3519 * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
3520 FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
3521 (ix86_preferred_output_reload_class): Ditto.
3522 * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
3523 clobber constraint to "=&f".
3524 (fix_truncdi_i387): Ditto.
3525 (lrintxfdi2): Ditto.
3526 (fistdi2_<rounding>): Ditto.
3527 (fpremxf4_i387): Change "=u" constraint to "=f".
3528 (fprem1xf4_i387): Ditto.
3530 (fptanxf4_i387): Ditto.
3531 (fxtractxf3_i387): Ditto.
3532 (fscalexf4_i387): Ditto.
3533 (atan2xf3): Change "u" constraint to "f".
3534 (fyl2xxf3_i387): Ditto.
3535 (fyl2xp1xf3_i387): Ditto.
3537 2018-09-26 Uros Bizjak <ubizjak@gmail.com>
3540 * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
3541 for removed I387_MASK_PM entity.
3544 2018-09-26 Jeff Law <law@redhat.com>
3546 2018-09-26 Alexey Neyman <stilor@att.net>
3548 * graphite.h: Include <isl/id.h> and <isl/space.h>; these
3549 headers are no longer pulled in by <isl/val.h>.
3551 2018-09-26 Richard Biener <rguenther@suse.de>
3554 * dwarf2out.c (gen_lexical_block_die): Do not equate inline
3555 or concrete instance DIE to the tree. Create abstract origin
3556 attributes also for concrete instances.
3558 2018-09-26 Alexey Neyman <stilor@att.net>
3560 * graphite.h: Include <isl/id.h> and <isl/space.h>; these
3561 headers are no longer pulled in by <isl/val.h>.
3563 2018-09-26 Matthew Malcomson <matthew.malcomson@arm.com>
3565 * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
3566 Use new helper functions.
3567 * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
3568 Use new helper functions.
3569 * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
3570 aarch_mm_needs_release): New declarations.
3571 * config/arm/aarch-common.c (aarch_mm_needs_acquire,
3572 aarch_mm_needs_release): New.
3574 2018-09-26 Eric Botcazou <ebotcazou@adacore.com>
3576 * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
3577 (arm32_output_mi_thunk): Deal with long calls.
3579 2018-09-26 Richard Biener <rguenther@suse.de>
3583 * tree-inline.c (expand_call_inline): When the location
3584 of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
3585 or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
3586 the inserted BLOCK to make inlined_function_outer_scope_p
3588 * dwarf2out.c (add_call_src_coords_attributes): Do not add
3589 coords for reserved locations.
3591 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
3593 * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
3594 (*call_indirect_nonlocal_sysv<mode>): Ditto.
3595 (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
3596 (*sibcall_nonlocal_sysv<mode>): Ditto.
3597 (*sibcall_value_nonlocal_sysv<mode>): Ditto.
3598 (<bd>_<mode>): Ditto.
3599 (<bd>tf_<mode>): Ditto.
3601 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
3603 * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
3604 control string as a list of templates instead of as C code.
3605 (*altivec_movti): Ditto.
3606 * config/rs6000/darwin.md (movdf_low_di): Ditto.
3608 2018-09-25 Jim Wilson <jimw@sifive.com>
3610 * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
3611 when target symbol is weak.
3613 2018-09-25 Bernd Edlinger <bernd.edlinger@hotmail.de>
3616 * builtins.c (unterminated_array): Simplify.
3617 * expr.c (string_constant): Handle SSA_NAME. Add more exceptions
3618 where pointer arithmetic is safe.
3620 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
3623 * config/rs6000/altivec.md (altivec_vspltb): Use
3624 const_0_to_15_operand instead of u5bit_cint_operand.
3625 (*altivec_vspltb_internal): Ditto.
3626 (altivec_vspltb_direct): Ditto.
3627 (altivec_vsplth): Use const_0_to_7_operand instead of
3629 (*altivec_vsplth_internal): Ditto.
3630 (altivec_vsplth_direct): Ditto.
3631 (altivec_vspltw): Use const_0_to_3_operand instead of
3633 (*altivec_vspltw_internal): Ditto.
3634 (altivec_vspltw_direct): Ditto.
3635 (altivec_vspltsf): Ditto.
3636 (*altivec_vspltsf_internal): Ditto.
3637 * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
3638 various splats with the proper size immediate. Reorder the various
3639 cases by ascending size of immediate, and put all such together.
3641 2018-09-25 Richard Biener <rguenther@suse.de>
3644 * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
3646 (maybe_create_die_with_external_ref): Do not create
3647 DW_TAG_imported_unit here.
3648 (add_abstract_origin_attribute): Handle external BLOCK refs.
3649 (dwarf2out_abstract_function): Simplify LTO case.
3650 (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
3651 rather than using maybe_create_die_with_external_ref.
3653 2018-09-25 Uros Bizjak <ubizjak@gmail.com>
3656 * config/i386/i386.md (frndintxf2_mask_pm): Remove.
3657 (frndintxf2_mask_pm_i387): Ditto.
3658 (nearbyintxf2): Rewrite expander pattern to match rintxf2.
3659 Enable for !flag_trapping_math.
3660 (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
3661 Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
3662 Change operand 1 predicate to nonimmediate_operand.
3663 (attr "i387_cw"): Remove mask_pm.
3664 * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
3665 (enum ix86_entity): Remove I387_MASK_PM.
3666 * config/i386/i386.c (ix86_i387_mode_needed): Do not
3667 handle I387_MASK_PM.
3668 (ix86_mode_needed): Ditto.
3669 (ix86_mode_after): Ditto.
3670 (ix86_mode_entry): Ditto.
3671 (ix86_mode_exit): Ditto.
3672 (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
3674 2018-09-25 Jakub Jelinek <jakub@redhat.com>
3676 * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
3677 to_update_switch_stmts to vNULL instead of calling create on them
3680 2018-09-25 Richard Biener <rguenther@suse.de>
3682 PR tree-optimization/87402
3683 * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
3684 (visit_phi): Re-instantiate handling of supposed to be VARYING
3685 but non-VARYING backedge value.
3687 2018-09-25 Richard Biener <rguenther@suse.de>
3690 * dwarf2out.c (struct sym_off_pair): New.
3691 (external_die_map): New global.
3692 (lookup_decl_die): When in LTO create DIEs lazily from the
3694 (lookup_block_die): New function, create DIEs lazily in LTO.
3695 (equate_block_to_die): New function.
3696 (dwarf2out_die_ref_for_decl): During WPA get the association
3697 from the external DIE map.
3698 (dwarf2out_register_external_die): Record mapping into the
3700 (maybe_create_die_with_external_ref): New function split out from
3701 DIE generation part of old dwarf2out_register_external_die.
3702 (add_abstract_origin_attribute): Do not return the DIE. When
3703 in LTO reference externals directly.
3704 (dwarf2out_abstract_function): When in LTO ignore calls for
3705 decls with external DIEs (already present abstract instances).
3706 (gen_call_site_die): Adjust.
3707 (add_high_low_attributes): Likewise.
3708 (gen_lexical_block_die): Likewise.
3709 (gen_inlined_subroutine_die): Likewie.
3710 (gen_block_die): Likewise.
3711 (dwarf2out_inline_entry): Likewise.
3712 (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
3715 2018-09-25 Martin Liska <mliska@suse.cz>
3717 * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
3718 integers and not by a float value.
3720 2018-09-25 Martin Liska <mliska@suse.cz>
3723 * dbgcnt.c (dbg_cnt_process_single_pair): Return false
3725 * dumpfile.c (dump_enable_all): Remove extra parenthesis.
3726 * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
3727 * godump.c (go_format_type): Remove extra parenthesis.
3729 2018-09-25 Martin Liska <mliska@suse.cz>
3731 * alias.c (set_dest_equal_p): Remove unused function.
3732 * config/i386/i386.c (def_builtin_pure2): Likewise.
3733 * diagnostic-show-locus.c (class layout): Remove
3735 (layout::layout): Likewise here.
3736 * dump-context.h (class temp_dump_context): Likewise.
3737 * dwarf2out.c (add_AT_fde_ref): Remove unused function.
3738 (add_AT_loclistsptr): Likewise.
3739 (add_AT_offset): Likewise.
3740 (get_AT_hi_pc): Likewise.
3741 (is_comdat_die): Likewise.
3742 (type_is_enum): Likewise.
3743 (ceiling): Likewise.
3744 (add_AT_vms_delta): Likewise.
3745 (is_class_die): Likewise.
3746 * edit-context.c (class line_event): Remove unused field.
3747 * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
3749 * ipa-cp.c (ipa_get_vr_lat): Likewise.
3750 * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
3751 (ok_for_base_p_nonstrict): Likewise.
3752 * tree-chrec.c (is_not_constant_evolution): Likewise.
3753 (chrec_fold_poly_cst): Likewise.
3754 * tree-if-conv.c (has_pred_critical_p): Likewise.
3755 * tree-ssa-coalesce.c (print_exprs): Likewise.
3756 * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
3757 * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
3758 * tree-vrp.c (value_ranges_intersect_p): Likewise.
3759 (value_range_nonnegative_p): Likewise.
3761 2018-09-25 Martin Liska <mliska@suse.cz>
3763 * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
3764 Do not handle "GNU Pascal".
3765 * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
3767 * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
3768 from documentation. Likewise.
3769 * dbxout.c (dbxout_range_type): Likewise.
3770 * doc/cpp.texi: Likewise.
3771 * doc/extend.texi: Likewise.
3772 * doc/frontends.texi: Likewise.
3773 * doc/invoke.texi: Remove Pascal entry.
3774 * tree.def (CLEANUP_POINT_EXPR): Likewise.
3775 * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
3777 2018-09-25 Martin Liska <mliska@suse.cz>
3780 * doc/invoke.texi: Document all parameters and remove default
3783 2018-09-25 Ilya Leoshkevich <iii@linux.ibm.com>
3786 * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
3787 contains HOST_WIDE_INTs when computing its size.
3789 2018-09-24 Jim Wilson <jimw@sifive.com>
3792 * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
3794 (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
3796 2018-09-24 Andrew Pinski <apinski@marvell.com>
3798 * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
3799 access prev before checking it for NULLness in the
3800 AARCH64_FUSE_CMP_BRANCH case.
3802 2018-09-24 H.J. Lu <hongjiu.lu@intel.com>
3805 * config/i386/i386.c (rest_of_insert_endbranch): Set
3806 endbr_queued_at_entrance to true and don't insert ENDBR if
3807 x86_function_profiler will be called.
3808 (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
3810 * config/i386/i386.h (machine_function): Add
3811 endbr_queued_at_entrance.
3813 2018-09-24 Ilya Leoshkevich <iii@linux.ibm.com>
3815 * genattrtab.c (mk_attr_alt): Use alternative_mask.
3816 (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
3818 (check_attr_test): Use alternative_mask.
3819 (get_attr_value): Likewise.
3820 (compute_alternative_mask): Use alternative_mask and XWINT.
3821 (make_alternative_compare): Use alternative_mask.
3822 (attr_alt_subset_p): Use XWINT.
3823 (attr_alt_subset_of_compl_p): Likewise.
3824 (attr_alt_intersection): Use alternative_mask and XWINT.
3825 (attr_alt_union): Likewise.
3826 (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
3827 (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
3828 (simplify_test_exp): Use alternative_mask and XWINT.
3829 (write_test_expr): Use alternative_mask and XWINT, adjust bit
3830 number calculation to support 64 bits. Generate code that
3831 checks 64-bit masks.
3832 (main): Use alternative_mask.
3833 * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
3835 2018-09-24 Ilya Leoshkevich <iii@linux.ibm.com>
3838 * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
3839 RETURN+USE when returning via %r14.
3841 2018-09-24 Martin Liska <mliska@suse.cz>
3843 * gcov.c (output_lines): Print colorization legend
3844 for both flag_use_colors and flag_use_hotness_colors.
3847 2018-09-24 Martin Liska <mliska@suse.cz>
3849 * coverage.c (get_coverage_counts): Use warning_at
3850 with current_function_decl location. Use %qD in warning
3853 2018-09-24 Martin Liska <mliska@suse.cz>
3855 * memory-block.h (memory_block_pool::release): Annotate with
3856 valgrind that the memory is not accessible.
3858 2018-09-24 Martin Liska <mliska@suse.cz>
3861 * asan.c: Make asan_handled_variables extern.
3863 * cfgexpand.c (expand_stack_vars): Make sure
3864 a representative is unpoison if another
3865 variable in the partition is handled by
3866 use-after-scope sanitization.
3868 2018-09-24 Richard Biener <rguenther@suse.de>
3870 PR tree-optimization/63155
3871 * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
3872 the worklist when the edge of the respective argument isn't
3875 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
3877 * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
3878 ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
3879 (MASK_CLASS_P): Update for rename.
3880 (MAYBE_MASK_CLASS_P): Ditto.
3881 (REG_CLASS_NAMES): Update.
3882 (REG_CLASS_CONTENT): Update.
3883 * config/i386/i386.c (regclass_map): Update for MASK_REG
3884 and ALL_MASK_REGS rename.
3885 * config/i386/constraints.md (Yk): Update for rename.
3888 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
3890 * config/i386/i386.h (enum reg_class): Remove
3891 EVEX_SSE_REGS and MOD4_SSE_REGS.
3892 (REG_CLASS_NAMES): Update.
3893 (REG_CLASS_CONTENT): Update.
3894 * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
3895 registers as ALL_SSE_REGS.
3896 (ix86_additional_allocno_class_p): Remove.
3897 (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
3898 (ix86_register_priority): Lower priority of EVEX SSE registers.
3899 Use IN_RANGE macro where appropriate.
3900 (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
3901 AVX-5124VNNIW checks.
3902 (ix86_modes_tieable_p): Tie 512-bit SSE modes.
3903 * config/i386/sse.md (avx5124fmaddps_4fmaddps)
3904 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
3905 (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
3906 (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
3907 (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
3908 (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
3909 (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
3910 (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
3911 (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
3912 (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
3913 * config/i386/constraints.md (Yh): Remove.
3915 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
3917 * config/i386/i386.c (regclass_map): Declare integer REX registers
3920 2018-09-23 Gerald Pfeifer <gerald@pfeifer.com>
3922 * doc/service.texi (Service): Switch the fsf.org link to https.
3924 2018-09-22 Chung-Ju Wu <jasonwucj@gmail.com>
3927 * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3928 Define to speculation_safe_value_not_needed.
3930 2018-09-21 Florian Weimer <fweimer@redhat.com>
3933 * doc/extend.texi (Common Function Attributes): Mention that
3934 noreturn suppresses tail call optimization.
3936 2018-09-21 Jeff Law <law@redhat.com>
3938 * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
3939 vr_values::cleanup_edges_and_switches.
3940 * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
3942 (identify_jump_threads): Remove EDGE_IGNORE handling.
3943 (execute_vrp): Move handling of to_remove_edges and
3944 to_update_switch_stmts into vr_values class member functions.
3945 * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
3946 (to_update_switch_stmts): Likewise.
3947 * vr-values.c: Include cfghooks.h.
3948 (vr_values::vr_values): Initialize to_remove_edges and
3949 to_update_switch_stmts.
3950 (vr_values::~vr_values): Verify to_remove_edges and
3951 to_update_switch_stmts are empty.
3952 (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
3953 (vr_values::cleanup_edges_and_switches): New member function.
3954 * vr-values.h (vr_values): Add cleanup_edges_and_switches member
3955 function. Add new data members.
3957 2018-09-21 David Malcolm <dmalcolm@redhat.com>
3959 PR tree-optimization/87309
3960 * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
3961 calls with pflags and alt_flags.
3962 (selftest::test_capture_of_dump_calls): Add test of interaction of
3963 MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
3965 2018-09-21 Olivier Hainque <hainque@adacore.com>
3967 * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
3969 2018-09-21 Olivier Hainque <hainque@adacore.com>
3971 * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
3973 2018-09-21 Olivier Hainque <hainque@adacore.com>
3975 * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
3976 Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
3977 (VXWORKS_LIBS_RTP): Minor reordering.
3979 2018-09-21 Olivier Hainque <hainque@adacore.com>
3981 * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
3982 (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
3984 2018-09-21 Olivier Hainque <hainque@adacore.com>
3986 * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
3987 (PTRDIFF_TYPE): Likewise.
3989 2018-09-21 Olivier Hainque <hainque@adacore.com>
3991 * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
3992 triplet, similar to support for VxWorks7.
3993 * config/vxworks-dummy.h: Provide a default definition
3994 of TARGET_VXWORKS64 to 0.
3996 2018-09-21 Olivier Hainque <hainque@adacore.com>
3998 * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
3999 * config/vxworks-dummy.h: here.
4001 2018-09-21 Olivier Hainque <hainque@adacore.com>
4003 * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
4005 2018-09-21 Shaokun Zhang <zhangshaokun@hisilicon.com>
4006 Bo Zhou <zbo.zhou@hisilicon.com>
4008 * config/aarch64/aarch64-cores.def (tsv110): New CPU.
4009 * config/aarch64/aarch64-tune.md: Regenerated.
4010 * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
4011 * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
4012 * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
4014 2018-09-21 Andrew Stubbs <ams@codesourcery.com>
4015 Julian Brown <julian@codesourcery.com>
4017 * builtins.c (get_builtin_sync_mem): Handle address spaces.
4019 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
4021 * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
4022 if the call takes a static chain.
4024 2018-09-21 Martin Liska <mliska@suse.cz>
4026 * auto-profile.c (autofdo_source_profile::read): Do not
4028 (read_profile): Do not add working sets.
4029 (read_autofdo_file): Remove sum_all.
4030 (afdo_callsite_hot_enough_for_early_inline): Remove const
4032 * coverage.c (struct counts_entry): Remove gcov_summary.
4033 (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
4034 do not support GCOV_TAG_PROGRAM_SUMMARY.
4035 (get_coverage_counts): Remove summary and expected
4037 * coverage.h (get_coverage_counts): Likewise.
4038 * doc/gcov-dump.texi: Remove -w option.
4039 * gcov-dump.c (dump_working_sets): Remove.
4040 (main): Do not support '-w' option.
4041 (print_usage): Likewise.
4042 (tag_summary): Likewise.
4043 * gcov-io.c (gcov_write_summary): Do not dump
4045 (gcov_read_summary): Likewise.
4046 (gcov_histo_index): Remove.
4047 (gcov_histogram_merge): Likewise.
4048 (compute_working_sets): Likewise.
4049 * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
4051 (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
4052 (GCOV_TAG_SUMMARY_LENGTH): Adjust.
4053 (GCOV_HISTOGRAM_SIZE): Remove.
4054 (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
4055 (struct gcov_summary): Simplify rapidly just
4056 to runs and sum_max fields.
4057 (gcov_histo_index): Remove.
4058 (NUM_GCOV_WORKING_SETS): Likewise.
4059 (compute_working_sets): Likewise.
4060 * gcov-tool.c (print_overlap_usage_message): Remove
4061 trailing empty line.
4062 * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
4063 (output_lines): Remove program related line.
4064 * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
4065 * lto-cgraph.c (output_profile_summary): Do not stream GCOV
4067 (input_profile_summary): Do not read it.
4068 (merge_profile_summaries): And do not merge it.
4069 (input_symtab): Do not call removed function.
4070 * modulo-sched.c (sms_schedule): Do not print sum_max.
4071 * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
4072 removed when histogram method was invented.
4073 (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
4075 * postreload-gcse.c (eliminate_partially_redundant_load): Fix
4077 * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
4078 and dump selected value.
4079 * profile.c (add_working_set): Remove.
4080 (get_working_sets): Likewise.
4081 (find_working_set): Likewise.
4082 (get_exec_counts): Do not work with working sets.
4083 (read_profile_edge_counts): Do not inform as sum_max is removed.
4084 (compute_branch_probabilities): Likewise.
4085 (compute_value_histograms): Remove argument for call of
4086 get_coverage_counts.
4087 * profile.h: Do not make gcov_summary const.
4089 2018-09-21 Monk Chiang <sh.chiang04@gmail.com>
4091 * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
4093 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
4095 PR tree-optimization/86990
4096 * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
4097 Check that the entire merged store group is made of constants only for
4100 2018-09-20 Allan Sandfeld Jensen <allan.jensen@qt.io>
4102 * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
4103 (VTABLE_VERIFICATION_SPEC): Likewise.
4104 (SANITIZER_EARLY_SPEC): Likewise.
4105 (SANITIZER_SPEC): Likewise.
4106 * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
4107 * doc/invoke.texi (Link Options): Document -r.
4109 2018-09-20 Richard Biener <rguenther@suse.de>
4112 * gimplify.c (gimplify_expr): Retain alignment of
4113 addressable lvalue in dereference.
4115 2018-09-20 Alexandre Oliva <aoliva@redhat.com>
4118 * configure.ac: Check for .loc is_stmt support.
4119 * configure, config.in: Rebuilt.
4120 * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
4123 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
4125 * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
4126 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
4129 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
4131 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
4133 * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
4136 2018-09-20 Marek Polacek <polacek@redhat.com>
4138 * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
4140 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
4142 PR tree-optimization/87288
4143 * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
4144 into account when determining PEELING_FOR_NITERS.
4146 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
4148 PR tree-optimization/86877
4149 * tree-vect-loop.c (vect_analyze_loop_2): Call
4150 vect_verify_datarefs_alignment.
4152 2018-09-19 Marek Polacek <polacek@redhat.com>
4154 * doc/invoke.texi: Document -Wclass-conversion.
4156 2018-09-19 John David Anglin <danglin@gcc.gnu.org>
4158 * config/pa/pa.c (pa_adjust_priority): Delete.
4159 (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
4161 * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
4162 (atomic_storehi): Likewise.
4163 (atomic_storesi): Likewise.
4164 (atomic_loaddi): Restore compare and swap exchange loop code.
4166 2018-09-19 Segher Boessenkool <segher@kernel.crashing.org>
4168 PR rtl-optimization/86902
4169 * combine.c (try_combine): When changing the CC mode used, don't change
4170 an unrelated mode in other_insn to that new CC mode.
4172 2018-09-19 David Malcolm <dmalcolm@redhat.com>
4174 * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
4175 with %T in place of calls to dump_generic_expr.
4176 (prune_runtime_alias_test_list): Likewise.
4177 (create_runtime_alias_checks): Likewise.
4178 * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
4179 (vect_analyze_data_ref_dependence): Likewise.
4180 (vect_slp_analyze_data_ref_dependence): Likewise.
4181 (vect_record_base_alignment): Likewise. Use %G in place of call
4182 to dump_gimple_stmt.
4183 (vect_compute_data_ref_alignment): Likewise.
4184 (verify_data_ref_alignment): Likewise.
4185 (vect_find_same_alignment_drs): Likewise.
4186 (vect_analyze_group_access_1): Likewise.
4187 (vect_analyze_data_ref_accesses): Likewise.
4188 (dependence_distance_ge_vf): Likewise.
4189 (dump_lower_bound): Likewise.
4190 (vect_prune_runtime_alias_test_list): Likewise.
4191 (vect_find_stmt_data_reference): Likewise.
4192 (vect_analyze_data_refs): Likewise.
4193 (vect_create_addr_base_for_vector_ref): Likewise.
4194 (vect_create_data_ref_ptr): Likewise.
4195 * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
4196 (vect_can_advance_ivs_p): Likewise.
4197 (vect_update_ivs_after_vectorizer): Likewise.
4198 (vect_gen_prolog_loop_niters): Likewise.
4199 (vect_prepare_for_masked_peels): Likewise.
4200 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
4201 (vect_determine_vectorization_factor): Likewise.
4202 (vect_is_simple_iv_evolution): Likewise.
4203 (vect_analyze_scalar_cycles_1): Likewise.
4204 (vect_analyze_loop_operations): Likewise.
4205 (report_vect_op): Likewise.
4206 (vect_is_slp_reduction): Likewise.
4207 (check_reduction_path): Likewise.
4208 (vect_is_simple_reduction): Likewise.
4209 (vect_create_epilog_for_reduction): Likewise.
4210 (vect_finalize_reduction:): Likewise.
4211 (vectorizable_induction): Likewise.
4212 (vect_transform_loop_stmt): Likewise.
4213 (vect_transform_loop): Likewise.
4214 (optimize_mask_stores): Likewise.
4215 * tree-vect-patterns.c (vect_pattern_detected): Likewise.
4216 (vect_split_statement): Likewise.
4217 (vect_recog_over_widening_pattern): Likewise.
4218 (vect_recog_average_pattern): Likewise.
4219 (vect_determine_min_output_precision_1): Likewise.
4220 (vect_determine_precisions_from_range): Likewise.
4221 (vect_determine_precisions_from_users): Likewise.
4222 (vect_mark_pattern_stmts): Likewise.
4223 (vect_pattern_recog_1): Likewise.
4224 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
4225 (vect_record_max_nunits): Likewise.
4226 (vect_build_slp_tree_1): Likewise.
4227 (vect_build_slp_tree_2): Likewise.
4228 (vect_print_slp_tree): Likewise.
4229 (vect_analyze_slp_instance): Likewise.
4230 (vect_detect_hybrid_slp_stmts): Likewise.
4231 (vect_detect_hybrid_slp_1): Likewise.
4232 (vect_slp_analyze_operations): Likewise.
4233 (vect_slp_analyze_bb_1): Likewise.
4234 (vect_transform_slp_perm_load): Likewise.
4235 (vect_schedule_slp_instance): Likewise.
4236 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
4237 (vect_mark_stmts_to_be_vectorized): Likewise.
4238 (vect_init_vector_1): Likewise.
4239 (vect_get_vec_def_for_operand): Likewise.
4240 (vect_finish_stmt_generation_1): Likewise.
4241 (vect_check_load_store_mask): Likewise.
4242 (vectorizable_call): Likewise.
4243 (vectorizable_conversion): Likewise.
4244 (vectorizable_operation): Likewise.
4245 (vectorizable_load): Likewise.
4246 (vect_analyze_stmt): Likewise.
4247 (vect_is_simple_use): Likewise.
4248 (vect_get_vector_types_for_stmt): Likewise.
4249 (vect_get_mask_type_for_stmt): Likewise.
4250 * tree-vectorizer.c (increase_alignment): Likewise.
4252 2018-09-19 Andrew Stubbs <ams@codesourcery.com>
4254 * doc/rtl.texi: Adjust vec_select description.
4255 * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
4256 non-constant selectors.
4258 2018-09-19 Matthew Malcomson <matthew.malcomson@arm.com>
4260 * config/aarch64/aarch64-protos.h
4261 (aarch64_offset_9bit_signed_unscaled_p): New declaration.
4262 * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
4263 (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
4264 * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
4265 (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
4266 (AARCH64_FL_PROFILE): Move index so flags are ordered.
4267 (AARCH64_ISA_RCPC8_4): New flag.
4268 * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
4269 to aarch64_offset_9bit_signed_unscaled_p.
4270 * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
4272 * config/aarch64/constraints.md (Ust): New constraint.
4273 * config/aarch64/predicates.md.
4274 (aarch64_9bit_offset_memory_operand): New predicate.
4275 (aarch64_rcpc_memory_operand): New predicate.
4277 2018-09-19 Eric Botcazou <ebotcazou@adacore.com>
4279 PR rtl-optimization/87361
4280 * rtlanal.c (nonzero_bits1): Revert accidental change.
4282 2018-09-19 Richard Biener <rguenther@suse.de>
4284 PR tree-optimization/87349
4285 PR tree-optimization/87342
4286 * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
4288 2018-09-18 Marek Polacek <polacek@redhat.com>
4290 P1064R0 - Allowing Virtual Function Calls in Constant Expressions
4291 * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
4293 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
4295 * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
4297 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
4299 PR rtl-optimization/86882
4300 * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
4302 2018-09-18 Uros Bizjak <ubizjak@gmail.com>
4304 * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
4305 *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
4307 2018-09-18 Jonathan Wakely <jwakely@redhat.com>
4310 * doc/invoke.texi (Link Options): Fix formatting and grammar.
4312 2018-09-18 Richard Biener <rguenther@suse.de>
4315 * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
4316 (compute_samebase_partition_bases): Likewise.
4317 (coalesce_ssa_name): Always use compute_optimized_partition_bases.
4318 (gimple_can_coalesce_p): Simplify.
4320 2018-09-18 Hans-Peter Nilsson <hp@bitrange.com>
4322 Handle a library implementation of ffs calling __builtin_ffs.
4323 * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
4324 (mmix_init_libfuncs): New function: make __builtin_ffs expand
4327 2018-09-17 David Malcolm <dmalcolm@redhat.com>
4329 * diagnostic-show-locus.c (class layout_range): Add field
4331 (layout_range::layout_range): Add "original_idx" param and use it
4332 to initialize new field.
4333 (make_range): Use 0 for original_idx.
4334 (layout::layout): Pass in index to calls to
4335 maybe_add_location_range.
4336 (layout::maybe_add_location_range): Add param "original_idx" and
4337 pass it on to layout_range.
4338 (layout::print_any_labels): Pass on range->m_original_idx to
4340 (gcc_rich_location::add_location_if_nearby): Use 0 for
4342 * gcc-rich-location.h (text_range_label::get_text): Update for new
4344 (range_label_for_type_mismatch::get_text): Likewise.
4346 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
4348 * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
4350 2018-09-17 David Malcolm <dmalcolm@redhat.com>
4352 * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
4353 format_string_diagnostic_t.
4354 (fmtwarn_n): Likewise.
4355 * substring-locations.c
4356 (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
4357 (format_warning_n_va): Convert to...
4358 (format_string_diagnostic_t::emit_warning_n_va): ...this.
4359 (format_warning_va): Convert to...
4360 (format_string_diagnostic_t::emit_warning_va): ...this.
4361 (format_warning_at_substring): Convert to...
4362 (format_string_diagnostic_t::emit_warning): ...this.
4363 (format_warning_at_substring_n): Convert to...
4364 (format_string_diagnostic_t::emit_warning_n): ...this.
4365 * substring-locations.h (class format_string_diagnostic_t): New
4367 (format_warning_va): Convert to
4368 format_string_diagnostic_t::emit_warning_va.
4369 (format_warning_n_va): Convert to
4370 format_string_diagnostic_t::emit_warning_n_va.
4371 (format_warning_at_substring): Convert to
4372 format_string_diagnostic_t::emit_warning.
4373 (format_warning_at_substring_n): Convert to
4374 format_string_diagnostic_t::emit_warning_n.
4376 2018-09-17 Cesar Philippidis <cesar@codesourcery.com>
4377 Bernd Schmidt <bernds_cb1@t-online.de>
4379 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
4382 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
4384 * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
4385 operand 0 predicate to nonimmediate operand.
4386 (rint<mode>2_frndint): Remove insn pattern.
4387 (rint<mode>2): Change operand 1 predicate to general_operand.
4388 Extend operand 1 to XFmode and generate rintxf2 insn.
4389 (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
4390 Do not use X87MODEF mode macro.
4391 (frndintxf2_<rounding>_i387): Rename from
4392 frndint<mode>2_<rounding>_i387. Do not use X87MODEF mode macro.
4393 (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
4394 to XFmode and generate significandxf3 insn.
4396 2018-09-17 Richard Biener <rguenther@suse.de>
4398 PR tree-optimization/87328
4399 * tree-ssa-sccvn.c (process_bb): Remove assertion about not
4400 visiting unexecutable backedges when not iterating.
4401 (do_rpo_vn): Mark all edges not executable even when not
4404 2018-09-17 Martin Jambor <mjambor@suse.cz>
4407 * doc/invoke.texi (Warning Options): Likewise.
4409 2018-09-17 Richard Biener <rguenther@suse.de>
4411 PR tree-optimization/87301
4412 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
4413 clean EH info from leftover copy assignments.
4415 2018-09-17 Martin Liska <mliska@suse.cz>
4417 PR gcov-profile/85871
4418 * gcov.c (output_intermediate_file): Fix out of bounds
4421 2018-09-17 Vineet Gupta <vgupta@synopsys.com>
4423 * config/arc/arc.c: Object attributes for core4 not reflected
4425 * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
4428 2018-09-17 Alexey Brodkin <abrodkin@synopsys.com>
4430 * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
4432 2018-09-17 Martin Liska <mliska@suse.cz>
4434 * doc/gcov.texi: Document new option --use-hotness-colors.
4435 * gcov.c (struct source_info): Declare new field.
4436 (source_info::source_info): Set default for maximum_count.
4437 (print_usage): Add new -q option.
4438 (process_args): Process it.
4439 (accumulate_line_info): Save src->maximum_count.
4440 (output_line_beginning): Make color line number if
4441 flag_use_hotness_colors is set.
4442 (output_line_details): Pass default argument value.
4443 (output_lines): Pass src->maximum_count.
4445 2018-09-17 Martin Liska <mliska@suse.cz>
4447 * common/config/i386/i386-common.c (ix86_get_valid_option_values):
4448 Use processor_names table.
4449 * config/i386/i386.c (ix86_default_align): Use
4450 processor_cost_table for alignment values.
4451 (ix86_option_override_internal): Use processor_names.
4452 (ix86_function_specific_print): Likewise.
4453 * config/i386/i386.h (struct processor_costs):
4454 Add alignment values.
4455 (struct ptt): Remove and replace with const char *.
4456 * config/i386/x86-tune-costs.h (struct processor_costs):
4457 Declare default alignments for all costs.
4459 2018-09-17 Aldy Hernandez <aldyh@redhat.com>
4461 * tree-vrp.c (extract_range_from_unary_expr): Do not special case
4462 symbolics or VR_VARYING ranges for ABS_EXPR.
4463 * wide-int-range.cc (wide_int_range_abs): Return positive numbers
4464 when range will wrap.
4466 2018-09-15 Eric Botcazou <ebotcazou@adacore.com>
4469 * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
4470 before and after a JUMP_TABLE_DATA.
4472 2018-09-14 John David Anglin <danglin@gcc.gnu.org>
4475 * dojump.c (do_compare_and_jump): Canonicalize function pointers
4476 when one operand is a function pointer. Use POINTER_TYPE_P and
4477 FUNC_OR_METHOD_TYPE_P.
4478 * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
4479 * fold-const.c (build_range_check): Likewise.
4480 * match.pd (simple_comparison): Likewise.
4482 2018-09-14 David Malcolm <dmalcolm@redhat.com>
4485 * spellcheck.c (get_edit_distance_cutoff): New function.
4486 (selftest::test_edit_distance_unit_test_oneway): Rename to...
4487 (selftest::test_get_edit_distance_one_way): ...this.
4488 (selftest::test_get_edit_distance_unit): Rename to...
4489 (selftest::test_get_edit_distance_both_ways): ...this.
4490 (selftest::test_edit_distances): Move tests to this new function,
4491 and test some more pairs of strings. Update for above renaming.
4492 (selftest::get_old_cutoff): New function.
4493 (selftest::test_get_edit_distance_cutoff): New function.
4494 (selftest::assert_suggested_for): New function.
4495 (ASSERT_SUGGESTED_FOR): New macro.
4496 (selftest::assert_not_suggested_for): New function.
4497 (ASSERT_NOT_SUGGESTED_FOR): New macro.
4498 (selftest::test_suggestions): New function.
4499 (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
4500 tests to selftest::test_edit_distances and call it. Add calls to
4501 selftest::test_get_edit_distance_cutoff and
4502 selftest::test_suggestions.
4503 * spellcheck.h (get_edit_distance_cutoff): New function declaration.
4504 (best_match::consider): Replace hard-coded cutoff calculation with
4506 (best_match::get_cutoff): New declaration.
4507 (best_match::get_best_meaningful_candidate): Likewise.
4509 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
4511 * builtins.c (fold_builtin_strlen): Remove TODO comment.
4513 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
4516 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
4518 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
4519 terminated string literal.
4521 2018-09-14 Martin Sebor <msebor@redhat.com>
4523 * builtins.c (unterminated_array): Handle ARRAY_REF.
4524 (expand_builtin_stpcpy_1): Detect unterminated char arrays.
4525 * builtins.h (unterminated_array): Declare extern.
4526 * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
4528 (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
4531 2018-09-14 Martin Sebor <msebor@redhat.com>
4532 Jeff Law <law@redhat.com>
4534 * builtins.c (unterminated_array): New.
4535 (expand_builtin_strcpy): Adjust.
4536 (expand_builtin_strcpy_args): Detect unterminated arrays.
4537 * gimple-fold.c (get_maxval_strlen): Add argument. Detect
4538 unterminated arrays.
4539 * gimple-fold.h (get_maxval_strlen): Add argument.
4540 (gimple_fold_builtin_strcpy): Detec unterminated arrays.
4542 * gimple-fold.c (get_range_strlen): Add argument.
4543 (get_maxval_strlen): Adjust.
4544 * gimple-fold.h (get_range_strlen): Add argument.
4546 2018-09-14 Wei Xiao <wei3.xiao@intel.com>
4548 * config/i386/movdirintrin.h: Fix copyright year.
4550 2018-09-14 Uros Bizjak <ubizjak@gmail.com>
4552 * reg-stack.c: Include regs.h.
4553 (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
4554 (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
4555 FIRST_STACK_REG, not DFmode.
4556 (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
4557 FIRST_STACK_REG, not XFmode. Explicitly construct swap RTX.
4558 (change stack): Default register mode to the reg_raw_mode of
4559 FIRST_STACK_REG, not DFmode.
4560 * config/i386/i386.md (*swap<mode>): Remove insn pattern.
4561 (*swapxf): Rename from swapxf.
4563 2018-09-14 Carl Love <cel@us.ibm.com>
4565 * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
4566 * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
4568 2018-09-14 Segher Boessenkool <segher@kernel.crashing.org>
4571 * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
4574 2018-09-14 Sam Tebbs <sam.tebbs@arm.com>
4577 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
4579 2018-09-14 Jason Merrill <jason@redhat.com>
4581 Fix --enable-gather-detailed-mem-stats.
4582 * hash-table.c (hash_table_usage): Change from variable to function.
4583 * hash-table.h: Adjust.
4584 * Makefile.in: Add missing dependencies on hash-table.h.
4586 2018-09-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4588 PR tree-optimization/87259
4590 (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
4591 execute_cse_reciprocals_1 has tried transforming.
4593 2018-09-14 Aldy Hernandez <aldyh@redhat.com>
4595 * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
4596 VR_VARYING for PLUS/MINUS_EXPR.
4598 2018-09-14 Ilya Leoshkevich <iii@linux.ibm.com>
4600 * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
4603 2018-09-14 Richard Biener <rguenther@suse.de>
4606 * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
4607 bits for the merged partition.
4609 2018-09-13 Martin Sebor <msebor@redhat.com>
4610 Bernd Edlinger <bernd.edlinger@hotmail.de>
4612 * builtins.h (c_srlen): Add argument.
4613 * builtins.c (warn_string_no_nul): New function.
4614 (c_strlen): Add argument and use it. Update recursive calls.
4615 Pass DECL argument to string_constant to get info on non
4616 terminated strings. Update *NONSTR as needed.
4617 (fold_builtin_strlen): Add argument to calls to c_strlen.
4618 Warn for unterminated arrays.
4619 (warn_string_no_null): Add prototype.
4620 * expr.c (string_constant): Update arguments. Update recursive
4621 calls appropriately. Detect missing NUL terminator and outermost
4622 declaration its missing in.
4623 Improve checks for arrays with nonzero lower bound or elements
4624 that are not a single byte. Simplify offset computation.
4625 Simplify checks for non-NUL terminated strings.
4626 * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
4627 * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
4629 2018-09-13 Bernd Edlinger <bernd.edlinger@hotmail.de>
4631 * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
4633 * fold-const.c (c_getstr): Fix function comment. Remove unused third
4634 argument. Fix range checks.
4635 * fold-const.h (c_getstr): Adjust protoype.
4636 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
4637 string is constant but contains no NUL byte.
4639 * expr.c (string_constant): Adjust function comment.
4640 Remove bogus check for zero termination.
4642 * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
4644 * varasm.c (compare_constant): Compare type size of STRING_CSTs.
4645 (get_constant_size): Don't make STRING_CSTs larger than they are.
4646 (check_string_literal): New check function for STRING_CSTs.
4647 (output_constant): Use it.
4649 2018-09-13 Eric Botcazou <ebotcazou@adacore.com>
4652 * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
4654 2018-09-13 Richard Biener <rguenther@suse.de>
4656 PR tree-optimization/87263
4657 * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
4658 (struct unwind_state): Add max_rpo field.
4659 (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
4660 Compute max_rpo, the max RPO number a block can be backwards reached
4661 from. Re-write non-iterating mode to a RPO ordered worklist approach,
4662 separating it from the iterating mode.
4664 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
4666 * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
4667 (rfs_decision): New scheduling decision.
4669 2018-09-13 Richard Biener <rguenther@suse.de>
4672 * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
4673 (vn_nary_op_insert_pieces_predicated): Do not write useless
4674 valid_dominated_by_p entry outside of the allocated storage.
4676 2018-09-13 Omar Sandoval <osandov@osandov.com>
4677 Tom de Vries <tdevries@suse.de>
4680 * dwarf2out.c (is_c): New function.
4681 (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
4683 2018-09-13 Sam Tebbs <sam.tebbs@arm.com>
4686 * config/aarch64/aarch64.md (*aarch64_bfxil):
4688 * config/aarch64/constraints.md (Ulc): Define.
4689 * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
4691 * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
4694 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
4696 * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
4697 * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
4698 aarch64_layout_frame call.
4699 (aarch64_expand_epilogue): Likewise.
4700 (aarch64_initial_elimination_offset): Likewise.
4701 (aarch64_get_separate_components): Likewise.
4702 (aarch64_use_return_insn_p): Likewise.
4703 (aarch64_layout_frame): Remove unneeded check.
4705 2018-09-13 Jakub Jelinek <jakub@redhat.com>
4707 * configure.ac: Only append
4708 " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
4709 gcc_config_arguments if it was never reconfigured or last reconfigure
4710 was with different arguments.
4711 * configure: Regenerated.
4713 2018-09-13 Jakub Jelinek <jakub@redhat.com>
4714 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4717 * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
4718 (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
4720 2018-09-13 Jakub Jelinek <jakub@redhat.com>
4722 PR tree-optimization/87287
4723 * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
4724 X % C == 0 to X % (unsigned) C == 0 optimization to ...
4725 * match.pd (X % C == 0): ... here. New optimization.
4727 2018-09-12 Jakub Jelinek <jakub@redhat.com>
4730 * expr.h (maybe_optimize_mod_cmp): Declare.
4731 * expr.c (mod_inv): New function.
4732 (maybe_optimize_mod_cmp): New function.
4733 (do_store_flag): Use it.
4734 * cfgexpand.c (expand_gimple_cond): Likewise.
4736 2018-09-09 Cesar Philippidis <cesar@codesourcery.com>
4737 Julian Brown <julian@codesourcery.com>
4740 * gimplify.c (gimplify_scan_omp_clauses): Set
4741 target_firstprivatize_array_bases in OpenACC parallel and kernels
4742 region contexts. Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
4743 OpenACC data regions.
4745 2018-09-12 Uros Bizjak <ubizjak@gmail.com>
4747 * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
4748 (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
4750 2018-09-12 Richard Biener <rguenther@suse.de>
4752 PR tree-optimization/87280
4753 * tree-ssa-sccvn.c (process_bb): Handle the case of executable
4754 edge but unreachable target.
4755 (do_rpo_vn): For conservatively handling a PHI only mark
4756 the backedge executable but not the block reachable.
4758 2018-09-12 Richard Biener <rguenther@suse.de>
4760 PR tree-optimization/87266
4761 * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
4764 2018-09-12 Andreas Krebbel <krebbel@linux.ibm.com>
4766 * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
4768 ("trunc<BFP:mode><DFP_ALL:mode>2")
4769 ("trunc<DFP_ALL:mode><BFP:mode>2")
4770 ("extend<BFP:mode><DFP_ALL:mode>2")
4771 ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
4772 according to the target operand type.
4774 2018-09-12 Jakub Jelinek <jakub@redhat.com>
4775 Andreas Krebbel <krebbel@linux.ibm.com>
4777 PR tree-optimization/86844
4778 * gimple-ssa-store-merging.c
4779 (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
4780 there are any overlapping stores in between them, make sure they are
4781 also coalesced or we give up completely.
4783 2018-09-12 Jakub Jelinek <jakub@redhat.com>
4786 * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
4787 BIT_AND_EXPR's second operand is a power of two. Formatting fix.
4789 2018-09-12 Tom de Vries <tdevries@suse.de>
4791 * common.opt (gdescribe-dies): Add option.
4792 * dwarf2out.c (add_name_and_src_coords_attributes): Add description
4793 attribute for artifical and nameless decls.
4794 (dwarf2out_register_external_die): Add description attribute to
4795 external reference die.
4796 (add_desc_attribute): New functions.
4797 (gen_subprogram_die): Add description attribute to
4798 DW_TAG_call_site_parameter.
4799 * tree-pretty-print.c (print_generic_expr_to_str): New function.
4800 * tree-pretty-print.h (print_generic_expr_to_str): Declare.
4801 * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
4803 (@item -gdescribe-dies): Add.
4805 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
4807 * tree-vrp.c (vrp_shift_undefined_p): Remove.
4808 (extract_range_from_binary_expr_1: Call
4809 wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
4810 * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
4813 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
4815 * gimple-ssa-warn-alloca.c
4816 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
4817 field for ALLOCA_BOUND_*_LARGE.
4819 2018-09-11 Nathan Sidwell <nathan@acm.org>
4821 * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
4823 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
4825 * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
4826 for clobbers. Remove obsolete comment.
4828 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
4830 * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
4831 mpxchk, mpxld and mpxst types.
4832 (define_attr length_immediate): Remove all processing of mpx types.
4833 (define_attr prefix_0f): Ditto.
4834 (define_attr memory): Ditto.
4836 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
4838 * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
4839 (log<mode>2): Change operand 1 predicate to general_operand.
4840 Extend operand 1 to XFmode and generate logxf3 insn.
4841 (log10<mode>2): Change operand 1 predicate to general_operand.
4842 Extend operand 1 to XFmode and generate log10xf3 insn.
4843 (log2<mode>2): Change operand 1 predicate to general_operand.
4844 Extend operand 1 to XFmode and generate log2xf3 insn.
4845 (fyl2xp1_extend<mode>xf3_i387): Remove.
4846 (log1p<mode>2): Change operand 1 predicate to general_operand.
4847 Extend operand 1 to XFmode and generate log1pxf3 insn.
4848 (fxtract_extend<mode>xf3_i387): Remove.
4849 (logb<mode>2): Change operand 1 predicate to general_operand.
4850 Extend operand 1 to XFmode and generate logbxf3 insn.
4851 (ilogb<mode>2): Change operand 1 predicate to general_operand.
4852 Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
4853 (significand<mode>2): Change operand 1 predicate to general_operand.
4854 Extend operand 1 to XFmode and generate significandxf3 insn.
4856 2018-09-11 Nathan Sidwell <nathan@acm.org>
4858 * gcc.c (perror_with_name, pfatal_with_name): Delete.
4859 (load_specs): Use fatal_error.
4860 (DELETE_IF_ORDINARY, process_command): Use error.
4861 (execute, run_attempt): Use fatal_error.
4863 2018-09-11 Andrew Stubbs <ams@codesourcery.com>
4865 * diagnostic-core.h (sorry_at): New prototype.
4866 * diagnostic.c (sorry_at): New function.
4868 2018-09-11 Aldy Hernandez <aldyh@redhat.com>
4870 * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
4871 by zero as VR_UNDEFINED.
4873 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
4875 * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
4876 (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
4877 (<sincos>mode2): New expander.
4878 (sincos_extend<mode>xf3_i387): Remove insn pattern.
4879 (sincos -> sin, cos splitters): Remove splitter patterns.
4880 (sincos<mode>3): Change operand 2 predicate to general_operand.
4881 Extend operand 2 to XFmode and generate sincosxf3 insn.
4882 (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
4883 Change operand 3 predicate to const1_operand.
4884 (fptan_extend<mode>xf4_i387): Remove insn pattern.
4885 (tanxf2): Update operands in the call to fptanxf4_i387.
4886 (tan<mode>2): Change operand 1 predicate to general_operand.
4887 Extend operand 1 to XFmode and generate tanxf3 insn.
4888 (atan2xf3): Rename from *fpatanxf3_i387.
4889 (fpatan_extend<mode>xf3_i387): Remove insn pattern.
4890 (atan2xf3): Remove expander.
4891 (atan2<mode<3): Change operand 1 and 2 predicates to general_operand.
4892 Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
4893 (atan<mode>2): Change operand 1 predicate to general_operand.
4894 Extend operand 1 to XFmode and generate atanxf3 insn.
4896 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
4898 * config/i386/i386.md (x87/SSE constant load splitter): Use
4899 memory_operand instead of nonimmediate_operand for input operand
4902 2018-09-09 Uros Bizjak <ubizjak@gmail.com>
4904 * config/i386/i386.md (float partial SSE register stall splitter): Move
4905 splitter near its instruction pattern.
4906 (float_extend partial SSE register stall splitter): Ditto.
4907 (float_truncate partial SSE register stall splitter): Ditto.
4909 2018-09-09 Hans-Peter Nilsson <hp@bitrange.com>
4912 * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
4913 to speculation_safe_value_not_needed.
4916 * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
4917 non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
4918 (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
4919 leaf_function_p, instead use has_hard_reg_initial_val.
4921 2018-09-09 Nathan Sidwell <nathan@acm.org>
4923 * gcc.h (pfatal_with_name): Don't declare here.
4924 * gcc.c (pfatal_with_name): Make static.
4926 2018-09-09 Xianmiao Qu <xianmiao_qu@c-sky.com>
4928 * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
4931 2018-09-08 John David Anglin <danglin@gcc.gnu.org>
4933 PR rtl-optimization/85458
4934 * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
4935 priority hook to reduce the priority of EXPR.
4937 2018-09-07 Uros Bizjak <ubizjak@gmail.com>
4939 * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
4940 DImode for x87 on 32bit targets. Conditionally disable x87 modes
4941 with X87_ENABLE_FLOAT. Remove preparation code.
4942 (*float<SWI48:mode><MODEF:mode>2): Rename from
4943 *float<SWI48:mode><MODEF:mode>2_mixed. Handle x87, SSE and mixed
4944 math using "enabled" attribute.
4945 (*floatdi<MODEF:mode>2_i387): Rename from
4946 *float<SWI48x:mode><MODEF:mode>2_i387. Handle only DImode and
4947 enable for 32bit targets only.
4948 (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
4950 (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
4951 as operand 1 predicate. Rewrite as define_insn_and_split.
4952 (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
4954 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
4956 * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
4957 to fallthru to FLOAT case.
4959 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
4962 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
4963 around folding of vec_sl to handle out of range shift values.
4965 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
4967 * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
4968 Update callers to gen_fix_trunc<mode>_i387_fisttp
4969 (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
4970 nonimmediate_operand.
4971 (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
4972 and corresponding splitters.
4973 (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
4974 (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
4975 (fix_truncdi_i387_with_temp): Remove insn pattern
4976 and corresponding splitters.
4977 (fix_trunc<mode>_i387): Change operand 0 predicate to
4978 nonimmediate_operand.
4979 (fix_trunc<mode>_i387_with_temp): Remove insn pattern
4980 and corresponding splitters.
4981 (*fistdi2_1): Remove.
4983 (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
4984 (lrintxfdi2): Remove expander. Reimplement as define_insn.
4985 (*fist<mode>2_1): Remove.
4986 (fist<mode>2): Ditto.
4987 (fist<mode>2_with_temp): Remove insn pattern and corresponding
4989 (lrintxf<mode>2): Remove expander. Reimplement as define_insn.
4990 (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
4991 (fistdi2_<rounding>): Change operand 0 predicate to
4992 nonimmediate_operand.
4993 (fistdi2_<rounding>_with_temp): Remove insn pattern
4994 and corresponding splitters.
4995 (fist<mode>2_<rounding>): Change operand 0 predicate to
4996 nonimmediate_operand.
4997 (fist<mode>2_<rounding>_with_temp): Remove insn pattern
4998 and corresponding splitters.
5000 (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
5002 2018-09-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
5004 * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
5007 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
5009 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
5010 early gimple folding of vec_splat().
5011 * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
5012 * gimple-fold.h: Add an extern define for tree_vec_extract().
5014 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
5016 * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
5017 wrappers around TREE_TYPE comparisons.
5019 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
5022 * config/s390/predicates.md: Add nonsym_memory_operand.
5023 * config/s390/s390.c (s390_legitimize_cs_operand): If operand
5024 contains a SYMBOL_REF, load it into an intermediate pseudo.
5025 (s390_emit_compare_and_swap): Legitimize operand.
5026 * config/s390/s390.md: Use the new nonsym_memory_operand
5027 with UNSPECV_CAS patterns.
5029 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
5032 * config/s390/s390-passes.def: New file.
5033 * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
5035 (make_pass_s390_early_mach): Add declaration.
5036 * config/s390/s390.c (make_pass_s390_early_mach):
5037 (s390_option_override): Remove dynamic registration.
5038 * config/s390/t-s390: Add s390-passes.def.
5040 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
5042 * config/s390/s390.c (s390_decompose_constant_pool_ref):
5043 Remove UNSPEC_LTREL_BASE check.
5044 (annotate_constant_pool_refs): Likewise.
5045 (find_constant_pool_ref): Likewise.
5046 (find_ltrel_base): Removed.
5047 (replace_ltrel_base): Removed.
5048 (s390_mainpool_finish): Remove replace_ltrel_base call.
5049 (s390_chunkify_start): Remove pending LTREL_BASE logic.
5050 (s390_chunkify_finish): Remove replace_ltrel_base call.
5051 * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
5053 2018-09-06 Hans-Peter Nilsson <hp@axis.com>
5056 * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
5057 to speculation_safe_value_not_needed.
5059 2018-09-05 Cesar Philippidis <cesar@codesourcery.com>
5060 Bernd Schmidt <bernds_cb1@t-online.de>
5062 * config/nvptx/nvptx-opts.h: New file.
5063 * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
5064 * config/nvptx/nvptx.h: Include "nvptx-opts.h".
5066 (TARGET_SM35): New macro.
5067 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
5069 * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
5071 (misa=): New option.
5072 * doc/invoke.texi (Nvidia PTX Options): Document -misa.
5074 2018-09-05 Uros Bizjak <ubizjak@gmail.com>
5076 * config/i386/i386.md (truncdfsf2): Remove expander.
5077 (truncdfsf2_with_temp): Ditto.
5078 (truncxf<mode>2): Ditto.
5079 (*truncdfsf_fast_mixed): Remove insn pattern.
5080 (*truncdfsf_fast_i387): Ditto.
5081 (*truncdfsf_mixed): Ditto.
5082 (*truncdfsf_i387): Ditto.
5083 (*truncdfsf2_i387_1): Ditto.
5084 (*truncxfsf2_mixed): Ditto.
5085 (*truncxfdf2_mixed): Ditto.
5086 (*truncxf<mode>2_i387_noop): Ditto. Update callers
5087 to call gen_truncxf<mode>2 instead.
5088 (*truncxf<mode>2_i387): Remove.
5089 (reg->reg splitters): Remove splitter pattern.
5090 (reg->mem splitters): Ditto.
5092 (truncdfsf2): New insn pattern.
5093 (truncxf<mode>2): Ditto.
5095 2018-09-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5097 * tree-ssa-math-opts.c (is_mult_by): New function.
5098 (is_square_of): Use the above.
5099 (optimize_recip_sqrt): New function.
5100 (pass_cse_reciprocals::execute): Use the above.
5102 2018-09-05 Richard Biener <rguenther@suse.de>
5105 * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
5106 to zero-init the emplaced vec.
5108 2018-09-05 Martin Liska <mliska@suse.cz>
5110 PR tree-optimization/87205
5111 * tree-switch-conversion.c (pass_lower_switch::execute):
5112 Group cases for switch statements.
5114 2018-09-05 Richard Biener <rguenther@suse.de>
5116 PR tree-optimization/87217
5117 * tree-ssa-sccvn.c (vuse_valueize): New.
5118 (vn_reference_lookup_pieces): Use it.
5119 (vn_reference_lookup): Likewise.
5121 2018-09-05 Nathan Sidwell <nathan@acm.org>
5124 * stor-layout.c (place_field): Scan forwards to check last
5125 bitfield when ms_bitfield_placement is in effect.
5127 2018-09-05 Richard Biener <rguenther@suse.de>
5130 * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
5133 2018-09-05 Siddhesh Poyarekar <siddhesh@sourceware.org>
5134 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
5136 * config/aarch64/falkor-tag-collision-avoidance.c: New file.
5137 * config.gcc (extra_objs): Build it.
5138 * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
5140 * config/aarch64/aarch64-passes.def
5141 (pass_tag_collision_avoidance): New pass.
5142 * config/aarch64/aarch64.c (qdf24xx_tunings): Add
5143 AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
5144 (aarch64_classify_address): Remove static qualifier.
5145 (aarch64_address_info, aarch64_address_type): Move to...
5146 * config/aarch64/aarch64-protos.h: ... here.
5147 (make_pass_tag_collision_avoidance): New function.
5148 * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
5151 2018-09-05 Martin Liska <mliska@suse.cz>
5153 * doc/gcov.texi: Update documentation of humar
5155 * gcov.c (format_count): Print one decimal place, it provides
5156 more fine number of situations like '1G' vs. '1.4G'.
5158 2018-09-05 Martin Liska <mliska@suse.cz>
5161 * config/rs6000/rs6000.opt: Mark the option as Deprecated.
5162 * optc-gen.awk: Allow 'Var' for Deprecated options in order
5163 to generate a MASK value.
5165 2018-09-04 H.J. Lu <hongjiu.lu@intel.com>
5168 * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
5169 if frame pointer isn't used.
5170 (compute_frame_pointer_to_fb_displacement): Likewise.
5172 2018-09-04 Jakub Jelinek <jakub@redhat.com>
5175 * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
5176 OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
5177 OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
5178 (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
5179 and OPTION_MASK_ISA_XSAVEC_UNSET.
5181 2018-09-04 Max Filippov <jcmvbkbc@gmail.com>
5183 * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
5184 XOR operations in NAND case.
5186 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
5188 * wide-int-range.cc (wide_int_range_convert): New.
5189 * wide-int-range.h (wide_int_range_convert): New.
5190 * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
5191 code into wide_int_range_convert.
5192 (extract_range_into_wide_ints): Do not munge anti range constants
5193 into the entire domain. Just return the range back.
5195 2018-09-04 Martin Liska <mliska@suse.cz>
5197 * genmatch.c (output_line_directive): Add new argument
5199 (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
5201 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
5203 * doc/invoke.texi (Option Summary): Add whitespace.
5205 * doc/invoke.texi (Option Summary): Add -Waligned-new.
5207 2018-09-04 Richard Biener <rguenther@suse.de>
5209 PR tree-optimization/87211
5210 * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
5211 backedge value we're supposed to treat as VARYING also number
5212 the PHI to VARYING in case it got a different value-number already.
5214 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
5216 * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
5217 (extract_range_from_binary_expr_1): Do not call
5218 vrp_can_optimize_bit_op.
5219 * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
5221 (wide_int_range_get_mask_and_bounds): New.
5222 (wide_int_range_optimize_bit_op): New.
5223 (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
5224 (wide_int_range_bit_and): Same.
5225 * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
5226 (wide_int_range_optimize_bit_op): New.
5227 (wide_int_range_get_mask_and_bounds): New.
5229 2018-09-04 Richard Biener <rguenther@suse.de>
5231 PR tree-optimization/87176
5232 * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
5233 variable. When value-numbering a virtual PHI node make sure
5234 to not value-number to the backedge value.
5236 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
5238 * doc/extend.texi (Long Long, Hex Floats): Document support for
5239 long long and hex floats in more recent versions of ISO C++.
5241 2018-09-03 Richard Biener <rguenther@suse.de>
5243 PR tree-optimization/87177
5244 * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
5247 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
5249 * bb-reorder.c (edge_order): Convert to C-qsort-style
5250 tri-state comparator.
5251 (reorder_basic_blocks_simple): Change std::stable_sort to
5254 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
5256 * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
5257 tri-state comparator.
5258 (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
5260 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
5262 * sort.cc (struct sort_ctx): New field 'nlim'. Use it...
5263 (mergesort): ... here as maximum count for using netsort.
5264 (gcc_qsort): Set nlim to 3 if stable sort is requested.
5265 (gcc_stablesort): New.
5266 * system.h (gcc_stablesort): Declare.
5268 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
5270 * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
5271 * system.h (qsort): Always redirect to gcc_qsort. Update comment.
5272 * vec.c (qsort_chk): Do not call gcc_qsort. Update comment.
5274 2018-09-03 Segher Boessenkool <segher@kernel.crashing.org>
5276 * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
5277 lxsdx and stxsdx alternatives.
5278 (*mov<mode>_hardfloat64): Ditto.
5279 * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
5281 2018-09-03 Richard Biener <rguenther@suse.de>
5283 PR tree-optimization/87200
5284 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
5287 2018-09-03 Martin Liska <mliska@suse.cz>
5289 PR tree-optimization/87201
5290 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
5291 Fix parenthesis in an expression.
5293 2018-09-03 Richard Biener <rguenther@suse.de>
5295 PR tree-optimization/87197
5296 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
5297 visited. CSE the VN_INFO hashtable lookup.
5299 PR tree-optimization/87169
5300 * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
5301 iterating make sure there's no extra backedges from irreducible
5302 regions feeding the header. Mark the destination block
5305 2018-09-03 Martin Liska <mliska@suse.cz>
5308 * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
5309 * common/common-targhooks.c (default_get_valid_option_values):
5311 * common/common-targhooks.h (default_get_valid_option_values):
5313 * common/config/i386/i386-common.c: Move processor_target_table
5315 (ix86_get_valid_option_values): New function.
5316 (TARGET_GET_VALID_OPTION_VALUES): New macro.
5317 * config/i386/i386.c (struct ptt): Move to i386-common.c.
5318 (PTA_*): Move all defined masks into i386-common.c.
5319 (ix86_function_specific_restore): Use new processor_cost_table.
5320 * config/i386/i386.h (struct ptt): Moved from i386.c.
5321 (struct pta): Likewise.
5322 * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
5323 * doc/tm.texi.in: Likewise.
5324 * opt-suggestions.c (option_proposer::suggest_option):
5325 Pass prefix to build_option_suggestions.
5326 (option_proposer::get_completions): Likewise.
5327 (option_proposer::build_option_suggestions): Use the new target
5329 * opts.c (struct option_help_tuple): New struct.
5330 (print_filtered_help): Use the new target hook.
5332 2018-09-03 Martin Liska <mliska@suse.cz>
5335 * predict.c (set_even_probabilities): Add likely_edges
5336 argument and handle cases where we have precisely one
5338 (combine_predictions_for_bb): Catch also likely_edges.
5339 (tree_predict_by_opcode): Handle gswitch statements.
5340 * tree-cfg.h (find_case_label_for_value): New declaration.
5341 (find_taken_edge_switch_expr): Likewise.
5342 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
5343 Find pivot in decision tree based on probabily, not by number of
5346 2018-09-02 Gerald Pfeifer <gerald@pfeifer.com>
5348 * doc/standards.texi (Standards): Update Objective-C reference.
5350 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
5352 * doc/install.texi (Prerequisites): Update link for MPC.
5354 2018-09-01 Michael Matz <matz@suse.de>
5356 PR tree-optimization/87074
5357 * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
5358 PHIs for outer-loop uses.
5360 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
5362 * doc/generic.texi (OpenMP): Adjust link to openmp.org.
5363 * doc/invoke.texi (C Dialect Options): Ditto.
5365 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
5367 * doc/install.texi (Prerequisites): Adjust link mpfr.org.
5369 2018-08-31 Richard Biener <rguenther@suse.de>
5371 PR tree-optimization/87168
5372 * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
5373 (rpo_elim::eliminate_avail): When OP was not visited it must
5376 2018-08-31 David Malcolm <dmalcolm@redhat.com>
5378 * tree-vrp.c (copy_value_range): Convert param "from" from
5379 "value_range *" to "const value_range *".
5380 (range_is_null): Likewise for param "vr".
5381 (range_int_cst_p): Likewise.
5382 (range_int_cst_singleton_p): Likewise.
5383 (symbolic_range_p): Likewise.
5384 (value_ranges_intersect_p): Likewise for both params.
5385 (value_range_nonnegative_p): Likewise for param "vr".
5386 (value_range_constant_singleton): Likewise.
5387 (vrp_set_zero_nonzero_bits): Likewise for param "ar".
5388 (extract_range_into_wide_ints): Likewise for param "vr".
5389 (extract_range_from_multiplicative_op): Likewise for params "vr0"
5391 (vrp_can_optimize_bit_op): Likewise.
5392 (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
5394 (extract_range_from_unary_expr): Likewise.
5395 (debug_value_range): Likewise for param "vr".
5396 (value_range::dump): Add "const" qualifier.
5397 (vrp_prop::check_array_ref): Convert local "vr" from
5398 "value_range *" to "const value_range *".
5399 (vrp_prop::check_mem_ref): Likewise.
5400 (vrp_prop::visit_stmt): Likewise for local "old_vr".
5401 (vrp_intersect_ranges_1): Likewise for param "vr_1".
5402 (vrp_intersect_ranges): Likewise.
5403 (simplify_stmt_for_jump_threading): Likewise for local "vr".
5404 (vrp_prop::vrp_finalize): Likewise.
5405 * tree-vrp.h (value_range::dump): Add "const" qualifier.
5406 (vrp_intersect_ranges): Add "const" qualifier to params as above.
5407 (extract_range_from_unary_expr): Likewise.
5408 (value_range_constant_singleton): Likewise.
5409 (symbolic_range_p): Likewise.
5410 (copy_value_range): Likewise.
5411 (extract_range_from_binary_expr_1): Likewise.
5412 (range_int_cst_p): Likewise.
5413 (vrp_set_zero_nonzero_bits): Likewise.
5414 (range_int_cst_singleton_p): Likewise.
5416 2018-08-31 Vlad Lazar <vlad.lazar@arm.com>
5418 * config/aarch64/arm_neon.h (vabsd_s64): New.
5419 (vnegd_s64): Likewise.
5421 2018-08-31 Martin Jambor <mjambor@suse.cz>
5423 * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
5425 2018-08-31 Martin Liska <mliska@suse.cz>
5427 * ipa-icf.c (sem_item::add_type): Use
5428 sem_item::m_type_hash_cache.
5429 * ipa-icf.h: Move the cache from sem_item_optimizer
5432 2018-08-31 Nathan Sidwell <nathan@acm.org>
5434 * doc/extend.texi (Backwards Compatibility): Remove implicit
5435 extern C leeway of () being (...).
5437 2018-08-31 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5439 * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
5441 2018-08-31 Segher Boessenkool <segher@kernel.crashing.org>
5445 * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
5447 2018-08-31 Jakub Jelinek <jakub@redhat.com>
5450 * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
5451 gen_int_mode. Formatting fixes.
5453 2018-08-30 Sandra Loosemore <sandra@codesourcery.com>
5455 * target.def (custom_function_descriptors): Improve documentation.
5456 * doc/tm.texi.in (Trampolines): Expand discussion of function
5457 descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
5458 beginning of the section.
5459 * doc/tm.texi: Regenerated.
5461 2018-08-30 Jose E. Marchesi <jose.marchesi@oracle.com>
5463 * cfg.h (class auto_edge_flag): Spell out the template-id of the
5464 base class in the initializer list. This is a workaround for
5465 building with older GCC.
5466 (class auto_bb_flag): Likewise.
5468 2018-08-30 Aaron Sawdey <acsawdey@linux.ibm.com>
5470 * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
5471 (altivec_vcmpequ<VI_char>_p): Remove star.
5472 * config/rs6000/rs6000-string.c (do_load_for_compare): Support
5474 (expand_strncmp_vec_sequence): New function.
5475 (emit_final_str_compare_vec): New function.
5476 (expand_strn_compare): Add support for vector strncmp.
5477 * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
5478 length specification to bytes.
5479 * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
5480 (vcmpnezb_p): New pattern.
5481 * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
5482 for option -mstring-compare-inline-limit.
5484 2018-08-30 Thiago Macieira <thiago.macieira@intel.com>
5486 * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
5487 (PTA_SKYLAKE): Add PTA_AES.
5488 (PTA_GOLDMONT): Likewise.
5490 2018-08-29 Jan Hubicka <jh@suse.cz>
5493 * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
5494 * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
5496 2018-08-29 Jan Hubicka <jh@suse.cz>
5498 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
5500 (hash_tree): Do not visit TYPE_STUB_DECL.
5501 * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
5502 stream TYPE_STUB_DECL.
5503 * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
5504 * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
5505 after free_lang_data.
5506 (type_in_anonymous_namespace_p): Likewise.
5508 2018-08-29 Jan Hubicka <jh@suse.cz>
5510 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
5511 comment that it has to be even number.
5512 (class sreal): Change m_sig type to int32_t.
5513 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
5514 int64_t for temporary calculations.
5515 (sreal_verify_basics): Drop one bit from minimum and maximum.
5517 2018-08-30 Richard Biener <rguenther@suse.de>
5519 PR tree-optimization/87147
5520 * tree-ssa-sccvn.c (SSA_VISITED): New function.
5521 (visit_phi): When the degenerate result is from the backedge and
5522 we didn't visit its definition yet drop to VARYING.
5523 (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
5525 2018-08-29 Jan Hubicka <jh@suse.cz>
5527 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
5529 (hash_tree): Likewise.
5531 2018-08-29 Jan Hubicka <jh@suse.cz>
5533 * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
5534 and TYPE_NEXT_REF_TO.
5536 2018-08-29 Jan Hubicka <jh@suse.cz>
5538 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
5539 comment that it has to be even number.
5540 (class sreal): Change m_sig type to int32_t.
5541 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
5542 int64_t for temporary calculations.
5543 (sreal_verify_basics): Drop one bit from minimum and maximum.
5545 2018-08-30 Tamar Christina <tamar.christina@arm.com>
5547 * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
5549 2018-08-30 Vlad Lazar <vlad.lazar@arm.com>
5552 * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
5553 if to_add is negative.
5555 2018-08-29 Bernd Edlinger <bernd.edlinger@hotmail.de>
5558 * builtins.c (c_strlen): Improve range checks.
5560 2018-08-29 Martin Sebor <msebor@redhat.com>
5561 Jeff Law <law@redhat.com>
5563 PR tree-optimization/86714
5564 PR tree-optimization/86711
5565 * builtins.c (c_strlen): Add arguments to call to string_constant.
5566 * expr.c (string_constant): Add argument. Detect missing nul
5567 terminator and outermost declaration it's missing in.
5568 * expr.h (string_constant): Add argument.
5569 * fold-const.c (read_from_constant_string): Add arguments to call to
5571 (c_getstr): Likewise.
5572 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
5574 * tree-ssa-strlen.c (get_stridx): Likewise.
5576 2018-08-29 Jan Hubicka <jh@suse.cz>
5578 * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
5579 Do not stream DECL_VINDEX.
5580 * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
5581 * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
5582 (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
5584 2018-08-29 Richard Biener <rguenther@suse.de>
5586 * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
5587 virtual operands that are not default defs to honor region
5589 (rpo_vn_valueize): Remove ineffective code here.
5591 2018-08-29 Richard Biener <rguenther@suse.de>
5593 PR tree-optimization/87132
5594 * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
5595 when skipping defs reachable over backedges.
5597 2018-08-29 Richard Biener <rguenther@suse.de>
5599 * tree-core.h: Document use of deprecated_flag in SSA_NAME.
5600 * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
5601 * tree-into-ssa.c (pass_build_ssa::execute): Initialize
5602 function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
5603 * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
5604 (vn_reference_lookup_3): Remove use of const_parms.
5605 (free_rpo_vn): Do not free const_parms.
5606 (do_rpo_vn): Do not call init_const_parms.
5607 * tree-ssa-alias.c (refs_may_alias_p_1): Honor
5608 SSA_NAME_POINTS_TO_READONLY_MEMORY.
5609 (call_may_clobber_ref_p_1): Likewise.
5611 2018-08-29 Alexander Monakov <amonakov@ispras.ru>
5614 * invoke.texi (Optimization Options): List -ftree-scev-cprop.
5616 (-ftree-scev-cprop): Document.
5618 2018-08-29 Jan Hubicka <jh@suse.cz>
5620 * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
5622 (sreal constructor): Update.
5623 * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
5624 sreal:operator/): Update.
5626 2018-08-29 Martin Liska <mliska@suse.cz>
5628 * tree-switch-conversion.c (switch_conversion::expand):
5629 Strenghten assumption about gswitch statements.
5631 2018-08-29 Richard Biener <rguenther@suse.de>
5633 PR tree-optimization/87117
5634 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
5635 re-value-number released SSA VDEFs.
5637 2018-08-29 Richard Biener <rguenther@suse.de>
5639 PR tree-optimization/87126
5640 * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
5642 2018-08-28 Jim Wilson <jimw@sifive.com>
5644 * config/riscv/pic.md: Rewrite.
5645 * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
5647 * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
5648 (SOFTF, default_load, softload, softstore): New.
5650 2018-08-28 Jeff Law <law@redhat.com>
5652 * fold-const.c (fold_binary_loc): Remove recently added assert.
5654 2018-08-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
5656 * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
5657 to OP parmeter of generated function.
5659 2018-08-28 MCC CS <deswurstes@users.noreply.github.com>
5661 PR tree-optimization/87009
5662 * match.pd: Add boolean optimizations.
5664 2018-08-28 Martin Sebor <msebor@redhat.com>
5667 * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
5668 * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
5669 (pass_walloca::gate): Use it.
5670 (alloca_call_type): Same.
5671 (pass_walloca::execute): Same.
5672 * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
5674 2018-08-28 David Malcolm <dmalcolm@redhat.com>
5676 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
5677 GCC_VERSION for usage of "__gcc_dump_printf__" format from
5680 2018-08-28 Richard Biener <rguenther@suse.de>
5682 PR tree-optimization/87124
5683 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
5684 constants before looking up avail.
5686 2018-08-28 Jakub Jelinek <jakub@redhat.com>
5689 * calls.c (maybe_warn_nonstring_arg): Punt early if
5690 warn_stringop_overflow is zero. Don't call get_range_strlen
5691 on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
5692 Swap comparison operands to have constants on rhs. Only use
5693 lenrng[1] if non-NULL and INTEGER_CST. Don't uselessly
5694 increment lenrng[0].
5696 2018-08-28 Richard Sandiford <richard.sandiford@arm.com>
5698 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
5699 use of tree_to_shwi. Remove duplicated test for the size being
5700 a whole number of bytes.
5702 2018-08-28 Richard Biener <rguenther@suse.de>
5704 PR tree-optimization/87117
5705 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
5706 Handle removed stmt without LHS (GIMPLE_NOP).
5708 2018-08-28 Richard Biener <rguenther@suse.de>
5710 PR tree-optimization/87117
5711 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
5712 void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
5714 2018-08-28 Richard Biener <rguenther@suse.de>
5716 PR tree-optimization/87117
5717 * tree-ssa-pre.c (compute_avail): Do not make expressions
5718 with predicated values available.
5719 (get_expr_value_id): Assert we do not run into predicated value
5722 2018-08-28 Richard Biener <rguenther@suse.de>
5724 PR tree-optimization/87117
5725 * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
5726 get virtual operands.
5727 (get_expr_operands): Handle STRING_CST like other decls.
5729 2018-08-28 Martin Liska <mliska@suse.cz>
5731 * tree.h: Update documentation of fndecl_built_in_p
5735 2018-08-27 Jeff Law <law@redhat.com>
5736 PR tree-optimization/87110
5737 * tree-ssa-dse.c (compute_trims): Handle non-constant
5740 2018-08-27 Martin Sebor <msebor@redhat.com>
5742 PR tree-optimization/86914
5743 * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
5745 2018-08-27 Martin Sebor <msebor@redhat.com>
5747 PR tree-optimization/87112
5748 * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
5749 the type of the bound argument.
5751 2018-08-27 Jeff Law <law@redhat.com>
5753 * tree-ssa-dse.c (compute_trims): Handle case where the reference's
5754 type does not have a TYPE_SIZE_UNIT.
5756 2018-08-27 Steve Ellcey <sellcey@cavium.com>
5758 * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
5759 with include of backend.h.
5761 2018-08-27 Richard Biener <rguenther@suse.de>
5763 PR tree-optimization/86927
5764 * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
5765 use const cond reduction code.
5767 2018-08-27 Alexander Monakov <amonakov@ispras.ru>
5769 PR tree-optimization/85758
5770 * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
5772 2018-08-27 David Malcolm <dmalcolm@redhat.com>
5775 * diagnostic-show-locus.c (class layout_range): Update for
5776 conversion of show_caret_p to a tri-state.
5777 (layout_range::layout_range): Likewise.
5778 (make_range): Likewise.
5779 (layout::maybe_add_location_range): Likewise.
5780 (layout::should_print_annotation_line_p): Don't show annotation
5781 lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
5782 (layout::get_state_at_point): Update for conversion of
5783 show_caret_p to a tri-state. Bail out early for
5784 SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
5785 underlining or source colorization.
5786 (gcc_rich_location::add_location_if_nearby): Update for conversion
5787 of show_caret_p to a tri-state.
5788 (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
5789 (selftest::test_one_liner_fixit_replace_equal_secondary_range):
5791 (selftest::test_one_liner_labels): Likewise.
5792 * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
5793 conversion of show_caret_p to a tri-state.
5794 * pretty-print.c (text_info::set_location): Likewise.
5795 * pretty-print.h (text_info::set_location): Likewise.
5796 * substring-locations.c (format_warning_n_va): Likewise.
5797 * tree-diagnostic.c (default_tree_printer): Likewise.
5798 * tree-pretty-print.c (newline_and_indent): Likewise.
5800 2018-08-27 David Malcolm <dmalcolm@redhat.com>
5803 * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
5804 line above for line-insertion fix-it hints.
5805 (selftest::test_fixit_insert_containing_newline): Update the
5806 expected results, and add a test with line-numbering enabled.
5808 2018-08-27 Martin Liska <mliska@suse.cz>
5811 * sanopt.c (sanitize_rewrite_addressable_params): Ignore
5812 params with DECL_HAS_VALUE_EXPR_P.
5814 2018-08-27 Martin Liska <mliska@suse.cz>
5816 * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
5817 selected expansion strategy.
5819 2018-08-27 Martin Liska <mliska@suse.cz>
5821 * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
5822 * builtins.c (is_builtin_fn): Likewise.
5823 * attribs.c (diag_attr_exclusions): Use new function
5824 fndecl_built_in_p and remove check for FUNCTION_DECL if
5826 (builtin_mathfn_code): Likewise.
5827 (fold_builtin_expect): Likewise.
5828 (fold_call_expr): Likewise.
5829 (fold_builtin_call_array): Likewise.
5830 (fold_call_stmt): Likewise.
5831 (set_builtin_user_assembler_name): Likewise.
5832 (is_simple_builtin): Likewise.
5833 * calls.c (gimple_alloca_call_p): Likewise.
5834 (maybe_warn_nonstring_arg): Likewise.
5835 * cfgexpand.c (expand_call_stmt): Likewise.
5836 * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
5837 (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
5838 (cgraph_node::verify_node): Likewise.
5839 * cgraphclones.c (build_function_decl_skip_args): Likewise.
5840 (cgraph_node::create_clone): Likewise.
5841 * config/arm/arm.c (arm_insert_attributes): Likewise.
5842 * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
5843 * dse.c (scan_insn): Likewise.
5844 * expr.c (expand_expr_real_1): Likewise.
5845 * fold-const.c (operand_equal_p): Likewise.
5846 (fold_binary_loc): Likewise.
5847 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
5848 * gimple-low.c (lower_stmt): Likewise.
5849 * gimple-pretty-print.c (dump_gimple_call): Likewise.
5850 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
5852 * gimple.c (gimple_build_call_from_tree): Likewise.
5853 (gimple_call_builtin_p): Likewise.
5854 (gimple_call_combined_fn): Likewise.
5855 * gimplify.c (gimplify_call_expr): Likewise.
5856 (gimple_boolify): Likewise.
5857 (gimplify_modify_expr): Likewise.
5858 (gimplify_addr_expr): Likewise.
5859 * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
5860 * ipa-cp.c (determine_versionability): Likewise.
5861 * ipa-fnsummary.c (compute_fn_summary): Likewise.
5862 * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
5863 * ipa-split.c (visit_bb): Likewise.
5864 (split_function): Likewise.
5865 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
5866 * lto-cgraph.c (input_node): Likewise.
5867 * lto-streamer-out.c (write_symbol): Likewise.
5868 * omp-low.c (setjmp_or_longjmp_p): Likewise.
5869 (lower_omp_1): Likewise.
5870 * predict.c (strip_predict_hints): Likewise.
5871 * print-tree.c (print_node): Likewise.
5872 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
5873 * trans-mem.c (is_tm_irrevocable): Likewise.
5874 (is_tm_load): Likewise.
5875 (is_tm_simple_load): Likewise.
5876 (is_tm_store): Likewise.
5877 (is_tm_simple_store): Likewise.
5878 (is_tm_abort): Likewise.
5879 (tm_region_init_1): Likewise.
5880 * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
5881 * tree-cfg.c (verify_gimple_call): Likewise.
5882 (move_stmt_r): Likewise.
5883 (stmt_can_terminate_bb_p): Likewise.
5884 * tree-eh.c (lower_eh_constructs_2): Likewise.
5885 * tree-if-conv.c (if_convertible_stmt_p): Likewise.
5886 * tree-inline.c (remap_gimple_stmt): Likewise.
5887 (copy_bb): Likewise.
5888 (estimate_num_insns): Likewise.
5889 (fold_marked_statements): Likewise.
5890 * tree-sra.c (scan_function): Likewise.
5891 * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
5892 (optimize_stack_restore): Likewise.
5893 (pass_fold_builtins::execute): Likewise.
5894 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
5895 (mark_all_reaching_defs_necessary_1): Likewise.
5896 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
5897 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
5898 (pass_forwprop::execute): Likewise.
5899 * tree-ssa-loop-im.c (stmt_cost): Likewise.
5900 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
5901 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
5902 * tree-ssa-strlen.c (get_string_length): Likewise.
5903 * tree-ssa-structalias.c (handle_lhs_call): Likewise.
5904 (find_func_aliases_for_call): Likewise.
5905 * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
5906 * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
5907 * tree-tailcall.c (find_tail_calls): Likewise.
5908 * tree.c (need_assembler_name_p): Likewise.
5909 (free_lang_data_in_decl): Likewise.
5910 (get_call_combined_fn): Likewise.
5911 * ubsan.c (is_ubsan_builtin_p): Likewise.
5912 * varasm.c (incorporeal_function_p): Likewise.
5913 * tree.h (DECL_BUILT_IN): Remove and replace with
5915 (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
5916 (fndecl_built_in_p): New.
5918 2018-08-27 Martin Liska <mliska@suse.cz>
5920 PR tree-optimization/86847
5921 * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
5922 Dump also subtree probability.
5923 (switch_decision_tree::do_jump_if_equal): New function.
5924 (switch_decision_tree::emit_case_nodes): Handle special
5925 situations in balanced tree that can be emitted much simpler.
5926 Fix calculation of probabilities that happen in tree expansion.
5927 * tree-switch-conversion.h (struct cluster): Add
5929 (struct simple_cluster): Likewise.
5930 (struct case_tree_node): Add new function has_child.
5931 (do_jump_if_equal): New.
5933 2018-08-27 Martin Liska <mliska@suse.cz>
5935 * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
5936 Add new argument to bit_test_cluster constructor.
5937 (bit_test_cluster::emit): Set bits really number of values
5939 (bit_test_cluster::hoist_edge_and_branch_if_true): Add
5940 probability argument.
5941 * tree-switch-conversion.h (struct bit_test_cluster):
5942 Add m_handles_entire_switch.
5944 2018-08-27 Martin Liska <mliska@suse.cz>
5946 PR tree-optimization/86702
5947 * tree-switch-conversion.c (jump_table_cluster::emit):
5948 Make probabilities even for values in jump table
5949 according to number of cases handled.
5950 (switch_decision_tree::compute_cases_per_edge): Pass
5951 argument to reset_out_edges_aux function.
5952 (switch_decision_tree::analyze_switch_statement): Likewise.
5953 * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
5956 2018-08-27 Martin Liska <mliska@suse.cz>
5958 * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
5959 cfun argument explicitly.
5960 * gimple-pretty-print.c (dump_gimple_switch): Likewise.
5961 * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
5962 function gimple_switch_default_bb.
5963 (convert_switch_statements):
5965 * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
5966 * stmt.c (label_to_block_fn): Use label_to_block and pass
5967 cfun argument explicitly and use gimple_switch_label_bb.
5968 (expand_case): Likewise.
5969 * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
5970 cfun argument explicitly. Likewise.
5971 (make_edges_bb): Likewise.
5972 (make_cond_expr_edges): Likewise.
5973 (get_cases_for_edge): Likewise.
5974 (make_gimple_switch_edges): Likewise.
5975 (label_to_block_fn): Likewise.
5976 (label_to_block): Likewise.
5977 (make_goto_expr_edges): Likewise.
5978 (make_gimple_asm_edges): Likewise.
5979 (main_block_label): Likewise.
5980 (group_case_labels_stmt): Likewise.
5981 (find_taken_edge_computed_goto): Likewise.
5982 (find_taken_edge_switch_expr): Likewise.
5983 (gimple_verify_flow_info): Likewise.
5984 (gimple_redirect_edge_and_branch): Likewise.
5985 (gimple_switch_label_bb): New function.
5986 (gimple_switch_default_bb): Likewise.
5987 (gimple_switch_edge): Likewise.
5988 (gimple_switch_default_edge): Likewise.
5989 * tree-cfg.h (label_to_block_fn): Remove and replace ...
5990 (label_to_block): ... with this.
5991 (gimple_switch_label_bb): New.
5992 (gimple_switch_default_bb): Likewise.
5993 (gimple_switch_edge): Likewise.
5994 (gimple_switch_default_edge): Likewise.
5995 * tree-cfgcleanup.c (convert_single_case_switch): Use
5996 new gimple functions and pass new argument to label_to_block.
5997 (cleanup_control_flow_bb):
5998 * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
5999 cfun argument explicitly.
6000 (make_eh_edges): Likewise.
6001 (redirect_eh_dispatch_edge): Likewise.
6002 (lower_resx): Likewise.
6003 (lower_eh_dispatch): Likewise.
6004 (maybe_remove_unreachable_handlers): Likewise.
6005 (unsplit_eh): Likewise.
6006 (cleanup_empty_eh): Likewise.
6007 (verify_eh_edges): Likewise.
6008 (verify_eh_dispatch_edge): Likewise.
6009 * tree-ssa-dom.c (record_edge_info): Likewise.
6010 * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
6011 * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
6012 (thread_through_normal_block): Likewise.
6013 * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
6014 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
6015 * tree-switch-conversion.c (switch_conversion::collect): Use new
6017 (switch_conversion::check_final_bb): Likewise.
6018 (switch_conversion::gather_default_values): Pass new argument
6020 (switch_conversion::build_constructors): Likewise.
6021 (switch_decision_tree::compute_cases_per_edge): Use new
6022 gimple_switch_edge function.
6023 (switch_decision_tree::analyze_switch_statement): Pass new argument
6025 (switch_decision_tree::try_switch_expansion): Use
6026 gimple_switch_default_edge.
6027 * tree-vrp.c (find_switch_asserts): Pass new argument
6029 * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
6030 (vr_values::simplify_switch_using_ranges): Likewise.
6032 2018-08-27 Richard Biener <rguenther@suse.de>
6034 * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
6035 * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
6037 * tree-ssa-sccvn.h (struct vn_pval): New structure.
6038 (struct vn_nary_op_s): Add unwind_to member. Add
6039 predicated_values flag and put result into a union together
6040 with a linked list of vn_pval.
6041 (struct vn_ssa_aux): Add name member to make maintaining
6042 a map of SSA name to vn_ssa_aux possible. Remove no longer
6043 needed info, dfsnum, low, visited, on_sccstack, use_processed
6044 and range_info_anti_range_p members.
6045 (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
6046 (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
6048 (vn_valueize): New global.
6049 (vn_context_bb): Likewise.
6050 (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
6051 VN_INFO_PTR_INFO): Remove.
6052 * tree-ssa-sccvn.c: ... (rewrite)
6053 (pass_fre::execute): For -O2+ initialize loops and run
6054 RPO VN in optimistic mode (iterating). For -O1 and -Og
6055 run RPO VN in non-optimistic mode.
6056 * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
6057 (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
6058 * doc/invoke.texi (sccvn-max-scc-size): Remove.
6059 (rpo-vn-max-loop-depth): Document.
6060 * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
6061 when valuezing the VUSE signals we walked out of the region.
6062 * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
6063 (phi_translate): Set VN context block to use for availability
6065 (compute_avail): Likewise.
6066 (pre_valueize): New function.
6067 (pass_pre::execute): Adjust to the RPO VN API.
6069 * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
6070 (propagate_constants_for_unrolling): Remove.
6071 (tree_unroll_loops_completely): Perform value-numbering
6072 on the unrolled bodies loop parent.
6074 2018-08-27 Richard Biener <rguenther@suse.de>
6076 * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
6077 for partial antic compute.
6079 2018-08-27 Jakub Jelinek <jakub@redhat.com>
6081 PR rtl-optimization/87065
6082 * combine.c (simplify_if_then_else): Formatting fix.
6083 (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
6085 (known_cond): Don't return const_true_rtx for vector modes. Use
6086 CONST0_RTX instead of const0_rtx. Formatting fixes.
6088 2018-08-27 Martin Liska <mliska@suse.cz>
6090 PR gcov-profile/87069
6091 * gcov.c (process_file): Record files already processed
6092 and warn about a file being processed multiple times.
6094 2018-08-27 Martin Liska <mliska@suse.cz>
6097 * config/aarch64/aarch64.c (aarch64_override_options_internal):
6098 Set default values for x_aarch64_*_string strings.
6099 * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
6100 prefix. For -mabi do not print '=ABI' in help and use
6101 <option_value> format for -msve-vector-bits and -moverride
6104 2018-08-26 Jeff Law <law@redhat.com>
6106 * config/mips/frame-header-opt.c: Include "backend.h" rather than
6109 2018-08-26 Marek Polacek <polacek@redhat.com>
6111 PR c++/87029, Implement -Wredundant-move.
6112 * doc/invoke.texi: Document -Wredundant-move.
6114 2018-08-25 Martin Sebor <msebor@redhat.com>
6116 PR tree-optimization/87059
6117 * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
6118 to the same type as the other.
6119 * fold-const.c (fold_binary_loc): Assert expectation.
6121 2018-08-25 Iain Sandoe <iain@sandoe.co.uk>
6123 * config/darwin.c (machopic_legitimize_pic_address): Clean up
6124 extraneous parentheses, dead code section and formatting.
6126 2018-08-24 David Malcolm <dmalcolm@redhat.com>
6129 * diagnostic-show-locus.c (layout::layout): Ensure the margin is
6130 wide enough for jumps in the line-numbering to be visible.
6131 (layout::print_gap_in_line_numbering): New member function.
6132 (layout::calculate_line_spans): When using line numbering, merge
6133 line spans that are only 1 line apart.
6134 (diagnostic_show_locus): When printing line numbers, show gaps in
6135 line numbering directly, rather than printing headers.
6136 (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
6137 line-numbering with multiple line spans.
6138 (selftest::test_fixit_insert_containing_newline_2): Add test of
6139 line-numbering, in which the spans are close enough to be merged.
6141 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
6143 * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
6144 to range_includes_zero_p. Do not special case VR_ANTI_RANGE.
6145 * tree-vrp.c (range_is_nonnull): Remove.
6146 (range_includes_zero_p): Accept value_range instead of min/max.
6147 (extract_range_from_binary_expr_1): Do not early bail on
6149 Use range_includes_zero_p instead of range_is_nonnull.
6150 (extract_range_from_unary_expr): Use range_includes_zero_p instead
6151 of range_is_nonnull.
6152 (vrp_meet_1): Pass value_range to range_includes_zero_p. Do not
6153 special case VR_ANTI_RANGE.
6154 (vrp_finalize): Same.
6155 * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
6157 (range_is_nonnull): Remove.
6158 * vr-values.c (vrp_stmt_computes_nonzero): Use
6159 range_includes_zero_p instead of range_is_nonnull.
6160 (extract_range_basic): Pass value_range to range_includes_zero_p
6161 instead of range_is_nonnull.
6163 2018-08-24 Uros Bizjak <ubizjak@gmail.com>
6165 * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
6166 * emit-rtl.h (rtl_data): Remove return_bnd.
6167 * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
6168 * function.c (diddle_return_value): Do not handle crtl->return_bnd.
6169 * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
6170 (POINTER_BOUNDS_MODE): Remove definition.
6171 (make_pointer_bounds_mode): Remove.
6172 (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
6173 * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
6174 (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
6175 * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
6176 * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
6177 * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
6178 * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
6180 * config/i386/i386-modes.def (BND32, BND64): Remove.
6181 * config/i386/i386.c (dbx_register_map): Remove bound registers.
6182 (dbx64_register_map): Ditto.
6183 (svr4_dbx_register_map): Ditto.
6184 (indirect_thunk_bnd_needed): Remove.
6185 (indirect_thunks_bnd_used): Ditto.
6186 (indirect_return_bnd_needed): Ditto.
6187 (indirect_return_via_cx_bnd): Ditto.
6188 (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
6189 (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
6190 (output_indirect_thunk): Ditto. Remove need_prefix argument.
6191 (output_indirect_thunk_function): Remove handling of
6192 indirect_return_bnd_needed, indirect_return_via_cx_bnd,
6193 indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
6194 (ix86_save_reg): Remove handling of crtl->return_bnd.
6195 (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
6196 (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
6197 and UNSPEC_BNDLX_ADDR.
6198 (ix86_output_indirect_branch_via_reg): Remove handling of
6199 indirect_thunk_prefix_bnd.
6200 (ix86_output_indirect_branch_via_push): Ditto.
6201 (ix86_output_function_return): Ditto.
6202 (ix86_output_indirect_function_return): Ditto.
6203 (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
6204 * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
6205 (CALL_USED_REGISTERS): Ditto.
6206 (REG_ALLOC_ORDER): Update for removal of bound registers.
6207 (HI_REGISTER_NAMES): Ditto.
6208 * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
6209 (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
6210 (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
6211 (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
6212 (FIRST_PSEUDO_REG): Update.
6213 (BND): Remove mode iterator.
6214 * config/i386/predicates.md (bnd_mem_operator): Remove.
6216 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
6218 * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
6221 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
6223 * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
6224 the case in which the permute needs only a single element and
6225 repeats for every vector of the result. Extend that case to
6226 handle variable-length vectors.
6227 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
6229 2018-08-24 H.J. Lu <hongjiu.lu@intel.com>
6232 * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
6233 on DW_MACINFO_start_file for -gsplit-dwarf -g3.
6235 2018-08-24 Richard Biener <rguenther@suse.de>
6237 * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
6238 bb_flags_allocated members.
6239 (auto_flag): New RAII class for allocating flags.
6240 (auto_edge_flag): New RAII class for allocating edge flags.
6241 (auto_bb_flag): New RAII class for allocating bb flags.
6242 * cfgloop.c (verify_loop_structure): Allocate temporary edge
6244 * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
6245 in favor of temporarily allocated BB flag.
6246 * hsa-brig.c: Re-order includes.
6247 * hsa-dump.c: Likewise.
6248 * hsa-regalloc.c: Likewise.
6249 * print-rtl.c: Likewise.
6250 * profile-count.c: Likewise.
6252 2018-08-24 Segher Boessenkool <segher@kernel.crashing.org>
6255 * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
6258 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
6261 * wide-int-range.cc (wide_int_range_div): Do not ignore result
6262 from wide_int_range_multiplicative_op.
6264 2018-08-23 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6266 * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
6269 2018-08-23 Giuliano Belinassi <giuliano.belinassi@usp.br>
6271 * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
6274 2018-08-23 Alexander Monakov <amonakov@ispras.ru>
6276 * tree-scalar-evolution.c (final_value_replacement_loop): Dump
6277 full GENERIC expression used for replacement.
6279 2018-08-23 Aldy Hernandez <aldyh@redhat.com>
6281 * tree-vrp.c (abs_extent_range): Remove.
6282 (extract_range_into_wide_ints): Pass wide ints by reference.
6283 (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
6284 Pass wide ints by reference in all calls to
6285 extract_range_into_wide_ints.
6286 * wide-int-range.cc (wide_int_range_div): New.
6287 * wide-int-range.h (wide_int_range_div): New.
6288 (wide_int_range_includes_zero_p): New.
6289 (wide_int_range_zero_p): New.
6291 2018-08-23 Matthew Malcomson <matthew.malcomson@arm.com>
6293 * config/aarch64/aarch64.md (arches): New enum.
6294 (arch): New enum attr.
6295 (arch_enabled): New attr.
6296 (enabled): Now uses arch_enabled only.
6297 (simd, sve, fp16): Removed attribute.
6298 (fp): Attr now defined in terms of 'arch'.
6299 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
6300 *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
6301 <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
6302 <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
6303 attributes into 'arch'.
6304 (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
6305 subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
6306 *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
6307 *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
6308 *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
6309 'simd' attribute into 'arch'.
6310 (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
6311 store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
6312 Convert use of 'fp' attribute to 'arch'.
6313 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
6314 move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
6316 (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
6317 (different modes) Merge 'fp' and 'simd' into 'arch'.
6318 (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
6321 2018-08-23 Segher Boessenkool <segher@kernel.crashing.org>
6323 PR rtl-optimization/87026
6324 * expmed.c (canonicalize_comparison): If we can no longer create
6325 pseudoregisters, don't.
6327 2018-08-23 Richard Earnshaw <rearnsha@arm.com>
6330 * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
6332 * config/arm/arm.c (speculation_barrier_libfunc): New static
6334 (arm_init_libfuncs): Initialize it.
6335 (arm_emit_speculation_barrier): New function.
6336 * config/arm/arm.md (speculation_barrier): Call
6337 arm_emit_speculation_barrier for architectures that do not have
6339 (speculation_barrier_insn): Only match on Armv7 or later.
6341 2018-08-23 Richard Biener <rguenther@suse.de>
6344 * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
6347 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
6349 * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
6350 of single-vector TBLs.
6351 (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
6354 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
6357 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
6358 aarch64_evpc_tbl guard.
6360 2018-08-22 Bernd Edlinger <bernd.edlinger@hotmail.de>
6362 * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
6365 2018-08-22 Martin Sebor <msebor@redhat.com>
6368 * tree-pretty-print.c (pretty_print_string): Add argument.
6369 (dump_generic_node): Call to pretty_print_string with string size.
6371 2018-08-22 Segher Boessenkool <segher@kernel.crashing.org>
6373 PR rtl-optimization/86771
6374 * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
6375 of two SETs into those two SETs, one to be placed at i2, if that SETs
6376 destination is modified between i2 and i3.
6378 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
6380 PR tree-optimization/86725
6381 * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
6383 (vect_analyze_scalar_cycles_1): Check it.
6385 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
6387 PR tree-optimization/86725
6388 * tree-vect-loop.c (vect_is_simple_reduction): When treating
6389 an outer loop phi as a double reduction, make sure that the
6390 single user of the phi result is an inner loop phi.
6392 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
6394 * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
6395 grouped stores with gaps to a strided group.
6397 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
6399 * tree-vect-stmts.c (get_group_load_store_type)
6400 (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
6401 first statement in a group.
6403 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
6405 * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
6406 the sequence used in gcc/gcc.c.
6408 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
6411 * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
6414 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
6416 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
6417 Darwin10-specific unwinder-shim.
6418 * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
6419 * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
6420 New to cater for Darwin10 Rosetta.
6422 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
6424 * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
6425 specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
6427 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
6432 * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
6433 (dwarf2out_switch_text_section): Generate a local label for the
6434 second function sub-section and apply it as the second FDE start
6436 * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
6437 second sub-section start.
6439 2018-08-22 Richard Biener <rguenther@suse.de>
6441 PR tree-optimization/86988
6442 * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
6444 2018-08-22 Richard Biener <rguenther@suse.de>
6446 PR tree-optimization/86945
6447 * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
6449 2018-08-22 Alexandre Oliva <oliva@adacore.com>
6451 * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
6452 a comment about how uses of r2 for .sdata2 come about.
6454 2018-08-22 Alexandre Oliva <aoliva@redhat.com>
6456 * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
6458 2018-08-21 Marek Polacek <polacek@redhat.com>
6460 PR c++/86981, Implement -Wpessimizing-move.
6461 * doc/invoke.texi: Document -Wpessimizing-move.
6463 2018-08-21 Jan Hubicka <jh@suse.cz>
6465 * tree.c (find_decls_types_r): Do not check for redundant typedefs.
6466 * tree.h (is_redundant_typedef): Remove.
6467 * dwarf2out.c (is_redundant_typedef): Turn into static function.
6469 2018-08-21 Jan Hubicka <jh@suse.cz>
6471 * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
6474 2018-08-21 Tamar Christina <tamar.christina@arm.com>
6476 * expmed.c (extract_low_bits): Reject invalid subregs early.
6478 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
6481 * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
6484 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
6486 * config/vxworks.h: Guard vxworks_asm_out_constructor and
6487 vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
6488 * config/vxworks.c: Likewise.
6490 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
6492 * config/vxworks.c: Set targetm.have_ctors_dtors
6493 if HAVE_INITFINI_ARRAY_SUPPORT.
6494 * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
6495 if HAVE_INITFINI_ARRAY_SUPPORT.
6497 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
6499 * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
6500 default search path for VxWorks < 7.
6502 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
6504 * gimple-ssa-sprintf.c (decl_constant_value): Remove.
6505 (get_format_string): Refer to c_getstr.
6507 2018-08-21 Tom de Vries <tdevries@suse.de>
6509 * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
6510 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
6511 (debuginfo_early_stop): Declare.
6512 * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
6513 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
6514 (debuginfo_early_stop): New function.
6515 (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
6516 and debuginfo_early_stop.
6517 * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
6518 * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
6519 (general_init): Call debuginfo_early_init.
6520 (finalize): Call debuginfo_fini.
6521 (do_compile): Call debuginfo_init.
6522 * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
6524 (@item -fdump-debug, @item -fdump-earlydebug): Add.
6526 2018-08-21 Tom de Vries <tdevries@suse.de>
6528 * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
6529 flag_dump_noaddr and flag_dump_unnumbered.
6531 2018-08-21 Aldy Hernandez <aldyh@redhat.com>
6533 * wide-int-range.cc (wide_int_range_abs): New.
6534 (wide_int_range_order_set): Rename from wide_int_range_min_max.
6535 * wide-int-range.h (wide_int_range_abs): New.
6536 (wide_int_range_min_max): New.
6537 * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
6538 case to call wide_int_range_abs.
6539 Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
6540 (extract_range_from_abs_expr): Delete.
6542 2018-08-20 Michael Meissner <meissner@linux.ibm.com>
6545 * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
6546 from 'Y' to 'YZ' to enable the LWAX instruction to be generated
6549 2018-08-20 Nathan Sidwell <nathan@acm.org>
6550 Jeff Law <law@redhat.com>
6552 * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
6553 * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
6555 2018-08-20 David Malcolm <dmalcolm@redhat.com>
6558 * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
6559 (decl_attributes): Likewise.
6560 * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
6562 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
6563 * diagnostic-core.h (class auto_diagnostic_group): New class.
6564 * diagnostic.c (diagnostic_initialize): Initialize the new fields.
6565 (diagnostic_report_diagnostic): Handle the first diagnostics within
6567 (emit_diagnostic): Add auto_diagnostic_group instance.
6569 (inform_n): Likewise.
6570 (warning): Likewise.
6571 (warning_at): Likewise.
6572 (warning_n): Likewise.
6573 (pedwarn): Likewise.
6574 (permerror): Likewise.
6576 (error_n): Likewise.
6577 (error_at): Likewise.
6579 (fatal_error): Likewise.
6580 (internal_error): Likewise.
6581 (internal_error_no_backtrace): Likewise.
6582 (auto_diagnostic_group::auto_diagnostic_group): New ctor.
6583 (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
6584 * diagnostic.h (struct diagnostic_context): Add fields
6585 "diagnostic_group_nesting_depth",
6586 "diagnostic_group_emission_count", "begin_group_cb",
6588 * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
6589 Add auto_diagnostic_group instance(s).
6590 (find_explicit_erroneous_behavior): Likewise.
6591 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
6592 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
6593 * gimplify.c (warn_implicit_fallthrough_r): Likewise.
6594 (gimplify_va_arg_expr): Likewise.
6595 * hsa-gen.c (HSA_SORRY_ATV): Likewise.
6596 (HSA_SORRY_AT): Likewise.
6597 * ipa-devirt.c (compare_virtual_tables): Likewise.
6598 (warn_odr): Likewise.
6599 * multiple_target.c (expand_target_clones): Likewise.
6600 * opts-common.c (cmdline_handle_error): Likewise.
6601 * reginfo.c (globalize_reg): Likewise.
6602 * substring-locations.c (format_warning_n_va): Likewise.
6603 * tree-inline.c (expand_call_inline): Likewise.
6604 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
6605 * tree-ssa-loop-niter.c
6606 (do_warn_aggressive_loop_optimizations): Likewise.
6607 * tree-ssa-uninit.c (warn_uninit): Likewise.
6608 * tree.c (warn_deprecated_use): Likewise.
6610 2018-08-20 H.J. Lu <hongjiu.lu@intel.com>
6613 * config/i386/i386.md (eh_return): Always update EH return
6614 address in word_mode.
6616 2018-08-20 Chung-Lin Tang <cltang@codesourcery.com>
6618 * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
6619 TARGET_SPLIT_COMPLEX_ARG is defined.
6621 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
6623 * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
6625 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
6628 * expr.c (expand_assignment): Assert that bitpos is positive.
6629 (store_field): Likewise
6630 (expand_expr_real_1): Make sure that bitpos is positive.
6631 * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
6634 2018-08-20 Nathan Sidwell <nathan@acm.org>
6636 * Makefile.in (CPP_ID_DATA_H): Delete.
6637 (CPP_INTERNAL_H): Don't add it.
6638 (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
6639 * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
6641 2018-08-20 Richard Biener <rguenther@suse.de>
6643 PR tree-optimization/78655
6644 * tree-vrp.c (extract_range_from_binary_expr_1): Make
6645 pointer + offset nonnull if either operand is nonnull work.
6647 2018-08-20 Tom de Vries <tdevries@suse.de>
6649 * dwarf2out.c (add_scalar_info): Don't add reference to existing die
6650 unless the referenced die describes the added property using
6651 DW_AT_location or DW_AT_const_value. Fall back to exprloc case.
6652 Otherwise, add a DW_AT_location to the referenced die.
6654 2018-08-19 Uros Bizjak <ubizjak@gmail.com>
6657 * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
6658 register_operand when calling ix86_set_reg_reg_cost.
6659 [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
6660 Set *total to 0 for operands that satisfy x86_64_immediate_operand
6661 predicate and to 1 otherwise.
6663 2018-08-18 Iain Sandoe <iain@sandoe.co.uk>
6665 * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
6666 emit a diagnostic that it is not supported and reset the option.
6667 * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
6668 supported and consume the option. (ASM_FINAL_SPEC): New.
6670 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
6672 * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
6675 2018-08-17 Sandra Loosemore <sandra@codesourcery.com>
6677 C-SKY port: Documentation
6679 * doc/extend.texi (C-SKY Function Attributes): New section.
6680 * doc/invoke.texi (Option Summary): Add C-SKY options.
6681 (C-SKY Options): New section.
6682 * doc/md.texi (Machine Constraints): Document C-SKY constraints.
6684 2018-08-17 Jojo <jijie_rong@c-sky.com>
6685 Huibin Wang <huibin_wang@c-sky.com>
6686 Sandra Loosemore <sandra@codesourcery.com>
6687 Chung-Lin Tang <cltang@codesourcery.com>
6689 C-SKY port: Backend implementation
6691 * config/csky/*: New.
6692 * common/config/csky/*: New.
6694 2018-08-17 Jojo <jijie_rong@c-sky.com>
6695 Huibin Wang <huibin_wang@c-sky.com>
6696 Sandra Loosemore <sandra@codesourcery.com>
6697 Chung-Lin Tang <cltang@codesourcery.com>
6698 Andrew Jenner <andrew@codesourcery.com>
6700 C-SKY port: Configury
6702 * config.gcc (csky-*-*): New.
6703 * configure.ac: Add csky to targets for dwarf2 debug_line support.
6704 * configure: Regenerated.
6706 2018-08-17 David Malcolm <dmalcolm@redhat.com>
6708 * dump-context.h: Include "dumpfile.h".
6709 (dump_context::dump_printf_va): Convert final param from va_list
6710 to va_list *. Convert from ATTRIBUTE_PRINTF to
6711 ATTRIBUTE_GCC_DUMP_PRINTF.
6712 (dump_context::dump_printf_loc_va): Likewise.
6713 * dumpfile.c: Include "stringpool.h".
6714 (make_item_for_dump_printf_va): Delete.
6715 (make_item_for_dump_printf): Delete.
6716 (class dump_pretty_printer): New class.
6717 (dump_pretty_printer::dump_pretty_printer): New ctor.
6718 (dump_pretty_printer::emit_items): New member function.
6719 (dump_pretty_printer::emit_any_pending_textual_chunks): New member
6721 (dump_pretty_printer::emit_item): New member function.
6722 (dump_pretty_printer::stash_item): New member function.
6723 (dump_pretty_printer::format_decoder_cb): New member function.
6724 (dump_pretty_printer::decode_format): New member function.
6725 (dump_context::dump_printf_va): Reimplement in terms of
6726 dump_pretty_printer.
6727 (dump_context::dump_printf_loc_va): Convert final param from va_list
6729 (dump_context::begin_scope): Reimplement call to
6730 make_item_for_dump_printf.
6731 (dump_printf): Update for change to dump_printf_va.
6732 (dump_printf_loc): Likewise.
6733 (selftest::test_capture_of_dump_calls): Convert "stmt" from
6734 greturn * to gimple *. Add a test_decl. Add tests of dump_printf
6735 with %T, %E, and %G.
6736 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
6737 (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
6738 ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
6739 (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
6740 ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
6741 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
6742 use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
6743 within a dump_printf_loc call to "%wu".
6744 (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
6745 converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd". Add a
6746 missing space after "=".
6747 * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
6748 call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
6749 * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
6750 convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
6751 * tree-vectorizer.c (try_vectorize_loop_1): Likewise. Remove
6752 duplicate "vectorized" from message.
6754 2018-08-17 Szabolcs Nagy <szabolcs.nagy@arm.com>
6756 * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
6757 polyNxK_t element's TYPE_STRING_FLAG.
6759 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
6761 * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
6762 (they were unnamed before). Fix comments.
6764 2018-08-17 Nathan Sidwell <nathan@acm.org>
6766 * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
6768 2018-08-17 Richard Biener <rguenther@suse.de>
6770 PR tree-optimization/86841
6771 * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
6773 2018-08-17 Martin Liska <mliska@suse.cz>
6775 * common.opt: Remove Warn, Init and Report for options with
6776 Ignore/Deprecated flag. Warning is done automatically for
6778 * config/i386/i386.opt: Likewise.
6779 * config/ia64/ia64.opt: Likewise.
6780 * config/rs6000/rs6000.opt: Likewise.
6781 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
6782 Remove usage of flag_check_pointer_bounds.
6783 * lto-wrapper.c (merge_and_complain): Do not handle
6784 OPT_fcheck_pointer_bounds.
6785 (append_compiler_options): Likewise.
6786 * opt-functions.awk: Do not handle Deprecated.
6787 * optc-gen.awk: Check that Var, Report and Init are not
6788 used for an option with Ignore/Deprecated flag.
6789 * opts-common.c (decode_cmdline_option): Do not report
6791 (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
6793 * opts.h (struct cl_option): Remove cl_deprecated flag.
6794 (CL_ERR_DEPRECATED): Remove error enum value.
6796 2018-08-17 Richard Biener <rguenther@suse.de>
6799 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
6800 across a va-arg-pack using call adjust its return value accordingly.
6802 2018-08-16 Martin Sebor <msebor@redhat.com>
6804 PR tree-optimization/86853
6805 * gimple-ssa-sprintf.c (struct format_result): Rename member.
6806 (struct fmtresult): Add member and initialize it in ctors.
6807 (format_character): Handle %C. Extend range to NUL. Set MAYFAIL.
6808 (format_string): Handle %S the same as %ls. Set MAYFAIL.
6809 (format_directive): Set POSUNDER4K when MAYFAIL is set.
6810 (parse_directive): Handle %C same as %c.
6811 (sprintf_dom_walker::compute_format_length): Adjust.
6812 (is_call_safe): Adjust.
6814 2018-08-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
6816 * builtins.c (c_strlen): Add new parameter eltsize. Use it
6817 for determining how to count the elements.
6818 * builtins.h (c_strlen): Adjust prototype.
6819 * expr.c (string_constant): Add new parameter mem_size.
6820 Set *mem_size appropriately.
6821 * expr.h (string_constant): Adjust protoype.
6822 * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
6823 * gimple-fold.h (get_range_strlen): Adjust prototype.
6824 * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
6825 (format_string): Call get_string_length with eltsize.
6827 2018-08-16 David Malcolm <dmalcolm@redhat.com>
6829 * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
6830 to emit the span, rather than setting it as the prefix.
6832 2018-08-16 David Malcolm <dmalcolm@redhat.com>
6834 * diagnostic-show-locus.c (layout::start_annotation_line): Add
6835 "margin_char" parameter, defaulting to space. Use it in place
6836 of pp_space for the initial part of the margin.
6837 (layout::print_leading_fixits): Use '+' when filling the margin
6838 of line-insertion fix-it hints.
6840 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
6842 * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
6845 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
6847 * config/rs6000/altivec.md: Don't set length attribute to the default
6849 * config/rs6000/darwin.md: Ditto.
6850 * config/rs6000/dfp.md: Ditto.
6851 * config/rs6000/htm.md: Ditto.
6852 * config/rs6000/rs6000.md: Ditto.
6853 * config/rs6000/sync.md: Ditto.
6854 * config/rs6000/vsx.md: Ditto.
6856 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
6858 * config/rs6000/altivec.md: Don't set length attribute to the default
6859 value, for branch instructions.
6860 * config/rs6000/darwin.md: Ditto.
6861 * config/rs6000/rs6000.md: Ditto.
6863 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
6865 * config/rs6000/rs6000.md (length): Always define as const_int 4.
6866 (unnamed conditional branch define_insn): Set length to 4 or 8
6867 depending on offset.
6868 (<bd>_<mode>): Similar, for alternative 0.
6869 (<bd>tf_<mode>): Ditto.
6871 2018-08-16 Tamar Christina <tamar.christina@arm.com>
6873 * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
6875 2018-08-16 Matthew Malcomson <matthew.malcomson@arm.com>
6877 * doc/rtl.texi: Replace old RTX class names with new names.
6880 2018-08-16 Vlad Lazar <vlad.lazar@arm.com>
6882 * expmed.h (canonicalize_comparison): New declaration.
6883 * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
6884 * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
6885 * optabs.c (prepare_cmp_insn): Likewise.
6886 * rtl.h (unsigned_condition_p): New function which checks if a
6887 comparison operator is unsigned.
6889 2018-08-16 Nathan Sidwell <nathan@acm.org>
6891 * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
6892 * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
6894 2018-08-16 Tamar Christina <tamar.christina@arm.com>
6897 * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
6898 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
6899 (mov<mov>): ..this and enable unconditionally.
6901 2018-08-16 Tamar Christina <tamar.christina@arm.com>
6903 * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
6905 2018-08-16 Sam Tebbs <sam.tebbs@arm.com>
6907 * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
6908 (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
6909 "Common" with "Target".
6911 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
6913 * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
6914 * doc/invoke.texi (mmitigate-rop): Remove.
6915 * config/i386/i386.c: Do not include "regrename.h".
6916 (ix86_rop_should_change_byte_p, reg_encoded_number)
6917 (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
6919 (ix86_reorg): Remove call to ix86_mitigate_rop.
6920 * config/i386/i386.md (attr "modrm_class"): Remove.
6921 (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
6922 (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
6923 (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
6925 2018-08-15 Will Schmidt <will_schmidt@vnet.ibm.com>
6927 * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
6928 allow folding of mergeh() and mergel() for the float and double types.
6929 (fold_mergehl_helper): Rework to handle building a permute tree
6932 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
6934 * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
6937 2018-08-15 David Malcolm <dmalcolm@redhat.com>
6939 * common.opt (fdiagnostics-show-labels): New option.
6940 * diagnostic-show-locus.c (class layout_range): Add field
6942 (class layout): Add field "m_show_labels_p".
6943 (layout_range::layout_range): Add param "label" and use it to
6945 (make_range): Pass in NULL for new "label" param of layout_range's
6947 (layout::layout): Initialize m_show_labels_p.
6948 (layout::maybe_add_location_range): Pass in loc_range->m_label
6949 when constructing layout_range instances.
6950 (struct line_label): New struct.
6951 (layout::print_any_labels): New member function.
6952 (layout::print_line): Call it if label-printing is enabled.
6953 (selftest::test_one_liner_labels): New test.
6954 (selftest::test_diagnostic_show_locus_one_liner): Call it.
6955 * diagnostic.c (diagnostic_initialize): Initialize
6956 context->show_labels_p.
6957 * diagnostic.h (struct diagnostic_context): Add field
6959 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
6960 -fno-diagnostics-show-labels.
6961 * dwarf2out.c (gen_producer_string): Add
6962 OPT_fdiagnostics_show_labels to the ignored options.
6963 * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
6965 (gcc_rich_location::maybe_add_expr): Likewise.
6966 * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
6967 label" param, defaulting to NULL.
6968 (gcc_rich_location::add_expr): Add "label" param.
6969 (gcc_rich_location::maybe_add_expr): Likewise.
6970 (class text_range_label): New class.
6971 (class range_label_for_type_mismatch): New class.
6972 * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
6973 of format_warning_va.
6974 (fmtwarn_n): Likewise for new params of format_warning_n_va.
6975 * lto-wrapper.c (merge_and_complain): Add
6976 OPT_fdiagnostics_show_labels to the "pick one setting" options.
6977 (append_compiler_options): Likewise to the dropped options.
6978 (append_diag_options): Likewise to the passed-on options.
6979 * opts.c (common_handle_option): Handle the new option.
6980 * selftest-diagnostic.c
6981 (test_diagnostic_context::test_diagnostic_context): Enable
6983 * substring-locations.c: Include "gcc-rich-location.h".
6984 (format_warning_n_va): Add "fmt_label" and "param_label" params
6985 and use them as appropriate.
6986 (format_warning_va): Add "fmt_label" and "param_label" params,
6987 passing them on to format_warning_n_va.
6988 (format_warning_at_substring): Likewise.
6989 (format_warning_at_substring_n): Likewise.
6990 * substring-locations.h (format_warning_va): Add "fmt_label" and
6991 "param_label" params.
6992 (format_warning_n_va): Likewise.
6993 (format_warning_at_substring): Likewise.
6994 (format_warning_at_substring_n): Likewise.
6995 * toplev.c (general_init): Initialize global_dc->show_labels_p.
6997 2018-08-15 Qing Zhao <qing.zhao@oracle.com>
7000 * builtins.c (expand_builtin_memcmp): Do not expand the call
7001 when overflow is detected.
7003 2018-08-15 Martin Sebor <msebor@redhat.com>
7005 PR tree-optimization/71625
7006 * config/aarch64/aarch64-builtins.c
7007 (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
7009 2018-08-15 Ilya Leoshkevich <iii@linux.ibm.com>
7011 * config/s390/s390.c (s390_reorg): Remove loop.
7013 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
7016 (darwin_function_switched_text_sections): Delete.
7017 * gcc/config/darwin.h
7018 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
7020 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
7023 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
7024 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
7025 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
7027 2018-08-15 Martin Liska <mliska@suse.cz>
7029 PR tree-optimization/86925
7030 * predict.c (expr_expected_value_1): When taking
7031 later predictor, assign also probability.
7032 Use fold_build2_initializer_loc in order to fold
7033 the expression in -frounding-math.
7035 2018-08-14 Allan Sandfeld Jensen <allan.jensen@qt.io>
7037 * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
7039 (expand_vec_perm_1): Try the new method.
7041 2018-08-14 Ilya Leoshkevich <iii@linux.ibm.com>
7044 * lra-lives.c (remove_some_program_points_and_update_live_ranges):
7045 Check whether lra_live_max_point is 0 before dividing.
7047 2018-08-14 Martin Sebor <msebor@redhat.com>
7049 PR tree-optimization/86650
7050 * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
7051 (vrp_prop::check_mem_ref): Same.
7053 2018-08-13 Liu Hao <lh_mouse@126.com>
7055 * pretty-print.c (eat_esc_sequence): Swap the foreground and
7056 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
7057 and clear it thereafter, as it only works for DBCS.
7059 2018-08-13 Liu Hao <lh_mouse@126.com>
7061 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
7062 handle returned by _get_osf_handle().
7064 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
7066 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
7067 for folding vec_perm.
7069 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
7071 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
7072 Add support for gimple-folding of vec_pack() and vec_unpack()
7075 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
7077 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
7078 vec_xst variants to the list.
7079 (rs6000_gimple_fold_builtin): Add support for folding unaligned
7080 vector loads and stores.
7082 2018-08-13 David Edelsohn <dje.gcc@gmail.com>
7084 * config.gcc (rs6000-ibm-aix4.x): Delete.
7085 (rs6000-ibm-aix5.1): Delete.
7086 (rs6000-ibm-aix5.2): Delete.
7087 (rs6000-ibm-aix5.3): Delete.
7088 * config/rs6000/aix43.h: Delete.
7089 * config/rs6000/aix51.h: Delete.
7090 * config/rs6000/aix52.h: Delete.
7091 * config/rs6000/t-aix43: Delete.
7093 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
7095 * config/s390/s390.c (s390_decompose_constant_pool_ref):
7097 (s390_decompose_address): Factor out constant pool ref
7100 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
7102 * config/nds32/nds32-predicates.c
7103 (nds32_can_use_bclr_p): Change return type as bool.
7104 (nds32_can_use_bset_p): Ditto.
7105 (nds32_can_use_btgl_p): Ditto.
7106 (nds32_can_use_bitci_p): Ditto.
7107 * config/nds32/nds32-protos.h
7108 (nds32_can_use_bclr_p): Change declaration.
7109 (nds32_can_use_bset_p): Ditto.
7110 (nds32_can_use_btgl_p): Ditto.
7111 (nds32_can_use_bitci_p): Ditto.
7113 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
7115 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
7116 Support -msched-prolog-epilog option.
7117 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
7119 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
7121 * common/config/nds32/nds32-common.c
7122 (nds32_option_optimization_table): Enalbe -malways-align.
7124 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
7126 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
7128 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
7129 OPT_misr_secure_ case.
7130 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
7131 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
7132 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
7134 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
7135 (nds32_isr_info): New field security_level.
7136 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
7137 * config/nds32/nds32.md (return_internal): Consider critical attribute.
7138 * config/nds32/nds32.opt (misr-secure): New option.
7139 * config/nds32/nds32_init.inc: New file.
7140 * config/nds32/nds32_isr.h: New file.
7142 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
7144 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
7145 Update comment for atomic instructions.
7146 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
7147 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
7149 (atomic_loaddi): Revise fence expansion to only emit fence prior to
7150 load for __ATOMIC_SEQ_CST model.
7151 (atomic_loaddi_1): Remove float register target.
7152 (atomic_storedi): Handle CONST_INT values.
7153 (atomic_storedi_1): Remove float register source. Add special case
7155 (memory_barrier): New expander and insn.
7157 2018-08-11 Jakub Jelinek <jakub@redhat.com>
7159 PR tree-optimization/86835
7160 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
7161 new_stmt after def_gsi, make sure to insert new_square_stmt after
7162 that stmt, not 2 stmts before it.
7164 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
7167 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
7170 2018-08-10 Martin Liska <mliska@suse.cz>
7173 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
7175 * builtins.c (expand_builtin_expect_with_probability):
7177 (expand_builtin_expect_with_probability): New function.
7178 (build_builtin_expect_predicate): Add new argumnet probability
7179 for BUILT_IN_EXPECT_WITH_PROBABILITY.
7180 (fold_builtin_expect):
7183 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
7184 * builtins.h (fold_builtin_expect): Set new argument.
7185 * doc/extend.texi: Document __builtin_expect_with_probability.
7186 * doc/invoke.texi: Likewise.
7187 * gimple-fold.c (gimple_fold_call): Pass new argument.
7188 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
7189 also BUILT_IN_EXPECT_WITH_PROBABILITY.
7190 * predict.c (get_predictor_value): New function.
7191 (expr_expected_value): Add new argument probability. Assume
7192 that predictor and probability are always non-null.
7193 (expr_expected_value_1): Likewise. For __builtin_expect and
7194 __builtin_expect_with_probability set probability. Handle
7195 combination in binary expressions.
7196 (tree_predict_by_opcode): Simplify code by simply calling
7197 get_predictor_value.
7198 (pass_strip_predict_hints::execute): Add handling of
7199 BUILT_IN_EXPECT_WITH_PROBABILITY.
7200 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
7202 * tree.h (DECL_BUILT_IN_P): New function.
7204 2018-08-10 Martin Liska <mliska@suse.cz>
7206 PR tree-optimization/85799
7207 * passes.def: Add argument for pass_strip_predict_hints.
7208 * predict.c (class pass_strip_predict_hints): Add new argument
7210 (strip_predictor_early): New function.
7211 (pass_strip_predict_hints::execute): Call the function to
7213 (strip_predict_hints): New function.
7214 * predict.def: Fix comment.
7216 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
7218 * Makefile.in: Clarify which tm.texi to copy over to assert the
7219 right to grant a GFDL license for all.
7221 2018-08-09 Jeff Law <law@redhat.com>
7223 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
7226 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
7228 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
7231 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
7233 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
7234 reductions for variable-length vectors.
7236 2018-08-09 David Malcolm <dmalcolm@redhat.com>
7239 * common.opt (fdiagnostics-show-line-numbers): New option.
7240 * diagnostic-show-locus.c (class layout): Add fields
7241 "m_show_line_numbers_p" and "m_linenum_width";
7242 (num_digits): New function.
7243 (test_num_digits): New function.
7244 (layout::layout): Initialize new fields. Update m_x_offset
7245 logic to handle any left margin.
7246 (layout::print_source_line): Print line number when requested.
7247 (layout::start_annotation_line): New member function.
7248 (layout::print_annotation_line): Call it.
7249 (layout::print_leading_fixits): Likewise.
7250 (layout::print_trailing_fixits): Likewise. Update calls to
7251 move_to_column for new parameter.
7252 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
7253 it to potentially call start_annotation_line.
7254 (layout::show_ruler): Call start_annotation_line.
7255 (selftest::test_line_numbers_multiline_range): New selftest.
7256 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
7257 and selftest::test_line_numbers_multiline_range.
7258 * diagnostic.c (diagnostic_initialize): Initialize
7259 show_line_numbers_p.
7260 * diagnostic.h (struct diagnostic_context): Add field
7261 "show_line_numbers_p".
7262 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
7263 -fno-diagnostics-show-line-numbers.
7264 * dwarf2out.c (gen_producer_string): Add
7265 OPT_fdiagnostics_show_line_numbers to the ignored options.
7266 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
7267 one setting" options.
7268 (append_compiler_options): Likewise to the dropped options.
7269 (append_diag_options): Likewise to the passed-on options.
7270 * opts.c (common_handle_option): Handle the new option.
7271 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
7273 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
7275 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
7276 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
7277 third argument of type "const signed char" to descriptions of
7278 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
7279 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
7280 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
7281 __builtin_bcdsub_ov functions.
7283 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
7285 PR tree-optimization/86858
7286 * tree-vect-loop.c (vect_is_simple_reduction): Restore
7287 flow_bb_inside_loop_p calls.
7289 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
7291 PR tree-optimization/86871
7292 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
7293 instead of gimple_assign_lhs.
7295 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
7298 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
7299 register constraint to operand 0.
7300 (add<mode>3_carryinC): Likewise.
7301 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
7303 2018-08-09 Martin Liska <mliska@suse.cz>
7306 * common.opt: Remove extra line.
7308 2018-08-09 Martin Liska <mliska@suse.cz>
7310 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
7311 at the end of a line, make first letter capital and end up
7312 a sentence with a dot.
7313 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
7314 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
7315 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
7316 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
7317 (PARAM_MAX_ISL_OPERATIONS): Likewise.
7318 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
7319 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
7320 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
7321 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
7322 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
7323 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
7324 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
7325 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
7326 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
7327 (PARAM_TREE_REASSOC_WIDTH): Likewise.
7328 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
7329 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
7330 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
7332 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
7335 * config/s390/s390.c (s390_option_override_internal): Reduce the
7336 stack-clash-protection-probe-interval param if it would be too big
7339 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
7342 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
7343 * config/m68k/m68k.c (handle_move_double): Don't call
7344 m68k_final_prescan_insn.
7345 (m68k_adjust_decorated_operand): Renamed from
7346 m68k_final_prescan_insn, remove first and third operand and
7348 (print_operand): Call it.
7349 (print_operand_address): Call it.
7351 2018-08-08 Nathan Sidwell <nathan@acm.org>
7353 * diagnostic.c (diagnostic_report_current_module): Use
7354 linemap_included_from & linemap_included_from_linemap.
7356 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
7358 * config/aarch64/aarch64-cores.def: Add phecda core.
7359 * config/aarch64/aarch64-tune.md: Regenerate.
7360 * doc/invoke.texi: Add phecda core.
7362 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
7365 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
7367 * config/s390/s390.md ("movti"): Add more alternatives for
7368 constant to GPR copies.
7370 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
7372 * config/s390/s390.c: Fix whitespace damage throughout the file.
7373 * config/s390/s390.h: Likewise.
7374 * config/s390/tpf.h: Likewise.
7376 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
7378 * config/s390/s390.c (s390_loadrelative_operand_p):
7379 Remove TARGET_CPU_ZARCH usages.
7380 (s390_rtx_costs): Likewise.
7381 (s390_legitimate_constant_p): Likewise.
7382 (s390_cannot_force_const_mem): Likewise.
7383 (legitimate_reload_constant_p): Likewise.
7384 (s390_preferred_reload_class): Likewise.
7385 (legitimize_pic_address): Likewise.
7386 (legitimize_tls_address): Likewise.
7387 (s390_split_branches): Removed.
7388 (s390_add_execute): Removed.
7389 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
7390 (s390_mainpool_start): Likewise.
7391 (s390_mainpool_finish): Likewise.
7392 (s390_mainpool_cancel): Removed.
7393 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
7394 (s390_chunkify_cancel): Likewise.
7395 (s390_return_addr_rtx): Likewise.
7396 (s390_register_info): Remove split_branches_pending_p uages.
7397 (s390_optimize_register_info): Likewise.
7398 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
7399 split_branches_pending_p usages.
7400 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
7401 (s390_load_got): Likewise.
7402 (s390_expand_split_stack_prologue): Likewise.
7403 (output_asm_nops): Likewise.
7404 (s390_function_profiler): Likewise.
7405 (s390_emit_call): Likewise.
7406 (s390_conditional_register_usage): Likewise.
7407 (s390_optimize_prologue): Likewise.
7408 (s390_reorg): Remove TARGET_CPU_ZARCH and
7409 split_branches_pending_p usages.
7410 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
7412 (s390_output_indirect_thunk_function): Likewise.
7413 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
7414 (TARGET_CPU_ZARCH_P): Removed.
7415 (struct machine_function): Remove split_branches_pending_p.
7416 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
7418 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
7420 * common/config/s390/s390-common.c (processor_flags_table):
7422 * config.gcc: Remove with_arch/with_tune support.
7423 * config/s390/2064.md: Remove cpu attribute comparisons.
7424 * config/s390/driver-native.c (s390_host_detect_local_cpu):
7426 * config/s390/linux.h (ASM_SPEC):
7427 Remove -march support.
7428 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
7429 Use a table to get an arch level.
7430 * config/s390/s390-opts.h (enum processor_type):
7432 * config/s390/s390.c
7433 (processor_table): Remove entries, add arch_level values.
7434 (s390_issue_rate): Remove cases.
7435 (s390_option_override): Adjust
7436 s390_option_override_internal() call.
7437 (s390_option_override_internal): Remove deprecation warning.
7438 (s390_valid_target_attribute_tree): Adjust
7439 s390_option_override_internal() call.
7440 * config/s390/s390.h (struct s390_processor):
7441 Share with s390-c.c, add arch_level field.
7442 * config/s390/s390.md:
7443 Remove occurrences in cpu attribute.
7444 * config/s390/s390.opt: Remove -march/-mtune support.
7445 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
7446 * doc/invoke.texi: Remove deprecation warning.
7448 2018-08-08 Luis Machado <luis.machado@linaro.org>
7450 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
7452 (qdf24xx_tunings): Set vector cost structure to
7453 qdf24xx_vector_cost.
7455 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
7456 <register_sextend>: Set to 3.
7458 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
7461 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
7462 * config/aarch64/aarch64-simd.md
7463 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
7464 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
7465 (aarch64_frecpx<mode>): ...this new pattern.
7466 * config/aarch64/aarch64-simd-builtins.def: Remove comment
7467 about aarch64_frecp<FRECP:frecp_suffix><mode>.
7469 2018-08-07 Martin Liska <mliska@suse.cz>
7472 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
7473 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
7474 * predict.def (PRED_MALLOC_NONNULL): New predictor.
7475 * doc/extend.texi: Document that malloc attribute adds
7478 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
7481 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7482 Define to speculation_safe_value_not_needed.
7484 2018-08-06 Jeff Law <law@redhat.com>
7486 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
7487 the vr_values instance to cprop_into_stmt.
7488 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
7489 (cprop_operand): Also query EVRP to determine if OP is a constant.
7491 2018-08-06 Nathan Sidwell <nathan@acm.org>
7493 * diagnostic.c (diagnostic_report_current_module): Reroll
7494 included-at loop. Translate text.
7496 2018-08-06 David Malcolm <dmalcolm@redhat.com>
7498 * function-tests.c (selftest::test_expansion_to_rtl): Call
7499 free_after_compilation.
7501 2018-08-06 Alan Hayward <alan.hayward@arm.com>
7503 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
7505 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
7507 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
7508 loops with memory block operations from getting unrolled.
7510 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
7513 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7514 Define to speculation_safe_value_not_needed.
7516 2018-08-06 Jeff Law <law@redhat.com>
7518 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
7521 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
7524 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
7525 with all enabled __intN types.
7527 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
7529 2018-08-06 Alan Hayward <alan.hayward@arm.com>
7531 * alias.c (record_set): Check for clobber high.
7532 * cfgexpand.c (expand_gimple_stmt): Likewise.
7533 * combine-stack-adj.c (single_set_for_csa): Likewise.
7534 * combine.c (find_single_use_1): Likewise.
7535 (set_nonzero_bits_and_sign_copies): Likewise.
7536 (get_combine_src_dest): Likewise.
7537 (is_parallel_of_n_reg_sets): Likewise.
7538 (try_combine): Likewise.
7539 (record_dead_and_set_regs_1): Likewise.
7540 (reg_dead_at_p_1): Likewise.
7541 (reg_dead_at_p): Likewise.
7542 * dce.c (deletable_insn_p): Likewise.
7543 (mark_nonreg_stores_1): Likewise.
7544 (mark_nonreg_stores_2): Likewise.
7545 * df-scan.c (df_find_hard_reg_defs): Likewise.
7546 (df_uses_record): Likewise.
7547 (df_get_call_refs): Likewise.
7548 * dwarf2out.c (mem_loc_descriptor): Likewise.
7549 * haifa-sched.c (haifa_classify_rtx): Likewise.
7550 * ira-build.c (create_insn_allocnos): Likewise.
7551 * ira-costs.c (scan_one_insn): Likewise.
7552 * ira.c (equiv_init_movable_p): Likewise.
7553 (rtx_moveable_p): Likewise.
7554 (interesting_dest_for_shprep): Likewise.
7555 * jump.c (mark_jump_label_1): Likewise.
7556 * postreload-gcse.c (record_opr_changes): Likewise.
7557 * postreload.c (reload_cse_simplify): Likewise.
7558 (struct reg_use): Add source expr.
7559 (reload_combine): Check for clobber high.
7560 (reload_combine_note_use): Likewise.
7561 (reload_cse_move2add): Likewise.
7562 (move2add_note_store): Likewise.
7563 * print-rtl.c (print_pattern): Likewise.
7564 * recog.c (decode_asm_operands): Likewise.
7565 (store_data_bypass_p): Likewise.
7566 (if_test_bypass_p): Likewise.
7567 * regcprop.c (kill_clobbered_value): Likewise.
7568 (kill_set_value): Likewise.
7569 * reginfo.c (reg_scan_mark_refs): Likewise.
7570 * reload1.c (maybe_fix_stack_asms): Likewise.
7571 (eliminate_regs_1): Likewise.
7572 (elimination_effects): Likewise.
7573 (mark_not_eliminable): Likewise.
7574 (scan_paradoxical_subregs): Likewise.
7575 (forget_old_reloads_1): Likewise.
7576 * reorg.c (find_end_label): Likewise.
7577 (try_merge_delay_insns): Likewise.
7578 (redundant_insn): Likewise.
7579 (own_thread_p): Likewise.
7580 (fill_simple_delay_slots): Likewise.
7581 (fill_slots_from_thread): Likewise.
7582 (dbr_schedule): Likewise.
7583 * resource.c (update_live_status): Likewise.
7584 (mark_referenced_resources): Likewise.
7585 (mark_set_resources): Likewise.
7586 * rtl.c (copy_rtx): Likewise.
7587 * rtlanal.c (reg_referenced_p): Likewise.
7588 (single_set_2): Likewise.
7589 (noop_move_p): Likewise.
7590 (note_stores): Likewise.
7591 * sched-deps.c (sched_analyze_reg): Likewise.
7592 (sched_analyze_insn): Likewise.
7594 2018-08-06 Alan Hayward <alan.hayward@arm.com>
7596 * cse.c (invalidate_reg): New function extracted from...
7597 (invalidate): ...here.
7598 (canonicalize_insn): Check for clobber high.
7599 (invalidate_from_clobbers): invalidate clobber highs.
7600 (invalidate_from_sets_and_clobbers): Likewise.
7601 (count_reg_usage): Check for clobber high.
7602 (insn_live_p): Likewise.
7603 * cselib.c (cselib_expand_value_rtx_1):Likewise.
7604 (cselib_invalidate_regno): Check for clobber in setter.
7605 (cselib_invalidate_rtx): Pass through setter.
7606 (cselib_invalidate_rtx_note_stores):
7607 (cselib_process_insn): Check for clobber high.
7608 * cselib.h (cselib_invalidate_rtx): Add operand.
7610 2018-08-06 Alan Hayward <alan.hayward@arm.com>
7612 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
7613 (mark_not_eliminable): Likewise.
7614 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
7615 * lra-lives.c (process_bb_lives): Check for clobber high.
7616 * lra.c (new_insn_reg): Remember clobber highs.
7617 (collect_non_operand_hard_regs): Check for clobber high.
7618 (lra_set_insn_recog_data): Likewise.
7619 (add_regs_to_insn_regno_info): Likewise.
7620 (lra_update_insn_regno_info): Likewise.
7622 2018-08-06 Alan Hayward <alan.hayward@arm.com>
7624 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
7625 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
7627 2018-08-06 Alan Hayward <alan.hayward@arm.com>
7629 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
7630 (copy_insn_1): Likewise.
7631 (gen_hard_reg_clobber_high): New gen function.
7632 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
7633 * genemit.c (gen_exp): Likewise.
7634 (gen_emit_seq): Pass through info.
7635 (gen_insn): Check for CLOBBER_HIGH.
7636 (gen_expand): Pass through info.
7637 (gen_split): Likewise.
7638 (output_add_clobbers): Likewise.
7639 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
7640 (remove_clobbers): Likewise.
7641 * rtl.h (gen_hard_reg_clobber_high): New declaration.
7643 2018-08-06 Alan Hayward <alan.hayward@arm.com>
7645 * doc/rtl.texi (clobber_high): Add.
7646 (parallel): Add in clobber high
7647 * rtl.c (rtl_check_failed_code3): Add function.
7648 * rtl.def (CLOBBER_HIGH): Add expression.
7649 * rtl.h (RTL_CHECKC3): Add macro.
7650 (rtl_check_failed_code3): Add declaration.
7651 (XC3EXP): Add macro.
7653 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
7656 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
7657 cfun->machine->max_used_stack_alignment if needed.
7659 2018-08-04 Martin Sebor <msebor@redhat.com>
7661 PR tree-optimization/86571
7662 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
7665 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
7668 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7671 2018-08-03 Jeff Law <law@redhat.com>
7674 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7675 Define to speculation_safe_value_not_needed.
7677 2018-08-03 David Malcolm <dmalcolm@redhat.com>
7679 * doc/gcov.texi (-x): Remove duplicate "to".
7680 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
7681 (-Wif-not-aligned): Remove duplicate "is".
7682 (-flto): Remove duplicate "the".
7683 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
7684 duplicate "v5.00.b".
7685 (MSP430 Options): Remove duplicate "and" from the description
7686 of "-mgprel-sec=regexp".
7687 (x86 Options): Remove duplicate copies of "vmldLog102" and
7688 vmlsLog104 from description of "-mveclibabi=type".
7690 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
7692 * internal-fn.h (first_commutative_argument): Declare.
7693 * internal-fn.c (first_commutative_argument): New function.
7694 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
7695 restrictions for pattern statements. Use first_commutative_argument
7696 to look for commutative operands in calls to internal functions.
7698 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
7700 * Makefile.in (wide-int-range.o): New.
7701 * tree-vrp.c: Move all the wide_int_* functions to...
7702 * wide-int-range.cc: ...here.
7703 * tree-vrp.h: Move all the wide_int_* prototypes to...
7704 * wide-int-range.h: ...here.
7706 2018-08-03 Tom de Vries <tdevries@suse.de>
7708 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
7710 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
7711 * except.c (output_function_exception_table): Do early exit if
7712 targetm_common.except_unwind_info (&global_options) == UI_NONE.
7714 2018-08-03 Martin Liska <mliska@suse.cz>
7716 * predict.c (dump_prediction): Change to 2 digits
7719 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
7721 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
7723 (falkor_am_1_gtov_gtov): ... a new insn reservation.
7725 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
7727 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
7728 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
7729 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
7730 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
7731 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
7733 2018-08-02 David Malcolm <dmalcolm@redhat.com>
7735 * diagnostic-show-locus.c (diagnostic_show_locus): Use
7736 pp_take_prefix when saving the existing prefix.
7737 * diagnostic.c (diagnostic_append_note): Likewise.
7738 * langhooks.c (lhd_print_error_function): Likewise.
7739 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
7740 param's type. Free the existing prefix.
7741 (pp_take_prefix): New function.
7742 (pretty_printer::pretty_printer): Drop the prefix parameter.
7743 Rename the length parameter to match the comment.
7744 (pretty_printer::~pretty_printer): Free the prefix.
7745 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
7747 (struct pretty_printer): Drop the "const" from "prefix" field's
7748 type and clarify memory management.
7749 (pp_set_prefix): Drop the "const" from the 2nd param.
7750 (pp_take_prefix): New decl.
7752 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
7754 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
7755 for word_mode_ok here instead of passing as argument.
7756 (expand_block_compare): Change select_block_compare_mode() call.
7757 (expand_strncmp_gpr_sequence): New function.
7758 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
7760 2018-08-02 Jeff Law <law@redhat.com>
7763 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7764 Define to speculation_safe_value_not_needed.
7767 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7768 Define to speculation_safe_value_not_needed.
7770 2018-08-02 Tom de Vries <tdevries@suse.de>
7773 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
7774 function. Return UI_TARGET unconditionally.
7775 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
7776 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
7778 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
7780 * genemit.c (print_overload_test): Fix typo.
7782 2018-08-02 Richard Biener <rguenther@suse.de>
7784 PR tree-optimization/86816
7785 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
7786 which checks for value availability before querying it.
7787 (gvn_uses_equal): Use it.
7788 (same_succ_hash): Likewise.
7789 (gimple_equal_p): Likewise.
7791 2018-08-02 Nick Clifton <nickc@redhat.com>
7794 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7795 Define to speculation_safe_value_not_needed.
7798 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7799 Define to speculation_safe_value_not_needed.
7802 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7803 Define to speculation_safe_value_not_needed.
7806 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7807 Define to speculation_safe_value_not_needed.
7810 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7811 Define to speculation_safe_value_not_needed.
7814 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7815 Define to speculation_safe_value_not_needed.
7818 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7819 Define to speculation_safe_value_not_needed.
7822 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7823 Define to speculation_safe_value_not_needed.
7826 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
7827 speculation_safe_value_not_needed.
7830 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
7831 to speculation_safe_value_not_needed.
7833 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
7835 * doc/md.texi: Expand the documentation of instruction names
7836 to mention port-local uses. Document '@' in pattern names.
7837 * read-md.h (overloaded_instance, overloaded_name): New structs.
7839 (md_reader::handle_overloaded_name): New member function.
7840 (md_reader::get_overloads): Likewise.
7841 (md_reader::m_first_overload): New member variable.
7842 (md_reader::m_next_overload_ptr): Likewise.
7843 (md_reader::m_overloads_htab): Likewise.
7844 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
7845 m_next_overload_ptr and m_overloads_htab.
7846 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
7847 (get_mode_token, get_code_token, get_int_token): New functions.
7848 (map_attr_string): Add an optional argument that passes back
7849 the associated iterator.
7850 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
7851 (md_reader::handle_overloaded_name, add_overload_instance): New
7853 (apply_iterators): Handle '@' names. Report an error if '@'
7854 is used without iterators.
7855 (initialize_iterators): Initialize the new iterator_group fields.
7856 * genopinit.c (handle_overloaded_code_for)
7857 (handle_overloaded_gen): New functions.
7858 (main): Use them to print declarations of maybe_code_for_* and
7859 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
7860 * genemit.c (print_overload_arguments, print_overload_test)
7861 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
7862 (main): Use it to print definitions of maybe_code_for_* and
7863 maybe_gen_* functions.
7864 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
7865 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
7866 instead of explicit mode checks.
7867 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
7868 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
7869 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
7870 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
7871 (aarch64_expand_compare_and_swap): Likewise
7872 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
7873 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
7874 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
7875 (aarch64_constant_pool_reload_icode): Delete.
7876 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
7877 instead of aarch64_constant_pool_reload_icode. Use
7878 code_for_aarch64_reload_mov instead of explicit mode checks.
7879 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
7880 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
7881 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
7882 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
7883 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
7884 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
7885 (aarch64_atomic_load_op_code): Delete.
7886 (aarch64_emit_atomic_load_op): Likewise.
7887 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
7888 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
7889 instead of aarch64_emit_atomic_load_op.
7890 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
7891 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
7892 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
7893 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
7894 character before the pattern name.
7895 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
7896 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
7897 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
7898 (aarch64_frecps<mode>): Likewise.
7899 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
7900 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
7901 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
7902 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
7903 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
7905 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
7907 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
7908 Allow HFmode constants if TARGET_FP_F16INST.
7910 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
7913 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
7914 No longer check last store for clobber of address register.
7916 2018-08-02 Martin Liska <mliska@suse.cz>
7918 PR gcov-profile/86817
7919 * gcov.c (process_all_functions): New function.
7921 (process_file): Move functions processing to
7922 process_all_functions.
7924 2018-08-02 David Malcolm <dmalcolm@redhat.com>
7926 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
7927 "const" to the "gimple *" and "rtx_insn *" parameters.
7928 * dumpfile.h (dump_user_location_t::dump_user_location_t):
7930 (dump_location_t::dump_location_t): Likewise.
7932 2018-08-01 Martin Sebor <msebor@redhat.com>
7934 PR tree-optimization/86650
7935 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
7936 rather than a "gcall *". Directly pass the data of interest
7937 to percent_K_format, rather than building a temporary CALL_EXPR
7939 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
7940 (gimple_fold_builtin_strncat): Adjust.
7941 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
7942 gcall* argument with gimple*.
7943 * gimple-ssa-warn-restrict.c (check_call): Same.
7944 (wrestrict_dom_walker::before_dom_children): Same.
7945 (builtin_access::builtin_access): Same.
7946 (check_bounds_or_overlap): Same
7947 (maybe_diag_overlap): Same.
7948 (maybe_diag_offset_bounds): Same.
7949 * tree-diagnostic.c (default_tree_printer): Move usage of
7950 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
7952 * tree-pretty-print.c (percent_K_format): Add argument.
7953 * tree-pretty-print.h: Add argument.
7954 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
7955 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
7956 (maybe_diag_stxncpy_trunc): Same.
7957 (handle_builtin_stxncpy): Same.
7958 (handle_builtin_strcat): Same.
7960 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
7962 * match.pd: Optimise pointer range checks.
7964 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
7966 PR tree-optimization/86758
7967 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
7968 to remove pattern statements.
7970 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
7972 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
7973 result of dfs_enumerate_from when constructing stmt_vec_infos,
7974 instead of additionally calling get_loop_body.
7976 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
7978 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
7980 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
7981 When creating an iv, assert that the step is not known to be zero.
7982 (vect_setup_realignment): Update call accordingly.
7983 * tree-vect-stmts.c (vectorizable_store): Likewise.
7984 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
7986 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
7988 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
7989 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
7990 (vectorizable_reduction): Likewise.
7991 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7992 (vect_detect_hybrid_slp_stmts): Likewise.
7993 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
7995 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
7997 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
7998 (wide_int_set_zero_nonzero_bits): ...this.
7999 (zero_nonzero_bits_from_vr): Rename to...
8000 (vrp_set_zero_nonzero_bits): ...this.
8001 (extract_range_from_multiplicative_op_1): Abstract wide int
8003 (wide_int_range_multiplicative_op): ...here.
8004 (extract_range_from_binary_expr_1): Extract wide int binary
8005 operations into their own functions.
8006 (wide_int_range_lshift): New.
8007 (wide_int_range_can_optimize_bit_op): New.
8008 (wide_int_range_shift_undefined_p): New.
8009 (wide_int_range_bit_xor): New.
8010 (wide_int_range_bit_ior): New.
8011 (wide_int_range_bit_and): New.
8012 (wide_int_range_trunc_mod): New.
8013 (extract_range_into_wide_ints): New.
8014 (vrp_shift_undefined_p): New.
8015 (extract_range_from_multiplicative_op): New.
8016 (vrp_can_optimize_bit_op): New.
8017 * tree-vrp.h (value_range::dump): New.
8018 (wide_int_range_multiplicative_op): New.
8019 (wide_int_range_lshift):New.
8020 (wide_int_range_shift_undefined_p): New.
8021 (wide_int_range_bit_xor): New.
8022 (wide_int_range_bit_ior): New.
8023 (wide_int_range_bit_and): New.
8024 (wide_int_range_trunc_mod): New.
8025 (zero_nonzero_bits_from_bounds): Rename to...
8026 (wide_int_set_zero_nonzero_bits): ...this.
8027 (zero_nonzero_bits_from_vr): Rename to...
8028 (vrp_set_zero_nonzero_bits): ...this.
8029 (range_easy_mask_min_max): Rename to...
8030 (wide_int_range_can_optimize_bit_op): this.
8031 * vr-values.c (simplify_bit_ops_using_ranges): Rename
8032 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
8034 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
8036 * tree-vectorizer.h (vect_orig_stmt): New function.
8037 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
8038 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
8039 (vect_create_epilog_for_reduction): Likewise.
8040 (vectorizable_live_operation): Likewise.
8041 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
8042 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
8043 * tree-vect-stmts.c (vectorizable_call): Likewise.
8044 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
8046 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
8048 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
8050 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
8051 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
8052 (vect_transform_loop): Likewise.
8053 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
8055 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
8057 * tree-vectorizer.h (vect_schedule_slp): Return void.
8058 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
8059 (vect_schedule_slp): Likewise.
8061 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
8063 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
8065 (vect_transform_loop): Update calls accordingly. Schedule SLP
8066 instances before the main loop, if any exist.
8068 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
8070 PR tree-optimization/86749
8071 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
8072 If the lhs is used in a COND_EXPR, check that it is being used
8073 as the "then" or "else" value.
8075 2018-08-01 Tom de Vries <tdevries@suse.de>
8078 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
8079 speculation_safe_value_not_needed.
8081 2018-08-01 Richard Biener <rguenther@suse.de>
8083 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
8086 2018-08-01 Martin Liska <mliska@suse.cz>
8088 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
8089 format how successful transformation is dumped.
8090 (gimple_mod_pow2_value_transform): Likewise.
8091 (gimple_mod_subtract_transform): Likewise.
8092 (gimple_stringops_transform): Likewise.
8094 2018-08-01 Martin Liska <mliska@suse.cz>
8097 * value-prof.c (interesting_stringop_to_profile_p):
8098 Simplify the code and add BUILT_IN_MEMMOVE.
8099 (gimple_stringops_transform): Likewise.
8101 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
8103 * config/aarch64/aarch64-simd.md
8104 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
8105 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
8106 use GPI iterator instead of SI mode.
8108 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
8110 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
8111 rs6000_speculation_barrier.
8112 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
8113 new barrier pattern name.
8115 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
8117 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
8118 (speculation_barrier): New insn.
8120 2018-08-01 Richard Biener <rguenther@suse.de>
8123 * graphite.h: Include isl/id.h and isl/space.h to allow build
8126 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
8129 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
8130 mode for COFF targets.
8131 * defaults.h (TARGET_COFF): Define.
8132 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
8133 TARGET_COFF): Define.
8134 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
8135 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
8136 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
8138 2018-07-31 Alexandre Oliva <oliva@adacore.com>
8140 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
8141 * gimple-streamer-out.c (output_bb): Save it.
8142 * lto-streamer-in.c (input_struct_function_base): Restore
8143 instance discriminator if available. Create map on demand.
8144 * lto-streamer-out.c (output_struct_function_base): Save it if
8146 * final.c (decl_to_instance_map): Document LTO strategy.
8148 2018-07-31 Alexandre Oliva <oliva@adacore.com>
8149 Olivier Hainque <hainque@adacore.com>
8151 * debug.h (decl_to_instance_map_t): New type.
8152 (decl_to_instance_map): Declare.
8153 (maybe_create_decl_to_instance_map): New inline function.
8154 * final.c (bb_discriminator, last_bb_discriminator): New statics,
8155 to track basic block discriminators.
8156 (final_start_function_1): Initialize them.
8157 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
8159 (decl_to_instance_map): New variable.
8160 (map_decl_to_instance, maybe_set_discriminator): New functions.
8161 (notice_source_line): Set discriminator.
8163 2018-07-31 Ian Lance Taylor <iant@golang.org>
8165 * targhooks.c (default_have_speculation_safe_value): Add
8168 2018-07-31 David Malcolm <dmalcolm@redhat.com>
8170 * dump-context.h: Include "pretty-print.h".
8171 (dump_context::refresh_dumps_are_enabled): New decl.
8172 (dump_context::emit_item): New decl.
8173 (class dump_context): Add fields "m_test_pp" and
8175 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
8176 (temp_dump_context::get_dumped_text): New decl.
8177 (class temp_dump_context): Add field "m_pp".
8178 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
8179 (dump_context::refresh_dumps_are_enabled): ...and add a test for
8181 (set_dump_file): Update for above change.
8182 (set_alt_dump_file): Likewise.
8183 (dump_loc): New overload, taking a pretty_printer *.
8184 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
8185 to any test pretty-printer.
8186 (make_item_for_dump_gimple_stmt): New function, adapted from
8187 optinfo::add_gimple_stmt.
8188 (dump_context::dump_gimple_stmt): Call it, and use the result,
8189 eliminating the direct usage of dump_file and alt_dump_file in
8190 favor of indirectly using them via emit_item.
8191 (make_item_for_dump_gimple_expr): New function, adapted from
8192 optinfo::add_gimple_expr.
8193 (dump_context::dump_gimple_expr): Call it, and use the result,
8194 eliminating the direct usage of dump_file and alt_dump_file in
8195 favor of indirectly using them via emit_item.
8196 (make_item_for_dump_generic_expr): New function, adapted from
8198 (dump_context::dump_generic_expr): Call it, and use the result,
8199 eliminating the direct usage of dump_file and alt_dump_file in
8200 favor of indirectly using them via emit_item.
8201 (make_item_for_dump_printf_va): New function, adapted from
8202 optinfo::add_printf_va.
8203 (make_item_for_dump_printf): New function.
8204 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
8205 and use the result, eliminating the direct usage of dump_file and
8206 alt_dump_file in favor of indirectly using them via emit_item.
8207 (make_item_for_dump_dec): New function.
8208 (dump_context::dump_dec): Call it, and use the result,
8209 eliminating the direct usage of dump_file and alt_dump_file in
8210 favor of indirectly using them via emit_item.
8211 (make_item_for_dump_symtab_node): New function, adapted from
8212 optinfo::add_symtab_node.
8213 (dump_context::dump_symtab_node): Call it, and use the result,
8214 eliminating the direct usage of dump_file and alt_dump_file in
8215 favor of indirectly using them via emit_item.
8216 (dump_context::begin_scope): Reimplement, avoiding direct usage
8217 of dump_file and alt_dump_file in favor of indirectly using them
8219 (dump_context::emit_item): New member function.
8220 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
8221 Set up test pretty-printer on the underlying context. Call
8222 refresh_dumps_are_enabled.
8223 (temp_dump_context::~temp_dump_context): Call
8224 refresh_dumps_are_enabled.
8225 (temp_dump_context::get_dumped_text): New member function.
8226 (selftest::verify_dumped_text): New function.
8227 (ASSERT_DUMPED_TEXT_EQ): New macro.
8228 (selftest::test_capture_of_dump_calls): Run all tests twice, with
8229 and then without optinfo enabled. Add uses of
8230 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
8231 * dumpfile.h: Update comment for the dump_* API.
8232 * optinfo-emit-json.cc
8233 (selftest::test_building_json_from_dump_calls): Update for new
8234 param for temp_dump_context ctor.
8235 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
8236 and "m_owned" field.
8237 (optinfo_item::~optinfo_item): Likewise.
8238 (optinfo::add_item): New member function.
8239 (optinfo::emit): Update comment.
8240 (optinfo::add_string): Delete.
8241 (optinfo::add_printf): Delete.
8242 (optinfo::add_printf_va): Delete.
8243 (optinfo::add_gimple_stmt): Delete.
8244 (optinfo::add_gimple_expr): Delete.
8245 (optinfo::add_tree): Delete.
8246 (optinfo::add_symtab_node): Delete.
8247 (optinfo::add_dec): Delete.
8248 * optinfo.h (class dump_context): New forward decl.
8249 (optinfo::add_item): New decl.
8250 (optinfo::add_string): Delete.
8251 (optinfo::add_printf): Delete.
8252 (optinfo::add_printf_va): Delete.
8253 (optinfo::add_gimple_stmt): Delete.
8254 (optinfo::add_gimple_expr): Delete.
8255 (optinfo::add_tree): Delete.
8256 (optinfo::add_symtab_node): Delete.
8257 (optinfo::add_dec): Delete.
8258 (optinfo::add_poly_int): Delete.
8259 (optinfo_item::optinfo_item): Remove "owned" param.
8260 (class optinfo_item): Remove field "m_owned".
8262 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
8265 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
8266 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
8267 requested variable alignment.
8268 (expand_one_ssa_partition): Likewise.
8269 (expand_one_var): Likewise.
8271 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
8273 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
8274 to speculation_safe_value_not_needed.
8276 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
8278 * targhooks.h (speculation_safe_value_not_needed): New prototype.
8279 * targhooks.c (speculation_safe_value_not_needed): New function.
8280 * target.def (have_speculation_safe_value): Update documentation.
8281 * doc/tm.texi: Regenerated.
8283 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
8285 * config/aarch64/iterators.md (ALLI_TI): New iterator.
8286 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
8288 (despeculate_copy<ALLI:mode>_insn): New insn.
8289 (despeculate_copyti_insn): New insn.
8290 (despeculate_simple<ALLI:mode>): New insn
8291 (despeculate_simpleti): New insn.
8292 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
8294 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
8295 aarch64_speculation_safe_value.
8296 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
8298 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
8300 * config/aarch64/aarch64-speculation.cc: New file.
8301 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
8302 before pass_reorder_blocks.
8303 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
8305 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
8306 X14 and X15 when tracking speculation.
8307 * config/aarch64/aarch64.md (register name constants): Add
8308 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
8309 (unspec): Add UNSPEC_SPECULATION_TRACKER.
8310 (speculation_barrier): New insn attribute.
8311 (cmp<mode>): Allow SP in comparisons.
8312 (speculation_tracker): New insn.
8313 (speculation_barrier): Add speculation_barrier attribute.
8314 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
8315 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
8316 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
8318 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
8320 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
8321 aarch64_track_speculation is true.
8322 (tb<optab><mode>1): Likewise.
8323 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
8324 generate CB[N]Z when tracking speculation.
8325 (aarch64_split_compare_and_swap): Likewise.
8326 (aarch64_split_atomic_op): Likewise.
8328 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
8330 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
8332 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
8334 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
8335 (speculation_barrier): New insn.
8337 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
8339 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
8340 * config/arm/arm.md (speculation_barrier): New expand.
8341 (speculation_barrier_insn): New pattern.
8343 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
8345 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
8346 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
8347 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
8348 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
8350 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
8351 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
8352 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
8353 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
8354 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
8355 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
8356 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
8357 * builtins.c (expand_speculation_safe_value): New function.
8358 (expand_builtin): Call it.
8359 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
8360 * doc/extend.texi: Document __builtin_speculation_safe_value.
8361 * doc/md.texi: Document "speculation_barrier" pattern.
8362 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
8363 TARGET_HAVE_SPECULATION_SAFE_VALUE.
8364 * doc/tm.texi: Regenerated.
8365 * target.def (have_speculation_safe_value, speculation_safe_value): New
8367 * targhooks.c (default_have_speculation_safe_value): New function.
8368 (default_speculation_safe_value): New function.
8369 * targhooks.h (default_have_speculation_safe_value): Add prototype.
8370 (default_speculation_safe_value): Add prototype.
8372 2018-07-31 David Malcolm <dmalcolm@redhat.com>
8374 * dump-context.h (dump_context::dump_loc): New decl.
8375 * dumpfile.c (dump_context::dump_loc): New member function.
8376 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
8377 and dump_gimple_stmt.
8378 (dump_context::dump_gimple_expr_loc): Likewise, using
8380 (dump_context::dump_generic_expr_loc): Likewise, using
8382 (dump_context::dump_printf_loc_va): Likewise, using
8384 (dump_context::begin_scope): Explicitly using the global function
8385 "dump_loc", rather than the member function.
8387 2018-07-31 Martin Sebor <msebor@redhat.com>
8389 PR tree-optimization/86741
8390 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
8392 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
8394 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
8395 depend on whether prefetch instructions will be emitted or not.
8396 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
8397 will be emitted or not.
8398 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
8399 (TARGET_SETMEM_PFD): New macros.
8401 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8403 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
8404 (NULL_STMT_VEC_INFO): Delete.
8405 (stmt_vec_info::operator*): Likewise.
8406 (stmt_vec_info::operator gimple *): Likewise.
8407 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
8408 of NULL_STMT_VEC_INFO.
8409 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
8410 (vect_reassociating_reduction_p): Likewise.
8411 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
8412 (vectorizable_store): Likewise.
8413 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
8414 (vec_info::free_stmt_vec_infos): Likewise.
8416 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8418 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
8419 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
8421 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8423 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
8424 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
8425 (vec_info::free_stmt_vec_info): New private member functions.
8426 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
8427 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
8428 * tree-parloops.c (gather_scalar_reductions): Remove calls to
8429 set_stmt_vec_info_vec and free_stmt_vec_infos.
8430 * tree-vect-loop.c (_loop_vec_info): Remove call to
8431 set_stmt_vec_info_vec.
8432 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
8433 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
8434 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
8435 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
8436 (vec_info::free_stmt_vec_info): ...these new functions. Remove
8437 assignments in {vec_info::,}new_stmt_vec_info that are redundant
8438 with the clearing in the xcalloc.
8439 (stmt_vec_info_vec): Delete.
8440 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
8441 (vectorize_loops): Likewise.
8442 (vec_info::~vec_info): Remove argument from call to
8443 free_stmt_vec_infos.
8444 (vec_info::add_stmt): Remove vinfo argument from call to
8447 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8449 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
8450 rather than a gimple stmt.
8451 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
8452 information for pattern statements when passed the original
8453 statement; instead wait to be passed the pattern statement itself.
8454 Don't call set_vinfo_for_stmt here.
8455 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
8456 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
8457 stmt_vec_infos here.
8458 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
8459 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
8460 stmt_vec_infos entry.
8462 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8464 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
8465 * tree-vectorizer.c (vec_info::replace_stmt): New function.
8466 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
8467 * tree-vect-stmts.c (vectorizable_call): Likewise.
8468 (vectorizable_simd_clone_call): Likewise.
8470 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8472 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
8473 * tree-vectorizer.c (vec_info::remove_stmt): New function.
8474 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
8475 * tree-vect-loop.c (vect_transform_loop): Likewise.
8476 * tree-vect-slp.c (vect_schedule_slp): Likewise.
8477 * tree-vect-stmts.c (vect_remove_stores): Likewise.
8479 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8481 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
8482 (vect_dr_stmt): Delete.
8483 * tree-vectorizer.c (vec_info::lookup_dr): New function.
8484 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
8486 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
8487 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
8488 (vect_verify_datarefs_alignment, vect_peeling_supportable)
8489 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
8490 (vect_analyze_data_refs): Likewise.
8491 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
8493 (vect_find_same_alignment_drs): Likewise.
8494 (vect_slp_analyze_node_dependences): Update calls accordingly.
8495 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
8496 instead of DR_VECT_AUX.
8497 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
8498 of a vector data references. Use vec_info::lookup_dr instead of
8500 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
8501 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
8502 instead of DR_VECT_AUX.
8504 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8506 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
8508 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
8510 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
8511 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
8512 (vect_gen_prolog_loop_niters): Likewise.
8514 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8516 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
8517 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
8518 (vect_known_alignment_in_bytes, vect_dr_behavior)
8519 (vect_get_scalar_dr_size): Take references as dr_vec_infos
8520 instead of data_references. Update calls to other routines for
8521 which the same change has been made.
8522 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
8523 dr_vec_infos instead of stmt_vec_infos.
8524 (vect_analyze_data_ref_dependence): Update call accordingly.
8525 (vect_slp_analyze_data_ref_dependence)
8526 (vect_record_base_alignments): Use DR_VECT_AUX.
8527 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
8528 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
8529 (vector_alignment_reachable_p, vect_get_data_access_cost)
8530 (vect_peeling_supportable, vect_analyze_group_access_1)
8531 (vect_analyze_group_access, vect_analyze_data_ref_access)
8532 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
8533 (vect_compile_time_alias, vect_small_gap_p)
8534 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
8535 (vect_supportable_dr_alignment): Take references as dr_vec_infos
8536 instead of data_references. Update calls to other routines for
8537 which the same change has been made.
8538 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
8539 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
8540 (vect_slp_analyze_and_verify_node_alignment)
8541 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
8542 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
8543 (vect_setup_realignment): Use dr_vec_infos. Update calls after
8545 (_vect_peel_info::dr): Replace with...
8546 (_vect_peel_info::dr_info): ...this new field.
8547 (vect_peeling_hash_get_most_frequent)
8548 (vect_peeling_hash_choose_best_peeling): Update accordingly.
8549 (vect_peeling_hash_get_lowest_cost):
8550 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
8551 routines for which the same change has been made.
8552 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
8554 * tree-vect-loop-manip.c (get_misalign_in_elems)
8555 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
8557 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
8558 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
8559 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
8560 (get_group_load_store_type, get_negative_load_store_type)
8561 (vect_get_data_ptr_increment, vectorizable_store)
8562 (vectorizable_load): Likewise.
8563 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
8564 Update calls to other routines for which the same change has been made.
8566 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8568 * tree-vectorizer.h (vec_info::move_dr): New member function.
8569 (dataref_aux): Rename to...
8570 (dr_vec_info): ...this and add "dr" and "stmt" fields.
8571 (_stmt_vec_info::dr_aux): Update accordingly.
8572 (_stmt_vec_info::data_ref_info): Delete.
8573 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
8574 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
8575 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
8576 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
8578 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
8579 (STMT_VINFO_DR_INFO): New macro.
8580 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
8581 (set_dr_misalignment): Update after rename of dataref_aux.
8582 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
8583 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
8584 initialization of STMT_VINFO_DATA_REF.
8585 * tree-vectorizer.c (vec_info::move_dr): New function.
8586 * tree-vect-patterns.c (vect_recog_bool_pattern)
8587 (vect_recog_mask_conversion_pattern)
8588 (vect_recog_gather_scatter_pattern): Use it.
8589 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
8590 the "dr" and "stmt" fields of dr_vec_info instead of
8591 STMT_VINFO_DATA_REF.
8593 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8595 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
8596 (is_pattern_stmt_p): Use it.
8597 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
8598 on pattern statements.
8600 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8602 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
8603 original stmt as a stmt_vec_info rather than a gimple stmt.
8604 (vect_pattern_recog_1): Take the statement directly as a
8605 stmt_vec_info, rather than via a gimple_stmt_iterator.
8606 Update call to vect_mark_pattern_stmts.
8607 (vect_pattern_recog): Update calls accordingly.
8609 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8611 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
8612 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
8613 a vect_def_type for the first argument.
8614 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
8615 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
8616 operand if it isn't defined by a vectorized statement.
8617 (vect_build_gather_load_calls): Remove the mask_dt argument and
8618 update calls to vect_get_vec_def_for_stmt_copy.
8619 (vectorizable_bswap): Likewise the dt argument.
8620 (vectorizable_call): Update calls to vectorizable_bswap and
8621 vect_get_vec_def_for_stmt_copy.
8622 (vectorizable_simd_clone_call, vectorizable_assignment)
8623 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
8624 (vectorizable_comparison): Update calls to
8625 vect_get_vec_def_for_stmt_copy.
8626 (vectorizable_store): Likewise. Remove now-unnecessary calls to
8628 (vect_get_loop_based_defs): Remove dt argument and update call
8629 to vect_get_vec_def_for_stmt_copy.
8630 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
8631 and vect_get_vec_def_for_stmt_copy.
8632 (vectorizable_load): Update calls to vect_build_gather_load_calls
8633 and vect_get_vec_def_for_stmt_copy.
8634 * tree-vect-loop.c (vect_create_epilog_for_reduction)
8635 (vectorizable_reduction, vectorizable_live_operation): Update calls
8636 to vect_get_vec_def_for_stmt_copy.
8638 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8640 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
8641 and gimple stmt arguments with a stmt_vec_info.
8642 (vect_record_base_alignments): Update calls accordingly.
8643 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
8644 and gimple stmt arguments with a stmt_vec_info.
8645 (vect_build_slp_tree_1): Remove vinfo argument and update call
8646 to vect_record_max_nunits.
8647 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
8648 and vect_record_max_nunits.
8650 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8652 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
8653 file and take a stmt_vec_info instead of a gimple stmt.
8654 (supportable_widening_operation, vect_finish_replace_stmt)
8655 (vect_finish_stmt_generation, vect_get_store_rhs)
8656 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
8657 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
8658 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
8659 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
8660 (vect_create_data_ref_ptr, bump_vector_ptr)
8661 (vect_permute_store_chain, vect_setup_realignment)
8662 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
8663 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
8664 (vectorizable_reduction, vectorizable_induction)
8665 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
8666 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
8667 than gimple stmts as arguments.
8668 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
8669 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
8670 (can_group_stmts_p, vect_check_gather_scatter)
8671 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
8672 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
8673 (vect_permute_load_chain, vect_shift_permute_load_chain)
8674 (vect_transform_grouped_load)
8675 (vect_record_grouped_load_vectors): Likewise.
8676 * tree-vect-loop.c (vect_fixup_reduc_chain)
8677 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
8678 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
8679 (vectorizable_reduction, vectorizable_induction)
8680 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
8681 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
8682 (vect_get_load_store_mask): Likewise.
8683 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
8684 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
8685 * tree-vect-stmts.c (vect_mark_relevant)
8686 (is_simple_and_all_uses_invariant)
8687 (exist_non_indexing_operands_for_use_p, process_use)
8688 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
8689 (vect_get_vec_def_for_operand, vect_get_vec_defs)
8690 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
8691 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
8692 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
8693 (get_negative_load_store_type, get_load_store_type)
8694 (vect_check_load_store_mask, vect_check_store_rhs)
8695 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
8696 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
8697 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
8698 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
8699 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
8700 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
8701 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
8702 (supportable_widening_operation): Likewise.
8704 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8706 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
8707 a stmt_vec_info instead of a gcall.
8708 (vect_check_gather_scatter): Update call accordingly.
8709 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
8711 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
8712 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
8713 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
8714 instead of a gimple stmt.
8715 (vect_transform_loop): Update calls accordingly.
8716 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
8717 stmt_vec_infos instead of gimple stmts.
8718 (vect_analyze_slp_instance): Update use accordingly.
8719 * tree-vect-stmts.c (read_vector_array, write_vector_array)
8720 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
8721 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
8722 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
8723 (vect_gen_widened_results_half, vect_get_loop_based_defs)
8724 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
8725 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
8726 down to subroutines.
8728 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8730 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
8731 of the worklist from a vector of gimple stmts to a vector of
8733 * tree-vect-stmts.c (vect_mark_relevant, process_use)
8734 (vect_mark_stmts_to_be_vectorized): Likewise
8736 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8738 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
8739 statement before passing it to vect_analyze_stmt.
8740 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
8741 the chain of phi vector definitions. Track the exit phi via its
8743 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
8744 STMT_VINFO_REDUC_DEF.
8745 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
8746 stmt_vec_infos to handle the statement chains.
8747 (vect_get_slp_defs): Record the first statement in the node
8748 using a stmt_vec_info.
8749 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
8750 statements here and pass their stmt_vec_info down to subroutines.
8751 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
8752 down to vect_finish_stmt_generation.
8753 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
8754 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
8755 stmt_vec_infos to subroutines.
8756 (vect_remove_stores): Use stmt_vec_infos to handle the statement
8759 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8761 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
8762 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
8763 (vect_permute_store_chain, vect_setup_realignment)
8764 (vect_permute_load_chain, vect_shift_permute_load_chain)
8765 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
8766 stmts internally, and when passing values to other vectorizer routines.
8767 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
8768 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
8769 (vect_analyze_loop_operations, get_initial_def_for_reduction)
8770 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
8771 (vectorizable_reduction, vectorizable_induction)
8772 (vectorizable_live_operation, vect_transform_loop_stmt)
8773 (vect_transform_loop): Likewise.
8774 * tree-vect-patterns.c (vect_reassociating_reduction_p)
8775 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
8776 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
8777 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8778 (vect_slp_analyze_node_operations_1): Likewise.
8779 * tree-vect-stmts.c (vect_mark_relevant, process_use)
8780 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
8781 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
8782 (vect_finish_stmt_generation_1, get_group_load_store_type)
8783 (get_load_store_type, vect_build_gather_load_calls)
8784 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
8785 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
8786 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
8787 (vectorizable_store, vectorizable_load, vectorizable_condition)
8788 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
8789 (supportable_widening_operation): Likewise.
8790 (vect_get_vector_types_for_stmt): Likewise.
8791 * tree-vectorizer.h (vect_dr_behavior): Likewise.
8793 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8795 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
8796 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
8797 (vect_permute_store_chain, vect_permute_load_chain)
8798 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
8799 repeated stmt_vec_info lookups.
8800 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
8801 (vect_update_ivs_after_vectorizer): Likewise.
8802 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
8803 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
8804 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
8805 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8806 (vect_bb_slp_scalar_cost): Likewise.
8807 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
8809 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8811 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
8812 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
8813 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
8814 to get gassigns and gcalls, rather than operating on generc gimple
8816 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
8817 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
8818 (vectorizable_load, vect_analyze_stmt): Likewise.
8819 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
8821 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8823 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
8824 return stmt_vec_infos rather than gimple stmts. Do not accept
8826 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
8828 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
8829 Update use of get_later_stmt.
8830 (vect_get_constant_vectors): Update call accordingly.
8831 (vect_schedule_slp_instance): Likewise
8832 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
8833 (vect_slp_analyze_instance_dependence): Likewise.
8834 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
8836 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8838 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
8839 (stmt_info_for_cost::stmt_info): ...this new field.
8840 (add_stmt_costs): Update accordingly.
8841 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
8842 (vect_get_known_peeling_cost): Likewise.
8843 (vect_estimate_min_profitable_iters): Likewise.
8844 * tree-vect-stmts.c (record_stmt_cost): Likewise.
8846 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8848 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
8849 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
8850 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
8852 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
8854 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8856 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
8857 a gimple stmt to a stmt_vec_info.
8858 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
8860 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8862 * tree-vectorizer.h (vec_info::grouped_stores): Change from
8863 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
8864 (_loop_vec_info::reduction_chains): Likewise.
8865 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
8867 * tree-vect-slp.c (vect_analyze_slp): Likewise.
8869 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8871 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
8872 a gimple stmt to a stmt_vec_info.
8873 (_stmt_vec_info::next_element): Likewise.
8874 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
8875 (vect_slp_analyze_and_verify_node_alignment)
8876 (vect_analyze_group_access_1, vect_analyze_group_access)
8877 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
8878 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
8879 (vect_supportable_dr_alignment): Update accordingly.
8880 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
8881 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
8882 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
8883 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
8884 * tree-vect-slp.c (vect_build_slp_tree_1)
8885 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
8886 (vect_split_slp_store_group, vect_analyze_slp_instance)
8887 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
8888 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
8889 (get_group_load_store_type, get_load_store_type)
8890 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
8891 (vect_transform_stmt, vect_remove_stores): Likewise.
8893 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8895 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
8897 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
8898 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
8899 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
8900 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
8901 (vector_alignment_reachable_p, vect_get_data_access_cost)
8902 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
8903 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
8904 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
8905 (vect_analyze_group_access_1, vect_analyze_group_access)
8906 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
8907 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
8908 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
8909 result of vect_dr_stmt and use the stmt_vec_info instead of
8910 the associated gimple stmt.
8911 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
8912 (vect_gen_prolog_loop_niters): Likewise.
8913 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
8915 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8917 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
8918 to a vec<stmt_vec_info>.
8919 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
8920 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
8922 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
8923 to a vec<stmt_vec_info>.
8924 (bst_traits::value_type, bst_traits::value_type): Likewise.
8925 (bst_traits::hash): Update accordingly.
8926 (vect_get_and_check_slp_defs): Change the stmts parameter from
8927 a vec<gimple *> to a vec<stmt_vec_info>.
8928 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
8929 (vect_build_slp_tree): Likewise.
8930 (vect_build_slp_tree_2): Likewise. Update uses of
8931 SLP_TREE_SCALAR_STMTS.
8932 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
8933 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
8934 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
8935 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
8936 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
8937 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
8938 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
8939 (vect_get_constant_vectors, vect_get_slp_defs)
8940 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
8941 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
8942 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
8943 instead of gimple stmts.
8944 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
8945 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
8946 (vect_slp_analyze_instance_dependence): Update uses of
8947 SLP_TREE_SCALAR_STMTS.
8948 (vect_slp_analyze_and_verify_node_alignment): Likewise.
8949 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
8950 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
8951 (get_initial_defs_for_reduction): Likewise.
8952 (vect_create_epilog_for_reduction): Likewise.
8953 (vectorize_fold_left_reduction): Likewise.
8954 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
8955 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
8956 (can_vectorize_live_stmts): Likewise.
8958 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8960 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
8961 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
8962 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
8964 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
8966 (gather_scalar_reductions): Update after above interface changes.
8967 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
8968 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
8970 (vect_force_simple_reduction): Likewise.
8971 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
8972 LOOP_VINFO_REDUCTIONS.
8973 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8975 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8977 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
8978 a gimple stmt to a stmt_vec_info.
8979 * tree-vect-loop.c (vect_active_double_reduction_p)
8980 (vect_force_simple_reduction, vectorizable_reduction): Update
8983 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8985 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
8986 vec<gimple *> to a vec<stmt_vec_info>.
8987 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
8988 the reduction_phis argument from a vec<gimple *> to a
8990 (vectorizable_reduction): Likewise the phis local variable that
8991 is passed to vect_create_epilog_for_reduction. Update for new type
8992 of SLP_TREE_VEC_STMTS.
8993 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
8994 (vectorizable_live_operation): Likewise.
8995 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
8996 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
8998 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9000 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
9001 a gimple stmt to a stmt_vec_info.
9002 (vectorizable_condition, vectorizable_live_operation)
9003 (vectorizable_reduction, vectorizable_induction): Pass back the
9004 vectorized statement as a stmt_vec_info.
9005 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
9006 use of STMT_VINFO_VEC_STMT.
9007 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
9008 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
9009 as stmt_vec_infos rather than gimple stmts.
9010 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
9012 (vectorizable_live_operation): Likewise.
9013 (vectorizable_reduction, vectorizable_induction): Likewise,
9014 updating use of STMT_VINFO_VEC_STMT.
9015 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
9016 of STMT_VINFO_VEC_STMT.
9017 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
9018 (vectorizable_simd_clone_call, vectorizable_conversion)
9019 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
9020 (vectorizable_store, vectorizable_load, vectorizable_condition)
9021 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
9022 from a gimple stmt to a stmt_vec_info.
9023 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
9024 pointer to a stmt_vec_info to the vectorizable_* routines.
9026 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9028 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
9029 a gimple stmt to a stmt_vec_info.
9030 (is_pattern_stmt_p): Update accordingly.
9031 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
9032 (vect_record_grouped_load_vectors): Likewise.
9033 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
9034 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
9035 (vect_model_reduction_cost): Likewise.
9036 (vect_create_epilog_for_reduction): Likewise.
9037 (vectorizable_reduction, vectorizable_induction): Likewise.
9038 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
9039 Return the stmt_vec_info for the pattern statement.
9040 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
9041 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
9042 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
9043 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
9044 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
9045 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
9046 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
9047 (free_stmt_vec_info, vect_is_simple_use): Likewise.
9049 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9051 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
9052 (vect_finish_stmt_generation): Likewise.
9053 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
9054 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
9055 (vect_build_gather_load_calls): Use the return value of the above
9056 functions instead of a separate call to vinfo_for_stmt. Use narrow
9057 scopes for the input gimple stmt and wider scopes for the associated
9058 stmt_vec_info. Use vec_info::lookup_def when setting these
9059 stmt_vec_infos from an SSA_NAME definition.
9060 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
9061 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
9062 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
9063 (vectorizable_store, vectorizable_load, vectorizable_condition)
9064 (vectorizable_comparison): Likewise.
9065 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
9066 (vectorizable_reduction): Likewise.
9068 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9070 * tree-vectorizer.h (vect_is_simple_use): Add an optional
9071 stmt_vec_info * parameter before the optional gimple **.
9072 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
9073 (process_use, vect_get_vec_def_for_operand_1): Update callers.
9074 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
9075 * tree-vect-loop.c (vectorizable_reduction): Likewise.
9076 (vectorizable_live_operation): Likewise.
9077 * tree-vect-patterns.c (type_conversion_p): Likewise.
9078 (vect_look_through_possible_promotion): Likewise.
9079 (vect_recog_rotate_pattern): Likewise.
9080 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
9082 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9084 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
9085 a typedef to a wrapper class.
9086 (NULL_STMT_VEC_INFO): New macro.
9087 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
9088 (stmt_vec_info::operator*): New function.
9089 (stmt_vec_info::operator gimple *): Likewise.
9090 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
9091 (add_stmt_costs): Likewise.
9092 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
9093 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
9094 (vect_get_known_peeling_cost): Likewise.
9095 (vect_estimate_min_profitable_iters): Likewise.
9096 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
9097 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
9098 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
9099 (vectorizable_store, free_stmt_vec_infos): Likewise.
9100 (new_stmt_vec_info): Change return type of xcalloc to
9103 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9105 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
9106 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
9107 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
9108 a single_imm_use-based sequence.
9109 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
9111 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9113 * tree-vectorizer.h (vec_info::lookup_def): Declare.
9114 * tree-vectorizer.c (vec_info::lookup_def): New function.
9115 * tree-vect-patterns.c (vect_get_internal_def): Use it.
9116 (vect_widened_op_tree): Likewise.
9117 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
9118 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
9119 (vectorizable_reduction): Likewise.
9120 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
9122 (vect_is_slp_reduction): Update calls accordingly. Use
9123 vec_info::lookup_def.
9124 (vect_is_simple_reduction): Likewise
9125 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
9127 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9129 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
9130 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
9131 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
9133 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
9134 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
9135 (vect_update_vf_for_slp, vect_analyze_loop_operations)
9136 (vect_is_slp_reduction, vectorizable_induction)
9137 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
9138 * tree-vect-patterns.c (vect_init_pattern_stmt):
9139 (vect_determine_min_output_precision_1, vect_determine_precisions)
9140 (vect_pattern_recog): Likewise.
9141 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
9142 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
9143 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
9144 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
9145 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
9146 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
9147 info field from a loop to a loop_vec_info.
9149 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9151 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
9152 (vec_info::add_stmt): Declare.
9153 * tree-vectorizer.c (vec_info::add_stmt): New function.
9154 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
9155 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
9156 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
9157 (vectorizable_induction): Likewise.
9158 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
9159 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
9160 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
9161 (vectorizable_load): Likewise.
9162 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
9163 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
9164 (vect_recog_gather_scatter_pattern): Likewise.
9165 (append_pattern_def_seq): Likewise. Remove a check that is
9166 performed by add_stmt itself.
9168 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9170 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
9171 which make_ssa_name was called with new_stmt before new_stmt
9174 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9176 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
9178 (vect_is_slp_reduction): ...here...
9179 (vect_is_simple_reduction): ...and here. Remove repetition of tests
9180 that are already known to be false.
9182 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9184 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
9185 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
9186 STMT_VINFO_NUM_SLP_USES when it's true.
9187 (vect_free_slp_instance): Add a final_p parameter and pass it to
9189 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
9190 (vect_analyze_slp_instance): Likewise.
9191 (vect_slp_analyze_operations): Likewise.
9192 (vect_slp_analyze_bb_1): Likewise.
9193 * tree-vectorizer.c (vec_info): Likewise.
9194 * tree-vect-loop.c (vect_transform_loop): Likewise.
9196 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9198 * tree-vect-loop.c (vectorizable_reduction): Assert that the
9199 function is not called for second and subsequent members of
9202 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9204 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
9205 cases for nested loops from here to ...
9206 (vect_create_epilog_for_reduction): ...here. Only call
9207 vect_is_simple_use for inner-loop reductions.
9209 2018-07-31 Martin Liska <mliska@suse.cz>
9211 PR gcov-profile/85338
9212 PR gcov-profile/85350
9213 PR gcov-profile/85372
9214 * profile.c (struct location_triplet): New.
9215 (struct location_triplet_hash): Likewise.
9216 (output_location): Do not output a BB that
9217 is already recorded for a line.
9218 (branch_prob): Use streamed_locations.
9220 2018-07-31 Martin Liska <mliska@suse.cz>
9222 PR gcov-profile/85370
9223 * coverage.c (coverage_begin_function): Do not mark target
9224 clones as artificial functions.
9226 2018-07-31 Martin Liska <mliska@suse.cz>
9228 PR gcov-profile/83813
9229 PR gcov-profile/84758
9230 PR gcov-profile/85217
9231 PR gcov-profile/85332
9232 * profile.c (branch_prob): Do not record GOTO expressions
9233 for GIMPLE statements which locations are already streamed.
9235 2018-07-31 Olivier Hainque <hainque@adacore.com>
9237 * gcc.c (handle_spec_function): Accept a soft_matched_part
9238 argument, as do_spec_1. Pass it down to ...
9239 (eval_spec_function): Accept a soft_matched_part argument,
9240 and pass it down to ...
9241 (do_spec_2): Accept a soft_matched_part argument, and pass
9242 it down to do_spec_1.
9243 (do_spec_1): Pass soft_matched_part to handle_spec_function.
9244 (handle_braces): Update call to handle_spec_function.
9245 (driver::set_up_specs): Update calls to do_spec_2.
9246 (compare_debug_dump_opt_spec_function): Likewise.
9247 (compare_debug_self_opt_spec_function): Likewise.
9249 2018-07-31 Olivier Hainque <hainque@adacore.com>
9251 * common.opt (nolibc): New option.
9252 * doc/invoke.texi (Link Options): Document it.
9253 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
9254 * config/alpha/linux.h: Likewise.
9255 * config/arc/elf.h: Likewise.
9256 * config/arm/uclinux-elf.h: Likewise.
9257 * config/arm/unknown-elf.h: Likewise.
9258 * config/avr/avrlibc.h: Likewise.
9259 * config/bfin/bfin.h: Likewise.
9260 * config/bfin/linux.h: Likewise.
9261 * config/bfin/uclinux.h: Likewise.
9262 * config/darwin.h: Likewise.
9263 * config/darwin10.h: Likewise.
9264 * config/darwin12.h: Likewise.
9265 * config/gnu-user.h: Likewise.
9266 * config/lm32/uclinux-elf.h: Likewise.
9267 * config/pa/pa-hpux11.h: Likewise.
9268 * config/pa/pa64-hpux.h: Likewise.
9269 * config/sparc/sparc.h: Likewise.
9271 2018-07-31 Olivier Hainque <hainque@adacore.com>
9273 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
9274 undefined variables.
9276 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
9279 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
9282 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
9284 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
9285 terminated string literal.
9287 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
9289 PR rtl-optimization/85160
9290 * combine.c (is_just_move): New function.
9291 (try_combine): Allow combining two instructions into two if neither of
9292 the original instructions was a move.
9294 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
9297 * doc/extend.texi (Global Register Variables): Discourage use of type
9299 (Local Register Variables): Likewise.
9301 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
9303 PR tree-optimization/86506
9304 * hwint.h (ceil_log2): Resync with hwint.c implementation.
9306 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
9309 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
9310 hard_regno, make sure no insn between `from` and `to` clobbers it.
9312 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
9313 Tom de Vries <tdevries@suse.de>
9315 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
9316 (PTX_DEFAULT_RUNTIME_DIM): ... this.
9317 (nvptx_goacc_validate_dims): Set default worker and gang dims to
9318 PTX_DEFAULT_RUNTIME_DIM.
9319 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
9321 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
9323 * config/pa/pa.c (pa_output_addr_vec): Align address table.
9324 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
9325 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
9327 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
9329 * config/rs6000/constraints.md (wG constraint): Delete, no longer
9331 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
9332 predicate to reflect toc fusion has been deleted.
9333 (toc_fusion_mem_raw): Delete, no longer used.
9334 (toc_fusion_mem_wrapped): Likewise.
9335 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
9337 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
9338 Delete, no longer used.
9339 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
9340 meant to be used for toc fusion.
9341 (rs6000_debug_print_mode): Delete toc fusion debugging.
9342 (rs6000_debug_reg_global): Likewise.
9343 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
9344 fusion and secondary reload support that were never used.
9345 (rs6000_option_override_internal): Delete TOC fusion, that was only
9346 partially defined, and it did not work unless you also used the
9348 (rs6000_legitimate_address_p): Delete TOC fusion support.
9349 (rs6000_opt_masks): Likewise.
9350 (fusion_wrap_memory_address): Delete function, no longer used.
9351 (fusion_split_address); Delete TOC fusion support.
9352 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
9353 longer used with toc fusion being deleted.
9354 (TARGET_TOC_FUSION_FP): Likewise.
9355 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
9357 (toc fusion spliter): Delete TOC fusion support.
9358 (toc_fusionload_<mode>): Likewise.
9359 (toc_fusionload_di): Likewise.
9360 (fusion_gpr_load_<mode>): Delete generator function, this insn no
9361 longer needs to be named. Rename predicate to delete TOC fusion.
9362 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
9363 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
9364 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
9365 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
9366 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
9368 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
9370 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
9371 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
9372 __int128 in built-in function prototypes.
9373 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
9374 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
9376 2018-07-27 Martin Sebor <msebor@redhat.com>
9378 PR tree-optimization/86696
9379 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
9380 types, including enums.
9381 (handle_char_store): Be prepared for the above function to fail.
9383 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
9385 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
9386 when optimization level is lower than 2 or optimize for size.
9388 2018-07-26 Martin Sebor <msebor@redhat.com>
9390 PR tree-optimization/86043
9391 PR tree-optimization/86042
9392 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
9393 (get_string_cst_length): Rename...
9394 (get_min_string_length): ...to this. Add argument.
9395 (handle_char_store): Extend to handle multi-character stores by
9397 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
9398 * tree.h (initializer_zerop): Add argument.
9400 2018-07-26 Jakub Jelinek <jakub@redhat.com>
9403 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
9404 declare target to variables if they have always,{to,from,tofrom} map
9407 2018-07-26 Martin Liska <mliska@suse.cz>
9410 * lto-wrapper.c: Add linker_output as prefix
9411 for ltrans_output_file.
9413 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
9415 PR rtl-optimization/85805
9416 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
9417 value for hard registers if that was written in the same mode.
9419 2018-07-26 Martin Liska <mliska@suse.cz>
9421 PR gcov-profile/86536
9422 * gcov.c (format_gcov): Use printf format %.*f directly
9423 and do not handle special values.
9425 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
9427 * common/config/arc/arc-common.c (arc_option_optimization_table):
9428 Update default optimizations for size.
9430 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
9432 * config/arc/arc.md (movsf_insn): Add short instruction selection.
9433 * config/arc/constraints.md (CfZ): New constraint.
9434 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
9435 (subsf3_fpu): Likewise.
9436 (cmpsf_fpu): Likewise.
9437 (cmpsf_fpu_uneq): Likewise.
9439 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
9441 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
9443 (arc_is_uncached_mem_p): uncached applies to both the variable and
9446 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
9448 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
9451 2018-07-25 David Malcolm <dmalcolm@redhat.com>
9453 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
9454 field "m_scopes" from vec to auto_vec.
9456 2018-07-25 Martin Liska <mliska@suse.cz>
9458 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
9461 2018-07-25 Richard Biener <rguenther@suse.de>
9464 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
9465 special wrt context_die late.
9466 (gen_subprogram_die): Re-use DIEs in local scope.
9468 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
9470 PR tree-optimization/86644
9471 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
9473 2018-07-25 Martin Liska <mliska@suse.cz>
9476 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
9477 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
9479 2018-07-25 Martin Liska <mliska@suse.cz>
9482 * params.def: Explain ASan abbreviation and provide
9483 a documentation link.
9485 2018-07-24 Martin Sebor <msebor@redhat.com>
9487 PR tree-optimization/86622
9488 PR tree-optimization/86532
9489 * builtins.h (string_length): Declare.
9490 * builtins.c (c_strlen): Correct handling of non-constant offsets.
9491 (check_access): Be prepared for non-constant length ranges.
9492 (string_length): Make extern.
9493 * expr.c (string_constant): Only handle the minor non-constant
9494 array index. Use string_constant to compute the length of
9495 a generic string constant.
9497 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
9499 PR tree-optimization/86618
9500 * tree-vect-stmts.c (vectorizable_call): Don't take the address
9501 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
9503 2018-07-24 David Malcolm <dmalcolm@redhat.com>
9505 PR tree-optimization/86636
9506 * json.cc (json::object::set): Fix comment. Add assertions.
9507 (json::array::append): Move here from json.h. Add comment and an
9509 (json::string::string): Likewise.
9510 * json.h (json::array::append): Move to json.cc.
9511 (json::string::string): Likewise.
9512 * optinfo-emit-json.cc
9513 (optrecord_json_writer::impl_location_to_json): Assert that we
9514 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
9515 wrapper around it. Expand the location once, rather than three
9517 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
9518 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
9520 (optrecord_json_writer::optinfo_to_json): Likewise, in four
9521 places. Fix some overlong lines.
9523 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
9525 * config/aarch64/aarch64-simd.md
9526 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
9527 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
9528 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
9529 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
9530 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
9531 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
9532 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
9533 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
9534 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
9536 2018-07-24 Jakub Jelinek <jakub@redhat.com>
9539 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
9540 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
9541 and abs_d == d, do the power of two handling if profitable.
9543 2018-07-24 Richard Biener <rguenther@suse.de>
9545 * match.pd: Add BIT_FIELD_REF canonicalizations.
9547 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
9550 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
9552 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
9554 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
9555 terminated STRING_CST object.
9557 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
9559 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
9561 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
9563 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
9564 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
9565 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
9566 the elements into a register.
9567 (rs6000_split_v4si_init_di_reg): Delete.
9568 (rs6000_split_v4si_init): Delete.
9569 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
9570 (vsx_init_v4si): Rewrite as a define_expand.
9572 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
9574 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
9575 zero_extend argument from memory): New.
9577 2018-07-22 Martin Sebor <msebor@redhat.com>
9580 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
9581 diagnosing calls with unknown arguments unless -Walloca-larger-than
9582 is restricted to less than PTRDIFF_MAX bytes.
9584 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
9586 * doc/gcov.texi (Invoking Gcov): Editorial changes.
9588 2018-07-20 David Malcolm <dmalcolm@redhat.com>
9590 * pretty-print.c (text_info::set_location): Remove redundant
9591 "line_table" parameter from call to rich_location::set_range.
9593 2018-07-20 Martin Sebor <msebor@redhat.com>
9596 * builtins.c (expand_builtin_alloca): Adjust.
9597 * calls.c (alloc_max_size): Simplify.
9598 * cgraphunit.c (cgraph_node::expand): Adjust.
9599 * common.opt (larger_than_size, warn_frame_larger_than): Remove
9601 (frame_larger_than_size): Same.
9602 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
9603 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
9605 * doc/invoke.texi (GCC Command Options): Document option arguments.
9606 Explain byte-size arguments and suffixes.
9607 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
9608 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
9609 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
9610 * doc/options.texi (UInteger): Expand.
9611 (Host_Wide_Int, ByteSize): Document new properties.
9612 * final.c (final_start_function_1): Include sizes in an error message.
9613 * function.c (frame_offset_overflow): Same.
9614 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
9615 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
9616 Diagnose unbounded alloca calls only for limits of less than
9618 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
9619 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
9621 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
9622 only for limits of less than PTRDIFF_MAX.
9623 * langhooks-def.h (lhd_handle_option): Change function argument
9625 * langhooks.c (lhd_handle_option): Same.
9626 * langhooks.h (handle_option): Same.
9627 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
9629 (var_type, var_type_struct): Same.
9630 (var_set): Handle ByteSize flag.
9631 * optc-gen.awk: Add comments to output to ease debugging. Make
9632 use of HOST_WIDE_INT where appropriate.
9633 * opts-gen-save.awk: Use %lx to format unsigned long.
9634 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
9635 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
9636 arguments. Parse bytes-size suffixes.
9637 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
9638 (enum_value_to_arg): Same.
9639 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
9640 (handle_option): Adjust.
9641 (generate_option): Change function argument to HOST_WIDE_INT.
9642 (cmdline_handle_error): Adjust.
9643 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
9644 (set_option): Change function argument to HOST_WIDE_INT.
9645 (option_enabled): Handle cl_host_wide_int.
9646 (get_option_state): Handle CLVC_SIZE.
9647 (control_warning_option): Same.
9648 * opts.c (common_handle_option): Change function argument to
9649 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
9650 OPT_Wvla_larger_than_.
9651 * opts.h (enum cl_var_type): Add an enumerator.
9652 * stor-layout.c (layout_decl): Print a more meaningful warning.
9653 * toplev.c (output_stack_usage): Adjust.
9655 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
9657 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
9658 call to inline_expand_builtin_string_cmp.
9659 (expand_builtin_strcmp): Likewise.
9660 (expand_builtin_strncmp): Likewise.
9661 (inline_string_cmp): Delete the last parameter, change char_type_node
9662 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
9664 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
9665 the inlining expansion on target where the type of the call has same or
9666 narrower precision than unsigned char.
9668 2018-07-20 David Malcolm <dmalcolm@redhat.com>
9670 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
9671 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
9672 * common.opt (fsave-optimization-record): New option.
9673 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
9674 * doc/invoke.texi (-fsave-optimization-record): New option.
9675 * dumpfile.c: Include "optinfo-emit-json.h".
9676 (struct kv_pair): Move to coretypes.h.
9677 (optgroup_options): Make non-static.
9678 (dump_context::end_scope): Call
9679 optimization_records_maybe_pop_dump_scope.
9680 * dumpfile.h (optgroup_options): New decl.
9681 * json.cc: New file.
9683 * optinfo-emit-json.cc: New file.
9684 * optinfo-emit-json.h: New file.
9685 * optinfo.cc: Include "optinfo-emit-json.h".
9686 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
9687 (optinfo_enabled_p): Check optimization_records_enabled_p.
9688 (optinfo_wants_inlining_info_p): Likewise.
9689 * optinfo.h: Update comment.
9690 * profile-count.c (profile_quality_as_string): New function.
9691 * profile-count.h (profile_quality_as_string): New decl.
9692 (profile_count::quality): New accessor.
9693 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
9694 and optinfo_emit_json_cc_tests.
9695 * selftest.h (selftest::json_cc_tests): New decl.
9696 (selftest::optinfo_emit_json_cc_tests): New decl.
9697 * toplev.c: Include "optinfo-emit-json.h".
9698 (compile_file): Call optimization_records_finish.
9699 (do_compile): Call optimization_records_start.
9700 * tree-ssa-live.c: Include optinfo.h.
9701 (remove_unused_scope_block_p): Retain inlining information if
9702 optinfo_wants_inlining_info_p returns true.
9704 2018-07-20 Richard Biener <rguenther@suse.de>
9707 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
9708 to cover -flto-partition=none.
9710 2018-07-20 Martin Liska <mliska@suse.cz>
9712 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
9713 (get_decl_source_range): Remove unused function.
9715 2018-07-20 Richard Biener <rguenther@suse.de>
9717 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
9718 (struct vn_phi_s): Likewise.
9719 (struct vn_reference_s): Likewise.
9720 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
9721 for searching the slot of an entry known to be in the hash itself.
9722 (vn_phi_hasher::equal): Likewise.
9723 (vn_reference_hasher::equal): Likewise.
9724 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
9726 (optimistic_info, current_info): Remove, keeping only valid_info.
9727 (vn_reference_lookup_1): Remove fallback lookup.
9728 (vn_reference_lookup_2): Likewise.
9729 (vn_nary_op_lookup_1): Likewise.
9730 (vn_phi_lookup): Likewise.
9731 (vn_nary_build_or_lookup_1): Make sure to not chain the built
9733 (vn_reference_insert): Adjust, chain the inserted hash element
9734 at last_inserted_ref.
9735 (vn_reference_insert_pieces): Likewise.
9736 (visit_reference_op_call): Likewise.
9737 (vn_nary_op_insert_into): Chain the inserted hash element at
9739 (vn_nary_op_insert_pieces): Adjust.
9740 (vn_nary_op_insert): Likewise.
9741 (vn_nary_op_insert_stmt): Likewise.
9742 (vn_phi_insert): Adjust, chain the inserted hash element at
9744 (process_scc): Remove clearing and copying the optimistic
9745 table. Instead remove elements inserted during an optimistic
9746 iteration from the single table we maintain.
9747 (init_scc_vn): Adjust.
9748 (free_scc_vn): Likewise.
9749 (sccvn_dom_walker::record_cond): Likewise.
9750 (sccvn_dom_walker::after_dom_children): Likewise.
9752 2018-07-19 Martin Sebor <msebor@redhat.com>
9754 PR tree-optimization/84047
9755 PR tree-optimization/83776
9756 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
9757 (check_array_bounds): Call it.
9759 2018-07-19 Martin Sebor <msebor@redhat.com>
9761 * align.h (align_flags): Use member initialization.
9763 2018-07-19 David Malcolm <dmalcolm@redhat.com>
9765 * Makefile.in (OBJS): Add optinfo.o.
9766 * coretypes.h (class symtab_node): New forward decl.
9767 (struct cgraph_node): New forward decl.
9768 (class varpool_node): New forward decl.
9769 * dump-context.h: New file.
9770 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
9772 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
9773 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
9774 (set_alt_dump_file): Likewise.
9775 (dump_context::~dump_context): New dtor.
9776 (dump_gimple_stmt): Move implementation to...
9777 (dump_context::dump_gimple_stmt): ...this new member function.
9778 Add the stmt to any pending optinfo, creating one if need be.
9779 (dump_gimple_stmt_loc): Move implementation to...
9780 (dump_context::dump_gimple_stmt_loc): ...this new member function.
9781 Start a new optinfo and add the stmt to it.
9782 (dump_gimple_expr): Move implementation to...
9783 (dump_context::dump_gimple_expr): ...this new member function.
9784 Add the stmt to any pending optinfo, creating one if need be.
9785 (dump_gimple_expr_loc): Move implementation to...
9786 (dump_context::dump_gimple_expr_loc): ...this new member function.
9787 Start a new optinfo and add the stmt to it.
9788 (dump_generic_expr): Move implementation to...
9789 (dump_context::dump_generic_expr): ...this new member function.
9790 Add the tree to any pending optinfo, creating one if need be.
9791 (dump_generic_expr_loc): Move implementation to...
9792 (dump_context::dump_generic_expr_loc): ...this new member
9793 function. Add the tree to any pending optinfo, creating one if
9795 (dump_printf): Move implementation to...
9796 (dump_context::dump_printf_va): ...this new member function. Add
9797 the text to any pending optinfo, creating one if need be.
9798 (dump_printf_loc): Move implementation to...
9799 (dump_context::dump_printf_loc_va): ...this new member function.
9800 Start a new optinfo and add the stmt to it.
9801 (dump_dec): Move implementation to...
9802 (dump_context::dump_dec): ...this new member function. Add the
9803 value to any pending optinfo, creating one if need be.
9804 (dump_context::dump_symtab_node): New member function.
9805 (dump_context::get_scope_depth): New member function.
9806 (dump_context::begin_scope): New member function.
9807 (dump_context::end_scope): New member function.
9808 (dump_context::ensure_pending_optinfo): New member function.
9809 (dump_context::begin_next_optinfo): New member function.
9810 (dump_context::end_any_optinfo): New member function.
9811 (dump_context::s_current): New global.
9812 (dump_context::s_default): New global.
9813 (dump_scope_depth): Delete global.
9814 (dumpfile_ensure_any_optinfo_are_flushed): New function.
9815 (dump_symtab_node): New function.
9816 (get_dump_scope_depth): Reimplement in terms of dump_context.
9817 (dump_begin_scope): Likewise.
9818 (dump_end_scope): Likewise.
9819 (selftest::temp_dump_context::temp_dump_context): New ctor.
9820 (selftest::temp_dump_context::~temp_dump_context): New dtor.
9821 (selftest::verify_item): New function.
9822 (ASSERT_IS_TEXT): New macro.
9823 (ASSERT_IS_TREE): New macro.
9824 (ASSERT_IS_GIMPLE): New macro.
9825 (selftest::test_capture_of_dump_calls): New test.
9826 (selftest::dumpfile_c_tests): Call it.
9827 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
9828 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
9829 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
9830 descriptive comment.
9831 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
9832 (dump_node, dump_bb): Move these unrelated decls.
9833 (class dump_manager): Add leading comment.
9834 * optinfo.cc: New file.
9835 * optinfo.h: New file.
9837 2018-07-19 Michael Collison <michael.collison@arm.com>
9838 Richard Henderson <rth@redhat.com>
9840 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
9841 (subti): Handle op1 zero.
9842 (subvti4, usub4ti4): New.
9843 (*sub<GPI>3_compare1_imm): New.
9844 (sub<GPI>3_carryinCV): New.
9845 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
9846 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
9848 2018-07-19 Michael Collison <michael.collison@arm.com>
9849 Richard Henderson <rth@redhat.com>
9851 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
9852 (addti3): Create simpler code if low part is already known to be 0.
9853 (addvti4, uaddvti4): New.
9854 (*add<GPI>3_compareC_cconly_imm): New.
9855 (*add<GPI>3_compareC_cconly): New.
9856 (*add<GPI>3_compareC_imm): New.
9857 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
9858 handle constants within this pattern..
9859 (*add<GPI>3_compareV_cconly_imm): New.
9860 (*add<GPI>3_compareV_cconly): New.
9861 (*add<GPI>3_compareV_imm): New.
9862 (add<GPI>3_compareV): New.
9863 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
9864 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
9865 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
9866 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
9867 with 'comparison' operator.
9868 (*add<GPI>3_compareV_cconly_imm): Ditto.
9869 (*add<GPI>3_compareV_cconly): Ditto.
9870 (*add<GPI>3_compareV_imm): Ditto.
9871 (add<GPI>3_compareV): Ditto.
9872 (add<mode>3_carryinC): Ditto.
9873 (*add<mode>3_carryinC_zero): Ditto.
9874 (*add<mode>3_carryinC): Ditto.
9875 (add<mode>3_carryinV): Ditto.
9876 (*add<mode>3_carryinV_zero): Ditto.
9877 (*add<mode>3_carryinV): Ditto.
9879 2018-07-19 Michael Collison <michael.collison@arm.com>
9880 Richard Henderson <rth@redhat.com>
9882 * config/aarch64/aarch64-modes.def (CC_V): New.
9883 * config/aarch64/aarch64-protos.h
9884 (aarch64_addti_scratch_regs): Declare
9885 (aarch64_subvti_scratch_regs): Declare.
9886 (aarch64_expand_subvti): Declare.
9887 (aarch64_gen_unlikely_cbranch): Declare
9888 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
9889 for signed overflow using CC_Vmode.
9890 (aarch64_get_condition_code_1): Handle CC_Vmode.
9891 (aarch64_gen_unlikely_cbranch): New function.
9892 (aarch64_addti_scratch_regs): New function.
9893 (aarch64_subvti_scratch_regs): New function.
9894 (aarch64_expand_subvti): New function.
9896 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
9898 * config/aarch64/aarch64-option-extensions.def: New entry for profile
9900 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
9901 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
9904 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
9907 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
9908 address check not strict.
9910 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
9912 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
9914 (store_pair_lanes<mode>): Likewise.
9915 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
9916 enum value 'ADDR_QUERY_LDP_STP_N'.
9917 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
9918 (aarch64_print_address_internal): Add declaration.
9919 (aarch64_print_ldpstp_address): Remove.
9920 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
9921 (aarch64_print_operand): Change printing of 'y'.
9922 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
9923 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
9924 'true' rather than '1'.
9925 * config/aarch64/constraints.md (Uml): Likewise.
9926 (Uml): Rename to Umn.
9929 2018-07-19 Richard Biener <rguenther@suse.de>
9931 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
9933 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
9934 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
9935 (vn_reference_hasher): Likewise.
9936 (struct vn_tables_s): Remove obstack and alloc-pool members.
9937 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
9938 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
9939 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
9940 (vn_reference_insert_pieces): Likewise.
9941 (alloc_vn_nary_op_noinit): Adjust.
9942 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
9943 (vn_phi_eq): Adjust.
9944 (shared_lookup_phiargs): Remove.
9945 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
9946 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
9947 (visit_reference_op_call): Likewise.
9948 (copy_nary, copy_phi, copy_reference): Remove.
9949 (process_scc): Rewind the obstack when iterating. Do not
9950 copy the elements to valid_info but just move them from one
9951 hashtable to the other.
9952 (allocate_vn_table): Adjust.
9953 (free_vn_table): Likewise.
9954 (init_scc_vn): Likewise.
9955 (free_scc_vn): Likewise.
9957 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
9960 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
9961 indirect_return as function type attribute.
9962 (ix86_attribute_table): Change indirect_return to function
9964 * doc/extend.texi: Update indirect_return attribute.
9966 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
9968 * wide-int.h (widest2_int): New.
9969 * gimple-fold.c (arith_overflowed_p): Use it.
9970 * tree.h (widest2_int_cst): New.
9971 * tree-vrp.c (wide_int_binop_overflow): Rename from
9972 vrp_int_const_binop.
9973 Rewrite to work on trees.
9974 (extract_range_from_multiplicative_op_1): Abstract code to...
9975 (wide_int_range_min_max): ...here.
9976 (wide_int_range_cross_product): ...and here.
9977 (extract_range_from_binary_expr_1): Abstract overflow code to...
9978 (wide_int_range_mult_wrapping): ...here.
9979 * tree-vrp.h (wide_int_range_cross_product): New.
9980 (wide_int_range_mult_wrapping): New.
9982 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
9983 Julia Koval <julia.koval@intel.com>
9985 * config/i386/x86-tune-costs.h (skylake_memcpy,
9986 skylake_memset): Replace rep_prefix with unrolling for size 512.
9988 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
9991 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
9992 comparision with EQ_EXPR in last stmt.
9994 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
9996 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
9997 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
9998 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
9999 previously known as "PowerPC AltiVec Built-in Functions". Move
10000 some material to new subsubsections "PowerPC AltiVec Built-in
10001 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
10003 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
10004 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
10005 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
10006 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
10008 2018-07-18 Richard Biener <rguenther@suse.de>
10010 PR tree-optimization/86557
10011 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
10014 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
10016 * config/s390/s390.c (s390_function_profiler): Generate CFI.
10018 2018-07-17 Jeff Law <law@redhat.com>
10020 * config/arm/arm.c (get_label_padding): Update for recent
10021 changes to label_to_alignment.
10023 PR tree-optimization/86010
10024 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
10026 * config/mips/mips.c (vr4130_align_insns): Update for recent
10027 changes to label_to_alignment.
10029 * config/frv/frv.c (frv_label_align): Update for recent changes
10030 to label_to_alignment.
10032 * config/nios2/nios2.c (nios2_label_align): Update for recent
10033 changes which dropped ALIGN_LABELS_LOG.
10035 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
10037 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
10038 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
10040 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
10042 * config/arc/arc.c (arc_label_align): Use align_labels instead of
10043 deprecated align_labels_log.
10045 2018-07-17 Richard Biener <rguenther@suse.de>
10048 * dwarf2out.c (init_sections_and_labels): Always generate
10049 a debug_line_str_section for early LTO debug.
10050 (dwarf2out_finish): Reset debug_line_str_hash output early.
10051 Bump counter for extra dwarf5 .debug_loc labels to not conflict
10053 (dwarf2out_early_finish): Output debug_line_str.
10055 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
10057 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
10058 index register on z196 or later.
10060 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
10062 * config/s390/s390.c (s390_default_align): Set default function
10064 (s390_override_options_after_change): Call s390_default align.
10065 (s390_option_override_internal): Call s390_default_align.
10066 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
10068 2018-07-17 Jakub Jelinek <jakub@redhat.com>
10070 PR middle-end/86542
10071 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
10072 to _looptemp_ clauses, other than the first two.
10074 2018-07-17 Martin Liska <mliska@suse.cz>
10076 * opts.c: Do not enable OPT_falign_* for -Os.
10078 2018-07-17 Martin Liska <mliska@suse.cz>
10080 * align.h (MAX_CODE_ALIGN): New.
10081 (MAX_CODE_ALIGN_VALUE): New.
10082 * common/config/i386/i386-common.c (ix86_handle_option):
10083 (MAX_CODE_ALIGN): Moved to align.h.
10084 * final.c (MAX_CODE_ALIGN): Likewise.
10085 * opts.c (parse_and_check_align_values):
10086 (MAX_CODE_ALIGN): Likewise.
10087 (MAX_CODE_ALIGN_VALUE): Likewise.
10089 2018-07-17 Martin Liska <mliska@suse.cz>
10091 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
10092 in order to fulfil coding style.
10093 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
10094 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10095 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
10096 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
10097 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
10098 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
10099 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
10100 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
10101 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10103 2018-07-17 Martin Liska <mliska@suse.cz>
10105 * align.h: New file.
10106 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
10108 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
10109 align_flags of label_to_alignment.
10110 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
10112 * config/m68k/m68k.c: Do not use removed align_labels_value and
10114 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
10115 (LOOP_ALIGN): Likewise.
10116 (LABEL_ALIGN): Likewise.
10117 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
10118 Remove not used macro.
10119 (rs6000_loop_align): Change return type to align_flags.
10120 (rs6000_loop_align_max_skip): Remove.
10121 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
10122 Change return type to align_flags.
10123 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
10124 Remove not used macro.
10125 (rs6000_loop_align): Change return type to align_flags.
10126 (rs6000_loop_align_max_skip): Remove.
10127 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
10128 * config/rx/rx-protos.h (rx_align_for_label): Make it
10130 * config/rx/rx.c (rx_align_for_label): Change return type
10132 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
10134 into align_flags class.
10135 (LABEL_ALIGN): Likewise.
10136 (LOOP_ALIGN): Likewise.
10137 * config/s390/s390.c (s390_label_align): Use align_flags
10139 (s390_asm_output_function_label): Likewise.
10140 * config/sh/sh.c (sh_override_options_after_change):
10141 Use align_flags class directly without macros.
10142 (find_barrier): Likewise.
10143 (barrier_align): Likewise.
10144 (sh_loop_align): Likewise.
10145 * config/spu/spu.c (spu_option_override):
10146 Use align_flags_tuple::get_value instead of removed macros.
10147 (spu_sched_init): Likewise.
10148 * config/spu/spu.h (GTY): Likewise.
10149 * config/visium/visium.c (visium_option_override):
10150 Set "8" as default secondary alignment.
10151 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
10152 in order to guarantee secondary alignment of 8.
10153 * coretypes.h: Include align.h header file.
10154 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
10155 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
10156 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
10157 * doc/tm.texi.in: Likewise.
10158 * final.c (struct label_alignment): Remove not used structure.
10159 (LABEL_ALIGN): Change type to align_flags.
10160 (LOOP_ALIGN): Likewise.
10161 (JUMP_ALIGN): Likewise.
10162 (default_loop_align_max_skip): Remove.
10163 (default_label_align_max_skip): Likewise.
10164 (default_jump_align_max_skip): Likewise.
10165 (default_label_align_after_barrier_max_skip):
10166 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
10167 (LABEL_TO_MAX_SKIP): Remove.
10168 (label_to_alignment): Return align_flags type instead of integer.
10169 (label_to_max_skip): Remove.
10170 (align_fuzz): Use align_flags type.
10171 (compute_alignments): Use align_flags type and use align_flags::max
10172 to combine multiple alignments.
10173 (grow_label_align): Grow vec instead of C array.
10174 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
10175 (shorten_branches): Use align_flags type and use align_flags::max
10176 to combine multiple alignments.
10177 (final_scan_insn_1): Remove usage of secondary alignment that comes
10178 from label alignment, but instead use proper secondary alignment
10179 which is computed in grow_label_align.
10180 * flags.h (struct align_flags_tuple): Move to align.h.
10181 (struct align_flags): Likewise.
10182 (state_align_loops): Rename to align_loops.
10183 (state_align_jumps): Rename to align_jumps.
10184 (state_align_labels): Rename to align_labels.
10185 (state_align_functions): Rename to align_functions.
10186 (align_loops_log): Remove.
10187 (align_jumps_log): Remove.
10188 (align_labels_log): Remove.
10189 (align_functions_log): Remove.
10190 (align_loops_max_skip): Remove.
10191 (align_jumps_max_skip): Remove.
10192 (align_labels_max_skip): Remove.
10193 (align_functions_max_skip): Remove.
10194 (align_loops_value): Remove.
10195 (align_jumps_value): Remove.
10196 (align_labels_value): Remove.
10197 (align_functions_value): Remove.
10198 * output.h (label_to_alignment): Change return type to align_flags.
10199 (label_to_max_skip): Remove.
10200 * target.def: Remove loop_align_max_skip, label_align_max_skip,
10201 jump_align_max_skip macros.
10202 * targhooks.h (default_loop_align_max_skip): Remove.
10203 (default_label_align_max_skip): Likewise.
10204 (default_jump_align_max_skip): Likewise.
10205 (default_label_align_after_barrier_max_skip): Remove.
10206 * toplev.c (read_log_maxskip): Use ::normalize function.
10207 (parse_N_M): Remove not used argument and also call ::normalize.
10208 (parse_alignment_opts): Do not pass unused arguments.
10209 * varasm.c (assemble_start_function): Use directly align_functions
10210 instead of removed macros.
10211 * system.h: Do not poison removed macros.
10213 2018-07-17 Jakub Jelinek <jakub@redhat.com>
10215 PR middle-end/86539
10216 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
10217 and cond temporaries don't have reference type if iterator has
10218 pointer type. For init use &for_pre_body instead of pre_p if
10219 for_pre_body is non-empty.
10221 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
10223 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
10224 double-double modes to SFmode directly directly.
10225 (trunc<mode>sf2_fprs): Delete.
10227 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
10229 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
10230 for conversions between IFmode and the decimal floating point modes.
10231 (init_float128_ieee): Use the correct names for conversions between
10232 KFmode and the decimal floating point modes.
10234 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
10236 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
10237 for the conversions between TDmode and IFmode.
10238 (init_float128_ieee): Use more correct names for the conversions
10239 between TDmode and KFmode.
10241 2018-07-16 Jakub Jelinek <jakub@redhat.com>
10243 PR tree-optimization/86526
10244 * builtins.c (expand_builtin_memcmp): Formatting fixes.
10245 (inline_expand_builtin_string_cmp): Likewise.
10246 (inline_string_cmp): Likewise. Use c_readstr instead of
10247 builtin_memcpy_read_str. Add unit_mode temporary.
10249 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
10251 PR middle-end/86528
10252 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
10253 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
10255 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
10257 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
10258 Alphabetize prototypes of built-in functions, separating out
10259 built-in functions that are listed in this section but should be
10260 described elsewhere.
10262 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
10265 * expmed.c (emit_store_flag): Do not emit setcc followed by a
10266 conditional move when trapping comparison was split to a
10267 non-trapping one (and vice versa).
10269 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
10271 * config/s390/s390.c (s390_function_profiler): Generate nops
10272 instead of profiler call sequences.
10273 * config/s390/s390.opt: Add the new option.
10275 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
10277 * config/s390/s390.c (s390_function_profiler): Generate
10278 __mcount_loc section.
10279 * config/s390/s390.opt: Add the new option.
10281 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
10283 * common.opt: Add the new warning.
10284 * config/s390/s390.c (s390_function_profiler): Emit "brasl
10285 %r0,__fentry__" when -mfentry is specified.
10286 (s390_option_override_internal): Disallow -mfentry for 31-bit
10288 * config/s390/s390.opt: Add the new option.
10290 2018-07-16 Richard Biener <rguenther@suse.de>
10293 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
10294 for function-local FUNCTION_DECL and RESULT_DECL immediately.
10296 2018-07-16 Martin Liska <mliska@suse.cz>
10299 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
10302 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
10304 * config/arc/arcHS.md: Update ARCHS scheduling rules.
10306 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
10308 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
10310 * config/arc/arc-cpus.def (hs4x): New cpu.
10312 * config/arc/arc-tables.opt: Regenerate.
10313 * config/arc/arc.c (arc_sched_issue_rate): New function.
10314 (TARGET_SCHED_ISSUE_RATE): Define.
10315 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
10316 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
10318 (attr tune): Add ARCHS4x tune values.
10319 (attr tune_dspmpy): Define.
10320 (*tst): Correct instruction type.
10321 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
10322 * config/arc/arcHS4x.md: New file.
10323 * config/arc/fpu.md: Update instruction type attributes.
10324 * config/arc/t-multilib: Regenerate.
10326 2018-07-16 Tom de Vries <tdevries@suse.de>
10329 * var-tracking.c (vt_initialize): Fix pre_dec handling.
10331 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10333 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
10336 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
10338 PR tree-optimization/86514
10339 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
10340 conversion to a boolean type from a type with greater precision.
10342 2018-07-16 Tom de Vries <tdevries@suse.de>
10344 * var-tracking.c (vt_initialize): Print adjusted insn slim if
10345 dump_flags request TDF_SLIM.
10347 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
10349 * fold-const.c (int_const_binop_1): Abstract...
10350 (wide_int_binop): ...wide int code here.
10351 (poly_int_binop): ...poly int code here.
10352 Abstract the rest of int_const_binop_1 into int_const_binop.
10353 * fold-const.h (wide_int_binop): New.
10354 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
10355 Remove useless PLUS/MINUS_EXPR case.
10356 (zero_nonzero_bits_from_vr): Move wide int code...
10357 (zero_nonzero_bits_from_bounds): ...here.
10358 (extract_range_from_binary_expr_1): Move mask optimization code...
10359 (range_easy_mask_min_max): ...here.
10360 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
10361 (range_easy_mask_min_max): New.
10363 2018-07-15 Jeff Law <law@redhat.com>
10366 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
10369 2018-07-14 Jim Wilson <jimw@sifive.com>
10371 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
10373 2018-07-14 Paul Koning <ni1d@arrl.net>
10375 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
10377 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
10379 * lto-streamer-out.c (copy_function_or_variable): Dump info about
10382 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
10383 Steve Munroe <munroesj52@gmail.com>
10385 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
10386 (_mm_andnot_si128): Likewise.
10387 (_mm_or_si128): Likewise.
10388 (_mm_xor_si128): Likewise.
10390 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
10392 PR middle-end/78809
10393 * builtins.c (expand_builtin_memcmp): Inline the calls first
10394 when result_eq is false.
10395 (expand_builtin_strcmp): Inline the calls first.
10396 (expand_builtin_strncmp): Likewise.
10397 (inline_string_cmp): New routine. Expand a string compare
10398 call by using a sequence of char comparison.
10399 (inline_expand_builtin_string_cmp): New routine. Inline expansion
10400 a call to str(n)cmp/memcmp.
10401 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
10403 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
10405 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
10407 * config/arm/driver-arm.c: Include arm-native.h.
10408 (host_detect_local_cpu): Use auto-generated data tables.
10409 (vendors, arm_cpu_table): Delete. Move part information to ...
10410 * config/arm/arm-cpus.in: ... here.
10411 * config/arm/parsecpu.awk (gen_native): New function.
10412 (vendor, part): New CPU fields.
10413 (END): Add support for building the native CPU detection tables.
10414 * config/arm/t-arm (arm-native.h): Add build rule.
10415 (driver-arm.o): Add dependency on arm-native.h.
10417 2018-07-13 Richard Biener <rguenther@suse.de>
10419 PR middle-end/85974
10420 * match.pd (addr1 - addr2): Allow either of the operand to
10423 2018-07-13 Tom de Vries <tdevries@suse.de>
10425 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
10428 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
10430 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
10431 arrays instead of numbered variables.
10433 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
10435 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
10436 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
10438 2018-07-13 Richard Biener <rguenther@suse.de>
10441 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
10442 instead of get_context_die.
10444 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
10445 Richard Biener <rguenther@suse.de>
10447 PR middle-end/86489
10448 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
10449 that the loop latch destination where phi is defined.
10451 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
10453 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
10454 (riscv_expand_epilogue): Add assertion to check interrupt mode.
10455 (riscv_set_current_function): Extract getting interrupt type to new
10457 (riscv_get_interrupt_type): New function.
10458 (riscv_merge_decl_attributes): New function, checking interrupt type is
10460 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
10462 2018-07-12 Paul Koning <ni1d@arrl.net>
10464 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
10467 2018-07-12 Paul Koning <ni1d@arrl.net>
10469 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
10470 zero reference, add doloop_end instead.
10471 * doc/md.texi (decrement_and_branch_until_zero): Remove.
10472 (Looping patterns): Remove decrement_and_branch_until_zero. Add
10473 detail for doloop_end.
10475 2018-07-12 Martin Sebor <msebor@redhat.com>
10478 * attribs.c (decl_attributes): Reject conflicting attributes before
10479 calling attribute handlers.
10481 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
10483 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
10485 (gcc::dump_manager::get_dump_file_name): likewise.
10486 (dump_begin): Likewise.
10487 * dumpfile.h (dump_begin): Update prototype.
10488 (gcc::dump_manager::get_dump_file_name,
10489 gcc::dump_manager::get_dump_file_name): Update prototype.
10491 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
10493 * internal-fn.h (vectorizable_internal_fn_p): New function.
10494 * tree-vect-slp.c (compatible_calls_p): Likewise.
10495 (vect_build_slp_tree_1): Remove nops argument. Handle calls
10496 to internal functions.
10497 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
10499 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
10501 * fold-const.h (inverse_conditions_p): Declare.
10502 * fold-const.c (inverse_conditions_p): New function.
10503 * match.pd: Use inverse_conditions_p. Add folds of view_converts
10504 that test the inverse condition of a conditional internal function.
10505 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
10506 * internal-fn.c (internal_fn_mask_index): Handle conditional
10507 internal functions.
10508 (vectorized_internal_fn_supported_p): New function.
10509 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
10510 (any_pred_load_store): Replace with...
10511 (need_to_predicate): ...this new variable.
10512 (redundant_ssa_names): New variable.
10513 (ifcvt_can_use_mask_load_store): Move initial checks to...
10514 (ifcvt_can_predicate): ...this new function. Handle tree codes
10515 for which a conditional internal function exists.
10516 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
10517 instead of ifcvt_can_use_mask_load_store. Update after variable
10519 (predicate_load_or_store): New function, split out from
10520 predicate_mem_writes.
10521 (check_redundant_cond_expr): New function.
10522 (value_available_p): Likewise.
10523 (predicate_rhs_code): Likewise.
10524 (predicate_mem_writes): Rename to...
10525 (predicate_statements): ...this. Use predicate_load_or_store
10526 and predicate_rhs_code.
10527 (combine_blocks, tree_if_conversion): Update after above name changes.
10528 (ifcvt_local_dce): Handle redundant_ssa_names.
10529 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
10530 general conditional functions.
10531 * tree-vect-stmts.c (vectorizable_call): Likewise.
10533 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
10534 Alan Hayward <alan.hayward@arm.com>
10535 David Sherwood <david.sherwood@arm.com>
10537 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
10538 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
10539 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
10540 plus and minus and convert them into IFN_COND_FMA-based sequences.
10541 (convert_mult_to_fma): Handle conditional plus and minus.
10543 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
10545 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
10546 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
10547 (cond_fnms_optab): New optabs.
10548 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
10549 internal functions.
10550 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
10551 * internal-fn.h (get_conditional_internal_fn): Declare.
10552 (get_unconditional_internal_fn): Likewise.
10553 * internal-fn.c (cond_ternary_direct): New macro.
10554 (expand_cond_ternary_optab_fn): Likewise.
10555 (direct_cond_ternary_optab_supported_p): Likewise.
10556 (FOR_EACH_COND_FN_PAIR): Likewise.
10557 (get_conditional_internal_fn): New function.
10558 (get_unconditional_internal_fn): Likewise.
10559 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
10560 (gimple_match_op::gimple_match_op): Add a new overload for 5
10562 (gimple_match_op::set_op): Likewise.
10563 (gimple_resimplify5): Declare.
10564 * genmatch.c (decision_tree::gen): Generate simplifications for
10566 * gimple-match-head.c (gimple_simplify): Define an overload for
10567 5 operands. Handle calls with 5 arguments in the top-level overload.
10568 (convert_conditional_op): Handle conversions from unconditional
10569 internal functions to conditional ones.
10570 (gimple_resimplify5): New function.
10571 (build_call_internal): Pass a fifth operand.
10572 (maybe_push_res_to_seq): Likewise.
10573 (try_conditional_simplification): Try converting conditional
10574 internal functions to unconditional internal functions.
10575 Handle 3-operand unconditional forms.
10576 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
10577 Define ternary equivalents of the current rules for binary conditional
10578 internal functions.
10579 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
10580 ternary operations.
10581 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
10582 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
10583 (optab): Handle them.
10584 (SVE_COND_FP_TERNARY): New int iterator.
10585 (sve_fmla_op, sve_fmad_op): New int attributes.
10586 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
10587 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
10588 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
10590 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
10592 * target.def (preferred_else_value): New target hook.
10593 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
10594 * doc/tm.texi: Regenerate.
10595 * targhooks.h (default_preferred_else_value): Declare.
10596 * targhooks.c (default_preferred_else_value): New function.
10597 * internal-fn.h (conditional_internal_fn_code): Declare.
10598 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
10599 (get_conditional_internal_fn): Use it.
10600 (conditional_internal_fn_code): New function.
10601 * gimple-match.h (gimple_match_cond): New struct.
10602 (gimple_match_op): Add a cond member function.
10603 (gimple_match_op::gimple_match_op): Update all forms to take a
10605 * genmatch.c (expr::gen_transform): Use the same condition as res_op
10606 for the suboperation, but don't specify a particular else_value.
10607 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
10608 (visit_nary_op, visit_reference_op_load): Pass
10609 gimple_match_cond::UNCOND to the gimple_match_op constructor.
10610 * gimple-match-head.c: Include tree-eh.h
10611 (convert_conditional_op): New function.
10612 (maybe_resimplify_conditional_op): Likewise.
10613 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
10614 (gimple_resimplify2): Likewise.
10615 (gimple_resimplify3): Likewise.
10616 (gimple_resimplify4): Likewise.
10617 (maybe_push_res_to_seq): Return null for conditional operations.
10618 (try_conditional_simplification): New function.
10619 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
10621 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
10623 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
10625 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
10627 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
10629 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
10631 (hash_tree): Do not hash DECL_FCONTEXT
10632 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
10633 Do not stream DECL_FCONTEXT.
10634 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
10635 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
10637 2018-07-12 Richard Biener <rguenther@suse.de>
10640 * dwarf2out.c (gen_block_die): Only output blocks when they have
10641 at least one !DECL_IGNORED_P variable.
10643 2018-07-12 Richard Biener <rguenther@suse.de>
10646 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
10647 Remove -mieee-fp handling.
10649 2018-07-12 Richard Biener <rguenther@suse.de>
10651 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
10652 left-over from last patch.
10654 2018-07-12 Jakub Jelinek <jakub@redhat.com>
10656 PR tree-optimization/86492
10657 * gimple-ssa-store-merging.c
10658 (imm_store_chain_info::coalesce_immediate_stores): Call
10659 check_no_overlap even for the merge_overlapping case. Formatting fix.
10661 2018-07-12 Richard Biener <rguenther@suse.de>
10663 PR middle-end/86479
10664 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
10665 move possibly trapping operations into the conditional.
10667 2018-07-12 Richard Biener <rguenther@suse.de>
10669 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
10670 (vn_lookup_simplify_result): Remove recursion limit applied
10672 (vn_nary_build_or_lookup_1): Adjust.
10673 (try_to_simplify): Likewise.
10674 * gimple-match-head.c (gimple_resimplify1): Instead apply one
10676 (gimple_resimplify2): Likewise.
10677 (gimple_resimplify3): Likewise.
10678 (gimple_resimplify4): Likewise.
10680 2018-07-11 Jakub Jelinek <jakub@redhat.com>
10682 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
10683 Use __mmask64 type instead of __mmask8 for __M argument.
10684 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
10685 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
10687 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
10688 __mmask16 for __M argument.
10689 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
10690 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
10691 to __mmask16 instead of __mmask8.
10692 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
10693 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
10694 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
10695 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
10696 instead of __mmask16 for __U argument.
10697 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
10698 __mmask16 instead of __mmask8 for __U argument.
10699 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
10701 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
10703 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
10705 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
10707 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
10709 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
10711 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
10713 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
10715 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
10717 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
10719 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
10720 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
10721 return type as well as __M argument type and all casts from __mmask8
10723 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
10724 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
10725 return type as well as __M argument type and all casts from __mmask8
10727 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
10728 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
10729 return type as well as __M argument type and all casts from __mmask8
10731 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
10732 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
10733 return type as well as __M argument type and all casts from __mmask8
10735 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
10736 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
10739 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
10741 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
10742 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
10745 2018-07-11 Paul Koning <ni1d@arrl.net>
10747 * doc/md.texi (define_subst): Document how multiple occurrences of
10748 the same argument in the replacement pattern are handled.
10750 2018-07-11 Paul Koning <ni1d@arrl.net>
10752 * doc/extend.texi (Common Variable Attributes): Move "mode" into
10753 alphabetical order.
10754 (Common Type Attributes): Add "mode" attribute.
10756 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
10758 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
10759 stream DECL_ORIGINAL_TYPE.
10760 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
10761 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
10762 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
10763 Do not walk original type.
10764 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
10766 (write_ts_decl_non_common_tree_pointers): Do not stream
10768 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
10769 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
10771 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
10773 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
10774 thread twice from the same starting edge.
10776 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
10778 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
10779 * gimple.c (gimple_call_nonnull_result_p): ...here...
10780 (gimple_call_nonnull_arg): ...and here.
10781 * gimple.h (gimple_call_nonnull_result_p): New.
10782 (gimple_call_nonnull_arg): New.
10784 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
10786 * config/arm/arm-cpus.in: Move information from fpu field of each
10787 cpu definition to the isa field.
10788 * config/arm/parsecpu.awk (fpu): Delete match rule.
10789 (gen_comm_data): Don't add bits from the CPU's FPU entry.
10791 2018-07-11 Richard Biener <rguenther@suse.de>
10794 * dwarf2out.c (init_sections_and_labels): Use
10795 output_asm_line_debug_info consistently.
10796 (dwarf2out_early_finish): Likewise.
10797 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
10800 2018-07-11 Richard Biener <rguenther@suse.de>
10802 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
10803 Rework father_bb setting in a way to avoid propagating constants
10804 multiple times on a loop body.
10806 2018-07-10 Mark Wielaard <mark@klomp.org>
10809 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
10811 2018-07-10 Richard Biener <rguenther@suse.de>
10813 * hash-map.h (hash_map::iterator::operator*): Return
10814 references to key and value.
10816 2018-07-10 Jakub Jelinek <jakub@redhat.com>
10819 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
10820 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
10821 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
10822 (gimplify_omp_for): For composite loops, move outer
10823 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
10824 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
10825 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
10826 TREE_LIST for both the original class iterator and the "last" helper
10827 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
10828 loop, remember has_decl_expr from outer composite loops for the
10829 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
10831 2018-07-09 Martin Sebor <msebor@redhat.com>
10833 PR middle-end/77357
10834 PR middle-end/86428
10835 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
10836 accessing implicitly initialized array elements.
10837 * expr.c (string_constant): Handle string initializers of
10838 character arrays within aggregates.
10839 * gimple-fold.c (fold_array_ctor_reference): Add argument.
10840 Store element offset. As a special case, handle zero size.
10841 (fold_nonarray_ctor_reference): Same.
10842 (fold_ctor_reference): Add argument. Store subobject offset.
10843 * gimple-fold.h (fold_ctor_reference): Add argument.
10845 2018-07-09 Paul Koning <ni1d@arrl.net>
10847 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
10848 (pdp11_insn_cost): New function.
10849 (pdp11_md_asm_adjust): New function.
10850 (TARGET_INVALID_WITHIN_DOLOOP): Define.
10851 (pdp11_rtx_costs): Update to match machine better.
10852 (output_addr_const_pdp11): Correct format mismatch warnings.
10853 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
10854 * config/pdp11/pdp11.md: General change to add base_cost and/or
10855 length attributes for use by new pdp11_insn_cost function.
10856 (MIN_BRANCH): Correct definition.
10858 (doloop_end): Use standard pattern name for looping pattern.
10859 (doloop_end_nocc): New.
10860 (movsf): Add another constraint alternative.
10861 (zero_extendqihi2): Add constraint alternatives for not in place
10863 (zero_extendhisi2): Remove.
10864 (shift patterns): Add CC handling variants.
10868 (define_peephole2): New peephole to recognize mov that sets CC for
10871 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10873 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
10874 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
10877 2018-07-09 Jakub Jelinek <jakub@redhat.com>
10880 * real.c (real_nextafter): Return true if result is denormal.
10882 2018-07-09 Martin Liska <mliska@suse.cz>
10884 * common.opt: Add back wrongly removed attribute.
10886 2018-07-09 Richard Biener <rguenther@suse.de>
10889 * dwarf2out.c (gen_block_die): For an early generated DIE
10890 always output high/low PC attributes.
10892 2018-07-09 Tom de Vries <tdevries@suse.de>
10894 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
10895 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
10896 onto VAR_DECL with abstract origin.
10898 2018-07-07 Jim Wilson <jimw@sifive.com>
10900 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
10902 2018-07-07 Tom de Vries <tdevries@suse.de>
10904 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
10906 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
10908 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
10910 (combine_bound): Use wide-int overflow calculation instead of
10912 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
10914 * fold-const.c (int_const_binop_2): Same.
10915 (extract_muldiv_1): Same.
10916 (fold_div_compare): Same.
10917 (fold_abs_const): Same.
10919 * poly-int.h (add): Same.
10923 * predict.c (predict_iv_comparison): Same.
10924 * profile-count.c (slow_safe_scale_64bit): Same.
10925 * simplify-rtx.c (simplify_const_binary_operation): Same.
10926 * tree-chrec.c (tree_fold_binomial): Same.
10927 * tree-data-ref.c (split_constant_offset_1): Same.
10928 * tree-if-conv.c (idx_within_array_bound): Same.
10929 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
10930 * tree-ssa-phiopt.c (minmax_replacement): Same.
10931 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
10932 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
10933 * vr-values.c (vr_values::adjust_range_with_scev): Same.
10934 * wide-int.cc (wi::add_large): Same.
10935 (wi::mul_internal): Same.
10936 (wi::sub_large): Same.
10937 (wi::divmod_internal): Same.
10938 * wide-int.h: Change overflow type to overflow_type for neg, add,
10939 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
10940 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
10941 mul_internal, divmod_internal.
10942 (overflow_type): New enum.
10943 (accumulate_overflow): New.
10945 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
10947 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
10948 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
10950 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
10952 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
10953 argument is checked for zero before entering loop, avoid checking again.
10955 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
10957 * gimplify.h (generic_expr_could_trap_p): Set as global function.
10958 * gimplify.h (generic_expr_could_trap_p): Likwise.
10959 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
10961 2018-07-06 Jakub Jelinek <jakub@redhat.com>
10963 PR tree-optimization/86401
10964 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
10965 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
10966 (fold_bit_and_mask): ... here. New helper function for match.pd.
10967 * fold-const.h (fold_bit_and_mask): Declare.
10968 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
10970 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
10973 * target.def (translate_mode_attribute): New hook.
10974 * targhooks.h (default_translate_mode_attribute): Declare.
10975 * targhooks.c (default_translate_mode_attribute): New function.
10976 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
10977 * doc/tm.texi: Regenerate.
10978 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
10979 (rs6000_translate_mode_attribute): New function.
10981 2018-07-06 Paul Koning <ni1d@arrl.net>
10983 * doc/md.texi (define_split): Document DONE and FAIL.
10984 (define_peephole2): Ditto.
10986 2018-07-05 Jeff Law <law@redhat.com>
10988 PR tree-optimization/86010
10989 * tree-ssa-dse.c (compute_trims): More aggressively trim at
10990 both the head and tail of mem* and str* calls.
10992 2018-07-05 Jim Wilson <jimw@sifive.com>
10994 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
10996 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
10998 * config/aarch64/aarch64-simd.md: correct flags text for
10999 MIN_EXPR replacement.
11001 2018-07-05 James Clarke <jrtc27@jrtc27.com>
11003 * configure: Regenerated.
11005 2018-07-05 Carl Love <cel@us.ibm.com>
11007 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
11008 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
11009 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
11010 VSX_BUILTIN_DOUBLEL_V4SF.
11012 2018-07-05 Martin Sebor <msebor@redhat.com>
11015 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
11016 than its domain to compute its the upper bound of a char array.
11018 2018-07-05 Nathan Sidwell <nathan@acm.org>
11020 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
11021 * doc/cpp.texi: Update comment.
11022 * doc/tm.texi: Rebuilt.
11023 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
11024 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
11025 * doc/extend.texi (Backwards Compatibility): Clarify it is system
11026 headers affected by extern "C".
11027 * system.h: Poison NO_IMPLICIT_EXTERN_C.
11028 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
11029 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
11030 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
11031 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
11032 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
11033 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
11034 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
11035 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
11036 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
11037 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
11038 config/sparc/sp64-elf.h, config/spu/spu.h,
11039 config/stormy16/stormy16.h, config/v850/v850.h,
11040 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
11041 define NO_IMPLICIT_EXTERN_C.
11042 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
11044 2018-07-05 Tamar Christina <tamar.christina@arm.com>
11047 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
11048 instead of GET_MODE_SIZE when comparing Units.
11050 2018-07-05 Tamar Christina <tamar.christina@arm.com>
11053 * rtlanal.c (set_noop_p): Constrain on mode change,
11054 include hard-reg-set.h
11056 2018-07-05 Tamar Christina <tamar.christina@arm.com>
11058 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
11060 2018-07-05 Jakub Jelinek <jakub@redhat.com>
11063 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
11066 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
11068 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
11070 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
11073 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
11075 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
11077 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
11079 PR middle-end/86380
11080 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
11082 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
11084 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
11085 neg_*_op* variables.
11087 2018-07-04 Martin Liska <mliska@suse.cz>
11089 * tree-switch-conversion.c: Define
11090 max_ratio_for_speed and max_ratio_for_size constants.
11092 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
11093 Martin Liska <mliska@suse.cz>
11095 PR middle-end/66240
11098 * common.opt: Rename align options with 'str_' prefix.
11099 * common/config/i386/i386-common.c (set_malign_value): New
11101 (ix86_handle_option): Use it to set -falign-* options/
11102 * config/aarch64/aarch64-protos.h (struct tune_params): Change
11103 type from int to string.
11104 * config/aarch64/aarch64.c: Update default values from int
11106 * config/alpha/alpha.c (alpha_override_options_after_change):
11108 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
11109 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
11110 max skip conditionally.
11111 * config/i386/freebsd.h (SUBALIGN_LOG): New.
11112 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
11113 max skip conditionally.
11114 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
11115 max skip conditionally.
11116 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
11117 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
11118 max skip conditionally.
11119 * config/i386/i386.c (struct ptt): Change type from int to
11121 (ix86_default_align): Set default values.
11122 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
11123 max skip conditionally.
11124 * config/i386/iamcu.h (SUBALIGN_LOG): New.
11125 (ASM_OUTPUT_MAX_SKIP_ALIGN):
11126 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
11127 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
11128 max skip conditionally.
11129 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
11130 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
11131 * config/i386/x86-64.h (SUBALIGN_LOG): New.
11132 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
11133 max skip conditionally.
11134 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
11135 * config/ia64/ia64.c (ia64_option_override): Set default values
11136 for alignment options.
11137 * config/m68k/m68k.c: Handle new str_align_* options.
11138 * config/mips/mips.c (mips_set_compression_mode): Change
11140 (mips_option_override): Set default values for options.
11141 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
11143 * config/rs6000/rs6000.c (rs6000_option_override_internal):
11145 * config/rx/rx.c (rx_option_override): Likewise.
11146 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
11147 (LABEL_ALIGN): Use align_labels_log.
11148 (LOOP_ALIGN): Use align_loops_align.
11149 * config/s390/s390.c (s390_asm_output_function_label): Use new
11151 * config/sh/sh.c (sh_override_options_after_change):
11152 Change type of constants.
11153 * config/spu/spu.c (spu_sched_init): Likewise.
11154 * config/sparc/sparc.c (sparc_option_override): Set default
11155 values for options.
11156 * config/visium/visium.c (visium_option_override): Likewise.
11157 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
11158 emit p2align format with last argument if it's not needed.
11159 * doc/invoke.texi: Document extended format of -falign-*.
11160 * final.c: Use align_labels alignment.
11161 * flags.h (struct target_flag_state): Change type to use
11163 (struct align_flags_tuple): New.
11164 (struct align_flags): Likewise.
11165 (align_loops_log): Redefine macro to use new types.
11166 (align_loops_max_skip): Redefine macro to use new types.
11167 (align_jumps_log): Redefine macro to use new types.
11168 (align_jumps_max_skip): Redefine macro to use new types.
11169 (align_labels_log): Redefine macro to use new types.
11170 (align_labels_max_skip): Redefine macro to use new types.
11171 (align_functions_log): Redefine macro to use new types.
11172 (align_loops): Redefine macro to use new types.
11173 (align_jumps): Redefine macro to use new types.
11174 (align_labels): Redefine macro to use new types.
11175 (align_functions): Redefine macro to use new types.
11176 (align_functions_max_skip): Redefine macro to use new types.
11177 (align_loops_value): New macro.
11178 (align_jumps_value): New macro.
11179 (align_labels_value): New macro.
11180 (align_functions_value): New macro.
11181 * function.c (invoke_set_current_function_hook): Propagate
11182 alignment values from flags to global variables default in
11184 * ipa-icf.c (sem_function::equals_wpa): Use
11185 cl_optimization_option_eq instead of memcmp.
11186 * lto-streamer.h (cl_optimization_stream_out): Support streaming
11188 (cl_optimization_stream_in): Likewise.
11189 * optc-save-gen.awk: Support strings in cl_optimization.
11190 * opth-gen.awk: Likewise.
11191 * opts.c (finish_options): Remove error checking of invalid
11193 (MAX_CODE_ALIGN): Remove.
11194 (MAX_CODE_ALIGN_VALUE): Likewise.
11195 (parse_and_check_align_values): New function.
11196 (check_alignment_argument): Likewise.
11197 (common_handle_option): Use check_alignment_argument.
11198 * opts.h (parse_and_check_align_values): Declare.
11199 * toplev.c (init_alignments): Remove.
11200 (read_log_maxskip): New.
11201 (parse_N_M): Likewise.
11202 (parse_alignment_opts): Likewise.
11203 (backend_init_target): Remove usage of init_alignments.
11204 * toplev.h (parse_alignment_opts): Declare.
11205 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
11207 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
11208 * tree.c (cl_option_hasher::equal): New.
11209 * varasm.c: Use new global macros.
11211 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
11213 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
11214 Use a simpler align directive also if MAXSKIP = ALIGN-1.
11215 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11216 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11217 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11218 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
11219 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
11220 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
11221 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
11222 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11223 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11224 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11225 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11227 2018-07-04 Martin Liska <mliska@suse.cz>
11228 Jonathan Wakely <jwakely@redhat.com>
11230 * coverage.c: Use correct type.
11231 * doc/invoke.texi: Language correction.
11233 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
11236 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
11237 ENDBRANCH for non-tail call which may return via indirect branch.
11238 * doc/extend.texi: Document indirect_return attribute.
11240 2018-07-03 Martin Sebor <msebor@redhat.com>
11242 PR tree-optimization/86274
11243 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
11245 (format_floating): Correct handling of infinities and NaNs.
11247 2018-07-03 Martin Sebor <msebor@redhat.com>
11249 * print-tree.c (print_real_cst): New function.
11250 (print_node_brief): Call it.
11251 (print_node): Ditto.
11253 2018-07-03 Jeff Law <law@redhat.com>
11255 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
11256 into a single pattern.
11258 * config/h8300/h8300.md (ors code_iterator): New.
11259 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
11260 a single pattern and single splitter.
11261 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
11262 (iorqi3_1, xorqi3_1): Likewise.
11263 (iorqi3, xorqi3 expanders): Similarly.
11265 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
11266 (movmd_internal) into a single pattern using the P mode iterator.
11267 (movmd splitters): Similarly.
11268 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
11269 (movsd splitters): Similarly.
11271 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
11272 ADDB, ADDW and ADDL into a single ADD attribute which selects the
11273 right table based on the size of the operand.
11274 * config/h8300/h8300.md (length_table): Corresponding changes. All
11275 references to "addb", "addw" and "addl" changed to "add".
11276 (btst patterns): Merge two variants into a single pattern.
11277 (tstqi, tsthi): Likewise.
11278 (addhi3_incdec, addsi3_incdec): Likewise.
11279 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
11280 (mulhi3, mulsi3): Likewise.
11281 (udivhi3, udivsi3): Likewise.
11282 (divhi3, divsi3): Likewise.
11283 (andorqi3, andorhi3, andorsi3): Likewise.
11285 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
11288 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
11289 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
11291 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
11293 PR tree-optimization/85694
11294 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
11295 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
11297 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
11298 (<u>avg<mode>3_ceil): New patterns.
11300 2018-07-03 David Malcolm <dmalcolm@redhat.com>
11302 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
11303 scan-tree-dump directive.
11304 * gcc.dg/vect/slp-perm-2.c: Likewise.
11305 * gcc.dg/vect/slp-perm-3.c: Likewise.
11306 * gcc.dg/vect/slp-perm-5.c: Likewise.
11307 * gcc.dg/vect/slp-perm-6.c: Likewise.
11308 * gcc.dg/vect/slp-perm-7.c: Likewise.
11309 * gcc.dg/vect/slp-perm-8.c: Likewise.
11311 2018-07-03 Marek Polacek <polacek@redhat.com>
11313 PR middle-end/86202
11314 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
11316 2018-07-03 Richard Biener <rguenther@suse.de>
11319 * tree-ssa-structalias.c (find_func_clobbers): Properly
11320 handle indirect calls.
11322 2018-07-03 Jeff Law <law@redhat.com>
11324 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
11325 (shifts): New code iterator.
11326 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
11327 expander. Fix HImode handling on H8/SX.
11328 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
11329 (subqi3, subhi3, subsi3 expanders): Likewise.
11330 (andqi3, andhi3, andsi3 expanders): Likewise.
11331 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
11332 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
11333 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
11334 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
11335 (zero_extendqihi2, zero_extendqisi2): Likewise.
11336 (extendqihi2, extendqisi2): Likewise.
11337 (rotlqi3, rotlhi3, rotlsi3): Likewise.
11338 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
11339 (rotlqi3_1, rotlhi3_1): Likewise.
11340 (logicalhi3_sn, logicalsi3_sn): Likewise.
11341 (logicalhi3, logicalsi3): Likewise.
11343 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
11345 * tree-vect-patterns.c (vect_recog_rotate_pattern)
11346 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
11347 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
11348 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
11349 type to append_pattern_def_seq instead of creating a stmt_vec_info
11351 (build_mask_conversion): Likewise. Remove vinfo argument.
11352 (vect_add_conversion_to_patterm): Likewise, renaming to...
11353 (vect_add_conversion_to_pattern): ...this.
11354 (vect_recog_mask_conversion_pattern): Update call to
11355 build_mask_conversion. Pass the vector type to
11356 append_pattern_def_seq here too.
11357 (vect_recog_gather_scatter_pattern): Update call to
11358 vect_add_conversion_to_pattern.
11360 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
11362 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
11363 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
11364 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
11365 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
11366 STMT_VINFO_PATTERN_DEF_SEQ to null here.
11367 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
11368 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
11369 append_pattern_def_seq instead of new_pattern_def_seq.
11370 (vect_recog_divmod_pattern): Do both of the above.
11371 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
11374 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
11376 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
11377 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
11378 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
11379 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
11380 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
11381 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
11382 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
11383 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
11384 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
11385 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
11386 parameter with a single stmt_vec_info.
11387 (vect_recog_func_ptr): Likewise.
11388 (vect_recog_gather_scatter_pattern): Likewise, folding in...
11389 (vect_try_gather_scatter_pattern): ...this.
11390 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
11391 the stmt_vec_info of the statement to be matched. Don't clear
11392 STMT_VINFO_RELATED_STMT.
11393 (vect_pattern_recog): Update call accordingly.
11395 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
11397 PR tree-optimization/85694
11398 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
11399 (uavgM3_ceil): Document new optabs.
11400 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
11401 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
11403 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
11404 (savg_ceil_optab): New optabs.
11405 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
11406 (vect_vect_recog_func_ptrs): Add it.
11407 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
11408 constant directly from the associated lhs.
11410 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
11412 * tree-vect-patterns.c (vect_split_statement): New function.
11413 (vect_convert_input): Use it to try to split an existing cast.
11415 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
11417 * poly-int.h (print_hex): New function.
11418 * dumpfile.h (dump_dec, dump_hex): Declare.
11419 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
11420 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
11421 min_input_precision, operation_precision and operation_sign.
11422 * tree-vect-patterns.c (vect_get_range_info): New function.
11423 (vect_same_loop_or_bb_p, vect_single_imm_use)
11424 (vect_operation_fits_smaller_type): Delete.
11425 (vect_look_through_possible_promotion): Add an optional
11426 single_use_p parameter.
11427 (vect_recog_over_widening_pattern): Rewrite to use new
11428 stmt_vec_info infomration. Handle one operation at a time.
11429 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
11430 (vect_truncatable_operation_p, vect_set_operation_type)
11431 (vect_set_min_input_precision): New functions.
11432 (vect_determine_min_output_precision_1): Likewise.
11433 (vect_determine_min_output_precision): Likewise.
11434 (vect_determine_precisions_from_range): Likewise.
11435 (vect_determine_precisions_from_users): Likewise.
11436 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
11437 (vect_vect_recog_func_ptrs): Put over_widening first.
11439 (vect_pattern_recog): Call vect_determine_precisions.
11441 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
11443 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
11444 statements that have been replaced by further pattern statements.
11445 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
11447 2018-07-03 Richard Biener <rguenther@suse.de>
11449 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
11450 always set *dt. Dump vectype in vectype overload.
11451 * dumpfile.h (dump_gimple_expr): New function.
11452 (dump_gimple_expr_loc): Likewise.
11453 * dumpfile.c (dump_gimple_expr): New function.
11454 (dump_gimple_expr_loc): Likewise.
11456 2018-07-02 Jeff Law <law@redhat.com>
11458 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
11459 the H8/300, H8/300H and H8/S variants into a single pattern.
11460 (movhi_h8300, movqi_h8300hs): Similarly.
11461 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
11462 (QHI mode iterator): New.
11464 * config/h8300/h8300.md: Remove trailing whitespace.
11466 2018-07-02 Jim Wilson <jimw@sifive.com>
11468 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
11469 instead of emit_insn for interrupt returns.
11470 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
11471 (riscv_sret, riscv_uret): Likewise.
11473 2018-07-02 David Malcolm <dmalcolm@redhat.com>
11475 * pretty-print.c (selftest::test_pp_format): Move save and restore
11476 of quotes to class auto_fix_quotes, and add an instance.
11477 * selftest.c: Include "intl.h".
11478 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
11479 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
11480 * selftest.h (selftest::auto_fix_quotes): New class.
11482 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
11484 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
11485 (aarch64_sve_prepare_conditional_op): Remove.
11486 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
11487 Allow aarch64_simd_reg_or_zero as select operand; remove
11488 the aarch64_sve_prepare_conditional_op call.
11489 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
11490 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
11491 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
11492 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
11493 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
11494 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
11495 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
11496 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
11497 and a splitters to match all of the *_any patterns.
11498 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
11500 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
11501 (SVE_COND_FP_BINARY_REV): Remove.
11502 (sve_int_op_rev, sve_fp_op_rev): New.
11503 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
11504 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
11505 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
11506 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
11507 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
11508 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
11509 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
11510 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
11511 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
11513 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
11514 Remove match_dup 1 from the inner unspec.
11515 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
11517 * config/aarch64/aarch64.md (movprfx): New attr.
11518 (length): Default movprfx to 8.
11519 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
11520 (*madd<SVE_I>, *msub<SVE_I): Likewise.
11521 (*<su>mul<SVE_I>3_highpart): Likewise.
11522 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
11523 (*v<ASHIFT><SVE_I>3): Likewise.
11524 (*<su><MAXMIN><SVE_I>3): Likewise.
11525 (*<su><MAXMIN><SVE_F>3): Likewise.
11526 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
11527 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
11528 (*div<SVE_F>4): Likewise.
11530 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
11532 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
11535 2018-07-02 Richard Biener <rguenther@suse.de>
11537 PR tree-optimization/86363
11538 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
11539 memset argument refers to a non-variable address.
11541 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
11543 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
11544 {PLUS,MINUS}_EXPR code to...
11545 (adjust_symbolic_bound): ...here,
11546 (combine_bound): ...here,
11547 (set_value_range_with_overflow): ...and here.
11549 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
11551 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
11553 (extract_range_from_abs_expr): ...here.
11555 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
11557 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
11558 -fno-omit-frame-pointer when not optimizing.
11560 2018-07-02 Martin Liska <mliska@suse.cz>
11563 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
11564 (propagate_nothrow): Likewise.
11566 2018-07-02 Martin Liska <mliska@suse.cz>
11569 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
11571 2018-07-02 David Malcolm <dmalcolm@redhat.com>
11573 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
11574 function in r262149, changing "loc" param from source_location to
11575 const dump_location_t &.
11576 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
11577 declaration, as above.
11579 2018-07-01 Paul Koning <ni1d@arrl.net>
11581 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
11582 -munit-asm, -mgnu-asm, -mdec-asm.
11583 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
11584 (pdp11_output_labelref): New.
11585 (pdp11_output_def): New.
11586 (pdp11_output_addr_vec_elt): New.
11587 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
11588 %# and %@ format codes.
11589 (pdp11_option_override): New.
11590 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
11591 (pdp11_output_ident): New.
11592 (pdp11_asm_named_section): New.
11593 (pdp11_asm_init_sections): New.
11594 (pdp11_file_start): New.
11595 (pdp11_file_end): New.
11596 (output_ascii): Use .ascii/.asciz for -mdec-asm.
11597 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
11598 %o, like %c but octal.
11599 (pdp11_option_override): New.
11600 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
11602 (DATA_SECTION_ASM_OP): Ditto.
11603 (READONLY_DATA_SECTION_ASM_OP): New.
11604 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
11605 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
11606 (ASM_OUTPUT_LABELREF): Ditto.
11607 (ASM_OUTPUT_DEF): Ditto.
11608 (ASM_OUTPUT_EXTERNAL): New.
11609 (ASM_OUTPUT_SOURCE_FILENAME): New.
11610 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
11611 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
11612 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
11613 %# and %@ format codes.
11614 * config/pdp11/pdp11.opt (mgnu-asm): New.
11615 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
11616 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
11617 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
11619 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
11621 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
11622 dereferencing path[] beyond its length.
11624 (debug_all_paths): New.
11625 (rewire_first_differing_edge): New.
11626 (adjust_paths_after_duplication): New.
11627 (duplicate_thread_path): Call adjust_paths_after_duplication.
11629 (thread_through_all_blocks): Add new argument to
11630 duplicate_thread_path.
11632 2018-06-30 Jim Wilson <jimw@sifive.com>
11634 * config/riscv/predicates.md (p2m1_shift_operand): New.
11635 (high_mask_shift_operand): New.
11636 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
11637 pattern using p2m1_shift_operand.
11638 (lshsi3_zero_extend_3+2): New combiner pattern using
11639 high_mask_shift_operand.
11641 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
11643 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
11645 (vect_recog_rotate_pattern): ...here.
11646 (vect_convert_input): Try to insert casts of invariants in the
11648 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
11649 preheader to be empty.
11651 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
11653 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
11654 vector type. If given, install it in the new statement's
11655 STMT_VINFO_VECTYPE.
11656 (vect_element_precision): New function.
11657 (vect_unpromoted_value): New struct.
11658 (vect_unpromoted_value::vect_unpromoted_value): New function.
11659 (vect_unpromoted_value::set_op): Likewise.
11660 (vect_look_through_possible_promotion): Likewise.
11661 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
11662 (vect_widened_op_tree, vect_convert_input): Likewise.
11663 (vect_convert_inputs, vect_convert_output): Likewise.
11664 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
11665 to handle the optional cast of the multiplication result and
11666 vect_widened_op_tree to detect the widened multiplication itself.
11667 Do not require the input and output of promotion casts to have
11668 the same sign, but base the signedness of the operation on the
11669 input rather than the result. If the pattern includes two
11670 promotions, check that those promotions have the same sign.
11671 Do not restrict the MULT_EXPR handling to a double-width result;
11672 handle quadruple-width results and wider. Use vect_convert_inputs
11673 to convert the inputs to the common type.
11674 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
11675 to handle the optional cast of the ABS result. Also allow a sign
11676 change or a sign extension between the ABS and MINUS.
11677 Use vect_widened_op_tree to detect the widened subtraction and use
11678 vect_convert_inputs to convert the inputs to the common type.
11679 (vect_handle_widen_op_by_const): Delete.
11680 (vect_recog_widen_op_pattern): New function.
11681 (vect_recog_widen_mult_pattern): Use it.
11682 (vect_recog_widen_shift_pattern): Likewise.
11683 (vect_recog_widen_sum_pattern): Use
11684 vect_look_through_possible_promotion to handle the promoted
11687 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
11689 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
11690 the containing gimple_seq *.
11691 * gimple-iterator.h (gsi_for_stmt): Declare it.
11692 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
11693 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
11694 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
11695 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
11696 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
11698 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
11700 (vect_mark_pattern_stmts): ...here. Handle cases in which the
11701 statement being replaced is part of an existing pattern
11702 definition sequence, inserting the new pattern statements before
11704 (vect_pattern_recog_1): Don't return a bool. If the statement
11705 is already part of a pattern, instead apply pattern matching
11706 to the pattern definition statements. Don't clear the
11707 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
11708 (vect_pattern_recog): Don't break after the first match;
11709 continue processing the pattern definition statements instead.
11710 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
11712 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
11714 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
11715 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
11716 (vect_recog_widen_sum_pattern): Use it.
11718 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
11720 * tree-vect-loop.c (vectorizable_reduction): Assert that the
11721 phi is not a pattern statement and has not been replaced by
11722 a pattern statement.
11723 * tree-vect-patterns.c (type_conversion_p): Don't check
11724 STMT_VINFO_IN_PATTERN_P.
11725 (vect_recog_vector_vector_shift_pattern): Likewise.
11726 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
11727 the pattern statement rather than the original statement; check
11728 directly for a WIDEN_MULT_EXPR here.
11729 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
11730 vect_is_simple_use to return the pattern statement rather
11731 than the original statement; use is_pattern_stmt_p to check
11732 for such a pattern statement.
11733 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
11734 to return the pattern statement rather than the original statement;
11735 don't do the same transformation here.
11736 (vect_is_simple_use): If the defining statement has been replaced
11737 by a pattern statement, return the pattern statement instead.
11738 Remove the corresponding (local) transformation from the vectype
11741 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
11743 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
11744 end and default to null.
11745 * tree-vect-loop.c (vect_create_epilog_for_reduction)
11746 (vectorizable_reduction): Update calls accordingly, dropping the
11747 gimple ** argument if the passed-back statement isn't needed.
11748 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
11749 (vect_recog_rotate_pattern): Likewise.
11750 (vect_recog_mask_conversion_pattern): Likewise.
11751 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
11752 (vect_mask_constant_operand_p): Likewise.
11753 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
11754 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
11755 (get_group_load_store_type, get_load_store_type): Likewise.
11756 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
11757 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
11758 (vectorizable_conversion, vectorizable_assignment): Likewise.
11759 (vectorizable_shift, vectorizable_operation): Likewise.
11760 (vectorizable_store, vect_is_simple_cond): Likewise.
11761 (vectorizable_condition, vectorizable_comparison): Likewise.
11762 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
11763 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
11764 and move it to the end. Cope with null def_stmt_outs.
11766 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
11768 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
11770 2018-06-29 Jeff Law <law@redhat.com>
11772 * config/v850/v850.c (v850_legitimate_address_p): Handle large
11773 displacements for TARGET_V850E2V3 and newer.
11774 (TARGET_LRA_P): Remove. Defaults to LRA now.
11775 * config/v850/v850.md (sign23byte_load): Remove.
11776 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
11777 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
11779 2018-06-29 Martin Liska <mliska@suse.cz>
11782 * coverage.c (coverage_init): Mangle full path name.
11783 * doc/invoke.texi: Document the change.
11784 * gcov-io.c (mangle_path): New.
11785 * gcov-io.h (mangle_path): Likewise.
11786 * gcov.c (mangle_name): Use mangle_path for path mangling.
11788 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11790 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
11791 if starting source register is not even.
11793 2018-06-29 Martin Liska <mliska@suse.cz>
11795 PR tree-optimization/86263
11796 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
11797 Make edge redirection.
11799 2018-06-29 David Malcolm <dmalcolm@redhat.com>
11801 * dumpfile.c (dump_loc): Add indentation based on scope depth.
11802 (dump_scope_depth): New variable.
11803 (get_dump_scope_depth): New function.
11804 (dump_begin_scope): New function.
11805 (dump_end_scope): New function.
11806 * dumpfile.h (get_dump_scope_depth): New declaration.
11807 (dump_begin_scope): New declaration.
11808 (dump_end_scope): New declaration.
11809 (class auto_dump_scope): New class.
11810 (AUTO_DUMP_SCOPE): New macro.
11811 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
11814 2018-06-29 Richard Biener <rguenther@suse.de>
11816 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
11817 compute_all_dependences succeeds.
11818 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
11819 exceed --param loop-max-datarefs-for-datadeps.
11821 2018-06-29 Jakub Jelinek <jakub@redhat.com>
11823 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
11825 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
11828 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
11829 alternative 0 in preferred_for_speed attribute.
11831 2018-06-28 Paul Koning <ni1d@arrl.net>
11833 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
11834 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
11835 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
11836 * config/pdp11/pdp11.md: Correct "length" attribute calculation
11837 for shift insn patterns.
11839 2018-06-28 David Malcolm <dmalcolm@redhat.com>
11841 * cgraph.c (cgraph_node::get_body): Replace assignments to
11842 "dump_file" with calls to set_dump_file.
11843 * dumpfile.c (alt_dump_file): Make static, and group with...
11844 (alt_flags): ...this definition.
11845 (dumps_are_enabled): New variable.
11846 (refresh_dumps_are_enabled): New function.
11847 (set_dump_file): New function.
11848 (set_alt_dump_file): New function.
11849 (gcc::dump_manager::dump_start): Replace assignments to
11850 "dump_file" and "alt_dump_file" with calls to set_dump_file and
11852 (gcc::dump_manager::dump_finish): Likewise.
11853 * dumpfile.h (alt_dump_file): Delete decl.
11854 (dumps_are_enabled): New variable decl.
11855 (set_dump_file): New function decl.
11856 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
11858 * tree-nested.c (lower_nested_functions): Replace assignments to
11859 "dump_file" with calls to set_dump_file.
11861 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
11863 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
11864 goto_locus of each outgoing edge of each basic block.
11866 2018-06-28 Richard Biener <rguenther@suse.de>
11868 * dwarf2out.c (decl_scope_table): Remove.
11869 (push_decl_scope): Likewise.
11870 (pop_decl_scope): Likewise.
11871 (gen_type_die_for_member): Do not call push/pop_decl_scope.
11872 (gen_struct_or_union_type_die): Likewise.
11873 (gen_tagged_type_die): Likewise.
11874 (dwarf2out_init): Do not initialize decl_scope_table.
11875 (dwarf2out_c_finalize): Do not free it.
11877 2018-06-28 Richard Biener <rguenther@suse.de>
11879 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
11880 deciding whether to not re-use a DIE.
11882 2018-06-28 Richard Biener <rguenther@suse.de>
11884 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
11885 DW_AT_abstract_origin attribute.
11887 2018-06-28 Martin Liska <mliska@suse.cz>
11889 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
11890 Use newly introduced constants.
11891 * tree-switch-conversion.h (struct jump_table_cluster):
11892 Define max_ratio_for_size and max_ratio_for_speed.
11894 2018-06-28 Martin Liska <mliska@suse.cz>
11896 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
11897 Add new checking assert to catch invalid state.
11898 (jump_table_cluster::can_be_handled): Handle single case
11900 (jump_table_cluster::is_beneficial): Bail out for such case.
11901 (bit_test_cluster::find_bit_tests):
11902 Add new checking assert to catch invalid state.
11903 (bit_test_cluster::can_be_handled): Handle single case
11905 (bit_test_cluster::is_beneficial): Bail out for such case.
11906 (switch_decision_tree::analyze_switch_statement):
11909 2018-06-28 Martin Liska <mliska@suse.cz>
11911 * common.opt: Introduce -completion option.
11912 * gcc.c (driver_handle_option): Handle it.
11913 (driver::main): Print completions if completion
11915 * opt-suggestions.c (option_proposer::get_completions):
11917 (option_proposer::suggest_completion): Likewise.
11918 (option_proposer::find_param_completions): Likewise.
11919 (verify_autocompletions): Likewise.
11920 (test_completion_valid_options): Likewise.
11921 (test_completion_valid_params): Likewise.
11922 (in_completion_p): Likewise.
11923 (empty_completion_p): Likewise.
11924 (test_completion_partial_match): Likewise.
11925 (test_completion_garbage): Likewise.
11926 (opt_proposer_c_tests): Likewise.
11927 * opt-suggestions.h: Declare new functions.
11928 * opts.c (common_handle_option): Handle OPT__completion_.
11929 * selftest-run-tests.c (selftest::run_tests): Add
11930 opt_proposer_c_tests.
11931 * selftest.c (assert_str_startswith): New.
11932 * selftest.h (assert_str_startswith): Likewise.
11933 (opt_proposer_c_tests): New.
11934 (ASSERT_STR_STARTSWITH): Likewise.
11936 2018-06-28 Martin Liska <mliska@suse.cz>
11938 * Makefile.in: Add opt-suggestions.o.
11939 * gcc-main.c: Include opt-suggestions.h.
11940 * gcc.c (driver::driver): Likewise.
11941 (driver::~driver): Remove m_option_suggestions.
11942 (driver::build_option_suggestions): Moved to option_proposer.
11943 (driver::suggest_option): Likewise.
11944 (driver::handle_unrecognized_options): Use option_proposer.
11945 * gcc.h (class driver): Add new memver m_option_proposer.
11946 * opt-suggestions.c: New file.
11947 * opt-suggestions.h: New file.
11949 2018-06-28 Martin Liska <mliska@suse.cz>
11951 * vec.h (class auto_string_vec): New (moved from auto_argvec).
11952 (auto_string_vec::~auto_string_vec): Likewise.
11954 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
11956 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
11957 prevent_decl_creation_for_types fields up and add reset_location field.
11958 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
11959 statement if id->reset_location is true.
11960 (copy_edges_for_bb): Do not set goto_locus on the new edges if
11961 id->reset_location is true.
11962 (copy_phis_for_bb): Force input_location on the arguments if
11963 id->reset_location is true.
11964 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
11965 is set on the function to be inlined.
11967 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
11969 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
11971 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
11973 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
11974 registers for Pmode.
11975 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
11976 hard registers for the clobbered pseudo.
11978 2018-06-27 Paul Koning <ni1d@arrl.net>
11980 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
11981 mutually exclusive options.
11982 * config/pdp11/constraints.md (h): New constraint.
11983 (O): Update definition to match shift code generation.
11984 (D): New constraint.
11985 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
11987 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
11989 (output_jump): Change arguments.
11990 (pdp11_fixed_cc_regs): New function.
11991 (pdp11_cc_mode): Ditto.
11992 (pdp11_expand_shift): Ditto.
11993 (pdp11_assemble_shift): Ditto.
11994 (pdp11_small_shift): Ditto.
11995 (pdp11_branch_cost): Remove.
11996 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
11998 (pdp11_register_move_cost): Update for CC registers.
11999 (pdp11_rtx_costs): Add case for LSHIFTRT.
12000 (pdp11_output_jump): Add CCNZ mode conditional branches.
12001 (notice_update_cc_on_set): Remove.
12002 (pdp11_cc_mode): New function.
12003 (simple_memory_operand): Correct pre/post decrement case.
12004 (no_side_effect_operand): New function.
12005 (pdp11_regno_reg_class): Add CC_REGS class.
12006 (pdp11_fixed_cc_regs): New function.
12007 (pdp11_small_shift): New function.
12008 (pdp11_expand_shift): New function to expand shift insns.
12009 (pdp11_assemble_shift): New function to output shifts.
12010 (pdp11_branch_cost): Remove.
12011 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
12012 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
12013 (WCHAR_TYPE): Ditto.
12014 (PTRDIFF_TYPE): Ditto.
12015 (ADJUST_INSN_LENGTH): New macro.
12016 (FIXED_REGISTERS): Add CC registers.
12017 (CALL_USED_REGISTERS): Ditto.
12018 (reg_class): Ditto.
12019 (REG_CLASS_NAMES): Ditto.
12020 (REG_CLASS_CONTENTS): Ditto.
12021 (SELECT_CC_MODE): Use new function.
12022 (TARGET_FLAGS_REGNUM): New macro.
12023 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
12024 (cc0_reg_rtx): Remove.
12025 (CC_STATUS_MDEP): Remove.
12026 (CC_STATUS_MDEFP_INIT): Remove.
12027 (CC_IN_FPU): Remove.
12028 (NOTICE_UPDATE_CC): Remove.
12029 (REGISTER_NAMES): Add CC registers.
12030 (BRANCH_COST): Change to constant 1.
12031 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
12033 * config/pdp11/pdp11.opt (mbcopy): Remove.
12034 (mbcopy-builtin): Remove.
12035 (mbranch-cheap): Remove.
12036 (mbranch-expensive): Remove.
12037 * config/pdp11/predicates.md (expand_shift_operand): Update to
12038 match shift code generation.
12039 (ccnz_operator): New predicate.
12040 * doc/invoke.texi (PDP-11 Options): Remove deleted options
12041 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
12042 Remove non-existent option -mabshi, -mno-abshi. Document mutually
12044 * doc/md.texi (PDP-11): Document new D and h constraints. Update
12045 description of O constraint.
12047 2018-06-27 Jeff Law <law@redhat.com>
12048 Austin Law <austinklaw@gmail.com>
12050 * config/v850/v850.md (addsi3_set_flags): New pattern.
12051 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
12052 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
12053 (zero_extendhisi2_v850_set_flags): Likewise.
12054 (zero_extendqisi2_v850_set_flags): Likewise.
12055 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
12056 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
12057 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
12059 * config/v850/v850-protos.h (notice_update_cc): Remove.
12060 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
12061 (v850_print_operand): Handle 'D' and "d".
12062 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
12063 Add handling of arithmetic/logical operations compared against zero.
12064 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
12065 Do not look at v850_compare_op, instead get mode from last argument.
12066 (v850_gen_compare): Remove
12067 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
12068 after reload for prologue insns.
12069 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
12071 (construct_save_jarl): Likewise.
12072 (TARGET_FLAGS_REGNUM): Define.
12073 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
12074 (NOTICE_UPDATE_CC): Remove.
12075 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
12076 than cc0. Conditionalize on reload_completed.
12077 (cmpsi_insn, setfcc_insn): Likewise.
12078 (tst1 splitter): Turn into define_and_split which sets the flags
12080 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
12081 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
12082 (cstoresf4, cstoredf4): Clobber the flags.
12083 (cmpsi, cmpsf, cmpdf): Remove expanders.
12084 (setf_insn): Remove pattern.
12085 (addsi3): Turn into define_and_split which clobbers the flags after
12086 reload and a suitable pattern (addsi3_clobber_flags) for use after
12088 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
12089 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
12090 (ashrsi3, ashrsi3_v850e2): Likewise.
12091 (bins): Clobber the flags.
12092 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
12093 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
12094 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
12095 (call_value_internal_short, call_value_internal_long): Likewise.
12096 (callt_save_interrupt, callt_return_interrupt): Likewise.
12097 (save_interrupt, return_interrupt): Likewise.
12098 (callt_save_all_interrupt, save_all_interrupt): Likewise.
12099 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
12100 (restore_all_interrupt, _restore_all_interrupt): Likewise.
12101 (All FP comparisons): Only allow after reload has completed.
12103 (divh, divhu): Tweak output template.
12104 (branch_z_normal, branch_z_invert): Remove
12105 (branch_nz_normal, branch_nz_invert): Likewise.
12106 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
12108 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
12109 * config/v850/v850.c (notice_update_cc): Remove.
12110 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
12111 (CC_NO_CARRY): Likewise.
12112 (NOTICE_UPDATE_CC): Define to nothing.
12113 * config/v850/v850.md: Remove block comment on cc0 handling
12114 Remove "cc" attribute from all patterns. Remove cc_status handling
12115 from all patterns. Minor formatting fixes.
12117 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12119 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
12120 (cortex-a76.cortex-a55): Likewise.
12121 * config/aarch64/aarch64-tune.md: Regenerate.
12122 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
12123 cortex-a76.cortex-a55.
12125 2018-06-27 Jeff Law <law@redhat.com>
12127 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
12128 (MULTILIB_DIRNAMES): Similarly.
12130 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
12132 * gimple.h (gimple_return_retbnd): Delete.
12133 (gimple_return_set_retbnd): Likewise.
12134 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
12135 gimple_return_set_retbnd.
12136 * gimple-pretty-print.c (dump_gimple_return): Remove call to
12137 gimple_return_retbnd and adjust.
12138 * tree-inline.h (struct copy_body_data): Remove retbnd field.
12139 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
12140 Explicitly return NULL in a couple more cases. Move assertion
12141 on debug statements and remove unreachable code.
12142 (reset_debug_binding): Do not test id->retbnd.
12143 (expand_call_inline): Do not set it.
12145 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
12147 * configure.ac: Add --disable-gcov option.
12148 * configure: Regenerate.
12149 * Makefile.in: Honour @enable_gcov@.
12150 * doc/install.texi: Document --disable-gcov.
12152 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12154 * config/arm/arm-cpus.in (cortex-a76): New entry.
12155 (cortex-a76.cortex-a55): Likewise.
12156 * config/arm/arm-tables.opt: Regenerate.
12157 * config/arm/arm-tune.md: Likewise.
12158 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
12159 * doc/invoke.texi (ARM Options): Document cortex-a76 and
12160 cortex-a76.cortex-a55.
12162 2018-06-27 Tamar Christina <tamar.christina@arm.com>
12165 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
12167 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
12169 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
12171 (EPILOGUE_USES): Likewise.
12173 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
12175 * tree-inline.c (remap_location): New function extracted from...
12176 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
12177 (copy_phis_for_bb): ...here. Call remap_location.
12178 (copy_cfg_body): Adjust call to copy_edges_for_bb.
12180 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
12182 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
12183 unaligned vsx for 16B memset.
12185 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
12188 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
12189 ieee128_float_type_node to long_double_type_node unless
12190 TARGET_LONG_DOUBLE_128 is set.
12192 2018-06-26 David Malcolm <dmalcolm@redhat.com>
12194 * cfgloop.c (get_loop_location): Convert return type from
12195 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
12196 by implicit construction from rtx_insn *, and using
12197 dump_user_location_t::from_function_decl for the fallback case.
12198 * cfgloop.h (get_loop_location): Convert return type from
12199 location_t to dump_user_location_t.
12200 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
12201 dump_printf_loc to pass in a dump_location_t rather than a
12202 location_t, via the gimple stmt.
12203 * coverage.c (get_coverage_counts): Update calls to
12204 dump_printf_loc to pass in dump_location_t rather than a
12206 * doc/optinfo.texi (Dump types): Convert example of
12207 dump_printf_loc from taking "locus" to taking "insn". Update
12208 description of the "_loc" calls to cover dump_location_t.
12209 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
12211 (dump_user_location_t::dump_user_location_t): New constructors,
12212 from gimple *stmt and rtx_insn *.
12213 (dump_user_location_t::from_function_decl): New function.
12214 (dump_loc): Make static.
12215 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
12216 const dump_location_t &.
12217 (dump_generic_expr_loc): Delete.
12218 (dump_printf_loc): Convert param "loc" from location_t to
12219 const dump_location_t &.
12220 (selftest::test_impl_location): New function.
12221 (selftest::dumpfile_c_tests): New function.
12222 * dumpfile.h: Include "profile-count.h".
12223 (class dump_user_location_t): New class.
12224 (struct dump_impl_location_t): New struct.
12225 (class dump_location_t): New class.
12226 (dump_printf_loc): Convert 2nd param from source_location to
12227 const dump_location_t &.
12228 (dump_generic_expr_loc): Delete.
12229 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
12230 const dump_location_t &.
12231 * gimple-fold.c (fold_gimple_assign): Update call to
12232 dump_printf_loc to pass in a dump_location_t rather than a
12233 location_t, via the gimple stmt.
12234 (gimple_fold_call): Likewise.
12235 * gimple-loop-interchange.cc
12236 (loop_cand::analyze_iloop_reduction_var): Update for change to
12237 check_reduction_path.
12238 (tree_loop_interchange::interchange): Update for change to
12239 find_loop_location.
12240 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
12241 change in return-type of find_loop_location.
12242 (graphite_regenerate_ast_isl): Likewise.
12243 * graphite-optimize-isl.c (optimize_isl): Likewise.
12244 * graphite.c (graphite_transform_loops): Likewise.
12245 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
12246 pass in a dump_location_t rather than a location_t, via the
12248 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
12249 * ipa.c (walk_polymorphic_call_targets): Likewise.
12250 * loop-unroll.c (report_unroll): Convert "locus" param from
12251 location_t to dump_location_t.
12252 (decide_unrolling): Update for change to get_loop_location's
12254 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
12255 location_t to dump_user_location_t.
12256 (grid_find_single_omp_among_assignments_1): Updates calls to
12257 dump_printf_loc to pass in a dump_location_t rather than a
12258 location_t, via the gimple stmt.
12259 (grid_parallel_clauses_gridifiable): Convert "tloc" from
12260 location_t to dump_location_t. Updates calls to dump_printf_loc
12261 to pass in a dump_location_t rather than a location_t, via the
12263 (grid_inner_loop_gridifiable_p): Likewise.
12264 (grid_dist_follows_simple_pattern): Likewise.
12265 (grid_gfor_follows_tiling_pattern): Likewise.
12266 (grid_target_follows_gridifiable_pattern): Likewise.
12267 (grid_attempt_target_gridification): Convert initialization
12268 of local "grid" from memset to zero-initialization; FIXME: does
12269 this require C++11? Update call to dump_printf_loc to pass in a
12270 optinfo_location rather than a location_t, via the gimple stmt.
12271 * profile.c (read_profile_edge_counts): Updates call to
12272 dump_printf_loc to pass in a dump_location_t rather than a
12274 (compute_branch_probabilities): Likewise.
12275 * selftest-run-tests.c (selftest::run_tests): Call
12277 * selftest.h (dumpfile_c_tests): New decl.
12278 * tree-loop-distribution.c (pass_loop_distribution::execute):
12279 Update for change in return type of find_loop_location.
12280 * tree-parloops.c (parallelize_loops): Likewise.
12281 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
12282 "locus" from location_t to dump_user_location_t.
12283 (canonicalize_loop_induction_variables): Likewise.
12284 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
12285 for change in return type of find_loop_location.
12286 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
12287 to dump_printf_loc to pass in a dump_location_t rather than a
12288 location_t, via the stmt.
12289 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
12291 * tree-vect-loop-manip.c (find_loop_location): Convert return
12292 type from source_location to dump_user_location_t.
12293 (vect_do_peeling): Update for above change.
12294 (vect_loop_versioning): Update for change in type of
12296 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
12297 from location_t to dump_user_location_t.
12298 (vect_estimate_min_profitable_iters): Update for change in type
12300 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
12301 location_t to dump_location_t.
12302 (vect_slp_bb): Update for change in type of vect_location.
12303 * tree-vectorizer.c (vect_location): Convert from source_location
12304 to dump_user_location_t.
12305 (try_vectorize_loop_1): Update for change in vect_location's type.
12306 (vectorize_loops): Likewise.
12307 (increase_alignment): Likewise.
12308 * tree-vectorizer.h (vect_location): Convert from source_location
12309 to dump_user_location_t.
12310 (find_loop_location): Convert return type from source_location to
12311 dump_user_location_t.
12312 (check_reduction_path): Convert 1st param from location_t to
12313 dump_user_location_t.
12314 * value-prof.c (check_counter): Update call to dump_printf_loc to
12315 pass in a dump_user_location_t rather than a location_t; update
12316 call to error_at for change in type of "locus".
12317 (check_ic_target): Update call to dump_printf_loc to
12318 pass in a dump_user_location_t rather than a location_t, via the
12321 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
12323 * config/s390/s390.h (enum processor_flags): Do not use
12324 default tune parameter when -march was specified.
12326 2018-06-26 Jakub Jelinek <jakub@redhat.com>
12329 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
12330 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
12333 2018-06-26 Richard Biener <rguenther@suse.de>
12335 PR tree-optimization/86287
12337 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
12338 (vect_analyze_loop): Initialize n_stmts.
12340 2018-06-26 Richard Biener <rguenther@suse.de>
12342 PR middle-end/86271
12343 * fold-const.c (fold_convertible_p): Pointer extension
12346 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
12349 * dwarf2out.c (loc_list_has_views): Adjust comments.
12350 (dw_loc_list): Split single cross-partition range with
12353 2018-06-25 Jeff Law <law@redhat.com>
12355 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
12356 on -mbig-switch by default.
12358 * config/v850/predicates.md (const_float_1_operand): Fix match_code
12360 (const_float_0_operand): Remove unused predicate.
12361 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
12362 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
12363 (recipsf2): New expander. Original pattern now called
12365 (recipdf2, recipdf2_insn): Similarly.
12366 (rsqrtsf2, rsqrtsf2_insn): Similarly
12367 (rsqrtdf2, rsqrtdf2_insn): Similarly
12369 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
12371 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
12372 Simplify logic for FreeBSD (twice).
12374 2018-06-25 Martin Sebor <msebor@redhat.com>
12376 PR tree-optimization/86204
12377 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
12378 a strnlen result if it's less than the length of the string.
12380 2018-06-25 Martin Sebor <msebor@redhat.com>
12382 PR tree-optimization/85700
12383 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
12384 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
12385 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
12387 2018-06-25 Martin Sebor <msebor@redhat.com>
12389 * doc/extend.texi (Zero-length arrays): Update and clarify.
12391 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
12393 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
12394 added IEEE/IBM long double multilib support on PowerPC little
12395 endian Linux systems.
12396 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
12397 (MULTILIB_DEFAULTS): Likewise.
12398 * config/rs6000/rs6000.c (rs6000_option_override_internal):
12400 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
12401 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
12402 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
12404 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
12406 PR middle-end/86311
12407 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
12408 (REORDER_45): Likewise.
12410 2018-06-25 Jeff Law <law@redhat.com>
12412 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
12413 dividend to 32 bits. Adjust length.
12414 (udivmodhi4): Cleanup output template. Fix length.
12416 2018-06-25 Carl Love <cel@us.ibm.com>
12418 * config/rs6000/vsx.md: Change word selector to prefered location.
12420 2018-06-25 Richard Biener <rguenther@suse.de>
12422 PR tree-optimization/86304
12423 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
12424 epilogue-if-converted loops as well.
12426 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
12428 * lto-section-out.c (lto_begin_section): Do not print section
12429 name for noaddr and unnumbered dumps.
12431 2018-06-25 Richard Biener <rguenther@suse.de>
12433 * tree-vectorizer.h (struct vec_info_shared): New structure
12434 with parts split out from struct vec_info and loop_nest from
12435 struct _loop_vec_info.
12436 (struct vec_info): Adjust accordingly.
12437 (struct _loop_vec_info): Likewise.
12438 (LOOP_VINFO_LOOP_NEST): Adjust.
12439 (LOOP_VINFO_DATAREFS): Likewise.
12440 (LOOP_VINFO_DDRS): Likewise.
12441 (struct _bb_vec_info): Likewise.
12442 (BB_VINFO_DATAREFS): Likewise.
12443 (BB_VINFO_DDRS): Likewise.
12444 (struct _stmt_vec_info): Add dr_aux member.
12445 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
12446 (DR_MISALIGNMENT_UNINITIALIZED): New.
12447 (set_dr_misalignment): Adjust.
12448 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
12449 (vect_analyze_loop): Adjust prototype.
12450 (vect_analyze_loop_form): Likewise.
12451 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
12452 Compute dependences lazily.
12453 (vect_record_base_alignments): Use shared datarefs/ddrs.
12454 (vect_verify_datarefs_alignment): Likewise.
12455 (vect_analyze_data_refs_alignment): Likewise.
12456 (vect_analyze_data_ref_accesses): Likewise.
12457 (vect_analyze_data_refs): Likewise.
12458 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
12459 constructor parameter for shared part.
12460 (vect_analyze_loop_form): Pass in shared part and adjust.
12461 (vect_analyze_loop_2): Pass in storage for the number of
12462 stmts. Move loop nest finding to the caller. Compute
12464 (vect_analyze_loop): Pass in shared part.
12465 (vect_transform_loop): Verify shared datarefs are unchanged.
12466 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
12467 constructor parameter for shared part.
12468 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
12469 (vect_slp_bb): Verify shared datarefs are unchanged before
12471 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
12473 (new_stmt_vec_info): Initialize DR_AUX misalignment to
12474 DR_MISALIGNMENT_UNINITIALIZED.
12475 * tree-vectorizer.c (vec_info::vec_info): Add constructor
12476 parameter for shared part.
12477 (vec_info::~vec_info): Adjust.
12478 (vec_info_shared::vec_info_shared): New.
12479 (vec_info_shared::~vec_info_shared): Likewise.
12480 (vec_info_shared::save_datarefs): Likewise.
12481 (vec_info_shared::check_datarefs): Likewise.
12482 (try_vectorize_loop_1): Construct shared part live for analyses
12483 of a single loop for multiple vector sizes.
12484 * tree-parloops.c (gather_scalar_reductions): Adjust.
12486 2018-06-25 Richard Biener <rguenther@suse.de>
12488 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
12489 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
12490 (vect_analyze_data_refs): Remove similar code from here and
12491 simplify accordingly.
12493 2018-06-25 Richard Biener <rguenther@suse.de>
12495 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
12496 for reverse storage order accesses rather than asserting
12497 they cannot happen here.
12499 2018-06-25 Tom de Vries <tdevries@suse.de>
12502 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
12503 Use data16 instead of .byte for insn prefix.
12505 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
12508 * parser.c (make_char_string_pack): Pass this literal chars
12509 through cpp_interpret_string.
12510 (cp_parser_userdef_numeric_literal): Check the result of
12511 make_char_string_pack.
12513 2018-06-24 Maya Rashish <coypu@sdf.org>
12515 * ginclude/stddef.h: Simplify conditions around avoiding
12516 re-definition of __size_t.
12518 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
12520 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
12521 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
12523 2018-06-22 Maya Rashish <coypu@sdf.org>
12525 * doc/invoke.texi (mno-fancy-math-387): Update for changes
12526 made to OpenBSD and NetBSD through the years.
12528 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
12530 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
12531 behavior of vec_pack (vector double, vector double) to match
12532 behavior of vec_float2 (vector double, vector double).
12534 2018-06-22 Olivier Hainque <hainque@adacore.com>
12536 * gimplify.c (gimplify_function_tree): Prevent creation
12537 of a trampoline for the address of the current function
12538 passed to entry/exit instrumentation hooks.
12540 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
12543 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
12546 2018-06-22 Martin Liska <mliska@suse.cz>
12548 PR tree-optimization/86263
12549 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
12550 Bail out if is_enabled is false.
12551 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
12553 (jump_table_cluster::is_enabled): New function.
12555 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
12557 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
12558 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
12559 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
12560 (lto_input_ts_binfo_tree_pointers): Likewise.
12561 * tree-streamer-out.c (streamer_write_tree_bitfields,
12562 write_ts_binfo_tree_pointers): Likewise.
12563 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
12565 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
12567 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
12569 2018-06-22 Martin Liska <mliska@suse.cz>
12571 * symbol-summary.h (get): Make it pure and inline move
12572 functionality from ::get function.
12573 (get): Remove and inline into ::get and ::get_create.
12574 (get_create): Move code from ::get function.
12576 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12579 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
12580 -x assembler-with-cpp.
12582 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12584 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
12585 _FILE_OFFSET_BITS=64 for C++.
12587 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
12589 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
12590 conversion insn that shows up when pr85657-3.c is compiled using
12591 IEEE 128-bit long double.
12592 (neg<mode>2_internal): Use the correct mode to check whether the
12593 mode is IBM extended.
12594 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
12595 multiply and divide external functions from being created more
12598 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
12600 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
12602 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
12603 the edge can be forwarded.
12604 (cfg_layout_merge_blocks): Likewise.
12606 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
12608 * except.c (finish_eh_generation): Commit edge insertions only after
12609 the EH edges have been redirected from post-landing to landing pads.
12611 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
12613 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
12614 create_tmp_var_for to create the FRAME decl.
12615 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
12617 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
12619 * tree-inline.c (copy_edges_for_bb): Minor tweak.
12620 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
12621 debug statement when resetting its value.
12622 (expand_call_inline): Copy the locus of the call onto the assignment
12623 of the return value, if any. Use local variable in more cases.
12625 2018-06-21 Martin Liska <mliska@suse.cz>
12627 * ipa-pure-const.c (propagate_nothrow): Use
12628 funct_state_summaries->get.
12629 (dump_malloc_lattice): Likewise.
12630 (propagate_malloc): Likewise.
12632 2018-06-21 Richard Biener <rguenther@suse.de>
12634 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
12635 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
12636 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
12638 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
12639 BLOCK_ABSTRACT_ORIGIN unconditionally.
12641 2018-06-21 David Malcolm <dmalcolm@redhat.com>
12643 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
12645 * ipa-reference.c (ipa_reference_c_finalize): Delete
12646 ipa_ref_opt_sum_summaries and set it to NULL.
12648 2018-06-21 Tom de Vries <tdevries@suse.de>
12650 PR tree-optimization/85859
12651 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
12652 test with comment from bb_no_side_effects_p.
12654 2018-06-21 Richard Biener <rguenther@suse.de>
12656 PR tree-optimization/86232
12657 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
12658 max for constant niter.
12660 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
12662 * config/aarch64/aarch64-simd.md
12663 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
12665 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
12667 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
12668 Make opernads of the unspec commutative.
12670 2018-06-21 Richard Biener <rguenther@suse.de>
12672 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
12673 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
12674 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
12675 (vect_analyze_data_ref_dependence): Re-order checks to deal with
12677 (vect_record_base_alignments): Do not record base alignment
12678 for gathers or scatters.
12679 (vect_compute_data_ref_alignment): Drop return value that is always
12680 true. Bail out early for gathers or scatters.
12681 (vect_enhance_data_refs_alignment): Bail out early for gathers
12683 (vect_find_same_alignment_drs): Likewise.
12684 (vect_analyze_data_refs_alignment): Remove dead code.
12685 (vect_slp_analyze_and_verify_node_alignment): Likewise.
12686 (vect_analyze_data_refs): For possible gathers or scatters do
12687 not create an alternate DR, just check their possible validity
12688 and mark them. Adjust DECL_NONALIASED handling to not rely
12689 on DR_BASE_ADDRESS.
12690 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
12691 update inits of gathers or scatters.
12692 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
12693 Also copy gather/scatter flag to pattern vinfo.
12695 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
12697 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
12698 behavior of vec_packsu (vector unsigned long long, vector unsigned
12699 long long) to match behavior of vec_packs with same signature.
12701 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
12702 Thomas Schwinge <thomas@codesourcery.com>
12703 Cesar Philippidis <cesar@codesourcery.com>
12705 * gimplify.c (gimplify_scan_omp_clauses): Add support for
12706 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
12707 (gimplify_adjust_omp_clauses): Likewise.
12708 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
12709 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
12710 (gimplify_omp_target_update): Update handling of acc update and
12712 * omp-low.c (install_var_field): Remove unused parameter
12713 base_pointers_restrict.
12714 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
12715 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
12717 (omp_target_base_pointers_restrict_p): Delete.
12718 (scan_omp_target): Update call to scan_sharing_clauses.
12719 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
12721 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
12722 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
12723 (convert_local_omp_clauses): Likewise.
12724 * tree-pretty-print.c (dump_omp_clause): Likewise.
12725 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
12727 (omp_clause_code_name): Likewise.
12729 2018-06-20 Jakub Jelinek <jakub@redhat.com>
12732 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
12735 PR tree-optimization/86231
12736 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
12737 anti-range don't overwrite *vr0min before using it to compute *vr0max.
12739 2018-06-20 Tom de Vries <tdevries@suse.de>
12741 PR tree-optimization/86097
12742 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
12743 iv type if signedness of iv type is not the same as that of *nit.
12745 2018-06-20 Jakub Jelinek <jakub@redhat.com>
12747 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
12748 EDGE_EH edges, verify they are all EDGE_EH.
12750 2018-06-20 Maya Rashish <coypu@sdf.org>
12752 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
12754 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12756 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
12757 * config/aarch64/aarch64.c (xgene1_tunings): Add
12758 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
12759 (aarch64_mode_valid_for_sched_fusion_p):
12760 Allow 16-byte modes.
12761 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
12762 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
12764 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
12766 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
12767 * config/aarch64/iterators.md (VQ2): New mode iterator.
12769 2018-06-20 Martin Liska <mliska@suse.cz>
12771 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
12772 Change default ratio from 10 to 8.
12774 2018-06-20 Martin Liska <mliska@suse.cz>
12776 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
12778 (bit_test_cluster::find_bit_tests): Likewise.
12779 (switch_decision_tree::analyze_switch_statement): Find clusters.
12780 * tree-switch-conversion.h (struct jump_table_cluster): Document
12783 2018-06-20 Martin Liska <mliska@suse.cz>
12785 * tree-switch-conversion.c (switch_conversion::collect):
12786 Record m_uniq property.
12787 (switch_conversion::expand): Bail out for special conditions.
12788 (group_cluster::~group_cluster): New.
12789 (group_cluster::group_cluster): Likewise.
12790 (group_cluster::dump): Likewise.
12791 (jump_table_cluster::emit): New.
12792 (switch_decision_tree::fix_phi_operands_for_edges): New.
12793 (struct case_node): Remove struct.
12794 (jump_table_cluster::can_be_handled): New.
12795 (case_values_threshold): Moved to header.
12796 (reset_out_edges_aux): Likewise.
12797 (jump_table_cluster::is_beneficial): New.
12798 (bit_test_cluster::can_be_handled): Likewise.
12799 (add_case_node): Remove.
12800 (bit_test_cluster::is_beneficial): New.
12801 (case_bit_test::cmp): New.
12802 (bit_test_cluster::emit): New.
12803 (expand_switch_as_decision_tree_p): Remove.
12804 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
12805 (fix_phi_operands_for_edge): Likewise.
12806 (switch_decision_tree::analyze_switch_statement): New.
12807 (compute_cases_per_edge): Move ...
12808 (switch_decision_tree::compute_cases_per_edge): ... here.
12809 (try_switch_expansion): Likewise.
12810 (switch_decision_tree::try_switch_expansion): Likewise.
12811 (record_phi_operand_mapping): Likewise.
12812 (switch_decision_tree::record_phi_operand_mapping): Likewise.
12813 (emit_case_decision_tree): Likewise.
12814 (switch_decision_tree::emit): Likewise.
12815 (balance_case_nodes): Likewise.
12816 (switch_decision_tree::balance_case_nodes): Likewise.
12817 (dump_case_nodes): Likewise.
12818 (switch_decision_tree::dump_case_nodes): Likewise.
12819 (emit_jump): Likewise.
12820 (switch_decision_tree::emit_jump): Likewise.
12821 (emit_cmp_and_jump_insns): Likewise.
12822 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
12823 (emit_case_nodes): Likewise.
12824 (switch_decision_tree::emit_case_nodes): Likewise.
12825 (conditional_probability): Remove.
12826 * tree-switch-conversion.h (enum cluster_type): New.
12828 (struct cluster): Likewise.
12829 (cluster::cluster): Likewise.
12830 (struct simple_cluster): Likewise.
12831 (simple_cluster::simple_cluster): Likewise.
12832 (struct group_cluster): Likewise.
12833 (struct jump_table_cluster): Likewise.
12834 (struct bit_test_cluster): Likewise.
12835 (struct min_cluster_item): Likewise.
12836 (struct case_tree_node): Likewise.
12837 (case_tree_node::case_tree_node): Likewise.
12838 (jump_table_cluster::case_values_threshold): Likewise.
12839 (struct case_bit_test): Likewise.
12840 (struct switch_decision_tree): Likewise.
12841 (struct switch_conversion): Likewise.
12842 (switch_decision_tree::reset_out_edges_aux): Likewise.
12844 2018-06-20 Martin Liska <mliska@suse.cz>
12846 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
12847 (hoist_edge_and_branch_if_true): Likewise.
12848 (expand_switch_using_bit_tests_p): Likewise.
12849 (struct case_bit_test): Likewise.
12850 (case_bit_test_cmp): Likewise.
12851 (emit_case_bit_tests): Likewise.
12852 (switch_conversion::switch_conversion): New class.
12853 (struct switch_conv_info): Remove old struct.
12854 (collect_switch_conv_info): More to ...
12855 (switch_conversion::collect): ... this.
12856 (check_range): Likewise.
12857 (switch_conversion::check_range): Likewise.
12858 (check_all_empty_except_final): Likewise.
12859 (switch_conversion::check_all_empty_except_final): Likewise.
12860 (check_final_bb): Likewise.
12861 (switch_conversion::check_final_bb): Likewise.
12862 (create_temp_arrays): Likewise.
12863 (switch_conversion::create_temp_arrays): Likewise.
12864 (free_temp_arrays): Likewise.
12865 (gather_default_values): Likewise.
12866 (switch_conversion::gather_default_values): Likewise.
12867 (build_constructors): Likewise.
12868 (switch_conversion::build_constructors): Likewise.
12869 (constructor_contains_same_values_p): Likewise.
12870 (switch_conversion::contains_same_values_p): Likewise.
12871 (array_value_type): Likewise.
12872 (switch_conversion::array_value_type): Likewise.
12873 (build_one_array): Likewise.
12874 (switch_conversion::build_one_array): Likewise.
12875 (build_arrays): Likewise.
12876 (switch_conversion::build_arrays): Likewise.
12877 (gen_def_assigns): Likewise.
12878 (switch_conversion::gen_def_assigns): Likewise.
12879 (prune_bbs): Likewise.
12880 (switch_conversion::prune_bbs): Likewise.
12881 (fix_phi_nodes): Likewise.
12882 (switch_conversion::fix_phi_nodes): Likewise.
12883 (gen_inbound_check): Likewise.
12884 (switch_conversion::gen_inbound_check): Likewise.
12885 (process_switch): Use the newly created class.
12886 (switch_conversion::expand): New.
12887 (switch_conversion::~switch_conversion): New.
12888 * tree-switch-conversion.h: New file.
12890 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
12892 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
12893 tree-vect-patterns.c.
12894 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
12895 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
12896 (vect_recog_sad_pattern): Likewise.
12897 (vect_recog_widen_sum_pattern): Likewise.
12898 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
12899 (vect_recog_widen_shift_pattern): Remove the type_in argument.
12900 (vect_recog_rotate_pattern): Likewise.
12901 (vect_recog_mult_pattern): Likewise.
12902 (vect_recog_vector_vector_shift_pattern): Likewise.
12903 (vect_recog_divmod_pattern): Likewise.
12904 (vect_recog_mixed_size_cond_pattern): Likewise.
12905 (vect_recog_bool_pattern): Likewise.
12906 (vect_recog_mask_conversion_pattern): Likewise.
12907 (vect_try_gather_scatter_pattern): Likewise.
12908 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
12909 (vect_recog_over_widening_pattern): Likewise.
12910 (vect_recog_gather_scatter_pattern): Likewise.
12911 (vect_recog_func_ptr): Move from tree-vectorizer.h
12912 (vect_vect_recog_func_ptrs): Move further down the file.
12913 (vect_recog_func): Likewise. Remove the third argument.
12914 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
12915 (vect_pattern_recog_1): Expect the pattern function to do any
12916 necessary target tests. Also expect it to provide a vector type.
12917 Remove the type_in handling.
12919 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
12921 * tree-vect-patterns.c (vect_pattern_detected): New function.
12922 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
12923 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
12924 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
12925 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
12926 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
12927 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
12928 (vect_recog_mask_conversion_pattern)
12929 (vect_try_gather_scatter_pattern): Likewise.
12931 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
12933 * tree-vect-patterns.c (vect_get_internal_def): New function.
12934 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
12935 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
12936 (search_type_for_mask_1): Use it.
12938 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
12940 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
12941 redundant WIDEN_SUM_EXPR handling.
12942 (vect_recog_sad_pattern): Likewise.
12944 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
12946 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
12947 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
12948 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
12949 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
12950 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
12952 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
12954 * tree-vect-stmts.c (vectorizable_call): Make sure that we
12955 use the stmt_vec_info of the original bb statement for the
12956 new zero assignment, even if the call is part of a pattern.
12958 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
12960 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
12961 that the sequence is attached to the original statement rather
12962 than the pattern statement.
12963 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
12964 PATTERN_DEF_SEQ from the original statement rather than
12965 the main pattern statement.
12966 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
12967 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
12968 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
12970 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
12972 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
12973 definition statements before the early exit for statements that aren't
12975 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
12977 (vect_transform_loop): ...here. Process pattern definition
12978 statements without first checking whether the main pattern
12979 statement is live or relevant.
12981 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
12983 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
12984 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
12986 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
12988 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
12989 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
12990 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
12991 (expand_block_compare): Change select_block_compare_mode call.
12992 (expand_strncmp_align_check): Use new functions, fix comment.
12993 (emit_final_str_compare_gpr): New function.
12994 (expand_strn_compare): Refactor and clean up code.
12995 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
12997 2018-06-19 Tony Reix <tony.reix@atos.com>
12998 Damien Bergamini <damien.bergamini@atos.com>
12999 David Edelsohn <dje.gcc@gmail.com>
13001 * collect2.c (static_obj): New variable.
13002 (static_libs): New variable.
13003 (is_in_list): Uncomment declaration.
13004 (main): Track AIX libraries linked statically.
13005 (is_in_list): Uncomment definition.
13006 (scan_prog_file): Don't add AIX shared libraries initializer
13007 to constructor list if linking statically.
13009 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
13011 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
13013 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
13015 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
13017 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
13020 2018-06-19 Martin Liska <mliska@suse.cz>
13022 * config/i386/i386.c (ix86_can_inline_p): Do not use
13023 ipa_fn_summaries::get_create.
13024 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
13026 (devirtualization_time_bonus): Likewise.
13027 (ipcp_propagate_stage): Likewise.
13028 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
13029 (edge_set_predicate): Likewise.
13030 (evaluate_conditions_for_known_args): Likewise.
13031 (evaluate_properties_for_edge): Likewise.
13032 (ipa_call_summary::reset): Tranform to ...
13033 (ipa_call_summary::~ipa_call_summary): ... this.
13034 (ipa_fn_summary::reset): Transform to ...
13035 (ipa_fn_summary::~ipa_fn_summary): ... this.
13036 (ipa_fn_summary_t::remove): Rename to ...
13037 (ipa_fn_summary_t::remove_callees): ... this.
13038 (ipa_fn_summary_t::duplicate): Use placement new
13039 instead of memory copy.
13040 (ipa_call_summary_t::duplicate): Likewise.
13041 (ipa_call_summary_t::remove): Remove.
13042 (dump_ipa_call_summary): Change get_create to get.
13043 (ipa_dump_fn_summary): Dump only when summary exists.
13044 (analyze_function_body): Use symbol_summary::get instead
13046 (compute_fn_summary): Likewise.
13047 (estimate_edge_devirt_benefit): Likewise.
13048 (estimate_edge_size_and_time): Likewise.
13049 (inline_update_callee_summaries): Likewise.
13050 (remap_edge_change_prob): Likewise.
13051 (remap_edge_summaries): Likewise.
13052 (ipa_merge_fn_summary_after_inlining): Likewise.
13053 (write_ipa_call_summary): Likewise.
13054 (ipa_fn_summary_write): Likewise.
13055 (ipa_free_fn_summary): Likewise.
13056 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
13057 (struct ipa_call_summary): Likewise.
13058 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
13060 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
13061 (estimate_size_after_inlining): Likewise.
13062 (estimate_growth): Likewise.
13063 (growth_likely_positive): Likewise.
13064 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
13065 (inline_call): Likewise.
13066 * ipa-inline.c (caller_growth_limits): Likewise.
13067 (can_inline_edge_p): Likewise.
13068 (can_inline_edge_by_limits_p): Likewise.
13069 (compute_uninlined_call_time): Likewise.
13070 (compute_inlined_call_time): Likewise.
13071 (want_inline_small_function_p): Likewise.
13072 (edge_badness): Likewise.
13073 (update_caller_keys): Likewise.
13074 (update_callee_keys): Likewise.
13075 (inline_small_functions): Likewise.
13076 (inline_to_all_callers_1): Likewise.
13077 (dump_overall_stats): Likewise.
13078 (early_inline_small_functions): Likewise.
13079 (early_inliner): Likewise.
13080 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
13081 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
13082 * ipa-pure-const.c (malloc_candidate_p): Likewise.
13083 * ipa-split.c (execute_split_functions): Likewise.
13084 * symbol-summary.h: Likewise.
13085 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
13087 2018-06-19 Richard Biener <rguenther@suse.de>
13089 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
13090 (vectorize_loops): ... here. Fix dbgcnt handling.
13091 (try_vectorize_loop): Wrap try_vectorize_loop_1.
13093 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
13096 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
13097 ieee128 argument takes up only one (vector) register, not two (floating
13100 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
13102 * gimplify.c (gimplify_init_constructor): Really never clear for an
13103 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
13105 2018-06-19 Richard Biener <rguenther@suse.de>
13107 PR tree-optimization/86179
13108 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
13109 after failed recognition.
13111 2018-06-18 Martin Sebor <msebor@redhat.com>
13113 PR middle-end/85602
13114 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
13115 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
13116 Handle integer subtraction.
13117 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
13118 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
13120 2018-06-18 David Malcolm <dmalcolm@redhat.com>
13122 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
13123 param from rtx to rtx_insn *.
13124 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
13126 (frv_ifcvt_modify_insn): Likwise.
13127 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
13128 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
13129 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
13130 as_a <rtx_insn *> cast to local "unprotected_region" once
13131 it's been established that it's not NULL or pc_rtx.
13132 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
13133 param "sethi" from rtx to rtx_insn *.
13134 (nds32_group_float_insns): Likewise for param "insn".
13135 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
13137 (vax_output_int_subtract): Likewise.
13138 * config/vax/vax.c (vax_output_int_add): Likewise for param
13140 (vax_output_int_subtract): Likewise.
13141 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
13142 (emit_pattern_after): Likewise for param "after".
13143 (emit_insn_after): Likewise.
13144 (emit_jump_insn_after): Likewise.
13145 (emit_call_insn_after): Likewise.
13146 (emit_debug_insn_after): Likewise.
13147 (emit_pattern_before): Likewise for param "before".
13148 (emit_insn_before): Likewise.
13149 (emit_jump_insn_before): Likewise.
13150 * final.c (get_insn_template): Likewise for param "insn", removing
13152 * output.h (get_insn_template): Likewise for 2nd param.
13153 * rtl.h (emit_insn_before): Likewise.
13154 (emit_jump_insn_before): Likewise.
13155 (emit_debug_insn_before_noloc): Likewise.
13156 (emit_insn_after): Likewise.
13157 (emit_jump_insn_after): Likewise.
13158 (emit_call_insn_after): Likewise.
13159 (emit_debug_insn_after): Likewise.
13160 (set_insn_deleted): Likewise for param.
13162 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
13165 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
13166 floating point modes, so that IFmode is numerically greater than
13167 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
13168 to declare the ordering. This prevents IFmode from being
13169 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
13170 machine. Include rs6000-modes.h to share the fractional values
13171 between genmodes* and the rest of the compiler.
13172 (IFmode): Likewise.
13173 (KFmode): Likewise.
13174 (TFmode): Likewise.
13175 * config/rs6000/rs6000-modes.h: New file.
13176 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
13177 meaning of rs6000_long_double_size so that 126..128 selects an
13178 appropriate 128-bit floating point type.
13179 (rs6000_option_override_internal): Likewise.
13180 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
13181 (TARGET_LONG_DOUBLE_128): Change the meaning of
13182 rs6000_long_double_size so that 126..128 selects an appropriate
13183 128-bit floating point type.
13184 (LONG_DOUBLE_TYPE_SIZE): Update comment.
13185 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
13186 source and destination to match the standard usage.
13187 (truncifkf2): Likewise.
13188 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
13189 ISA 2.07 to use an explicit clobber, instead of passing in a
13191 (copysign<mode>3_soft): Likewise.
13193 2018-06-18 David Malcolm <dmalcolm@redhat.com>
13195 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
13196 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
13197 (vect_slp_analyze_instance_dependence): Likewise.
13198 (vect_enhance_data_refs_alignment): Likewise.
13199 (vect_analyze_data_refs_alignment): Likewise.
13200 (vect_slp_analyze_and_verify_instance_alignment
13201 (vect_analyze_data_ref_accesses): Likewise.
13202 (vect_prune_runtime_alias_test_list): Likewise.
13203 (vect_analyze_data_refs): Likewise.
13204 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
13205 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13206 (vect_analyze_scalar_cycles_1): Likewise.
13207 (vect_get_loop_niters): Likewise.
13208 (vect_analyze_loop_form_1): Likewise.
13209 (vect_update_vf_for_slp): Likewise.
13210 (vect_analyze_loop_operations): Likewise.
13211 (vect_analyze_loop): Likewise.
13212 (vectorizable_induction): Likewise.
13213 (vect_transform_loop): Likewise.
13214 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
13215 * tree-vect-slp.c (vect_analyze_slp): Likewise.
13216 (vect_make_slp_decision): Likewise.
13217 (vect_detect_hybrid_slp): Likewise.
13218 (vect_slp_analyze_operations): Likewise.
13219 (vect_slp_bb): Likewise.
13220 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
13221 (vectorizable_bswap): Likewise.
13222 (vectorizable_call): Likewise.
13223 (vectorizable_simd_clone_call): Likewise.
13224 (vectorizable_conversion): Likewise.
13225 (vectorizable_assignment): Likewise.
13226 (vectorizable_shift): Likewise.
13227 (vectorizable_operation): Likewise.
13228 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
13230 2018-06-18 Martin Sebor <msebor@redhat.com>
13232 PR tree-optimization/81384
13233 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
13234 * builtins.c (expand_builtin_strnlen): New function.
13235 (expand_builtin): Call it.
13236 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
13237 * builtins.def (BUILT_IN_STRNLEN): New.
13238 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
13239 Warn for bounds in excess of maximum object size.
13240 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
13241 single-value ranges. Handle strnlen.
13242 (handle_builtin_strlen): Handle strnlen.
13243 (strlen_check_and_optimize_stmt): Same.
13244 * doc/extend.texi (Other Builtins): Document strnlen.
13246 2018-06-18 Maya Rashish <coypu@sdf.org>
13248 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
13249 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
13250 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
13252 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
13254 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
13256 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13258 * tree.c (escaped_string::escape): Replace cast to char * by
13259 const_cast<char *> (unescaped).
13261 2018-06-18 Nick Clifton <nickc@redhat.com>
13264 * tree.c (escaped_string): New class. Converts an unescaped
13265 string into its escaped equivalent.
13266 (warn_deprecated_use): Use the new class to convert the
13267 deprecation message, if present.
13268 (test_escaped_strings): New self test.
13269 (test_c_tests): Add test_escaped_strings.
13270 * doc/extend.texi (deprecated): Add a note that the
13271 deprecation message is affected by the -fmessage-length
13272 option, and that control characters will be escaped.
13273 (#pragma GCC error): Document this pragma.
13274 (#pragma GCC warning): Likewise.
13275 * doc/invoke.texi (-fmessage-length): Document this option's
13276 effect on the #warning and #error preprocessor directives and
13277 the deprecated attribute.
13279 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
13281 * tree.c (decl_value_expr_lookup): Revert latest change.
13282 (decl_value_expr_insert): Likewise.
13284 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
13286 * gimplify.c (nonlocal_vlas): Delete.
13287 (nonlocal_vla_vars): Likewise.
13288 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
13290 (gimplify_body): Do not create and destroy nonlocal_vlas.
13291 * tree-nested.c: Include diagnostic.h.
13292 (use_pointer_in_frame): Tweak.
13293 (lookup_field_for_decl): Add assertion and declare the transformation.
13294 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
13295 internal error when the reference is in a wrong context. Do not
13296 create a debug decl by default.
13297 (note_nonlocal_block_vlas): Delete.
13298 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
13299 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
13300 create a debug decl by default.
13301 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
13302 call is in a wrong context.
13303 (fixup_vla_decls): New function.
13304 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
13305 debug variables were created.
13306 * tree.c (decl_value_expr_lookup): Add checking assertion.
13307 (decl_value_expr_insert): Likewise.
13309 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
13311 PR middle-end/82479
13312 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
13313 * tree-scalar-evolution.c (interpret_expr): Likewise.
13314 (expression_expensive_p): Likewise.
13315 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
13316 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
13317 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
13318 (ssa_defined_by_minus_one_stmt_p): New.
13320 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
13322 PR middle-end/64946
13323 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
13324 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
13325 * dojump.c (do_jump): Likewise.
13326 * expr.c (expand_expr_real_2): Check operand type's sign.
13327 * fold-const.c (const_unop): Handle ABSU_EXPR.
13328 (fold_abs_const): Likewise.
13329 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
13330 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
13331 (strip_sign_op_1): Likesise.
13332 * match.pd: Add new pattern to generate ABSU_EXPR.
13333 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
13334 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
13335 * tree-eh.c (operation_could_trap_helper_p): Likewise.
13336 * tree-inline.c (estimate_operator_cost): Likewise.
13337 * tree-pretty-print.c (dump_generic_node): Likewise.
13338 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
13339 * tree.def (ABSU_EXPR): New.
13341 2018-06-16 Jakub Jelinek <jakub@redhat.com>
13343 PR middle-end/86095
13344 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
13345 documented as preserved for backward compatibility only.
13346 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
13348 PR rtl-optimization/86108
13349 * bb-reorder.c (create_forwarder_block): Renamed to ...
13350 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
13351 jump from new landing pad to the second part.
13352 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
13355 2018-06-15 Jakub Jelinek <jakub@redhat.com>
13357 PR middle-end/85878
13358 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
13359 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
13360 Only call store_expr for halves if the mode is the same.
13362 PR middle-end/86123
13363 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
13364 Fix up comment formatting.
13366 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
13368 * typed-splay-tree.h (typed_splay_tree::remove): New function.
13369 (typed_splay_tree::closure,
13370 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
13371 (typed_splay_tree::typed_splay_tree,
13372 typed_splay_tree::operator =): Declared private.
13373 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
13374 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
13375 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
13376 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
13377 typed_splay_tree::splay_tree_splay,
13378 typed_splay_tree::splay_tree_foreach_helper,
13379 typed_splay_tree::splay_tree_insert,
13380 typed_splay_tree::splay_tree_remove,
13381 typed_splay_tree::splay_tree_lookup,
13382 typed_splay_tree::splay_tree_predecessor,
13383 typed_splay_tree::splay_tree_successor,
13384 typed_splay_tree::splay_tree_min,
13385 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
13386 (typed_splay_tree::root, typed_splay_tree::comp,
13387 typed_splay_tree::delete_key,
13388 typed_splay_tree::delete_value): New data members.
13389 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
13390 typed_splay_tree::remove.
13392 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
13394 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
13395 -mginv and -mno-ginv to the assembler.
13396 * config/mips/mips.opt (-mcrc): New option.
13397 (-mginv): Likewise.
13398 * doc/invoke.text (-mcrc): Document.
13399 (-mginv): Likewise.
13401 2018-06-15 Nick Clifton <nickc@redhat.com>
13404 * tree.c (escaped_string): New class. Converts an unescaped
13405 string into its escaped equivalent.
13406 (warn_deprecated_use): Use the new class to convert the
13407 deprecation message, if present.
13408 (test_escaped_strings): New self test.
13409 (test_c_tests): Add test_escaped_strings.
13410 * doc/extend.texi (deprecated): Add a note that the
13411 deprecation message is affected by the -fmessage-length
13412 option, and that control characters will be escaped.
13413 (#pragma GCC error): Document this pragma.
13414 (#pragma GCC warning): Likewise.
13415 * doc/invoke.texi (-fmessage-length): Document this option's
13416 effect on the #warning and #error preprocessor directives and
13417 the deprecated attribute.
13419 2018-06-15 Richard Biener <rguenther@suse.de>
13421 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
13422 here, also noting vector size used.
13423 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
13424 size used in MSG_OPTIMIZED_LOCATIONS dump.
13425 (pass_slp_vectorize::execute): Adjust.
13427 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
13430 * config/arc/arc.c (arc_return_address_register): Fix
13433 2018-06-15 Richard Biener <rguenther@suse.de>
13435 PR middle-end/86159
13436 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
13437 leave useless conversion stripping to force_gimple_operand_gsi.
13438 (gimplify_build2): Likewise.
13439 (gimplify_build1): Likewise.
13441 2018-06-15 Richard Biener <rguenther@suse.de>
13443 PR middle-end/86076
13444 * tree-cfg.c (move_stmt_op): unshare invariant addresses
13445 before adjusting their block.
13447 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
13449 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
13450 multilibs for *-*-rtems*.
13451 * config/riscv/t-rtems: New file.
13453 2018-06-14 Jakub Jelinek <jakub@redhat.com>
13455 PR middle-end/86122
13456 * match.pd ((A +- CST1) +- CST2): Punt if last resort
13457 unsigned_type_for returns NULL.
13460 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
13461 subregs of multi-word pseudos unless the float mode has word size.
13463 2018-06-14 Richard Biener <rguenther@suse.de>
13465 PR middle-end/86139
13466 * tree-vect-generic.c (build_word_mode_vector_type): Remove
13467 duplicate and harmful type_hash_canon.
13468 * tree.c (type_hash_canon): Assert we didn't find ourselves.
13470 2018-06-14 Richard Biener <rguenther@suse.de>
13473 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
13476 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
13478 * config/rtems.h (STDINT_LONG32): Define.
13480 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
13481 Prachi Godbole <prachi.godbole@imgtec.com>
13483 * config/mips/mips-cpus.def: Define P6600.
13484 * config/mips/mips-tables.opt: Regenerate.
13485 * config/mips/mips.c (mips_ucbranch_type): New enum.
13486 (mips_rtx_cost_data): Add support for P6600.
13487 (mips_issue_rate): Likewise.
13488 (mips_multipass_dfa_lookahead): Likewise.
13489 (mips_avoid_hazard): Likewise.
13490 (mips_reorg_process_insns): Likewise.
13491 (mips_classify_branch_p6600): New function.
13492 * config/mips/mips.h (TUNE_P6600): New define.
13493 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
13494 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
13495 * config/mips/mips.md: Include p6600.md.
13496 (processor): Add p6600.
13497 * config/mips/p6600.md: New file.
13498 * doc/invoke.texi: Add p6600 to supported architectures.
13500 2018-06-13 Martin Sebor <msebor@redhat.com>
13502 PR tree-optimization/86114
13503 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
13505 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
13507 2018-06-13 Richard Biener <rguenther@suse.de>
13509 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
13510 Properly set vector type of the intermediate stmt.
13511 * tree-vect-stmts.c (vectorizable_operation): The destination
13512 var always has vectype_out type.
13514 2018-06-13 Jeff Law <law@redhat.com>
13516 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
13517 integer 0 for argument to print_rtl_with_bb.
13518 (rl78_reorg): Likewise.
13520 2018-06-13 David Malcolm <dmalcolm@redhat.com>
13522 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
13523 from rtx to rtx_insn *.
13524 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
13526 (add_sched_insns_for_speculation): Likewise for local "target",
13527 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
13528 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
13529 from rtx_insn ** to rtx_code_label **.
13530 (reorg_emit_nops): Likewise.
13531 (c6x_reorg): Likewise for local "call_labels".
13532 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
13534 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
13535 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
13536 the loops over LABEL_REFS.
13537 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
13539 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
13540 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
13541 (split_branches): Strengthen local "olabel" from rtx to
13542 rtx_insn *, adding a safe_as_a cast.
13543 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
13545 (add_insn_after): Likewise for first two params.
13546 (add_insn_before): Likewise.
13547 (remove_insn): Likewise for param.
13548 (emit_pattern_before_noloc): Likewise for second and third params.
13549 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
13550 (emit_call_insn_before_noloc): Likewise.
13551 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
13553 (emit_barrier_before): Likewise.
13554 (emit_label_before): Strengthen "label" param from "rtx" to
13555 "rtx_code_label *". Strengthen "before" param from "rtx" to
13557 (emit_insn_after_1): Strengthen "after" param from "rtx" to
13559 (emit_pattern_after_noloc): Likewise.
13560 (emit_insn_after_noloc): Likewise.
13561 (emit_jump_insn_after_noloc): Likewise.
13562 (emit_call_insn_after_noloc): Likewise.
13563 (emit_debug_insn_after_noloc): Likewise.
13564 (emit_barrier_after): Likewise.
13565 (emit_label_after): Likewise for both params.
13566 (emit_pattern_after_setloc): Likewise for "after" param. Convert
13567 "loc" param from "int" to "location_t".
13568 (emit_insn_after_setloc): Likewise.
13569 (emit_jump_insn_after_setloc): Likewise.
13570 (emit_call_insn_after_setloc): Likewise.
13571 (emit_debug_insn_after_setloc): Likewise.
13572 (emit_pattern_before_setloc): Likewise for "before" param. Convert
13573 "loc" param from "int" to "location_t".
13574 (emit_pattern_before): Convert NULL_RTX to NULL.
13575 (emit_insn_before_setloc): Convert "loc" param from "int" to
13577 (emit_jump_insn_before_setloc): Likewise.
13578 (emit_call_insn_before_setloc): Likewise.
13579 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
13580 rtx_insn *. Convert "loc" param from "int" to "location_t".
13581 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
13582 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
13583 Convert 3rd param from "int" to "location_t".
13584 (emit_barrier_before, emit_barrier_after, next_real_insn):
13585 Strengthen param from rtx to rtx_insn *.
13586 (emit_label_before): Strengthen 1st param from "rtx" to
13587 "rtx_code_label *". Strengthen 2nd param from "rtx" to
13589 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
13590 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
13591 Strengthen 2nd param from "rtx" to "rtx_insn *".
13592 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
13593 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
13594 Likewise. Convert 3rd param from "int" to "location_t".
13595 (emit_label_after): Strengthen 1st param from "rtx" to
13596 "rtx_code_label *".
13597 (next_real_insn, remove_insn): Strengthen param from "rtx" to
13599 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
13600 from "rtx" to "rtx_insn *".
13602 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
13604 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
13605 bodies streamed in with -Q.
13606 * dumpfile.c (dump_files): Add lto-stream-out dump file.
13607 * dumpfile.h (tree_dump_index): Add lto_stream_out.
13608 * gimple-streamer-out.c: Include gimple-pretty-print.h
13609 (output_bb): Dump stmts streamed.
13610 * lto-section-out.c: Include print-tree.h
13611 (lto_begin_section): Dump sections created.
13612 (lto_output_decl_index): Dump decl encoded.
13613 * lto-streamer-out.c: Include print-tree.h
13614 (create_output_block): Dump output block created.
13615 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
13616 (output_function): Dump function output.
13617 (output_constructor): Dump constructor streamed.
13618 (write_global_stream): Output indexes encoded.
13619 (produce_asm_for_decls): Dump streams encoded.
13620 * lto-streamer.c (streamer_dump_file): New global var.
13621 * lto-streamer.h (streamer_dump_file): Declare.
13622 * passes.c (ipa_write_summaries): Initialize streamer dump.
13623 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
13626 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
13629 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
13630 offsets for register save directives. Emit a second batch of save
13631 directives, if need be, when the function accesses prior frames.
13633 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
13635 * config/arc/fpu.md (fmasf4): Force operand to register.
13636 (fnmasf4): Likewise.
13638 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
13640 * config/arc/arc-protos.h (arc_pad_return): Remove.
13641 * config/arc/arc.c (machine_function): Remove force_short_suffix
13643 (arc_print_operand): Adjust printing of '&'.
13644 (arc_verify_short): Remove conditional printing of short suffix.
13645 (arc_final_prescan_insn): Remove reference to size_reason.
13646 (pad_return): New function.
13647 (arc_reorg): Call pad_return.
13648 (arc_pad_return): Remove.
13649 (arc_init_machine_status): Remove reference to force_short_suffix.
13650 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
13651 (attr length): When attribute iscompact is true force to 2
13652 regardless; in the case of maybe check if we want to force the
13653 instruction to have 4 bytes length.
13654 (nopv): Change it to generate 4 byte long nop as well.
13655 (blockage): New pattern.
13656 (simple_return): Remove call to arc_pad_return.
13657 (p_return_i): Likewise.
13659 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
13661 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
13663 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
13665 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
13668 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
13670 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
13671 for ARC700 and ARCv2.
13673 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
13676 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
13677 operands[2] instead of operands[1].
13680 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
13682 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
13683 case, check whether the outer register overlaps an unallocatable
13684 register, not just whether it fits the required class.
13686 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
13688 * poly-int.h (can_div_trunc_p): Add new overload in which all values
13690 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
13691 (memrefs_conflict_p): Likewise.
13692 (init_alias_analysis): Likewise.
13693 * cfgexpand.c (expand_debug_expr): Likewise.
13694 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
13695 * cse.c (fold_rtx): Likewise.
13696 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
13697 * expr.c (emit_block_move_hints): Likewise.
13698 (clear_storage_hints, push_block, emit_push_insn): Likewise.
13699 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
13700 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
13701 (emit_group_store): Likewise.
13702 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
13703 to read the PRE/POST_MODIFY increment.
13704 * calls.c (store_one_arg): Use strip_offset.
13705 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
13707 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
13709 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
13710 (simplify_binary_operation_1): Extend CONST_INT handling to
13712 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
13713 than a HOST_WIDE_INT.
13714 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
13716 (adjust_mems, add_stores): Update accodingly.
13717 (vt_canonicalize_addr): Track polynomial offsets.
13718 (emit_note_insn_var_location): Likewise.
13719 (vt_add_function_parameter): Likewise.
13720 (vt_initialize): Likewise.
13722 2018-06-12 Jeff Law <law@redhat.com>
13724 * config.gcc (alpha*-*-freebsd*): Remove.
13725 * config/alpha/freebsd.h: Remove.
13727 2018-06-12 David Malcolm <dmalcolm@redhat.com>
13730 * spellcheck-tree.c (levenshtein_distance): Rename to...
13731 (get_edit_distance): ...this, and update for underlying renaming.
13732 * spellcheck-tree.h (levenshtein_distance): Rename to...
13733 (get_edit_distance): ...this.
13734 * spellcheck.c (levenshtein_distance): Rename to...
13735 (get_edit_distance): ...this. Convert from Levenshtein distance
13736 to Damerau-Levenshtein distance by supporting transpositions of
13737 adjacent characters. Rename "v1" to "v_next" and "v0" to
13739 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
13740 (selftest::test_edit_distance_unit_test_oneway): ...this, and
13741 update for underlying renaming.
13742 (selftest::levenshtein_distance_unit_test): Rename to...
13743 (selftest::test_get_edit_distance_unit): ...this, and update for
13744 underlying renaming.
13745 (selftest::test_find_closest_string): Add example from PR 69968
13746 where transposition helps
13747 (selftest::test_metric_conditions): Update for renaming.
13748 (selftest::test_metric_conditions): Likewise.
13749 (selftest::spellcheck_c_tests): Likewise.
13750 * spellcheck.h (levenshtein_distance): Rename both overloads to...
13751 (get_edit_distance): ...this.
13752 (best_match::consider): Update for renaming.
13754 2018-06-12 Martin Sebor <msebor@redhat.com>
13756 PR tree-optimization/85259
13757 * builtins.c (compute_objsize): Handle constant offsets.
13758 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
13759 true iff a warning has been issued.
13760 * gimple.h (gimple_nonartificial_location): New function.
13761 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
13762 gimple_nonartificial_location and handle -Wno-system-headers.
13763 (handle_builtin_stxncpy): Same.
13765 2018-06-12 Martin Sebor <msebor@redhat.com>
13768 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
13770 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
13772 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
13773 BUILTIN_VEC_XST entries for pointer to double and long long.
13775 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
13778 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
13780 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
13783 2018-06-12 Martin Liska <mliska@suse.cz>
13785 * doc/options.texi: Document IntegerRange.
13787 2018-06-12 Martin Liska <mliska@suse.cz>
13789 * config/i386/i386.opt: Make MPX-related options as Deprecated.
13790 * opt-functions.awk: Handle Deprecated flag.
13791 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
13793 (read_cmdline_option): Report warning for a deprecated option.
13794 * opts.h (struct cl_option): Add new field cl_deprecated.
13795 (CL_ERR_DEPRECATED): New.
13797 2018-06-12 Martin Liska <mliska@suse.cz>
13799 * doc/options.texi: Document Deprecated option flag.
13801 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
13803 * config/arc/arc-arch.h (arc_extras): New enum.
13804 (arc_cpu_t):Add field extra.
13805 (arc_cpu_types): Consider the extras.
13806 * config/arc/arc-cpus.def: Add extras info.
13807 * config/arc/arc-opts.h (processor_type): Consider extra field.
13808 * config/arc/arc.c (arc_override_options): Handle extra field.
13810 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
13812 * config/arc/arc-arch.h: Update ARC_OPTX macro.
13813 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
13815 * config/arc/arc.c (arc_init): Update pic warning.
13816 (irq_range): Update irq range parsing warnings.
13817 (arc_override_options): Update various warning messages.
13818 (arc_handle_aux_attribute): Likewise.
13820 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
13822 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
13824 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
13826 * doc/sourcebuild.texi: Document usage of line number 0 in verify
13827 compiler messages directives.
13829 2018-06-12 Matthew Fortune <mfortune@gmail.com>
13831 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
13832 * config/mips/mips-tables.opt: Regenerate.
13833 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
13835 * doc/invoke.texi: Document -march=i6500.
13837 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
13839 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
13840 (i6400_gpmul): Add cpu_unit.
13841 (i6400_gpdiv): Likewise.
13842 (i6400_msa_add_d): Update reservations.
13843 (i6400_msa_int_add) Likewise.
13844 (i6400_msa_short_logic3) Likewise.
13845 (i6400_msa_short_logic2) Likewise.
13846 (i6400_msa_short_logic) Likewise.
13847 (i6400_msa_move) Likewise.
13848 (i6400_msa_cmp) Likewise.
13849 (i6400_msa_short_float2) Likewise.
13850 (i6400_msa_div_d) Likewise.
13851 (i6400_msa_long_logic1) Likewise.
13852 (i6400_msa_long_logic2) Likewise.
13853 (i6400_msa_mult) Likewise.
13854 (i6400_msa_long_float2) Likewise.
13855 (i6400_msa_long_float4) Likewise.
13856 (i6400_msa_long_float5) Likewise.
13857 (i6400_msa_long_float8) Likewise.
13858 (i6400_fpu_fadd): Include frint type.
13859 (i6400_fpu_store): New define_insn_reservation.
13860 (i6400_fpu_load): Likewise.
13861 (i6400_fpu_move): Likewise.
13862 (i6400_fpu_fcmp): Likewise.
13863 (i6400_fpu_fmadd): Likewise.
13864 (i6400_int_mult): Include imul3nc type and update reservation.
13865 (i6400_int_div): Include idiv3 type and update reservation.
13866 (i6400_int_load): Update to check type not move_type.
13867 (i6400_int_store): Likewise.
13868 (i6400_int_prefetch): Set zero latency.
13870 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
13872 * gcc.c: Document new %@{...} sequence.
13873 (LINK_COMMAND_SPEC): Use it for the -L switches.
13874 (cpp_unique_options): Use it for the -I switches.
13875 (at_file_argbuf): New global variable.
13876 (in_at_file): Likewise.
13877 (alloc_args): Create at_file_argbuf.
13878 (clear_args): Truncate at_file_argbuf.
13879 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
13880 (open_at_file): New function.
13881 (close_at_file): Likewise.
13882 (create_at_file): Delete.
13883 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
13886 (validate_switches_from_spec): Deal with %@{...} sequence.
13887 (validate_switches): Likewise.
13888 (driver::finalize): Call clear_args.
13890 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
13892 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
13894 2018-06-11 Martin Sebor <msebor@redhat.com>
13896 * doc/invoke.texi (-Wall): List -Wc++17-compat.
13897 (Wno-class-memaccess): Add @opindex.
13898 (Wno-templates, Wno-multiple-inheritance): Same.
13899 (Wno-virtual-inheritance, Wno-namespaces): Same.
13900 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
13901 (Wno-format-overflow, Wno-format-truncation): Same.
13902 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
13903 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
13904 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
13905 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
13906 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
13907 (Wno-misspelled-isr): Same.
13909 2018-06-11 Martin Sebor <msebor@redhat.com>
13911 * PR tree-optimization/86083
13912 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
13914 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
13916 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
13918 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
13921 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
13922 on the correct operand.
13923 (*movdi_internal64): Ditto.
13925 2018-06-11 Martin Liska <mliska@suse.cz>
13927 PR tree-optimization/86089
13928 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
13930 2018-06-11 Julia Koval <julia.koval@intel.com>
13932 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
13933 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
13934 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
13936 2018-06-11 Olivier Hainque <hainque@adacore.com>
13938 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
13939 for Ada with strict dwarf2.
13941 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
13944 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
13947 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
13949 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
13951 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
13953 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
13956 2018-06-08 Martin Liska <mliska@suse.cz>
13958 * tree-cfg.h (debug_function): Fix argument type to match
13961 2018-06-08 Martin Liska <mliska@suse.cz>
13963 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
13964 Remove usage of MPX-related (and removed) fields.
13965 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
13967 2018-06-08 David Malcolm <dmalcolm@redhat.com>
13969 * cfg.c (debug): Use TDF_NONE rather than 0.
13970 * cfghooks.c (debug): Likewise.
13971 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
13972 (struct dump_option_value_info): Convert to...
13973 (struct kv_pair): ...this template type.
13974 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
13976 (optinfo_verbosity_options): Likewise.
13977 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
13979 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
13980 than int for "optgroup_flags" param.
13981 (dump_generic_expr_loc): Use dump_flags_t rather than int for
13983 (dump_dec): Likewise.
13984 (dump_finish): Use TDF_NONE rather than 0.
13985 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
13986 rather than int for "optgroup_flags" param. Use TDF_NONE rather
13987 than 0. Update for change to option_ptr.
13988 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
13989 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
13990 0. Update for changes to optinfo_verbosity_options and
13992 (opt_info_switch_p): Convert optgroup_flags from int to
13994 (dump_basic_block): Use dump_flags_t rather than int
13995 for "dump_kind" param.
13996 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
13997 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
13998 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
13999 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
14000 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
14001 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
14002 TDF_NONE): Convert from macros to...
14003 (enum dump_flag): ...this new enum.
14004 (dump_flags_t): Update to use enum.
14005 (operator|, operator&, operator~, operator|=, operator&=):
14006 Implement for dump_flags_t.
14007 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
14008 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
14009 Convert from macros to...
14010 (enum optgroup_flag): ...this new enum.
14011 (optgroup_flags_t): New typedef.
14012 (operator|, operator|=): Implement for optgroup_flags_t.
14013 (struct dump_file_info): Convert field "alt_flags" to
14014 dump_flags_t. Convert field "optgroup_flags" to
14016 (dump_basic_block): Use dump_flags_t rather than int for param.
14017 (dump_generic_expr_loc): Likewise.
14018 (dump_dec): Likewise.
14019 (dump_register): Convert param "optgroup_flags" to
14021 (opt_info_enable_passes): Likewise.
14022 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
14024 * gimple-pretty-print.c (debug): Likewise.
14025 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
14026 (merged_store_group::apply_stores): Likewise.
14027 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
14028 * gimple.c (verify_gimple_pp): Likewise.
14029 * graphite-poly.c (print_pbb_body): Likewise.
14030 * passes.c (pass_manager::register_one_dump_file): Convert
14031 local "optgroup_flags" to optgroup_flags_t.
14032 * print-tree.c (print_node): Use TDF_NONE rather than 0.
14034 (debug_body): Likewise.
14035 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
14036 to optgroup_flags_t.
14037 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
14039 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
14040 (convert_mult_to_fma): Likewise.
14041 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
14042 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
14043 * tree-vect-data-refs.c (dump_lower_bound): Convert param
14044 "dump_kind" to dump_flags_t.
14046 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
14048 * config/rs6000/rs6000.c (min, max): Delete.
14050 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
14052 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
14053 -mabi=spe and -mabi=no-spe.
14055 2018-06-08 Martin Liska <mliska@suse.cz>
14057 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
14058 where we expect an existing summary.
14060 2018-06-08 Martin Liska <mliska@suse.cz>
14062 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
14063 * ipa-inline.h (estimate_edge_growth): Likewise.
14065 2018-06-08 Martin Liska <mliska@suse.cz>
14067 * cgraph.c (function_version_hasher::hash): Use
14068 cgraph_node::get_uid ().
14069 (function_version_hasher::equal):
14070 * cgraph.h (cgraph_node::get_uid): New method.
14071 * ipa-inline.c (update_caller_keys): Use
14072 cgraph_node::get_uid ().
14073 (update_callee_keys): Likewise.
14074 * ipa-utils.c (searchc): Likewise.
14075 (ipa_reduced_postorder): Likewise.
14076 * lto-cgraph.c (input_node): Likewise.
14077 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
14078 * symbol-summary.h (symtab_insertion): Likewise.
14079 (symtab_removal): Likewise.
14080 (symtab_duplication): Likewise.
14081 * tree-pretty-print.c (dump_function_header): Likewise.
14082 * tree-sra.c (convert_callers_for_node): Likewise.
14084 2018-06-08 Martin Liska <mliska@suse.cz>
14086 * cgraph.c (symbol_table::create_edge): Always assign a new
14088 (symbol_table::free_edge): Do not recycle numbers.
14089 * cgraph.h (cgraph_edge::get): New method.
14090 * symbol-summary.h (symtab_removal): Use it.
14091 (symtab_duplication): Likewise.
14092 (call_summary::hashable_uid): Remove.
14094 2018-06-08 Martin Liska <mliska@suse.cz>
14096 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
14097 (initialize_growth_caches): Remove.
14098 (free_growth_caches): Likewise.
14099 (do_estimate_edge_time): Use edge_growth_cache.
14100 (do_estimate_edge_size): Likewise.
14101 (do_estimate_edge_hints): Likewise.
14102 * ipa-inline.c (reset_edge_caches): Likewise.
14103 (recursive_inlining): Likewise.
14104 (inline_small_functions): Likewise.
14105 * ipa-inline.h (initialize_growth_caches): Remove.
14106 (estimate_edge_size): Likewise.
14107 (estimate_edge_time): Likewise.
14108 (estimate_edge_hints): Likewise.
14109 (reset_edge_growth_cache): Likewise.
14110 * symbol-summary.h (call_summary::remove): New method.
14112 2018-06-08 Martin Liska <mliska@suse.cz>
14114 * ipa-cp.c (class edge_clone_summary): New summary.
14115 (grow_edge_clone_vectors): Remove.
14116 (ipcp_edge_duplication_hook): Remove.
14117 (class edge_clone_summary_t): New call_summary class.
14118 (ipcp_edge_removal_hook): Remove.
14119 (edge_clone_summary_t::duplicate): New function.
14120 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
14121 (create_specialized_node): Likewise.
14122 (ipcp_driver): Initialize edge_clone_summaries and do not
14125 2018-06-08 Martin Liska <mliska@suse.cz>
14127 * symbol-summary.h (get): New function.
14128 (call_summary::m_initialize_when_cloning): New class member.
14130 2018-06-08 Martin Liska <mliska@suse.cz>
14132 * cgraph.c (cgraph_node::remove): Do not recycle uid.
14133 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
14134 (symbol_table::allocate_cgraph_symbol): Do not set uid.
14135 * passes.c (uid_hash_t): Record removed_nodes by their uids.
14136 (remove_cgraph_node_from_order): Use the removed_nodes set.
14137 (do_per_function_toporder): Likwise.
14138 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
14139 instead of summary_uid.
14140 (symtab_removal): Likewise.
14141 (symtab_duplication): Likewise.
14143 2018-06-08 Martin Liska <mliska@suse.cz>
14145 * ipa-cp.c (ipcp_store_bits_results): Use
14146 ipcp_transformation_sum.
14147 (ipcp_store_vr_results): Likewise.
14148 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
14150 (ipcp_transformation_initialize): ... this.
14151 (ipa_set_node_agg_value_chain):
14152 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
14153 (write_ipcp_transformation_info): Likewise.
14154 (read_ipcp_transformation_info): Likewise.
14155 (ipcp_update_bits): Likewise.
14156 (ipcp_update_vr): Likewise.
14157 (ipcp_transform_function): Likewise.
14158 * ipa-prop.h: Rename ipcp_transformation_summary to
14159 ipcp_transformation.
14160 (class ipcp_transformation_t): New function summary.
14161 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
14162 (ipa_get_agg_replacements_for_node): Likewise.
14164 2018-06-08 Martin Liska <mliska@suse.cz>
14166 * ipa-pure-const.c (struct funct_state_d): Do it class instead
14168 (class funct_state_summary_t): New function_summary class.
14169 (has_function_state): Remove.
14170 (get_function_state): Likewise.
14171 (set_function_state): Likewise.
14172 (add_new_function): Likewise.
14173 (funct_state_summary_t::insert): New function.
14174 (duplicate_node_data): Remove.
14175 (remove_node_data): Remove.
14176 (funct_state_summary_t::duplicate): New function.
14177 (register_hooks): Create new funct_state_summaries.
14178 (pure_const_generate_summary): Use it.
14179 (pure_const_write_summary): Likewise.
14180 (pure_const_read_summary): Likewise.
14181 (propagate_pure_const): Likewise.
14182 (propagate_nothrow): Likewise.
14183 (dump_malloc_lattice): Likewise.
14184 (propagate_malloc): Likewise.
14185 (execute): Do not register hooks, just remove summary
14187 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
14190 2018-06-08 Martin Liska <mliska@suse.cz>
14192 * ipa-reference.c (remove_node_data): Remove.
14193 (duplicate_node_data): Likewise.
14194 (class ipa_ref_var_info_summary_t): New class.
14195 (class ipa_ref_opt_summary_t): Likewise.
14196 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
14197 (get_reference_optimization_summary): Use
14198 ipa_ref_opt_sum_summaries.
14199 (set_reference_vars_info): Remove.
14200 (set_reference_optimization_summary): Likewise.
14201 (ipa_init): Create summaries.
14202 (init_function_info): Use function summary.
14203 (ipa_ref_opt_summary_t::duplicate): New function.
14204 (ipa_ref_opt_summary_t::remove): New function.
14205 (get_read_write_all_from_node): Fix GNU coding style.
14206 (propagate): Use function summary.
14207 (write_node_summary_p): Fix GNU coding style.
14208 (stream_out_bitmap): Likewise.
14209 (ipa_reference_read_optimization_summary): Use function summary.
14210 (ipa_reference_c_finalize): Do not release hooks.
14212 2018-06-08 Martin Liska <mliska@suse.cz>
14214 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
14215 (analyze_function_body): Extract multiple calls of get_create.
14216 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
14217 * ipa-inline.c (recursive_inlining): Use ::get method.
14218 * ipa-inline.h (estimate_edge_growth): Likewise.
14220 2018-06-08 Martin Liska <mliska@suse.cz>
14222 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
14224 (hsa_function_summary::hsa_function_summary): Use the new enum
14226 (hsa_gpu_implementation_p): Use hsa_summaries::get.
14227 * hsa-gen.c (hsa_get_host_function): Likewise.
14228 (get_brig_function_name): Likewise.
14229 * ipa-hsa.c (process_hsa_functions): Likewise.
14230 (ipa_hsa_write_summary): Likewise.
14231 * symbol-summary.h (symtab_duplication): Use ::get function/
14232 (get): New function.
14234 2018-06-08 Martin Liska <mliska@suse.cz>
14236 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
14238 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
14239 (hsa_register_kernel): Likewise.
14240 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
14241 * hsa-gen.c (hsa_get_host_function): Likewise.
14242 (get_brig_function_name): Likewise.
14243 (generate_hsa): Likewise.
14244 (pass_gen_hsail::execute): Likewise.
14245 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
14246 (devirtualization_time_bonus): Likewise.
14247 (ipcp_propagate_stage): Likewise.
14248 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
14249 (edge_set_predicate): Likewise.
14250 (evaluate_conditions_for_known_args): Likewise.
14251 (evaluate_properties_for_edge): Likewise.
14252 (ipa_fn_summary::reset): Likewise.
14253 (ipa_fn_summary_t::duplicate): Likewise.
14254 (dump_ipa_call_summary): Likewise.
14255 (ipa_dump_fn_summary): Likewise.
14256 (analyze_function_body): Likewise.
14257 (compute_fn_summary): Likewise.
14258 (estimate_edge_devirt_benefit): Likewise.
14259 (estimate_edge_size_and_time): Likewise.
14260 (estimate_calls_size_and_time): Likewise.
14261 (estimate_node_size_and_time): Likewise.
14262 (inline_update_callee_summaries): Likewise.
14263 (remap_edge_change_prob): Likewise.
14264 (remap_edge_summaries): Likewise.
14265 (ipa_merge_fn_summary_after_inlining): Likewise.
14266 (ipa_update_overall_fn_summary): Likewise.
14267 (read_ipa_call_summary): Likewise.
14268 (inline_read_section): Likewise.
14269 (write_ipa_call_summary): Likewise.
14270 (ipa_fn_summary_write): Likewise.
14271 (ipa_free_fn_summary): Likewise.
14272 * ipa-hsa.c (process_hsa_functions): Likewise.
14273 (ipa_hsa_write_summary): Likewise.
14274 (ipa_hsa_read_section): Likewise.
14275 * ipa-icf.c (sem_function::merge): Likewise.
14276 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
14277 (do_estimate_edge_time): Likewise.
14278 (estimate_size_after_inlining): Likewise.
14279 (estimate_growth): Likewise.
14280 (growth_likely_positive): Likewise.
14281 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
14282 (inline_call): Likewise.
14283 * ipa-inline.c (caller_growth_limits): Likewise.
14284 (can_inline_edge_p): Likewise.
14285 (can_inline_edge_by_limits_p): Likewise.
14286 (compute_uninlined_call_time): Likewise.
14287 (compute_inlined_call_time): Likewise.
14288 (want_inline_small_function_p): Likewise.
14289 (edge_badness): Likewise.
14290 (update_caller_keys): Likewise.
14291 (update_callee_keys): Likewise.
14292 (recursive_inlining): Likewise.
14293 (inline_small_functions): Likewise.
14294 (inline_to_all_callers_1): Likewise.
14295 (dump_overall_stats): Likewise.
14296 (early_inline_small_functions): Likewise.
14297 (early_inliner): Likewise.
14298 * ipa-inline.h (estimate_edge_growth): Likewise.
14299 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
14300 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
14301 * ipa-prop.h (IPA_NODE_REF): Likewise.
14302 (IPA_EDGE_REF): Likewise.
14303 * ipa-pure-const.c (malloc_candidate_p): Likewise.
14304 (propagate_malloc): Likewise.
14305 * ipa-split.c (execute_split_functions): Likewise.
14306 * symbol-summary.h: Rename get to get_create.
14308 (get_create): Likewise.
14309 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
14311 2018-06-08 Martin Liska <mliska@suse.cz>
14313 * symbol-summary.h (release): Move definition out of class
14315 (symtab_removal): Likewise.
14316 (symtab_duplication): Likewise.
14318 2018-06-08 Martin Liska <mliska@suse.cz>
14320 * symbol-summary.h (function_summary): Move constructor
14321 implementation out of class declaration.
14322 (release): Likewise.
14323 (symtab_insertion): Likewise.
14324 (symtab_removal): Likewise.
14325 (symtab_duplication): Likewise.
14328 2018-06-08 Martin Liska <mliska@suse.cz>
14330 * Makefile.in: Remove support for MPX (macros, related functions,
14331 fields in cgraph_node, ...).
14332 * builtin-types.def (BT_BND): Likewise.
14333 (BT_FN_BND_CONST_PTR): Likewise.
14334 (BT_FN_CONST_PTR_BND): Likewise.
14335 (BT_FN_VOID_PTR_BND): Likewise.
14336 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
14337 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
14338 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
14339 (expand_builtin_mempcpy_with_bounds): Likewise.
14340 (expand_builtin_memset_with_bounds): Likewise.
14341 (expand_builtin_memset_args): Likewise.
14342 (std_expand_builtin_va_start): Likewise.
14343 (expand_builtin): Likewise.
14344 (expand_builtin_with_bounds): Likewise.
14345 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
14346 (DEF_LIB_BUILTIN_CHKP): Likewise.
14347 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
14348 (DEF_CHKP_BUILTIN): Likewise.
14349 (BUILT_IN_MEMCPY): Likewise.
14350 (BUILT_IN_MEMMOVE): Likewise.
14351 (BUILT_IN_MEMPCPY): Likewise.
14352 (BUILT_IN_MEMSET): Likewise.
14353 (BUILT_IN_STPCPY): Likewise.
14354 (BUILT_IN_STRCAT): Likewise.
14355 (BUILT_IN_STRCHR): Likewise.
14356 (BUILT_IN_STRCPY): Likewise.
14357 (BUILT_IN_STRLEN): Likewise.
14358 (BUILT_IN_MEMCPY_CHK): Likewise.
14359 (BUILT_IN_MEMMOVE_CHK): Likewise.
14360 (BUILT_IN_MEMPCPY_CHK): Likewise.
14361 (BUILT_IN_MEMSET_CHK): Likewise.
14362 (BUILT_IN_STPCPY_CHK): Likewise.
14363 (BUILT_IN_STRCAT_CHK): Likewise.
14364 (BUILT_IN_STRCPY_CHK): Likewise.
14365 * calls.c (store_bounds): Likewise.
14366 (emit_call_1): Likewise.
14367 (special_function_p): Likewise.
14368 (maybe_warn_nonstring_arg): Likewise.
14369 (initialize_argument_information): Likewise.
14370 (finalize_must_preallocate): Likewise.
14371 (compute_argument_addresses): Likewise.
14372 (expand_call): Likewise.
14373 * cfgexpand.c (expand_call_stmt): Likewise.
14374 (expand_return): Likewise.
14375 (expand_gimple_stmt_1): Likewise.
14376 (pass_expand::execute): Likewise.
14377 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
14378 (cgraph_node::remove): Likewise.
14379 (cgraph_node::dump): Likewise.
14380 (cgraph_node::verify_node): Likewise.
14381 * cgraph.h (chkp_function_instrumented_p): Likewise.
14382 (symtab_node::get_alias_target): Likewise.
14383 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
14384 (cgraph_local_p): Likewise.
14385 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
14386 (cgraph_edge::rebuild_references): Likewise.
14387 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
14388 (walk_polymorphic_call_targets): Likewise.
14389 (cgraph_node::expand_thunk): Likewise.
14390 (symbol_table::output_weakrefs): Likewise.
14391 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
14392 (ix86_handle_option): Likewise.
14393 * config/i386/constraints.md: Likewise.
14394 * config/i386/i386-builtin-types.def (BND): Likewise.
14398 * config/i386/i386-builtin.def (BDESC_END): Likewise.
14399 (BDESC_FIRST): Likewise.
14401 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
14402 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
14403 * config/i386/i386.c (enum reg_class): Likewise.
14404 (ix86_target_string): Likewise.
14405 (ix86_option_override_internal): Likewise.
14406 (ix86_conditional_register_usage): Likewise.
14407 (ix86_valid_target_attribute_inner_p): Likewise.
14408 (ix86_set_indirect_branch_type): Likewise.
14409 (ix86_set_current_function): Likewise.
14410 (ix86_function_arg_regno_p): Likewise.
14411 (init_cumulative_args): Likewise.
14412 (ix86_function_arg_advance): Likewise.
14413 (ix86_function_arg): Likewise.
14414 (ix86_pass_by_reference): Likewise.
14415 (ix86_function_value_regno_p): Likewise.
14416 (ix86_function_value_1): Likewise.
14417 (ix86_function_value_bounds): Likewise.
14418 (ix86_return_in_memory): Likewise.
14419 (ix86_setup_incoming_vararg_bounds): Likewise.
14420 (ix86_va_start): Likewise.
14421 (indirect_thunk_need_prefix): Likewise.
14422 (print_reg): Likewise.
14423 (ix86_print_operand): Likewise.
14424 (ix86_expand_call): Likewise.
14425 (ix86_output_function_return): Likewise.
14426 (reg_encoded_number): Likewise.
14427 (BDESC_VERIFYS): Likewise.
14428 (ix86_init_mpx_builtins): Likewise.
14429 (ix86_init_builtins): Likewise.
14430 (ix86_emit_cmove): Likewise.
14431 (ix86_emit_move_max): Likewise.
14432 (ix86_expand_builtin): Likewise.
14433 (ix86_builtin_mpx_function): Likewise.
14434 (ix86_get_arg_address_for_bt): Likewise.
14435 (ix86_load_bounds): Likewise.
14436 (ix86_store_bounds): Likewise.
14437 (ix86_load_returned_bounds): Likewise.
14438 (ix86_store_returned_bounds): Likewise.
14439 (ix86_class_likely_spilled_p): Likewise.
14440 (ix86_hard_regno_mode_ok): Likewise.
14441 (x86_order_regs_for_local_alloc): Likewise.
14442 (ix86_mitigate_rop): Likewise.
14443 (ix86_bnd_prefixed_insn_p): Likewise.
14444 (ix86_mpx_bound_mode): Likewise.
14445 (ix86_make_bounds_constant): Likewise.
14446 (ix86_initialize_bounds): Likewise.
14447 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
14448 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
14449 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
14450 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
14451 (TARGET_CHKP_BOUND_MODE): Likewise.
14452 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
14453 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
14454 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
14455 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
14456 * config/i386/i386.h (TARGET_MPX): Likewise.
14457 (TARGET_MPX_P): Likewise.
14458 (VALID_BND_REG_MODE): Likewise.
14459 (FIRST_BND_REG): Likewise.
14460 (LAST_BND_REG): Likewise.
14461 (enum reg_class): Likewise.
14462 (BND_REG_P): Likewise.
14463 (BND_REGNO_P): Likewise.
14464 (BNDmode): Likewise.
14465 (ADJUST_INSN_LENGTH): Likewise.
14466 * config/i386/i386.md: Likewise.
14467 * config/i386/i386.opt: Likewise.
14468 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
14469 (defined): Likewise.
14470 (LINK_MPX): Likewise.
14471 (MPX_SPEC): Likewise.
14472 (LIBMPX_SPEC): Likewise.
14473 (LIBMPXWRAPPERS_SPEC): Likewise.
14474 (CHKP_SPEC): Likewise.
14475 * config/i386/predicates.md: Likewise.
14476 * dbxout.c (dbxout_type): Likewise.
14477 * doc/extend.texi: Likewise.
14478 * doc/invoke.texi: Likewise.
14479 * doc/md.texi: Likewise.
14480 * doc/tm.texi: Likewise.
14481 * doc/tm.texi.in: Likewise.
14482 * dwarf2out.c (is_base_type): Likewise.
14483 (gen_formal_types_die): Likewise.
14484 (gen_subprogram_die): Likewise.
14485 (gen_type_die_with_usage): Likewise.
14486 (gen_decl_die): Likewise.
14487 (dwarf2out_late_global_decl): Likewise.
14488 * expr.c (expand_assignment): Likewise.
14489 (emit_storent_insn): Likewise.
14490 (store_expr_with_bounds): Likewise.
14491 (store_expr): Likewise.
14492 (expand_expr_real_1): Likewise.
14493 * expr.h (store_expr_with_bounds): Likewise.
14494 * function.c (use_register_for_decl): Likewise.
14495 (struct bounds_parm_data): Likewise.
14496 (assign_parms_augmented_arg_list): Likewise.
14497 (assign_parm_find_entry_rtl): Likewise.
14498 (assign_parm_is_stack_parm): Likewise.
14499 (assign_parm_load_bounds): Likewise.
14500 (assign_bounds): Likewise.
14501 (assign_parms): Likewise.
14502 (expand_function_start): Likewise.
14503 * gcc.c (CHKP_SPEC): Likewise.
14504 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
14505 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
14506 (wrestrict_dom_walker::check_call): Likewise.
14507 * gimple.c (gimple_build_call_from_tree): Likewise.
14508 * gimple.h (enum gf_mask): Likewise.
14509 (gimple_call_with_bounds_p): Likewise.
14510 (gimple_call_set_with_bounds): Likewise.
14511 * gimplify.c (gimplify_init_constructor): Likewise.
14512 * ipa-cp.c (initialize_node_lattices): Likewise.
14513 (propagate_constants_across_call): Likewise.
14514 (find_more_scalar_values_for_callers_subset): Likewise.
14515 * ipa-hsa.c (process_hsa_functions): Likewise.
14516 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
14517 * ipa-icf.c (sem_function::merge): Likewise.
14518 * ipa-inline.c (early_inliner): Likewise.
14519 * ipa-pure-const.c (warn_function_noreturn): Likewise.
14520 (warn_function_cold): Likewise.
14521 (propagate_pure_const): Likewise.
14522 * ipa-ref.h (enum GTY): Likewise.
14523 * ipa-split.c (find_retbnd): Likewise.
14524 (consider_split): Likewise.
14525 (split_function): Likewise.
14526 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
14527 * ipa.c (walk_polymorphic_call_targets): Likewise.
14528 (symbol_table::remove_unreachable_nodes): Likewise.
14529 (process_references): Likewise.
14530 (cgraph_build_static_cdtor_1): Likewise.
14531 * lto-cgraph.c (lto_output_node): Likewise.
14532 (output_refs): Likewise.
14533 (compute_ltrans_boundary): Likewise.
14534 (input_overwrite_node): Likewise.
14535 (input_node): Likewise.
14536 (input_cgraph_1): Likewise.
14537 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
14538 * passes.c (pass_manager::execute_early_local_passes): Likewise.
14539 (class pass_chkp_instrumentation_passes): Likewise.
14540 (make_pass_chkp_instrumentation_passes): Likewise.
14541 * passes.def: Likewise.
14542 * rtl.h (struct GTY): Likewise.
14543 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
14544 * stor-layout.c (layout_type): Likewise.
14545 * symtab.c: Likewise.
14546 * target.def: Likewise.
14547 * targhooks.c (default_chkp_bound_type): Likewise.
14548 (default_chkp_bound_mode): Likewise.
14549 (default_builtin_chkp_function): Likewise.
14550 (default_chkp_function_value_bounds): Likewise.
14551 (default_chkp_make_bounds_constant): Likewise.
14552 (default_chkp_initialize_bounds): Likewise.
14553 * targhooks.h (default_chkp_bound_type): Likewise.
14554 (default_chkp_bound_mode): Likewise.
14555 (default_builtin_chkp_function): Likewise.
14556 (default_chkp_function_value_bounds): Likewise.
14557 (default_chkp_make_bounds_constant): Likewise.
14558 (default_chkp_initialize_bounds): Likewise.
14559 * toplev.c (compile_file): Likewise.
14560 (process_options): Likewise.
14561 * tree-core.h (DEF_BUILTIN): Likewise.
14562 (DEF_BUILTIN_CHKP): Likewise.
14563 * tree-inline.c (declare_return_variable): Likewise.
14564 (remap_gimple_stmt): Likewise.
14565 (copy_bb): Likewise.
14566 (initialize_inlined_parameters): Likewise.
14567 (expand_call_inline): Likewise.
14568 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
14569 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
14570 (make_pass_ipa_chkp_produce_thunks): Likewise.
14571 (make_pass_chkp): Likewise.
14572 (make_pass_chkp_opt): Likewise.
14573 (make_pass_chkp_instrumentation_passes): Likewise.
14574 * tree-pretty-print.c (dump_generic_node): Likewise.
14575 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
14576 * tree-ssa-dce.c (propagate_necessity): Likewise.
14577 (eliminate_unnecessary_stmts): Likewise.
14578 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
14579 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
14580 * tree-ssa-sccvn.h: Likewise.
14581 * tree-ssa-strlen.c (get_string_length): Likewise.
14582 (valid_builtin_call): Likewise.
14583 (adjust_last_stmt): Likewise.
14584 (handle_builtin_strchr): Likewise.
14585 (handle_builtin_strcpy): Likewise.
14586 (handle_builtin_stxncpy): Likewise.
14587 (handle_builtin_memcpy): Likewise.
14588 (handle_builtin_strcat): Likewise.
14589 (strlen_check_and_optimize_stmt): Likewise.
14590 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
14591 * tree-streamer-in.c: Likewise.
14592 * tree-streamer.c (record_common_node): Likewise.
14593 * tree.c (tree_code_size): Likewise.
14594 (wide_int_to_tree_1): Likewise.
14595 (type_contains_placeholder_1): Likewise.
14596 (build_common_tree_nodes): Likewise.
14597 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
14598 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
14599 (POINTER_BOUNDS_P): Likewise.
14600 (BOUNDED_TYPE_P): Likewise.
14601 (BOUNDED_P): Likewise.
14602 (CALL_WITH_BOUNDS_P): Likewise.
14603 (pointer_bounds_type_node): Likewise.
14604 * value-prof.c (gimple_ic): Likewise.
14605 * var-tracking.c (vt_add_function_parameters): Likewise.
14606 * varasm.c (make_decl_rtl): Likewise.
14607 (assemble_start_function): Likewise.
14608 (output_constant): Likewise.
14609 (maybe_assemble_visibility): Likewise.
14610 * varpool.c (ctor_for_folding): Likewise.
14611 * chkp-builtins.def: Remove.
14612 * ipa-chkp.c: Remove.
14613 * ipa-chkp.h: Remove.
14614 * rtl-chkp.c: Remove.
14615 * rtl-chkp.h: Remove.
14616 * tree-chkp-opt.c: Remove.
14617 * tree-chkp.c: Remove.
14618 * tree-chkp.h: Remove.
14620 2018-06-07 Carl Love <cel@us.ibm.com>
14622 * config/rs6000/vsx.md (vextract_fp_from_shorth,
14623 vextract_fp_from_shortl): Add BE support.
14625 2018-06-07 Paul Koning <ni1d@arrl.net>
14627 * compare-elim.c (try_merge_compare): Don't merge compare if
14628 address contains a side effect.
14629 (try_eliminate_compare): Likewise.
14631 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
14633 * config.gcc: Support "tremont".
14634 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
14635 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14637 * config/i386/i386.c (m_TREMONT): Define.
14638 (processor_target_table): Add "tremont".
14639 (PTA_TREMONT): Define.
14640 (ix86_lea_outperforms): Add TARGET_TREMONT.
14641 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
14642 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
14643 and M_INTEL_GOLDMONT_PLUS.
14644 (fold_builtin_cpu): Add "tremont".
14645 (ix86_add_stmt_cost): Add TARGET_TREMONT.
14646 (ix86_option_override_internal): Add "tremont".
14647 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
14648 (processor_type): Add PROCESSOR_TREMONT.
14649 * config/i386/x86-tune.def: Add m_TREMONT.
14650 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
14652 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
14654 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
14655 symbol defined for msp430i* devices to be lower case.
14657 2018-06-07 Richard Biener <rguenther@suse.de>
14659 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
14660 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
14661 Properly wrap signed arithmetic if overflow wraps.
14663 2018-06-07 Jakub Jelinek <jakub@redhat.com>
14665 PR tree-optimization/69615
14666 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
14667 of a cast from a same precision integral SSA_NAME in a bb dominated
14668 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
14669 cast to utype if rhs2 has already a compatible type.
14671 2018-06-07 Richard Biener <rguenther@suse.de>
14673 PR tree-optimization/85935
14674 * graphite-scop-detection.c (find_params_in_bb): Analyze
14675 condition operands with respect to the correct loop. Assert
14676 the analysis doesn't fail.
14678 2018-06-04 Carl Love <cel@us.ibm.com>
14680 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
14681 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
14682 as it is slightly cheaper.
14683 (first_match_or_eos_index_<mode>):
14684 Calculate index using natural element order.
14685 (first_match_index_<mode>):
14686 Calculate index using natural element order.
14687 (first_match_or_eos_index_<mode>):
14688 Calculate index using natural order.
14689 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
14690 for BE and LE modes.
14691 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
14692 P9V_BUILTIN_VCLZLSBB_V16QI.
14693 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
14696 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
14698 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
14699 indentation and line wrap for many prototypes. Add missing
14700 @smallexample directives around block of prototypes for vec_xl and
14703 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
14705 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
14706 track if we pass or return IEEE 128-bit floating point.
14707 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
14708 C++ mangling that is compatible with GCC 8.1.
14709 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
14710 (init_cumulative_args): Note if we pass or return IEEE 128-bit
14711 floating point types.
14712 (rs6000_function_arg_advance_1): Likewise.
14713 (rs6000_mangle_type): Optionally generate mangled names that match
14714 what GCC 8.1 generated for IEEE 128-bit floating point types.
14715 (rs6000_globalize_decl_name): If we have an external function that
14716 passes or returns IEEE 128-bit types, generate a weak reference
14717 from the mangled name used in GCC 8.1 to the current mangled
14719 (rs6000_init_builtins): Make __ibm128 use the long double type if
14720 long double is IBM extended double. Make __float128 use the long
14721 double type if long double is IEEE 128-bit.
14724 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
14725 macro for __ibm128 built-in functions.
14726 (PACK_IF): Add __ibm128 pack/unpack functions.
14727 (UNPACK_IF): Likewise.
14728 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
14729 enable long double built-in functions if long double is IEEE
14730 128-bit floating point.
14731 (rs6000_invalid_builtin): Update long double built-in function
14733 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
14734 functions, adjust the built-in function to use the long double
14735 built-in function if __ibm128 and long double are the same type.
14736 * doc/extend.texi (PowerPC builtins): Update documention for
14737 __builtin_{,un}pack_longdouble. Add documentation for
14738 __builtin_{,un}pack_ibm128.
14740 2018-06-06 Jim Wilson <jimw@sifive.com>
14742 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
14743 (struct machine_function): New field interrupt_mode.
14744 (riscv_handle_type_attribute): New function. Add forward declaration.
14745 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
14746 (riscv_expand_epilogue): Check interrupt_mode field.
14747 (riscv_set_current_function): Check interrupt attribute args and
14748 set interrupt_mode field.
14749 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
14750 (riscv_sret, riscv_uret): New.
14751 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
14752 new arguments to interrupt attribute.
14754 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
14757 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
14758 Don't handle -mcpu=power8 if -mpower9-vector is also used.
14760 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
14762 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
14763 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
14764 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
14765 several redundant entries.
14767 2018-06-06 David Malcolm <dmalcolm@redhat.com>
14769 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
14770 type from "rtx" to "rtx_insn *".
14771 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
14772 for local "call_insn", removing cast.
14773 (ix86_expand_call): Likewise, introducing a "call_insn" local.
14775 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
14777 PR tree-optimization/86066
14778 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
14779 for BIT_INSERT_EXPR stores.
14781 2018-06-06 Richard Biener <rguenther@suse.de>
14783 PR tree-optimization/86062
14784 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
14785 component refs ontop
14786 of to be offsetted base.
14788 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
14790 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
14791 to be static and remove check on interrupt attribute name.
14793 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
14795 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
14796 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
14798 2018-06-05 Steve Ellcey <sellcey@cavium.com>
14801 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
14803 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
14804 Remove second argument, change how error is called.
14805 (aarch64_layout_arg): Remove second argument from
14806 aarch64_err_no_fpadvsimd call.
14807 (aarch64_init_cumulative_args): Ditto.
14808 (aarch64_gimplify_va_arg_expr): Ditto.
14809 * config/aarch64/aarch64.md (mov<mode>): Ditto.
14811 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
14813 * config/i386/i386.md (simple_return_indirect_internal): New expander.
14814 (*simple_return_indirect_internal<mode>): Rename from
14815 simple_return_indirect_internal. Use W mode iterator.
14816 (rstorssp): New expander.
14817 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
14818 (clrssbsy): New expander.
14819 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
14821 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
14823 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
14825 (cmse_check_pointed_object): Likewise.
14827 2018-06-05 Martin Liska <mliska@suse.cz>
14829 PR gcov-profile/47618
14830 * doc/invoke.texi: Document how -fprofile-dir format
14833 2018-06-05 Richard Biener <rguenther@suse.de>
14835 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
14836 removal pretend DOM info isn't available so we do not update
14837 it and only remove edges, not dominated blocks. Actually free
14838 DOM info in case we removed something. Remove unreachable blocks.
14839 (mfb_keep_latches): Work with either DOM info or marked backedges.
14840 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
14841 first. Mark backedges if DOM info isn't available.
14842 (Re-)compute DOM info after cleanup_control_flow_pre.
14844 2018-06-05 Richard Biener <rguenther@suse.de>
14846 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
14847 (locus_discrim_hasher::hash): Adjust.
14848 (locus_discrim_hasher::equal): Likewise.
14849 (next_discriminator_for_locus): Work on line directly.
14850 (same_line_p): Pass in expanded locus1 as well.
14851 (assign_discriminators): Avoid redundant location expansions.
14853 2018-06-05 Richard Biener <rguenther@suse.de>
14855 PR tree-optimization/86046
14856 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
14857 if required after clearing TREE_ADDRESSABLE.
14859 2018-06-05 Richard Biener <rguenther@suse.de>
14861 PR tree-optimization/86047
14862 * tree-ssa-loop.c (for_each_index): Glob handling of all
14863 decls and constants and really handle all of them.
14865 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14868 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
14869 qualifier_void_pointer and qualifier_const_void_pointer.
14870 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
14871 (arm_init_builtins): Handle the above.
14872 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
14873 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
14874 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
14877 2018-06-05 Martin Liska <mliska@suse.cz>
14879 * auto-profile.c (read_autofdo_file): Do not use
14880 gcov_ctr_summary struct.
14881 (afdo_callsite_hot_enough_for_early_inline): Likewise.
14882 * coverage.c (struct counts_entry): Likewise.
14883 (read_counts_file): Read just single summary entry.
14884 (get_coverage_counts): Use gcov_summary struct.
14885 * coverage.h (get_coverage_counts): Likewise.
14886 * gcov-dump.c (dump_working_sets): Likewise.
14887 (tag_summary): Dump just single summary.
14888 * gcov-io.c (gcov_write_summary): Write just histogram
14890 (gcov_read_summary): Read just single summary.
14891 (compute_working_sets): Use gcov_summary struct.
14892 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
14893 of GCOV_COUNTERS_SUMMABLE.
14894 (GCOV_COUNTERS_SUMMABLE): Remove.
14895 (GCOV_FIRST_VALUE_COUNTER): Replace with
14896 GCOV_COUNTER_V_INTERVAL.
14897 (struct gcov_ctr_summary): Remove.
14898 (struct gcov_summary): Directly use fields of former
14900 (compute_working_sets): Use gcov_summary struct.
14901 * gcov.c (read_count_file): Do not use ctrs fields.
14902 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
14904 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
14906 * profile.c: Likewise.
14907 * profile.h: Likewise.
14909 2018-06-05 Martin Liska <mliska@suse.cz>
14911 PR gcov-profile/84846
14912 * gcov.c (output_lines): Print working directory only
14913 in intermediate format.
14915 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
14917 * config/s390/s390-builtin-types.def: Add void function type.
14918 * config/s390/s390-builtins.def: Use the function type for the
14921 2018-06-04 Jim Wilson <jimw@sifive.com>
14923 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
14925 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
14926 and maybe_eh_return. Change regno to unsigned int. Use new args to
14927 handle EH_RETURN_DATA_REGNO registers properly.
14928 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
14929 (riscv_expand_epilogue): Update comment. Change argument name and
14930 type. Update code to use new name and type. Pass new args to
14931 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
14933 * config/riscv/riscv.md (NORMAL_RETURN): New.
14934 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
14935 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
14936 (eh_return): Call gen_eh_return_internal and emit barrier.
14937 (eh_return_internal): Call riscv_expand_epilogue.
14939 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
14941 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
14942 bit_insertion field and declare can_be_merged_into method.
14943 (merged_store_group::can_be_merged_into): New method.
14944 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
14945 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
14946 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
14948 2018-06-04 Richard Biener <rguenther@suse.de>
14950 PR tree-optimization/85955
14951 * builtins.c (fold_builtin_sincos): Convert pointers to
14952 destination to appropriate type before dereferencing.
14954 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
14956 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
14958 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
14960 * expr.c (expand_expr_real_1): Force the operand into memory if
14961 its TYPE_MODE is BLKmode and if there is no integer mode for
14962 the number of bits being extracted.
14964 2018-06-04 Jakub Jelinek <jakub@redhat.com>
14968 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
14969 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
14971 2018-06-04 Richard Biener <rguenther@suse.de>
14973 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
14974 (cleanup_tree_cfg_noloop): ... single caller. Do
14975 start_recording_case_labels later.
14977 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
14979 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
14980 to _IMMINTRIN_H_INCLUDED.
14981 * config/i386/pconfigintrin.h: Ditto.
14982 * config/i386/waitpkgintrin.h: Ditto.
14983 * config/i386/immintrin.h: Add includes for sgxintrin.h,
14984 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
14985 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
14986 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
14987 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
14988 waitpkgintrin.h and cldemoteintrin.h.
14990 2018-06-04 Richard Biener <rguenther@suse.de>
14992 PR tree-optimization/86038
14993 * tracer.c (find_best_successor): Check probability for
14994 being initialized, bail out if not.
14996 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
14999 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
15000 of bits to ignore when comparing architectures.
15002 2018-06-04 Jakub Jelinek <jakub@redhat.com>
15004 PR tree-optimization/69615
15005 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
15006 maximum or minimum of the type, try to merge it also as if
15007 range1 is + [-, x - 1] or + [x + 1, -].
15010 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
15012 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
15014 PR tree-optimization/86034
15015 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
15016 the unsigned bitfield type in a bit insertion sequence if it does not
15017 have a larger precision than the bitfield size.
15018 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
15020 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
15022 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
15024 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
15026 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
15027 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
15028 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
15029 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
15031 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
15033 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
15034 Disable -fdelete-null-pointer-checks for ELF toolchain.
15036 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
15037 Kito Cheng <kito.cheng@gmail.com>
15039 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
15040 (nds32le-*-*, nds32be-*-*): Integrate checking process.
15041 (nds32*-*-*): Add glibc and uclibc conditions.
15042 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
15043 (TARGET_EXCEPT_UNWIND_INFO): Define.
15044 * config/nds32/elf.h: New file.
15045 * config/nds32/linux.h: New file.
15046 * config/nds32/nds32-elf.opt: New file.
15047 * config/nds32/nds32-linux.opt: New file.
15048 * config/nds32/nds32-fp-as-gp.c
15049 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
15050 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
15052 (nds32_asm_file_end): Ditto.
15053 (nds32_print_operand): Ditto.
15054 (nds32_insert_attributes): Ditto.
15055 (nds32_init_libfuncs): New function.
15056 (TARGET_HAVE_TLS): Define.
15057 (TARGET_INIT_LIBFUNCS): Define.
15058 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
15060 (TARGET_ELF): Apply different mcmodel setting.
15061 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
15062 been migrated into elf.h and linux.h files.
15063 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
15064 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
15065 (mcmodel): The content has been migrated into nds32-elf.opt and
15066 nds32-linux.opt files.
15067 * config/nds32/t-elf: New file.
15068 * config/nds32/t-linux: New file.
15070 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
15071 Shiva Chen <shiva0217@gmail.com>
15073 * config/nds32/constants.md (unspec_volatile_element): Add
15074 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
15075 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
15077 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
15078 (make_pass_nds32_fp_as_gp): Declare.
15079 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
15081 (nds32_asm_function_end_prologue): Remove unused asm output.
15082 (nds32_asm_function_begin_epilogue): Remove unused asm output.
15083 (nds32_asm_file_start): Output necessary fp_as_gp information.
15084 (nds32_option_override): Adjust register usage.
15085 (nds32_expand_prologue): Consider fp_as_gp situation.
15086 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
15087 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
15090 (simple_return): Ditto.
15091 (omit_fp_begin): Output special directive for fp_as_gp.
15092 (omit_fp_end): Output special directive for fp_as_gp.
15093 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
15094 mforbid-fp-as-gp): New options.
15096 2018-06-01 Mark Wielaard <mark@klomp.org>
15098 * dwarf2out.c (dwarf2out_finish): Remove generation of
15099 DW_AT_loclists_base.
15101 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
15103 * gimple-ssa-store-merging.c: Include gimple-fold.h.
15104 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
15105 (struct merged_store_group): Add bit_insertion field.
15106 (dump_char_array): Use standard hexadecimal format.
15107 (merged_store_group::merged_store_group): Set bit_insertion to false.
15108 (merged_store_group::apply_stores): Use optimal buffer size. Deal
15109 with BIT_INSERT_EXPR stores. Move up code updating the mask and
15110 also print the mask in the dump file.
15111 (pass_store_merging::gate): Minor tweak.
15112 (imm_store_chain_info::coalesce_immediate): Fix wrong association
15113 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
15114 stores with INTEGER_CST stores.
15115 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
15116 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
15117 and use it throughout. Generate bit insertion sequences if need be.
15118 (pass_store_merging::process_store): Remove redundant condition.
15119 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
15121 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
15123 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
15124 the 128-bit floating point types. Fix function comment.
15126 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15128 * config/aarch64/aarch64-simd.md
15129 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
15131 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
15134 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
15136 PR tree-optimization/85989
15137 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
15139 (backprop::intersect_uses): Check it when deciding whether this
15140 is a backedge reference.
15141 (backprop::process_block): Add each phi to m_visited_phis
15142 after visiting it, then clear it at the end.
15144 2018-06-01 Richard Biener <rguenther@suse.de>
15146 * tree-vectorizer.h (vect_dr_stmt): New function.
15147 (vect_get_load_cost): Adjust.
15148 (vect_get_store_cost): Likewise.
15149 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
15150 Use vect_dr_stmt instead of DR_SMTT.
15151 (vect_record_base_alignments): Likewise.
15152 (vect_calculate_target_alignment): Likewise.
15153 (vect_compute_data_ref_alignment): Likewise and make static.
15154 (vect_update_misalignment_for_peel): Likewise.
15155 (vect_verify_datarefs_alignment): Likewise.
15156 (vector_alignment_reachable_p): Likewise.
15157 (vect_get_data_access_cost): Likewise. Pass down
15158 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
15159 (vect_get_peeling_costs_all_drs): Likewise.
15160 (vect_peeling_hash_get_lowest_cost): Likewise.
15161 (vect_enhance_data_refs_alignment): Likewise.
15162 (vect_find_same_alignment_drs): Likewise.
15163 (vect_analyze_data_refs_alignment): Likewise.
15164 (vect_analyze_group_access_1): Likewise.
15165 (vect_analyze_group_access): Likewise.
15166 (vect_analyze_data_ref_access): Likewise.
15167 (vect_analyze_data_ref_accesses): Likewise.
15168 (vect_vfa_segment_size): Likewise.
15169 (vect_small_gap_p): Likewise.
15170 (vectorizable_with_step_bound_p): Likewise.
15171 (vect_prune_runtime_alias_test_list): Likewise.
15172 (vect_analyze_data_refs): Likewise.
15173 (vect_supportable_dr_alignment): Likewise.
15174 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
15175 (vect_gen_prolog_loop_niters): Likewise.
15176 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
15177 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
15179 (vect_recog_mask_conversion_pattern): Likewise.
15180 (vect_try_gather_scatter_pattern): Likewise.
15181 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
15182 to vect_get_store_cost.
15183 (vect_get_store_cost): Get stmt_info instead of DR.
15184 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
15185 (vect_get_load_cost): Get stmt_info instead of DR.
15187 2018-06-01 Richard Biener <rguenther@suse.de>
15189 PR middle-end/86017
15190 * gimple-fold.c (var_decl_component_p): Also allow offsetted
15191 vars wrapped in MEM_REFs.
15193 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
15195 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
15196 Fix subreg tests so that we only return a choice between
15197 GENERAL_REGS and FP_REGS if the original classes included both.
15199 2018-06-01 Richard Biener <rguenther@suse.de>
15202 * tree-ssa-structalias.c (get_function_part_constraint):
15203 Handle NULL fi->decl.
15204 (find_func_aliases_for_call): Properly handle indirect
15205 fi from direct call.
15206 (find_func_clobbers): Likewise.
15207 (ipa_pta_execute): Likewise.
15208 (create_variable_info_for): For functions that are ifunc_resolver
15209 resolve to a varinfo that contains the result of the resolver call.
15210 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
15213 2018-05-31 Michael Collison <michael.collison@arm.com>
15215 * config/aarch64/aarch64.md:
15216 (*fix_to_zero_extenddfdi2): New pattern.
15217 * gcc.target/aarch64/fix_extend1.c: New testcase.
15219 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
15221 PR middle-end/78809
15222 PR middle-end/83026
15223 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
15224 and BUILT_IN_STRNCMP_EQ.
15225 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
15226 BUILT_IN_STRNCMP_EQ.
15227 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
15228 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
15229 (gimple_fold_builtin): Likewise.
15230 * tree-ssa-strlen.c (compute_string_length): New function.
15231 (determine_min_obsize): New function.
15232 (handle_builtin_string_cmp): New function to handle calls to
15233 string compare functions.
15234 (strlen_optimize_stmt): Add handling to builtin string compare
15236 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
15237 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
15238 * tree.c (build_common_builtin_nodes): Add new defines of
15239 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
15241 2018-05-31 Jakub Jelinek <jakub@redhat.com>
15244 * bb-reorder.c (pass_partition_blocks::gate): Return false for
15245 functions with naked attribute.
15247 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
15249 * config/i386/sse.md (avx_vec_concat<mode>):
15250 Substitute concat_tg_mode mode attribute with xtg_mode.
15251 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
15252 (concat_tg_mode): Remove mode attribute.
15254 2018-05-31 Martin Sebor <msebor@redhat.com>
15257 * calls.c (alloc_max_size): Correct a logic error/typo.
15258 Treat excessive arguments as infinite. Warn for invalid arguments.
15259 * doc/invoke.texi (-Walloc-size-larger-than): Update.
15261 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
15264 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
15265 and movx for Haswell.
15267 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
15268 Cesar Philippidis <cesar@codesourcery.com>
15270 PR middle-end/85879
15271 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
15272 when emitting error on private/firstprivate reductions.
15273 * omp-low.c (lower_omp_target): Avoid reference-type processing
15274 on pointers for firstprivate clause.
15276 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
15278 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
15281 * config/aarch64/aarch64-simd.md
15282 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
15283 (aarch64_ld1_x3_<mode>): Likewise
15284 (aarch64_st1x2<VALLDIF:mode>): Likewise
15285 (aarch64_st1_x2_<mode>): Likewise
15286 (aarch64_st1x3<VALLDIF:mode>): Likewise
15287 (aarch64_st1_x3_<mode>): Likewise
15288 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
15289 (vld1_s8_x3): Likewise.
15290 (vld1_u16_x3): Likewise.
15291 (vld1_s16_x3): Likewise.
15292 (vld1_u32_x3): Likewise.
15293 (vld1_s32_x3): Likewise.
15294 (vld1_u64_x3): Likewise.
15295 (vld1_s64_x3): Likewise.
15296 (vld1_f16_x3): Likewise.
15297 (vld1_f32_x3): Likewise.
15298 (vld1_f64_x3): Likewise.
15299 (vld1_p8_x3): Likewise.
15300 (vld1_p16_x3): Likewise.
15301 (vld1_p64_x3): Likewise.
15302 (vld1q_u8_x3): Likewise.
15303 (vld1q_s8_x3): Likewise.
15304 (vld1q_u16_x3): Likewise.
15305 (vld1q_s16_x3): Likewise.
15306 (vld1q_u32_x3): Likewise.
15307 (vld1q_s32_x3): Likewise.
15308 (vld1q_u64_x3): Likewise.
15309 (vld1q_s64_x3): Likewise.
15310 (vld1q_f16_x3): Likewise.
15311 (vld1q_f32_x3): Likewise.
15312 (vld1q_f64_x3): Likewise.
15313 (vld1q_p8_x3): Likewise.
15314 (vld1q_p16_x3): Likewise.
15315 (vld1q_p64_x3): Likewise.
15316 (vst1_s64_x2): Likewise.
15317 (vst1_u64_x2): Likewise.
15318 (vst1_f64_x2): Likewise.
15319 (vst1_s8_x2): Likewise.
15320 (vst1_p8_x2): Likewise.
15321 (vst1_s16_x2): Likewise.
15322 (vst1_p16_x2): Likewise.
15323 (vst1_s32_x2): Likewise.
15324 (vst1_u8_x2): Likewise.
15325 (vst1_u16_x2): Likewise.
15326 (vst1_u32_x2): Likewise.
15327 (vst1_f16_x2): Likewise.
15328 (vst1_f32_x2): Likewise.
15329 (vst1_p64_x2): Likewise.
15330 (vst1q_s8_x2): Likewise.
15331 (vst1q_p8_x2): Likewise.
15332 (vst1q_s16_x2): Likewise.
15333 (vst1q_p16_x2): Likewise.
15334 (vst1q_s32_x2): Likewise.
15335 (vst1q_s64_x2): Likewise.
15336 (vst1q_u8_x2): Likewise.
15337 (vst1q_u16_x2): Likewise.
15338 (vst1q_u32_x2): Likewise.
15339 (vst1q_u64_x2): Likewise.
15340 (vst1q_f16_x2): Likewise.
15341 (vst1q_f32_x2): Likewise.
15342 (vst1q_f64_x2): Likewise.
15343 (vst1q_p64_x2): Likewise.
15344 (vst1_s64_x3): Likewise.
15345 (vst1_u64_x3): Likewise.
15346 (vst1_f64_x3): Likewise.
15347 (vst1_s8_x3): Likewise.
15348 (vst1_p8_x3): Likewise.
15349 (vst1_s16_x3): Likewise.
15350 (vst1_p16_x3): Likewise.
15351 (vst1_s32_x3): Likewise.
15352 (vst1_u8_x3): Likewise.
15353 (vst1_u16_x3): Likewise.
15354 (vst1_u32_x3): Likewise.
15355 (vst1_f16_x3): Likewise.
15356 (vst1_f32_x3): Likewise.
15357 (vst1_p64_x3): Likewise.
15358 (vst1q_s8_x3): Likewise.
15359 (vst1q_p8_x3): Likewise.
15360 (vst1q_s16_x3): Likewise.
15361 (vst1q_p16_x3): Likewise.
15362 (vst1q_s32_x3): Likewise.
15363 (vst1q_s64_x3): Likewise.
15364 (vst1q_u8_x3): Likewise.
15365 (vst1q_u16_x3): Likewise.
15366 (vst1q_u32_x3): Likewise.
15367 (vst1q_u64_x3): Likewise.
15368 (vst1q_f16_x3): Likewise.
15369 (vst1q_f32_x3): Likewise.
15370 (vst1q_f64_x3): Likewise.
15371 (vst1q_p64_x3): Likewise.
15373 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
15375 * config/msp430/msp430.c (msp430_output_labelref): Prepend
15376 user_label_prefix to name.
15378 * tree-core.h: Update comment about the format of NAME string
15379 passed to handler in attribute_spec.
15381 * config/msp430/msp430.md: Remove erroneous subreg expression from
15382 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
15383 zero_extend{q,h}isi2.
15385 2018-05-30 Borislav Petkov <bp@suse.de>
15387 * doc/extend.texi: Document some architecture specific
15388 constraints and sort entries.
15390 2018-05-30 Martin Sebor <msebor@redhat.com>
15392 PR middle-end/85369
15393 * builtins.c (expand_builtin_stpcpy_1): New function.
15394 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
15395 only if the former succeeds.
15397 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
15399 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
15402 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
15404 * doc/invoke.texi (-flinker-output): Document
15406 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
15408 * passes.c (ipa_write_summaries): Only modify statements if body
15410 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
15411 incrementally linking.
15412 (ipa_passes): Likewise.
15413 * lto-cgraph.c (lto_output_node): When incrementally linking do not
15414 pass down resolution info.
15415 * common.opt (flag_incremental_link): Update info.
15416 * gcc.c (plugin specs): Turn flinker-output=* to
15417 -plugin-opt=-linker-output-known
15418 * toplev.c (compile_file): Also cut compilation when doing incremental
15420 * flag-types. (enum lto_partition_model): Add
15421 LTO_LINKER_OUTPUT_NOLTOREL.
15422 (invoke.texi): Add -flinker-output docs.
15423 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
15424 link same way as WPA; do not stream in dead initializers.
15426 * dwarf2out.c (dwarf2out_die_ref_for_decl,
15427 darf2out_register_external_decl): Support incremental link.
15429 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
15431 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
15433 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
15435 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
15436 it down to simple_object_copy_lto_debug_sections.
15437 (run_gcc): Determine incremental LTO link time and configure
15438 lto1 into non-wpa mode, disable renaming of debug sections.
15440 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
15442 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
15443 descriptions of various incorrectly documented functions.
15445 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
15448 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
15449 address check not strict.
15451 2018-05-30 Richard Biener <rguenther@suse.de>
15453 PR tree-optimization/85964
15454 * tracer.c (better_p): Drop initialized count check, we only
15455 call the function with initialized counts now.
15456 (find_best_successor): Do find a best edge if one
15457 has uninitialized count.
15458 (find_best_predecessor): Likewise. Do BB frequency check only
15459 if count is initialized.
15461 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
15463 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
15464 (aarch64_ldrstr_offset_compare): New.
15465 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
15466 load/store orderings.
15467 (aarch64_gen_adjusted_ldpstp): Likewise.
15469 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
15471 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
15472 Check for subset of GENERAL_REGS and FP_REGS.
15473 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
15476 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
15478 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
15479 and wi::to_poly_offset. Add the current offset and then check
15480 whether the sum fits, rather than using an unchecked addition of
15481 a checked term. Check for a shwi rather than a uhwi.
15482 * expr.c (get_bit_range): Use tree_to_poly_uint64.
15483 (store_constructor): Use poly_int_tree_p.
15484 (expand_expr_real_1): Likewise.
15485 * function.c (assign_temp): Likewise.
15486 * fold-const.c (const_binop): Use poly_int_tree_p and
15487 wi::to_poly_offset.
15488 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
15490 * ipa-icf-gimple.c (func_checker::compare_operand): Use
15491 to_poly_offset for MEM offsets.
15492 * ipa-icf.c (sem_variable::equals): Likewise.
15493 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
15494 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
15495 wi::to_poly_offset for BIT_FIELD_REF offsets.
15496 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
15497 wi::to_poly_offset.
15498 * var-tracking.c (emit_note_insn_var_location): Use
15499 tree_to_poly_uint64.
15501 2018-05-29 Jim Wilson <jimw@sifive.com>
15503 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
15505 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
15508 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
15509 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
15511 (sse4_1_round<mode>2): Use nonimmediate_operand
15512 for operand 1 predicate.
15514 2018-05-29 Martin Sebor <msebor@redhat.com>
15515 Richard Biener <rguenther@suse.de>
15518 * calls.c (get_size_range): Call determine_value_range instead
15519 of get_value_range..
15520 * tree-vrp.h (determine_value_range): Declared new function.
15521 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
15523 2018-05-29 Richard Biener <rguenther@suse.de>
15525 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
15526 sure to use non-pattern stmts for get_earlier_stmt arguments.
15527 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
15528 called on pattern stmts.
15529 (get_later_stmt): Likewise.
15531 2018-05-29 Martin Liska <mliska@suse.cz>
15533 PR gcov-profile/85759
15534 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
15537 2018-05-29 Jakub Jelinek <jakub@redhat.com>
15539 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
15541 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
15545 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
15546 VEC_PACK_FLOAT_EXPR): New tree codes.
15547 * tree-pretty-print.c (op_code_prio): Handle
15548 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
15549 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
15550 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
15551 * tree-inline.c (estimate_operator_cost): Likewise.
15552 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
15553 * fold-const.c (const_binop): Likewise.
15554 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
15555 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
15556 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
15557 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
15558 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
15559 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
15560 * expr.c (expand_expr_real_2): Likewise.
15561 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
15562 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
15563 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
15565 * optabs.c (expand_widen_pattern_expr): For
15566 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
15567 sign from result type rather than operand's type.
15568 (expand_binop_directly): For vec_packu_float_optab and
15569 vec_packs_float_optab allow result type to be different from operand's
15571 * optabs-tree.c (optab_for_tree_code): Handle
15572 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
15573 VEC_PACK_FLOAT_EXPR. Formatting fixes.
15574 * tree-vect-generic.c (expand_vector_operations_1): Handle
15575 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
15576 VEC_PACK_FLOAT_EXPR.
15577 * tree-vect-stmts.c (supportable_widening_operation): Handle
15579 (supportable_narrowing_operation): Handle FLOAT_EXPR.
15580 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
15581 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
15582 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
15583 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
15585 (vec_pack<floatprefix>_float_<mode>): New expander.
15586 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
15588 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
15589 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
15590 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
15591 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
15592 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
15594 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
15595 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
15596 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
15597 VEC_PACK_FLOAT_EXPR): Document.
15599 2018-05-29 Richard Biener <rguenther@suse.de>
15601 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
15603 (stmt_vec_info_vec): Make pointer.
15604 (init_stmt_vec_info_vec): Remove.
15605 (free_stmt_vec_info_vec): Likewise.
15606 (set_stmt_vec_info_vec): New function.
15607 (free_stmt_vec_infos): Likewise.
15608 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
15609 (set_vinfo_for_stmt): Likewise.
15610 (get_earlier_stmt): Likewise.
15611 (get_later_stmt): Likewise.
15612 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
15613 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
15614 (vec_info::~vec_info): Free stmt_vec_infos.
15615 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
15616 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
15617 (pass_slp_vectorize::execute): Likewise.
15618 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
15619 (free_stmt_vec_info_vec): Likewise.
15620 (set_stmt_vec_info_vec): New function.
15621 (free_stmt_vec_infos): Likewise.
15622 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
15623 the global stmt_vec_info_vec.
15624 * tree-parloops.c (gather_scalar_reductions): Use
15625 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
15628 2018-05-29 Richard Biener <rguenther@suse.de>
15630 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
15632 2018-05-29 Martin Liska <mliska@suse.cz>
15633 David Malcolm <dmalcolm@redhat.com>
15635 * vec.c (test_reverse): New.
15636 (vec_c_tests): Add new test.
15637 * vec.h (vl_ptr>::reverse): New function.
15639 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
15641 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
15643 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
15646 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
15648 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
15650 2018-05-28 Richard Biener <rguenther@suse.de>
15652 PR tree-optimization/85933
15653 * tree-vect-data-refs.c (vect_record_base_alignments): Only
15654 look at stmts marked as vectorizable.
15656 2018-05-28 Richard Biener <rguenther@suse.de>
15658 PR tree-optimization/85934
15659 * tree-vect-generic.c (expand_vector_operations_1): Hoist
15660 vector boolean check before scalar optimization.
15662 2018-05-28 Jakub Jelinek <jakub@redhat.com>
15664 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
15667 2018-05-28 Mark Wielaard <mark@klomp.org>
15669 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
15670 if it is an expression containing a minus sign.
15672 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
15674 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
15676 2018-05-27 Paul Koning <ni1d@arrl.net>
15678 * config/pdp11/pdp11.md (truncsihi2): Remove.
15680 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
15681 Chung-Ju Wu <jasonwucj@gmail.com>
15683 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
15685 (unaligned_store_dw): Ditto.
15686 * config/nds32/nds32-memory-manipulation.c
15687 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
15688 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
15689 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
15690 (emit_setmem_word_loop): Rename to ...
15691 (emit_setmem_doubleword_loop): ... this.
15692 (nds32_gen_dup_4_byte_to_word_value): New function.
15693 (nds32_gen_dup_8_byte_to_double_word_value): New function.
15694 (nds32_expand_setmem_loop): Refine implementation.
15695 (nds32_expand_setmem_loop_v3m): Ditto.
15696 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
15699 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
15701 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
15703 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
15705 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
15706 (nds32_init_machine_status): Initialize machine->attr_naked_p and
15707 machine->attr_no_prologue_p.
15708 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
15709 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
15710 (nds32_expand_epilogue): Consider attr_naked_p.
15711 (nds32_expand_epilogue_v3pop): Likewise.
15712 (nds32_can_use_return_insn): Likewise.
15713 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
15714 attr_no_prologue_p fields.
15715 * config/nds32/nds32.opt (mret-in-naked-func): New option.
15717 2018-05-27 Jakub Jelinek <jakub@redhat.com>
15720 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
15722 * config/i386/sse.md
15723 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
15725 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
15727 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
15729 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
15731 (*<floatsuffix>floatv2div2sf2): Rename to ...
15732 (*float<floatunssuffix>v2div2sf2): ... this.
15733 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
15734 (float<floatunssuffix>v2div2sf2_mask): ... this.
15735 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
15736 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
15737 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
15739 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
15741 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
15743 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
15745 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
15747 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
15749 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
15750 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
15751 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
15752 gen_ufix_truncv8dfv8si2.
15753 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
15754 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
15755 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
15756 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
15757 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
15758 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
15759 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
15760 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
15762 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
15766 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
15768 2018-05-25 Jim Wilson <jimw@sifive.com>
15770 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
15771 * config/riscv/riscv.c (struct machine_function): Add
15772 interrupt_handler_p and attribute_checked_p fields.
15773 (riscv_attribute_table): Add interrupt.
15774 (riscv_interrupt_type_p): New.
15775 (riscv_save_reg_p): Save extra regs for interrupt handler.
15776 (riscv_use_save_libcall): Return false for interrupt handler.
15777 (riscv_first_stack_step): Add forward declaration.
15778 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
15779 for interrupt handler with large frame. Use it for saved reg list.
15780 (riscv_expand_prologue): Move flag_stack_usage_info support to
15781 eliminate duplication.
15782 (riscv_expand_epilogue): Generate mret for interrupt handler.
15783 (riscv_epilogue_uses): New.
15784 (riscv_can_use_return_insn): Return false for interrupt handler.
15785 (riscv_function_ok_for_sibcall): Likewise.
15786 (riscv_set_current_function): Add interrupt handler support.
15787 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
15788 * config/riscv/riscv.md (UNSPECV_MRET): New.
15790 (riscv_frflags, riscv_fsflags): Use tab after opcode.
15792 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
15794 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
15796 PR tree-optimization/85712
15797 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
15798 this candidate has already been replaced in-situ by a copy.
15800 2018-05-25 Jason Merrill <jason@redhat.com>
15802 PR c++/80485 - inline function non-zero address.
15803 * symtab.c (nonzero_address): Check DECL_COMDAT.
15805 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
15808 * config/alpha/alpha.md (ashlsi3): New insn pattern.
15809 (*ashlsi_se): Rename from *ashldi_se. Define as sign
15810 extension of SImode operation. Use const123_operand predicate.
15811 (*saddsi_1): Remove.
15812 (*saddl_se_1): Ditto.
15813 (*ssubsi_1): Ditto.
15814 (*ssubl_se_1): Ditto.
15815 * config/alpha/predicates.md (const123_operand): New predicate.
15816 * config/alpha/constraints.md (P): Use IN_RANGE.
15818 2018-05-25 Richard Biener <rguenther@suse.de>
15820 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
15822 (ref_maybe_used_by_stmt_p): Likewise.
15823 (stmt_may_clobber_ref_p): Likewise.
15824 (stmt_may_clobber_ref_p_1): Likewise.
15825 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
15827 (ref_maybe_used_by_stmt_p): Likewise.
15828 (stmt_may_clobber_ref_p): Likewise.
15829 (stmt_may_clobber_ref_p_1): Likewise.
15830 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
15831 the alias oracle to disambiguate DRs with stmts DR analysis
15833 (vect_analyze_data_refs): Do not give up on not analyzable
15834 DRs for BB vectorization. Remove code truncating the dataref
15837 2018-05-25 Jakub Jelinek <jakub@redhat.com>
15840 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
15841 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
15842 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
15844 2018-05-25 Richard Biener <rguenther@suse.de>
15846 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
15847 function, combining stmt data ref gathering and fatal analysis
15849 (vect_analyze_data_refs): Remove now redudnant code and simplify.
15850 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
15851 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
15852 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
15853 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
15855 2018-05-25 Bin Cheng <bin.cheng@arm.com>
15857 PR tree-optimization/85720
15858 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
15859 SCC if all partitions are builtins.
15860 (version_loop_by_alias_check): New parameter. Generate cancelable
15861 runtime alias check if all partitions are builtins.
15862 (distribute_loop): Update call to above function.
15864 2018-05-25 Bin Cheng <bin.cheng@arm.com>
15866 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
15867 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
15868 (parm_default_def_partition_arg): Ditto.
15869 (set_parm_default_def_partition): Ditto.
15870 (get_parm_default_def_partitions): Ditto and make it static.
15871 (get_undefined_value_partitions): Ditto and make it static.
15872 (remove_ssa_form): Refactor call to init_var_map here.
15873 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
15874 computation for loop region.
15875 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
15876 (register_default_def): Delete.
15877 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
15878 (parm_default_def_partition_arg): Ditto.
15879 (set_parm_default_def_partition): Ditto.
15880 (get_parm_default_def_partitions): Ditto and make it static.
15881 (get_undefined_value_partitions): Ditto and make it static.
15882 (coalesce_with_default, coalesce_with_default): Update comment.
15883 (create_coalesce_list_for_region): New func factored out from
15884 create_outofssa_var_map.
15885 (populate_coalesce_list_for_outofssa): New func factored out from
15886 create_outofssa_var_map and coalesce_ssa_name.
15887 (create_outofssa_var_map): Delete.
15888 (coalesce_ssa_name): Refactor to support live range computation.
15889 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
15890 (get_parm_default_def_partitions): Delete.
15891 (get_undefined_value_partitions): Ditto.
15892 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
15893 computation for loop region.
15894 (new_tree_live_info, loe_visit_block): Ditto.
15895 (live_worklist, set_var_live_on_entry): Ditto.
15896 (calculate_live_on_exit, verify_live_on_entry): Ditto.
15897 * tree-ssa-live.h (struct _var_map): New fields.
15898 (init_var_map): Change decl.
15899 (region_contains_p): New.
15901 2018-05-25 Bin Cheng <bin.cheng@arm.com>
15903 * tree-ssa-live.h (live_merge_and_clear): Delete.
15905 2018-05-25 Richard Biener <rguenther@suse.de>
15908 * tree-dump.c (dequeue_and_dump): Remove access to removed
15909 operand 2 of a SWITCH_EXPR.
15911 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
15913 * doc/sourcebuild.texi (vect_double_cond_arith): Include
15914 multiplication and division.
15915 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
15916 (cond_udiv@var{m}, cond_umod@var{m}): Document.
15917 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
15918 (cond_udiv_optab, cond_umod_optab): New optabs.
15919 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
15920 (IFN_COND_RDIV): New internal functions.
15921 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
15922 TRUNC_MOD_EXPR and RDIV_EXPR.
15923 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
15924 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
15926 (SVE_INT_BINARY): Include mult.
15927 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
15928 (optab, sve_int_op): Handle mult.
15929 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
15931 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
15932 for SVE_INT_BINARY_SD.
15934 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
15936 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
15937 (optab, sve_int_op): Handle div and udiv.
15938 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
15939 for SVE_INT_BINARY_SD.
15940 (*<optab><mode>3): New insn for the same.
15942 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
15944 * tree-vect-patterns.c: Include predict.h.
15945 (vect_recog_divmod_pattern): Restrict check for division support
15946 to when optimizing for size.
15948 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
15950 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
15951 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
15952 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
15953 (gimple_match_op::set_op): Likewise.
15954 (gimple_resimplify4): Declare.
15955 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
15956 (expr::gen_transform): Likewise.
15957 (decision_tree::gen): Generate a simplification routine for 4 operands.
15958 * gimple-match-head.c (gimple_simplify): Add an overload for
15959 4 operands. In the top-level function, handle up to 4 call
15960 arguments and call gimple_resimplify4.
15961 (gimple_resimplify4): New function.
15962 (build_call_internal): Pass a fourth operand.
15963 (maybe_push_to_seq): Likewise.
15964 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
15965 Fold VEC_COND_EXPRs of an operation and a default value into
15966 an IFN_COND_* function if possible.
15967 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
15969 (SVE_COND_FP_BINARY): Include them.
15970 (optab, sve_fp_op): Handle them.
15971 (SVE_INT_BINARY_REV): New code iterator.
15972 (SVE_COND_FP_BINARY_REV): New int iterator.
15973 (commutative): New int attribute.
15974 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
15976 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
15978 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
15979 (*cond_<optab><mode>): New patterns for reversed operands.
15981 2018-05-25 Richard Biener <rguenther@suse.de>
15983 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
15984 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
15985 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
15986 (STMT_VINFO_GROUPED_ACCESS): Adjust.
15987 * tree-vect-data-refs.c (everywhere): Adjust users.
15988 * tree-vect-loop.c (everywhere): Likewise.
15989 * tree-vect-slp.c (everywhere): Likewise.
15990 * tree-vect-stmts.c (everywhere): Likewise.
15991 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
15993 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15995 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
15997 (gcc_cv_as_section_exclude): ... this.
15998 Try Solaris as #exclude syntax.
15999 * configure: Regenerate.
16000 * config.in: Regenerate.
16001 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
16003 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
16004 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
16006 * varasm.c (default_elf_asm_named_section): Don't check if
16007 HAVE_GAS_SECTION_EXCLUDE is defined.
16009 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
16011 * doc/md.texi: Update the documentation of the cond_* optabs
16012 to mention the new final operand. Fix GET_MODE_NUNITS call.
16013 Describe the scalar case too.
16014 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
16015 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
16017 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
16018 (get_conditional_internal_fn): Update comment.
16019 * tree-vect-loop.c (vectorizable_reduction): Pass the original
16020 accumulator value as a final argument to conditional functions.
16021 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
16022 a define_expand and add an "else" operand. Assert for now that
16023 the else operand is equal to operand 2. Use SVE_INT_BINARY and
16024 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
16025 (*cond_<optab><mode>): New patterns.
16026 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
16027 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
16028 (UNSPEC_COND_EOR): Delete.
16029 (optab): Remove associated mappings.
16030 (SVE_INT_BINARY): New code iterator.
16031 (sve_int_op): Remove int attribute and add "minus" to the code
16033 (SVE_COND_INT_OP): Delete.
16034 (SVE_COND_FP_OP): Rename to...
16035 (SVE_COND_FP_BINARY): ...this.
16037 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
16039 * optabs.c (can_reuse_operands_p): New function.
16040 (maybe_legitimize_operands): Try to reuse the results for
16043 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
16045 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
16046 Add {q} suffix to insn mnemonic.
16048 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
16050 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
16051 (msp430_warn_func_return): New.
16053 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
16055 * fold-const.c (tree_nonzero_bits): New function.
16056 * fold-const.h (tree_nonzero_bits): Likewise.
16057 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
16058 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
16060 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
16064 * varasm.c (assemble_alias): Check ifunc_resolver only on
16067 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
16070 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
16071 when memory input operand is handled.
16073 2018-05-24 Luis Machado <luis.machado@linaro.org>
16075 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
16077 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
16079 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
16081 * match.pd: Delay FMA folds until after vectorization.
16083 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
16086 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
16087 address check not strict.
16089 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
16091 * gimple-match.h (gimple_match_op): New class.
16092 (mprts_hook): Replace parameters with a gimple_match_op *.
16093 (maybe_build_generic_op): Likewise.
16094 (gimple_simplified_result_is_gimple_val): Replace parameters with
16095 a const gimple_match_op *.
16096 (gimple_simplify): Replace code_helper * and tree * parameters with
16097 a gimple_match_op * parameter.
16098 (gimple_resimplify1): Replace code_helper *, tree and tree *
16099 parameters with a gimple_match_op * parameter.
16100 (gimple_resimplify2): Likewise.
16101 (gimple_resimplify3): Likewise.
16102 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
16103 parameters with a gimple_match_op * parameter.
16104 * gimple-match-head.c (gimple_simplify): Change prototypes of
16105 auto-generated functions to take a gimple_match_op * instead of
16106 separate code_helper * and tree * parameters. Make the same
16107 change in the top-level overload and update calls to the
16108 gimple_resimplify routines. Update calls to the auto-generated
16109 functions and to maybe_push_res_to_seq in the publicly-facing
16110 operation-specific gimple_simplify overloads.
16111 (gimple_match_op::MAX_NUM_OPS): Define.
16112 (gimple_resimplify1): Replace rcode and ops with a single res_op
16113 parameter. Update call to gimple_simplify.
16114 (gimple_resimplify2): Likewise.
16115 (gimple_resimplify3): Likewise.
16116 (mprts_hook): Replace parameters with a gimple_match_op *.
16117 (maybe_build_generic_op): Likewise.
16118 (build_call_internal): Replace type, nargs and ops with
16119 a gimple_match_op *.
16120 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
16121 with a single gimple_match_op *. Update calls to mprts_hook,
16122 build_call_internal and gimple_simplified_result_is_gimple_val.
16123 Factor out code that is common to the tree_code and combined_fn cases.
16124 * genmatch.c (expr::gen_transform): Replace tem_code and
16125 tem_ops with a gimple_match_op called tem_op. Update calls
16126 to the gimple_resimplify functions and maybe_push_res_to_seq.
16127 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
16128 res_ops. Update call to the gimple_resimplify functions.
16129 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
16130 (decision_tree::gen): Make the functions take a gimple_match_op *
16131 called res_op instead of separate res_code and res_ops parameters.
16132 Update call accordingly.
16133 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
16134 and ops with a single res_op parameter. Update calls to
16135 maybe_build_generic_op and maybe_push_res_to_seq.
16136 (fold_stmt_1): Update calls to gimple_simplify and
16137 replace_stmt_with_simplification.
16138 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
16139 and gimple_simplified_result_is_gimple_val.
16140 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
16142 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
16143 with a gimple_match_op *.
16144 (vn_nary_build_or_lookup): Likewise. Update call to
16145 vn_nary_build_or_lookup_1.
16146 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
16147 gimple_match_op *. Update calls to the gimple_resimplify routines
16148 and to gimple_simplified_result_is_gimple_val.
16149 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
16150 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
16151 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
16152 (visit_nary_op): Likewise.
16153 (visit_reference_op_load): Likewise.
16155 2018-05-23 Luis Machado <luis.machado@linaro.org>
16157 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
16158 modifier for printing the step amount.
16160 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
16163 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
16166 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
16168 * doc/sourcebuild.texi (Endianness): New subsubsection.
16170 2018-05-23 Luis Machado <luis.machado@linaro.org>
16172 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
16173 <prefetch_dynamic_strides>: New const bool field.
16174 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
16175 prefetch_dynamic_strides.
16176 (exynosm1_prefetch_tune): Likewise.
16177 (thunderxt88_prefetch_tune): Likewise.
16178 (thunderx_prefetch_tune): Likewise.
16179 (thunderx2t99_prefetch_tune): Likewise.
16180 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
16182 (aarch64_override_options_internal): Update to set
16183 PARAM_PREFETCH_DYNAMIC_STRIDES.
16184 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
16185 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
16186 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
16187 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
16188 prefetch-dynamic-strides setting.
16190 2018-05-23 Luis Machado <luis.machado@linaro.org>
16192 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
16193 <minimum_stride>: New const int field.
16194 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
16195 minimum_stride field defaulting to -1.
16196 (exynosm1_prefetch_tune): Likewise.
16197 (thunderxt88_prefetch_tune): Likewise.
16198 (thunderx_prefetch_tune): Likewise.
16199 (thunderx2t99_prefetch_tune): Likewise.
16200 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
16201 <default_opt_level>: Set to 3.
16202 (aarch64_override_options_internal): Update to set
16203 PARAM_PREFETCH_MINIMUM_STRIDE.
16204 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
16205 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
16206 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
16207 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
16208 stride is constant and is below the minimum stride threshold.
16210 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16212 * config/arm/arm-cpus.in (mode26): Delete.
16213 (armv4): Delete mode26 reference.
16214 * config/arm/arm.c (arm_configure_build_target): Delete use of
16217 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
16219 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
16221 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
16222 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
16224 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
16225 (floatunsdidf2): Ditto.
16227 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
16229 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
16230 (fixuns_trunc<mode>si2_avx512f): Ditto.
16231 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
16232 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
16233 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
16235 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
16237 PR rtl-optimization/79985
16238 * df-scan.c (df_insn_refs_collect): Remove special case for
16239 global registers and asm statements.
16241 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
16243 * extend.texi (Global Register Variables): Rewrite the bullet list.
16244 Note that the register is available for allocation. Note that access
16245 via inline asm must use constraints. Add note about async-signal
16246 handlers. Remove paragraph about automagic register selection.
16248 2018-05-23 Richard Biener <rguenther@suse.de>
16250 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
16251 of fixed offset from memset VN.
16253 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
16255 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
16256 first_interp field.
16257 (alloc_cand_and_find_basis): Initialize first_interp field.
16258 (slsr_process_mul): Modify first_interp field.
16259 (slsr_process_add): Likewise.
16260 (slsr_process_cast): Modify first_interp field for each new
16262 (slsr_process_copy): Likewise.
16263 (dump_candidate): Dump first_interp field.
16264 (replace_mult_candidate): Process all interpretations, not just
16266 (replace_rhs_if_not_dup): Likewise.
16267 (replace_one_candidate): Likewise.
16269 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
16271 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
16273 (aarch64_needs_frame_chain): New function.
16274 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
16276 2018-05-23 Sudakshina Das <sudi.das@arm.com>
16279 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
16280 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
16281 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
16282 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
16283 as true for strict-align.
16284 (aarch64_can_inline_p): Perform checks even when callee has no
16285 attributes to check for strict alignment.
16286 * doc/extend.texi (AArch64 Function Attributes): Document
16288 * doc/invoke.texi: (AArch64 Options): Likewise.
16290 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
16292 PR tree-optimization/85853
16293 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
16294 the handling of the root of the node to...
16295 (vect_slp_analyze_node_operations_1): ...this new function,
16296 and run the whole thing with the child nodes' def types
16297 set according to their SLP node's def type.
16299 2018-05-23 Richard Biener <rguenther@suse.de>
16301 PR middle-end/85874
16302 * tree-data-ref.c (create_runtime_alias_checks): Defer
16303 and ignore overflow warnings.
16305 2018-05-23 Yury Gribov <tetra2005@gmail.com>
16307 PR tree-optimization/85822
16308 * tree-vrp.c (is_masked_range_test): Fix handling of negative
16311 2018-05-23 Richard Biener <rguenther@suse.de>
16313 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
16314 memset constants via native_interpret_expr.
16316 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
16319 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
16321 (cgraph_node::create_alias): Likewise.
16322 (cgraph_node::get_availability): Check ifunc_resolver instead
16323 of looking up ifunc attribute.
16324 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
16325 * varasm.c (do_assemble_alias): Likewise.
16326 (assemble_alias): Likewise.
16327 (default_binds_local_p_3): Likewise.
16328 * cgraph.h (cgraph_node): Add ifunc_resolver.
16329 (cgraph_node::only_called_directly_or_aliased_p): Return false
16330 for IFUNC resolver.
16331 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
16333 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
16334 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
16335 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
16336 instead of looking up ifunc attribute.
16338 2018-05-22 Luis Machado <luis.machado@linaro.org>
16340 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
16342 2018-05-22 Martin Sebor <msebor@redhat.com>
16344 PR middle-end/85359
16345 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
16346 only when expasion succeeds.
16347 (expand_builtin_strcmp): Same.
16348 (expand_builtin_strncmp): Same.
16350 2018-05-22 Martin Sebor <msebor@redhat.com>
16352 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
16354 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
16355 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16357 * config/aarch64/aarch64-ldpstp.md: Replace uses of
16358 aarch64_mem_pair_operand with memory_operand and delete operand swapping
16360 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
16361 Add check for legitimate_address.
16362 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
16363 (aarch64_swap_ldrstr_operands): New.
16364 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
16367 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
16368 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16370 * config/aarch64/aarch64.md: New patterns to generate stp
16372 (store_pair_sw, store_pair_dw): New patterns to generate stp for
16373 single words and double words.
16374 (load_pair_sw, load_pair_dw): Likewise.
16375 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
16377 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
16379 * config/aarch64/aarch64-ldpstp.md: Modify peephole
16380 for different mode ldpstp and add peephole for merged zero stores.
16381 Likewise for loads.
16382 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
16384 (aarch64_gen_store_pair): Rename calls to match new patterns.
16385 (aarch64_gen_load_pair): Rename calls to match new patterns.
16386 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
16387 (load_pair<DREG:mode><DREG2:mode>): ... This.
16388 (store_pair<mode>): Rename to...
16389 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
16390 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
16391 New mode iterators.
16392 (V_INT_EQUIV): Handle SImode.
16393 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
16396 2018-05-22 Martin Sebor <msebor@redhat.com>
16399 * calls.c (maybe_warn_nonstring_arg): Use string length to set
16400 or ajust the presumed bound on an operation to avoid unnecessary
16403 2018-05-22 Martin Sebor <msebor@redhat.com>
16405 PR tree-optimization/85826
16406 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
16407 assuming that a DECL necesarily has a constant size.
16409 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
16411 PR middle-end/85862
16412 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
16414 2018-05-22 Richard Biener <rguenther@suse.de>
16416 PR tree-optimization/85834
16417 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
16418 non-constant and non-zero memset arguments.
16420 2018-05-22 Martin Liska <mliska@suse.cz>
16423 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
16425 2018-05-22 Richard Biener <rguenther@suse.de>
16427 PR tree-optimization/85863
16428 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
16429 comparisons when vectype is specified.
16430 (vectorizable_condition): Do not specify vectype for
16431 vect_is_simple_cond when SLP vectorizing.
16433 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
16436 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
16437 define __ibm128 as long double.
16438 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
16439 as a distinct type when IEEE 128-bit support is enabled.
16440 (init_float128_ieee): Fix up conversions between IFmode and IEEE
16441 128-bit types to use the correct functions.
16442 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
16443 convert between 128-bit floating point types that have different
16444 modes but the same representation, instead of using gen_lowpart to
16446 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
16448 (IFKF_reg): New attributes to give the register constraints for
16450 (extend<mode>tf2_internal): New insns to mark an explicit
16451 conversion between 128-bit floating point types that have a
16452 different mode but share the same representation.
16454 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
16456 PR tree-optimization/85814
16457 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
16458 a null return from get_strinfo when unsharing the next
16459 strinfo in the chain.
16461 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
16464 * varasm.c (weak_finish): Clean up weak_decls.
16466 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16468 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
16469 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
16470 UNSPEC_UADALP values.
16471 * config/aarch64/iterators.md (ABAL): New int iterator.
16474 (sur): Add mappings for the above.
16475 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
16477 (aarch64_<sur>abal<mode>_4): Likewise.
16478 (aarch64_<sur>adalp<mode>_3): Likewise.
16479 (<sur>sadv16qi): New define_expand.
16481 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
16483 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
16484 (*movdf_internal): Ditto.
16485 (*rcpsf2_sse): Ditto.
16486 (*rsqrtsf2_sse): Ditto.
16487 (*sqrt<mode>2_sse): Ditto.
16489 2018-05-21 Tamar Christina <tamar.christina@arm.com>
16491 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
16493 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
16494 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
16495 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
16496 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
16498 * config/aarch64/arm_neon.h: Likewise.
16499 * config/aarch64/iterators.md (VQ_I): New.
16501 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
16503 * config.gcc: Add arc/t-multilib-linux to tmake_file for
16505 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
16508 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
16510 * config/nds32/constraints.md (S): New constraint.
16511 * config/nds32/nds32.md (call_internal): Use constraint S.
16512 (call_value_internal): Likewise.
16513 (sibcall_internal): Likewise.
16514 (sibcall_value_internal): Likewise.
16516 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
16517 Chung-Ju Wu <jasonwucj@gmail.com>
16519 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
16520 into consideration.
16522 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
16523 Chung-Ju Wu <jasonwucj@gmail.com>
16525 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
16526 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
16527 (nds32_rtx_costs_impl): Simplify.
16528 (nds32_address_cost_impl): Simplify.
16529 (nds32_init_rtx_costs): New function.
16530 (nds32_rtx_costs_speed_prefer): Likewise.
16531 (nds32_rtx_costs_size_prefer): Likewise.
16532 (nds32_address_cost_speed_prefer): Likewise.
16533 (nds32_address_cost_speed_fwprop): Likewise.
16534 (nds32_address_cost_size_prefer): Likewise.
16535 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
16536 * config/nds32/nds32.c (nds32_option_override): Use
16537 nds32_init_rtx_costs function.
16539 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
16541 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
16542 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
16543 (TARGET_PIPELINE_N8): Likewise.
16544 (TARGET_PIPELINE_N10): Likewise.
16545 (TARGET_PIPELINE_N13): Likewise.
16546 (TARGET_PIPELINE_GRAYWOLF): Likewise.
16548 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
16550 * config/nds32/nds32-fpu.md: Update copyright year.
16552 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
16554 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
16556 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
16558 * config/nds32/nds32.c
16559 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
16560 * config/nds32/nds32.opt (minline-asm-r15): New option.
16562 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
16564 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
16566 * config/nds32/nds32.md (abssi2): New pattern.
16568 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
16570 * config/i386/i386.md (rex64namesuffix): New mode attribute.
16571 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
16572 Merge insn pattern from sse_cvtsi2ss<round_name> and
16573 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
16574 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
16575 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
16576 using SWI48 mode iterator.
16577 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
16578 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
16579 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
16580 pattern from sse_cvttss2si<round_saeonly_name>
16581 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
16582 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
16583 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
16584 using SWI48 mode iterator.
16585 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
16586 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
16587 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
16588 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
16589 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
16590 using SWI48 mode iterator.
16591 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
16592 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
16593 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
16594 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
16595 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
16596 SWI48 mode iterator.
16597 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
16598 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
16599 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
16600 pattern from sse_cvttsd2si<round_saeonly_name>
16601 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
16603 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
16605 * config/nds32/nds32-md-auxiliary.c
16606 (nds32_valid_smw_lwm_base_p): Refine.
16607 (nds32_output_smw_single_word): Refine.
16608 (nds32_output_smw_double_word): New.
16609 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
16611 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
16613 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
16614 (nds32_output_stack_pop): Refine.
16615 (nds32_expand_unaligned_load): Refine.
16616 (nds32_expand_unaligned_store): Refine.
16618 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
16619 Chung-Ju Wu <jasonwucj@gmail.com>
16621 * config/nds32/constants.md: Add TP_REGNUM constant.
16622 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
16623 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
16625 * config/nds32/nds32-doubleword.md: Consider flag_pic.
16626 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
16627 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
16628 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
16629 and PIC code generation.
16630 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
16632 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
16634 * config/nds32/nds32.md: Support TLS and PIC.
16635 * config/nds32/nds32.c: Support TLS and PIC.
16636 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
16637 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
16640 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
16642 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
16643 mode with E_ prefix.
16645 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
16646 Chung-Ju Wu <jasonwucj@gmail.com>
16648 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
16649 * config/nds32/nds32-md-auxiliary.c
16650 (symbolic_reference_mentioned_p): New.
16651 (nds32_legitimize_ict_address): New.
16652 (nds32_expand_ict_move): New.
16653 (nds32_indirect_call_referenced_p): New.
16654 (nds32_symbol_binds_local_p): Delete.
16655 (nds32_long_call_p): Modify.
16656 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
16657 * config/nds32/nds32-protos.h
16658 (symbolic_reference_mentioned_p): Declare.
16659 (nds32_legitimize_ict_address): Declare.
16660 (nds32_expand_ict_move): Declare.
16661 (nds32_indirect_call_referenced_p): Declare.
16662 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
16663 (nds32_relax_group): Use nds32_ict_const_p as condition.
16664 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
16665 (nds32_asm_file_start): Output ict_model directive in asm code.
16666 (nds32_legitimate_address_p): Consider indirect call.
16667 (nds32_print_operand): Consider indirect call.
16668 (nds32_print_operand_address): Consider indirect call.
16669 (nds32_insert_attributes): Handle "indirect_call" attribute.
16670 (TARGET_LEGITIMATE_ADDRESS_P): Define.
16671 (TARGET_LEGITIMATE_CONSTANT_P): Define.
16672 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
16673 (TARGET_DELEGITIMIZE_ADDRESS): Define.
16674 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
16675 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
16676 (TARGET_ICT_MODEL_SMALL): Define.
16677 (TARGET_ICT_MODEL_LARGE): Define.
16678 * config/nds32/nds32.md (movsi): Consider ict model.
16679 (call, call_value): Consider ict model.
16680 (sibcall, sibcall_value): Consider ict model.
16681 * config/nds32/nds32.opt (mict-model): New option.
16682 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
16685 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
16686 Monk Chiang <sh.chiang04@gmail.com>
16687 Jim Wilson <jimw@sifive.com>
16689 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
16690 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
16691 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
16692 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
16693 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
16694 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
16695 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
16696 compute save_libcall_adjustment properly.
16697 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
16698 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
16699 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
16700 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
16701 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
16702 (ABI_SPEC): Handle mabi=ilp32e.
16703 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
16704 (RVE): Add RVE mask.
16705 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
16706 <-march>: Add rv32e as an example.
16708 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
16711 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
16712 (intra_create_variable_infos): Handle C++ constructors.
16714 2018-05-18 Martin Liska <mliska@suse.cz>
16716 * passes.def: Remove a redundant pass.
16718 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
16721 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
16723 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16725 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
16727 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
16729 (armv2, armv2a, armv3, armv3m): Delete architectures.
16730 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
16731 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
16732 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
16734 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
16735 (*mulsidi3adddi): Likewise.
16736 (mulsidi3): Likewise.
16737 (*mulsidi3_nov6): Likewise.
16738 (umulsidi3): Likewise.
16739 (umulsidi3_nov6): Likewise.
16740 (umaddsidi4): Likewise.
16741 (*umulsidi3adddi): Likewise.
16742 (smulsi3_highpart): Likewise.
16743 (*smulsi3_highpart_nov6): Likewise.
16744 (umulsi3_highpart): Likewise.
16745 (*umulsi3_highpart_nov6): Likewise.
16746 * config/arm/arm.h (arm_arch3m): Delete.
16747 * config/arm/arm.c (arm_arch3m): Delete.
16748 (arm_option_override_internal): Update armv3-related comment.
16749 (arm_configure_build_target): Delete use of isa_bit_mode32.
16750 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
16751 (arm_rtx_costs_internal): Delete check of arm_arch3m.
16752 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
16753 (mulsa3): Likewise.
16754 (mulusa3): Likewise.
16755 * config/arm/arm-protos.h (arm_arch3m): Delete.
16756 * config/arm/arm-tables.opt: Regenerate.
16757 * config/arm/arm-tune.md: Likewise.
16758 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
16759 deleted architectures.
16761 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16763 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
16764 (armv5t, armv5te): New features.
16765 (ARMv5, ARMv5e): Delete fgroups.
16766 (ARMv5t, ARMv5te): Adjust for above changes.
16767 (ARMv6m): Likewise.
16768 (armv5, armv5e): Delete arches.
16769 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
16771 (*call_reg_arm): Likewise.
16772 (*call_value_reg_armv5): Likewise.
16773 (*call_value_reg_arm): Likewise.
16774 (*call_symbol): Likewise.
16775 (*call_value_symbol): Likewise.
16776 (*sibcall_insn): Likewise.
16777 (*sibcall_value_insn): Likewise.
16778 (clzsi2): Likewise.
16779 (prefetch): Likewise.
16780 (define_split and define_peephole2 dependent on arm_arch5):
16782 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
16784 (TARGET_ARM_QBIT): Likewise.
16785 (TARGET_DSP_MULTIPLY): Likewise.
16786 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
16787 (arm_arch5, arm_arch5e): Delete.
16788 (arm_arch5t, arm_arch5te): Declare.
16789 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
16790 (arm_arch5t): Declare.
16791 (arm_option_reconfigure_globals): Update for the above.
16792 (arm_options_perform_arch_sanity_checks): Update comment, replace
16793 use of arm_arch5 with arm_arch5t.
16794 (use_return_insn): Likewise.
16795 (arm_emit_call_insn): Likewise.
16796 (output_return_instruction): Likewise.
16797 (arm_final_prescan_insn): Likewise.
16798 (arm_coproc_builtin_available): Likewise.
16799 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
16800 arm_arch5e with arm_arch5t and arm_arch5te.
16801 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
16802 (arm_arch5t, arm_arch5te): Declare.
16803 * config/arm/arm-tables.opt: Regenerate.
16804 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
16805 * config/arm/t-multilib: Likewise.
16806 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
16807 instead of arm_arch5.
16808 (*call_reg_thumb1): Likewise.
16809 (*call_value_reg_thumb1_v5): Likewise.
16810 (*call_value_reg_thumb1): Likewise.
16811 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
16813 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
16815 2018-05-18 Martin Liska <mliska@suse.cz>
16817 PR gcov-profile/84846
16818 * doc/gcov.texi: Document -t option of gcov tool.
16820 2018-05-18 Martin Liska <mliska@suse.cz>
16822 PR gcov-profile/84846
16823 * gcov.c (print_usage): Add new -t option.
16824 (process_args): Handle the option.
16825 (generate_results): Use stdout as output when requested by
16828 2018-05-18 Martin Liska <mliska@suse.cz>
16830 PR gcov-profile/84846
16831 * coverage.c (coverage_init): Write PWD to .gcno file.
16832 * doc/gcov.texi: Document how working directory is printed.
16833 * gcov-dump.c (dump_gcov_file): Print PWD.
16834 * gcov.c (output_intermediate_file): Likewise.
16835 (read_graph_file): Read PWD string.
16836 (output_lines): Print PWD.
16838 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16840 PR middle-end/85817
16841 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
16842 for retval and return false if all args to phi are zero.
16844 2018-05-18 Richard Biener <rguenther@suse.de>
16846 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
16848 (evrp_dom_walker::before_dom_children): Call it.
16850 2018-05-18 Richard Biener <rguenther@suse.de>
16852 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
16853 results when processing array refs with variable index.
16855 2018-05-18 Toon Moene <toon@moene.org>
16857 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
16858 directly after that of -floop-interchange. Indicate that both
16859 options are enabled by default when specifying -O3.
16861 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16863 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
16864 iterator. Delete separate integer-mode vec_set<mode> expander.
16865 (aarch64_simd_vec_setv2di): Delete.
16866 (vec_setv2di): Delete.
16867 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
16868 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
16869 the "w, r" alternative.
16871 2018-05-18 Martin Liska <mliska@suse.cz>
16873 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
16874 * tree-pass.h (make_pass_lower_switch_O0): New function.
16875 * tree-switch-conversion.c (node_has_low_bound): Remove.
16876 (node_has_high_bound): Likewise.
16877 (node_is_bounded): Likewise.
16878 (class pass_lower_switch): Make it a template type and create
16880 (pass_lower_switch::execute): Add template argument.
16881 (make_pass_lower_switch): New function.
16882 (make_pass_lower_switch_O0): New function.
16883 (do_jump_if_equal): Remove.
16884 (emit_case_nodes): Simplify to just handle all 3 cases and leave
16885 all the hard work to tree optimization passes.
16887 2018-05-18 Martin Liska <mliska@suse.cz>
16889 * dbgcnt.c (limit_low): Renamed from limit.
16890 (limit_high): New variable.
16891 (dbg_cnt_is_enabled): Check for upper limit.
16892 (dbg_cnt): Adjust dumping.
16893 (dbg_cnt_set_limit_by_index): Add new argument for high
16895 (dbg_cnt_set_limit_by_name): Likewise.
16896 (dbg_cnt_process_single_pair): Parse new format.
16897 (dbg_cnt_process_opt): Use strtok.
16898 (dbg_cnt_list_all_counters): Remove 'value' and add
16900 * doc/invoke.texi: Document changes.
16902 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
16904 * doc/sourcebuild.texi (scalar_all_fma): Document.
16905 * tree.def (FMA_EXPR): Delete.
16906 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
16907 * internal-fn.c (ternary_direct): New macro.
16908 (expand_ternary_optab_fn): Likewise.
16909 (direct_ternary_optab_supported_p): Likewise.
16910 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
16911 * builtins.c (fold_builtin_fma): Delete.
16912 (fold_builtin_3): Don't call it.
16913 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
16914 * expr.c (expand_expr_real_2): Likewise.
16915 * fold-const.c (operand_equal_p): Likewise.
16916 (fold_ternary_loc): Likewise.
16917 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
16918 * gimple.c (DEFTREECODE): Likewise.
16919 * gimplify.c (gimplify_expr): Likewise.
16920 * optabs-tree.c (optab_for_tree_code): Likewise.
16921 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
16922 * tree-eh.c (operation_could_trap_p): Likewise.
16923 (stmt_could_throw_1_p): Likewise.
16924 * tree-inline.c (estimate_operator_cost): Likewise.
16925 * tree-pretty-print.c (dump_generic_node): Likewise.
16926 (op_code_prio): Likewise.
16927 * tree-ssa-loop-im.c (stmt_cost): Likewise.
16928 * tree-ssa-operands.c (get_expr_operands): Likewise.
16929 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
16930 * fold-const-call.h (fold_fma): Delete.
16931 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
16932 CFN_FNMA and CFN_FNMS.
16933 (fold_fma): Delete.
16934 * genmatch.c (combined_fn): New enum.
16935 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
16936 (commutative_op): New function.
16937 (commutate): Use it. Handle more than 2 operands.
16938 (dt_operand::gen_gimple_expr): Use commutative_op.
16939 (parser::parse_expr): Allow :c to be used with non-binary
16940 operators if the commutative operand is known.
16941 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
16942 CFN_FMS, CFN_FNMA and CFN_FNMS.
16943 (backprop::process_assign_use): Remove FMA_EXPR handling.
16944 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
16945 (gen_hsa_fma): New function.
16946 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
16947 IFN_FNMA and IFN_FNMS.
16948 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
16949 * gimple-fold.h (follow_all_ssa_edges): Declare.
16950 * gimple-fold.c (follow_all_ssa_edges): New function.
16951 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
16952 gimple_build interface and use follow_all_ssa_edges to fold the result.
16953 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
16954 instead of checking for optabs directly.
16955 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
16956 rather than FMA_EXPRs.
16957 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
16958 call to IFN_FMA instead of an FMA_EXPR.
16960 2018-05-17 Jim Wilson <jimw@sifive.com>
16962 * expr.c (do_tablejump): When converting index to Pmode, if we have a
16963 sign extended promoted subreg, and the range does not have the sign bit
16964 set, then do a sign extend.
16966 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
16967 test, check for sign extended subreg and/or constant operands, and
16968 do a sign extend in that case.
16970 2018-05-17 Steve Ellcey <sellcey@cavium.com>
16972 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
16973 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
16975 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
16976 Change logics_shift_reg to logics_shift_imm.
16977 (thunderx2t99_fp_loadpair_basic): Delete.
16978 (thunderx2t99_fp_storepair_basic): Delete.
16979 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
16980 (thunderx2t99_asimd_polynomial): Delete.
16981 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
16982 and neon_fp_mul_d_scalar_q.
16983 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
16984 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
16985 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
16986 (thunderx2t99_asimd_lut): Add missing tbl types.
16987 (thunderx2t99_asimd_ext): Delete.
16988 (thunderx2t99_asimd_load1_1_mult): Delete.
16989 (thunderx2t99_asimd_load1_2_mult): Delete.
16990 (thunderx2t99_asimd_load1_ldp): New.
16991 (thunderx2t99_asimd_load1): New.
16992 (thunderx2t99_asimd_load2): Add missing *load2* types.
16993 (thunderx2t99_asimd_load3): New.
16994 (thunderx2t99_asimd_load4): New.
16995 (thunderx2t99_asimd_store1_1_mult): Delete.
16996 (thunderx2t99_asimd_store1_2_mult): Delete.
16997 (thunderx2t99_asimd_store2_mult): Delete.
16998 (thunderx2t99_asimd_store2_onelane): Delete.
16999 (thunderx2t99_asimd_store_stp): New.
17000 (thunderx2t99_asimd_store1): New.
17001 (thunderx2t99_asimd_store2): New.
17002 (thunderx2t99_asimd_store3): New.
17003 (thunderx2t99_asimd_store4): New.
17005 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
17007 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
17008 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
17010 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
17011 Segher Boessenkool <segher@kernel.crashing.org>
17014 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
17017 2018-05-17 Richard Biener <rguenther@suse.de>
17019 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
17020 for pruning loop and prune defs feeding only already visited PHIs.
17022 2018-05-17 Richard Biener <rguenther@suse.de>
17024 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
17026 2018-05-17 Bin Cheng <bin.cheng@arm.com>
17027 Richard Biener <rguenther@suse.de>
17029 PR tree-optimization/85793
17030 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
17031 for VMAT_ELEMENTWISE.
17033 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
17035 * internal-fn.h (lookup_internal_fn): Declare
17036 * internal-fn.c (lookup_internal_fn): New function.
17037 * gimple.c (gimple_build_call_from_tree): Handle calls to
17038 internal functions.
17039 * gimple-pretty-print.c (dump_gimple_call): Print "." before
17040 internal function names.
17041 * tree-pretty-print.c (dump_generic_node): Likewise.
17042 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
17044 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
17046 * gimple-fold.h (gimple_build): Make the function forms take
17047 combined_fn rather than built_in_function.
17048 (gimple_simplify): Likewise.
17049 * gimple-match-head.c (gimple_simplify): Likewise.
17050 * gimple-fold.c (gimple_build): Likewise.
17051 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
17052 rather than gimple_build_call_internal.
17053 (get_initial_defs_for_reduction): Likewise.
17054 (vect_create_epilog_for_reduction): Likewise.
17055 (vectorizable_live_operation): Likewise.
17057 2018-05-17 Martin Liska <mliska@suse.cz>
17059 * gimple-ssa-sprintf.c (format_directive): Do not use
17060 space in between 'G_' and '('.
17062 2018-05-17 Jakub Jelinek <jakub@redhat.com>
17065 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
17066 even if the mask is not all ones.
17069 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
17071 (ix86_gimple_fold_builtin): Likewise.
17074 * config/i386/i386.c: Include tree-vector-builder.h.
17075 (ix86_vector_shift_count): New function.
17076 (ix86_fold_builtin): Fold shift builtins by scalar count.
17077 (ix86_gimple_fold_builtin): Likewise.
17079 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
17080 _mm512_setzero): New intrinsics.
17082 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
17083 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17085 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
17086 code generation for cases where splatting a value is not useful.
17087 * simplify-rtx.c (simplify_ternary_operation): Simplify
17088 vec_merge across a vec_duplicate and a paradoxical subreg forming
17089 a vector mode to a vec_concat.
17091 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
17093 * config.gcc: Support "goldmont-plus".
17094 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
17096 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17097 PROCESSOR_GOLDMONT_PLUS.
17098 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
17099 (processor_target_table): Add "goldmont-plus".
17100 (PTA_GOLDMONT_PLUS): Define.
17101 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
17102 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
17103 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
17104 (fold_builtin_cpu): Add "goldmont-plus".
17105 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
17106 (ix86_option_override_internal): Add "goldmont-plus".
17107 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
17108 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
17109 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
17110 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
17112 2018-05-17 Richard Biener <rguenther@suse.de>
17114 PR tree-optimization/85757
17115 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
17116 remove defs that only feed that PHI from further processing.
17118 2018-05-16 Jim Wilson <jimw@sifive.com>
17120 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
17122 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
17124 2018-05-16 Mark Wielaard <mark@klomp.org>
17126 * dwarf2out.c (count_index_strings): New function.
17127 (output_indirect_strings): Call count_index_strings and generate
17128 header for dwarf_version >= 5.
17130 2018-05-16 Mark Wielaard <mark@klomp.org>
17132 * dwarf2out.c (dwarf_FORM): New function.
17133 (set_indirect_string): Use dwarf_FORM.
17134 (reset_indirect_string): Likewise.
17135 (size_of_die): Likewise.
17136 (value_format): Likewise.
17137 (output_die): Likewise.
17138 (add_skeleton_AT_string): Likewise.
17139 (output_macinfo_op): Likewise.
17140 (index_string): Likewise.
17141 (output_index_string_offset): Likewise.
17142 (output_index_string): Likewise.
17143 (count_index_strings): Likewise.
17145 2018-05-16 Carl Love <cel@us.ibm.com>
17147 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
17148 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
17150 2018-05-16 Martin Jambor <mjambor@suse.cz>
17152 * ipa-prop.c (ipa_free_all_edge_args): Remove.
17153 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
17155 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
17157 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
17158 (fnma<mode>4): Likewise.
17159 (fms<mode>4): Likewise.
17160 (fnms<mode>4): Likewise.
17161 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
17162 (aarch64_fnma<mode>4): Likewise.
17163 (aarch64_fms<mode>4): Likewise.
17164 (aarch64_fnms<mode>4): Likewise.
17165 (aarch64_fnmadd<mode>4): Likewise.
17167 2018-05-16 Jason Merrill <jason@redhat.com>
17169 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
17171 2018-05-16 Richard Biener <rguenther@suse.de>
17173 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
17174 (dump_stmt_cost): Declare.
17175 (add_stmt_cost): Dump cost we add.
17176 (add_stmt_costs): New function.
17177 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
17178 No longer exported.
17179 (vect_analyze_stmt): Adjust prototype.
17180 (vectorizable_condition): Likewise.
17181 (vectorizable_live_operation): Likewise.
17182 (vectorizable_reduction): Likewise.
17183 (vectorizable_induction): Likewise.
17184 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
17185 cost vector to pass to vectorizable_ and record afterwards.
17186 (vect_model_reduction_cost): Take cost vector argument and adjust.
17187 (vect_model_induction_cost): Likewise.
17188 (vectorizable_reduction): Likewise.
17189 (vectorizable_induction): Likewise.
17190 (vectorizable_live_operation): Likewise.
17191 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
17192 SLP_TREE_NUMBER_OF_VEC_STMTS.
17193 (vect_analyze_slp_cost_1): Remove.
17194 (vect_analyze_slp_cost): Likewise.
17195 (vect_slp_analyze_node_operations): Take visited args and
17196 a target cost vector. Avoid processing already visited stmt sets.
17197 (vect_slp_analyze_operations): Use a local cost vector to gather
17198 costs and register those of non-discarded instances.
17199 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
17200 (vect_schedule_slp_instance): Remove copying of
17201 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
17203 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
17204 adding cost. Record cost entry location.
17205 (vect_prologue_cost_for_slp_op): Function to compute cost of
17206 a constant or invariant generated for SLP vect in the prologue,
17207 split out from vect_analyze_slp_cost_1.
17208 (vect_model_simple_cost): Make static. Adjust for SLP costing.
17209 (vect_model_promotion_demotion_cost): Likewise.
17210 (vect_model_store_cost): Likewise, make static.
17211 (vect_model_load_cost): Likewise.
17212 (vectorizable_bswap): Add cost vector arg and adjust.
17213 (vectorizable_call): Likewise.
17214 (vectorizable_simd_clone_call): Likewise.
17215 (vectorizable_conversion): Likewise.
17216 (vectorizable_assignment): Likewise.
17217 (vectorizable_shift): Likewise.
17218 (vectorizable_operation): Likewise.
17219 (vectorizable_store): Likewise.
17220 (vectorizable_load): Likewise.
17221 (vectorizable_condition): Likewise.
17222 (vectorizable_comparison): Likewise.
17223 (can_vectorize_live_stmts): Likewise.
17224 (vect_analyze_stmt): Likewise.
17225 (vect_transform_stmt): Adjust calls to vectorizable_*.
17226 * tree-vectorizer.c: Include gimple-pretty-print.h.
17227 (dump_stmt_cost): New function.
17229 2018-05-16 Richard Biener <rguenther@suse.de>
17231 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
17232 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
17233 * tree-ssa-dse.c: Include tree-ssa-loop.h.
17234 (check_name): New callback.
17235 (dse_classify_store): Track cycles via a visited bitmap of PHI
17236 defs and simplify handling of in-loop and across loop dead stores
17237 and properly fail for loop-variant refs. Handle byte-tracking with
17238 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
17241 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
17243 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
17244 (vect_get_mask_type_for_stmt): Likewise.
17245 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
17247 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
17248 to determine the statement's vector type and the vector type that
17249 should be used for calculating nunits. Deal with cases in which
17250 the type has to be deferred.
17251 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
17252 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
17253 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
17254 (vect_determine_vf_for_stmt): New functions, split out from...
17255 (vect_determine_vectorization_factor): ...here.
17256 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
17257 (vect_get_mask_type_for_stmt): New functions, split out from
17258 vect_determine_vectorization_factor.
17260 2018-05-16 Richard Biener <rguenther@suse.de>
17262 * tree-cfg.c (verify_gimple_assign_ternary): Properly
17263 verify the [VEC_]COND_EXPR embedded comparison.
17265 2018-05-15 Martin Sebor <msebor@redhat.com>
17267 PR tree-optimization/85753
17268 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
17269 RECORD_TYPE in addition to ARRAY_TYPE.
17271 2018-05-15 Martin Sebor <msebor@redhat.com>
17273 PR middle-end/85643
17274 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
17276 2018-05-15 Richard Biener <rguenther@suse.de>
17278 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
17279 add by_clobber_p one. Change algorithm to collect all defs
17280 representing uses we need to walk and try reducing them to
17281 a single one before failing.
17282 (dse_dom_walker::dse_optimize_stmt): Adjust.
17284 2018-05-13 Mark Wielaard <mark@klomp.org>
17286 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
17287 (size_of_loc_descr): Likewise.
17288 (output_loc_operands): Likewise.
17289 (output_loc_operands_raw): Likewise.
17290 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
17291 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
17292 (hash_loc_operands): Likewise.
17293 (compare_loc_operands): Likewise.
17295 2018-05-14 Mark Wielaard <mark@klomp.org>
17297 * dwarf2out.c (count_index_addrs): New function.
17298 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
17300 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17302 PR tree-optimization/83648
17303 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
17304 return value as malloc candidate.
17306 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17309 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
17310 param as true in call to suggest_attribute.
17312 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
17314 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
17315 -mreadonly-in-sdata.
17317 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17319 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
17321 (aarch64_crypto_aesd_fused): Likewise.
17323 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
17325 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
17326 (movsi_aarch64): Likewise.
17327 (load_pairsi): Likewise.
17328 (load_pairdi): Likewise.
17329 (store_pairsi): Likewise.
17330 (store_pairdi): Likewise.
17331 (load_pairsf): Likewise.
17332 (load_pairdf): Likewise.
17333 (store_pairsf): Likewise.
17334 (store_pairdf): Likewise.
17335 (zero_extend): Likewise.
17336 (trunc): Swap alternatives.
17337 (fcvt_target): Add '?' to prefer w over r.
17339 2018-05-14 Jakub Jelinek <jakub@redhat.com>
17342 * config/i386/i386.md: Disallow non-commutative arithmetics in
17343 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
17344 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
17345 in the peephole2 before it.
17347 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
17349 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
17350 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
17351 (ix86_handle_option): Handle -mcldemote.
17352 * config.gcc: New header.
17353 * config/i386/cldemoteintrin.h: New file.
17354 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
17355 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
17357 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17358 OPTION_MASK_ISA_CLDEMOTE.
17359 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
17360 (ix86_valid_target_attribute_inner_p): Ditto.
17361 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
17362 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
17363 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
17364 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
17365 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
17367 * config/i386/i386.opt: Add -mcldemote.
17368 * config/i386/x86intrin.h: New header.
17369 * doc/invoke.texi: Add -mcldemote.
17371 2018-05-14 Richard Biener <rguenther@suse.de>
17373 * doc/match-and-simplify.texi: Adjust :s documentation.
17375 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
17377 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
17378 intended memcpy size.
17379 (REORDER_45): Likewise.
17381 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
17383 * sort.cc: New file.
17384 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
17385 * vec.c (qsort_chk): Use gcc_qsort.
17386 * Makefile.in (OBJS-libcommon): Add sort.o.
17387 (build/sort.o): New target. Use it...
17388 (BUILD_RTL): ... here, and...
17389 (build/gencfn-macros): ... here, and...
17390 (build/genmatch): ... here.
17392 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
17393 Chung-Ju Wu <jasonwucj@gmail.com>
17395 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
17396 * config/nds32/nds32-graywolf.md: New file.
17397 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
17398 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
17400 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
17401 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
17402 * config/nds32/nds32.md (pipeline_model): Add graywolf.
17403 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
17404 * config/nds32/pipelines.md: Include n15 settings.
17406 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
17407 Chung-Ju Wu <jasonwucj@gmail.com>
17409 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
17410 * config/nds32/nds32-n13.md: New file.
17411 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
17412 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
17414 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
17415 * config/nds32/nds32.md (pipeline_model): Add n13.
17416 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
17417 * config/nds32/pipelines.md: Include n13 settings.
17419 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
17420 Chung-Ju Wu <jasonwucj@gmail.com>
17422 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
17423 * config/nds32/nds32-n10.md: New file.
17424 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
17425 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
17427 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
17428 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
17429 * config/nds32/nds32.md (pipeline_model): Add n10.
17430 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
17431 * config/nds32/pipelines.md: Include n10 settings.
17433 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
17434 Kito Cheng <kito.cheng@gmail.com>
17435 Chung-Ju Wu <jasonwucj@gmail.com>
17437 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
17438 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
17439 Add enum values for DSP extension instructions.
17440 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
17442 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
17443 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
17444 New code iterators.
17445 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
17446 * config/nds32/nds32-dspext.md: New file for DSP implementation.
17447 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
17448 * config/nds32/nds32-intrinsic.md: Likewise.
17449 * config/nds32/nds32_intrinsic.h: Likewise.
17450 * config/nds32/nds32-md-auxiliary.c: Likewise.
17451 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
17452 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
17453 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
17454 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
17455 * config/nds32/nds32-protos.h: New declarations for DSP extension.
17456 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
17457 TYPE_DMAC in switch statement.
17458 * config/nds32/nds32.c: New checking and implementation for DSP
17459 extension instructions.
17460 * config/nds32/nds32.h: Likewise.
17461 * config/nds32/nds32.md: Likewise.
17462 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
17463 * config/nds32/predicates.md: Implement new predicates for DSP
17466 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
17468 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
17469 Reformat alternatives and attributes so it is easier to identify
17470 which constraints/attributes go with which instruction.
17471 (mov<mode>_hardfloat32, FMOVE64): Likewise.
17472 (mov<mode>_softfloat32, FMOVE64): Likewise.
17473 (mov<mode>_hardfloat64, FMOVE64): Likewise.
17474 (mov<mode>_softfloat64, FMOVE64): Likewise.
17476 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
17478 * doc/extend.texi (PowerPC Built-in Functions): Rename this
17480 (Basic PowerPC Built-in Functions): The new name of the
17481 subsection previously known as "PowerPC Built-in Functions".
17482 (Basic PowerPC Built-in Functions Available on all Configurations):
17484 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
17485 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
17486 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
17487 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
17489 2018-05-11 Martin Jambor <mjambor@suse.cz>
17492 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
17495 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
17498 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
17500 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
17502 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
17503 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
17504 (ix86_handle_option): Handle -mwaitpkg.
17505 * config.gcc: New header.
17506 * config/i386/cpuid.h (bit_WAITPKG): New bit.
17507 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
17508 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
17510 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17511 OPTION_MASK_ISA_WAITPKG.
17512 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
17513 (ix86_option_override_internal): Add PTA_WAITPKG.
17514 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
17515 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
17516 IX86_BUILTIN_TPAUSE.
17517 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
17518 __builtin_ia32_umwait and __builtin_ia32_tpause.
17519 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
17520 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
17521 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
17522 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
17523 UNSPECV_TPAUSE): New.
17524 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
17525 * config/i386/i386.opt: Add -mwaitpkg.
17526 * config/i386/waitpkgintrin.h: New file.
17527 * config/i386/x86intrin.h: New header.
17528 * doc/invoke.texi: Add -mwaitpkg.
17530 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
17533 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
17535 (cortex-m0): Use armv6s-m isa.
17536 (cortex-m0plus): Likewise.
17537 (cortex-m1): Likewise.
17538 (cortex-m0.small-multiply): Likewise.
17539 (cortex-m0plus.small-multiply): Likewise.
17540 (cortex-m1.small-multiply): Likewise.
17542 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
17543 Jakub Jelinek <jakub@redhat.com>
17545 PR tree-optimization/85692
17546 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
17547 source permute as well.
17549 2018-05-11 Martin Liska <mliska@suse.cz>
17552 * doc/extend.texi: Document LLVM style format for no_sanitize
17555 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
17557 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
17558 mode_supports_vsx_dform_quad to mode_supports_dq_form.
17559 (mode_supports_vsx_dform_quad): Likewise.
17560 (mode_supports_vmx_dform): Move these functions to be next to the
17561 other mode_supports functions.
17562 (mode_supports_dq_form): Likewise.
17563 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
17564 mode_supports_dq_form.
17565 (reg_offset_addressing_ok_p): Likewise.
17566 (offsettable_ok_by_alignment): Likewise.
17567 (rs6000_legitimate_offset_address_p): Likewise.
17568 (legitimate_lo_sum_address_p): Likewise.
17569 (rs6000_legitimize_address): Likewise.
17570 (rs6000_legitimize_reload_address): Likewise.
17571 (rs6000_secondary_reload_inner): Likewise.
17572 (rs6000_preferred_reload_class): Likewise.
17573 (rs6000_output_move_128bit): Likewise.
17575 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
17577 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
17578 Generate SImode target register for null target.
17579 <case IX86_BUILTIN_XGETBV>: Ditto.
17580 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
17581 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
17583 2018-05-10 Carl Love <cel@us.ibm.com>
17585 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
17586 dcbtt and dcbtstt if operands[2] is 0.
17588 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
17591 * config/i386/sse.md (usadv64qi): New expander.
17593 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
17595 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
17596 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
17597 -maltivec=be support.
17598 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
17599 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
17600 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
17601 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
17602 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
17603 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
17604 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
17605 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
17606 altivec_vsumsws): Adjust.
17607 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
17608 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
17609 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
17611 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
17612 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
17613 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
17614 (altivec_lve<VI_char>x): Delete expand.
17615 (*altivec_lve<VI_char>x_internal): Rename to...
17616 (altivec_lve<VI_char>x): ... this.
17617 (altivec_lvxl_<mode>): Delete expand.
17618 (*altivec_lvxl_<mode>_internal): Rename to ...
17619 (altivec_lvxl_<mode>): ... this.
17620 (altivec_stvxl_<mode>): Delete expand.
17621 (*altivec_stvxl_<mode>_internal): Rename to ...
17622 (altivec_stvxl_<mode>): ... this.
17623 (altivec_stve<VI_char>x): Delete expand.
17624 (*altivec_stve<VI_char>x_internal): Rename to ...
17625 (altivec_stve<VI_char>x): ... this.
17626 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
17627 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
17628 reduc_plus_scal_<mode>): Adjust.
17629 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
17631 (rs6000_cpu_cpp_builtins): Adjust.
17632 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
17633 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
17634 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
17635 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
17636 -maltivec=be support.
17637 (rs6000_split_vec_extract_var): Adjust.
17638 (rs6000_split_v4si_init): Adjust.
17639 (swap_selector_for_mode): Delete.
17640 (altivec_expand_lvx_be, altivec_expand_stvx_be,
17641 altivec_expand_stvex_be): Delete.
17642 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
17643 -maltivec=be support.
17644 (rs6000_gimple_fold_builtin): Ditto.
17645 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
17647 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
17648 (TARGET_DIRECT_MOVE_64BIT): Adjust.
17649 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
17650 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
17651 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
17652 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
17653 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
17654 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
17655 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
17656 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
17657 anonymous split): Adjust.
17658 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
17659 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
17661 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
17663 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
17664 when --with-gxx-include-dir is also specified.
17665 * configure: Regenerate.
17667 2018-05-09 Jim Wilson <jimw@sifive.com>
17670 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
17671 * config/riscv/t-withmultilib: New.
17672 * config/riscv/withmultilib.h: New.
17673 * doc/install.texi: Document RISC-V --with-multilib-list support.
17675 2018-05-09 Richard Biener <rguenther@suse.de>
17677 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
17679 (vect_bb_vectorization_profitable_p): Adjust. Compute
17680 actual scalar cost using the cost vector and the add_stmt_cost
17683 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
17685 PR rtl-optimization/85645
17686 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
17687 in the REG_CFA_REGISTER note for LR, don't leave it empty.
17689 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
17691 PR rtl-optimization/85645
17692 * shrink-wrap.c (spread_components): Return a boolean saying if
17693 anything was changed.
17694 (try_shrink_wrapping_separate): Iterate spread_components until
17695 nothing changes anymore.
17697 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
17699 PR rtl-optimization/85645
17700 * regrename.c (build_def_use): Also kill the chains that include the
17701 destination of a REG_CFA_REGISTER note.
17703 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
17705 PR rtl-optimization/85645
17706 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
17707 insn that has a REG_CFA_REGISTER note.
17709 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
17711 * cfgexpand.c (expand_clobber): New function.
17712 (expand_gimple_stmt_1): Use it.
17713 * tree-vect-stmts.c (vect_clobber_variable): New function,
17715 (vectorizable_simd_clone_call): ...here.
17716 (vectorizable_store): Emit a clobber either side of an
17717 IFN_STORE_LANES sequence.
17718 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
17720 2018-05-09 Tom de Vries <tom@codesourcery.com>
17723 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
17724 (define_insn "trap_if_false"): Add exit after trap.
17726 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
17728 PR rtl-optimization/85638
17729 * bb-reorder.c: Include common/common-target.h.
17730 (create_forwarder_block): New function extracted from...
17731 (fix_up_crossing_landing_pad): ...here. Rename into...
17732 (dw2_fix_up_crossing_landing_pad): ...this.
17733 (sjlj_fix_up_crossing_landing_pad): New function.
17734 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
17735 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
17736 from both partitions and exit the loop after one iteration.
17738 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
17741 * doc/extend.texi (PowerPC Built-in Functions): Rename this
17743 (Basic PowerPC Built-in Functions): The new name of the
17744 subsection previously known as "PowerPC Built-in Functions".
17745 (Basic PowerPC Built-in Functions Available on all Configurations):
17747 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
17749 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
17751 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
17753 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
17756 2018-05-08 Jim Wilson <jimw@sifive.com>
17758 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
17759 (LD_EMUL_SUFFIX): New.
17760 (LINK_SPEC): Use it.
17762 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
17764 * doc/extend.texi (PowerPC Built-in Functions): Rename this
17766 (Basic PowerPC Built-in Functions): The new name of the
17767 subsection previously known as "PowerPC Built-in Functions".
17768 (Basic PowerPC Built-in Functions Available on all Configurations):
17770 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
17772 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
17774 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
17776 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
17779 2018-05-08 Jakub Jelinek <jakub@redhat.com>
17782 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
17783 after cmpelim optimization.
17785 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
17787 * config.gcc: Support "goldmont".
17788 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
17789 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17790 PROCESSOR_GOLDMONT.
17791 * config/i386/i386.c (m_GOLDMONT): Define.
17792 (processor_target_table): Add "goldmont".
17793 (PTA_GOLDMONT): Define.
17794 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
17795 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
17796 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
17797 (fold_builtin_cpu): Add "goldmont".
17798 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
17799 (ix86_option_override_internal): Add "goldmont".
17800 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
17801 (processor_type): Add PROCESSOR_GOLDMONT.
17802 * config/i386/i386.md: Add CPU "glm".
17803 * config/i386/glm.md: New file.
17804 * config/i386/x86-tune.def: Add m_GOLDMONT.
17805 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
17807 2018-05-08 Jakub Jelinek <jakub@redhat.com>
17810 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
17812 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
17813 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
17814 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
17817 * config/i386/i386.c (ix86_fold_builtin): Handle
17818 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
17821 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
17822 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
17824 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
17827 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
17828 (check_arch): Likewise.
17829 (check_fpu): Return the result rather than printing it.
17830 (end arch): Fix operator precedence.
17831 (end cpu): Likewise.
17832 (END): Print the result from check_fpu.
17834 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
17835 Alan Hayward <alan.hayward@arm.com>
17836 David Sherwood <david.sherwood@arm.com>
17838 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
17839 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
17840 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
17841 (*fcmuo<mode>_and): New patterns.
17843 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
17845 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
17846 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
17847 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
17848 (cmp_op, sve_imm_con): New code attributes.
17849 (SVE_COND_INT_CMP, imm_con): Delete.
17850 (cmp_op): Remove above unspecs from int attribute.
17851 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
17853 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
17854 comparison-specific unspecs.
17855 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
17856 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
17857 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
17858 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
17859 (*vec_fcm<cmp_op><mode>): Rename to...
17860 (*fcm<cmp_op><mode>): ...this and adjust likewise.
17861 (*vec_fcmuo<mode>): Rename to...
17862 (*fcmuo<mode>): ...this and adjust likewise.
17863 (*pred_fcm<cmp_op><mode>): New pattern.
17864 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
17865 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
17867 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
17869 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
17870 (aarch64_emit_sve_predicated_cond): New function.
17871 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
17872 (aarch64_emit_unspec_cond_or): Replace with...
17873 (aarch64_emit_sve_or_conds): ...this new function. Use
17874 aarch64_emit_sve_ptrue_op for the individual comparisons and
17875 aarch64_emit_binop to OR them together.
17876 (aarch64_emit_inverted_unspec_cond): Replace with...
17877 (aarch64_emit_sve_inverted_cond): ...this new function. Use
17878 aarch64_emit_sve_ptrue_op for the comparison and
17879 aarch64_emit_unop to invert the result.
17880 (aarch64_expand_sve_vec_cmp_float): Update after the above
17881 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
17883 2018-05-07 Nathan Sidwell <nathan@acm.org>
17885 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
17886 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
17887 (Backwards Compatibility): Likewise.
17889 2018-05-07 Luis Machado <luis.machado@linaro.org>
17893 2018-05-07 Luis Machado <luis.machado@linaro.org>
17895 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
17896 <prefetch_dynamic_strides>: New const bool field.
17897 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
17898 prefetch_dynamic_strides.
17899 (exynosm1_prefetch_tune): Likewise.
17900 (thunderxt88_prefetch_tune): Likewise.
17901 (thunderx_prefetch_tune): Likewise.
17902 (thunderx2t99_prefetch_tune): Likewise.
17903 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
17905 (aarch64_override_options_internal): Update to set
17906 PARAM_PREFETCH_DYNAMIC_STRIDES.
17907 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
17908 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
17909 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
17910 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
17911 prefetch-dynamic-strides setting.
17913 2018-05-07 Luis Machado <luis.machado@linaro.org>
17915 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
17916 <minimum_stride>: New const int field.
17917 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
17918 minimum_stride field.
17919 (exynosm1_prefetch_tune): Likewise.
17920 (thunderxt88_prefetch_tune): Likewise.
17921 (thunderx_prefetch_tune): Likewise.
17922 (thunderx2t99_prefetch_tune): Likewise.
17923 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
17924 (aarch64_override_options_internal): Update to set
17925 PARAM_PREFETCH_MINIMUM_STRIDE.
17926 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
17927 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
17928 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
17929 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
17930 stride is constant and is below the minimum stride threshold.
17932 2018-05-07 Luis Machado <luis.machado@linaro.org>
17934 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
17937 2018-05-07 Luis Machado <luis.machado@linaro.org>
17939 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
17940 <prefetch_dynamic_strides>: New const bool field.
17941 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
17942 prefetch_dynamic_strides.
17943 (exynosm1_prefetch_tune): Likewise.
17944 (thunderxt88_prefetch_tune): Likewise.
17945 (thunderx_prefetch_tune): Likewise.
17946 (thunderx2t99_prefetch_tune): Likewise.
17947 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
17949 (aarch64_override_options_internal): Update to set
17950 PARAM_PREFETCH_DYNAMIC_STRIDES.
17951 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
17952 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
17953 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
17954 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
17955 prefetch-dynamic-strides setting.
17957 2018-05-07 Luis Machado <luis.machado@linaro.org>
17959 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
17960 <minimum_stride>: New const int field.
17961 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
17962 minimum_stride field.
17963 (exynosm1_prefetch_tune): Likewise.
17964 (thunderxt88_prefetch_tune): Likewise.
17965 (thunderx_prefetch_tune): Likewise.
17966 (thunderx2t99_prefetch_tune): Likewise.
17967 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
17968 (aarch64_override_options_internal): Update to set
17969 PARAM_PREFETCH_MINIMUM_STRIDE.
17970 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
17971 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
17972 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
17973 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
17974 stride is constant and is below the minimum stride threshold.
17976 2018-05-06 Jakub Jelinek <jakub@redhat.com>
17979 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
17980 the type is addressable. Don't force op into register if it has
17983 2018-05-05 Roland McGrath <mcgrathr@google.com>
17986 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
17987 any section for which we don't know a specific type it should have,
17988 regardless of name. Previously this was done only for the exact
17989 names ".init_array", ".fini_array", and ".preinit_array".
17990 (default_elf_asm_named_section): Add comment about
17991 relationship with default_section_type_flags and SECTION_NOTYPE.
17992 (get_section): Don't consider it a type conflict if one side has
17993 SECTION_NOTYPE and the other doesn't, as long as neither has the
17994 SECTION_BSS et al used in the default_section_type_flags logic.
17996 2018-05-05 Tom de Vries <tom@codesourcery.com>
17999 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
18000 (workaround_barsyncs): New function.
18001 (nvptx_reorg): Use workaround_barsyncs.
18002 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
18003 (define_expand "nvptx_membar_cta"): New define_expand.
18004 (define_insn "*nvptx_membar_cta"): New insn.
18006 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
18008 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
18009 To improve optimization opportunities.
18010 * builtin-types.def: The new needed builtin types for the above.
18012 2018-05-04 Richard Biener <rguenther@suse.de>
18014 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
18015 * gimple-ssa-store-merging.c
18016 (imm_store_chain_info::output_merged_store): Remove redundant create,
18017 release split_store vector contents on failure.
18018 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
18019 scalar stmt vector on cache hit.
18021 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
18023 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
18025 * config.gcc (powerpc-xilinx-eabi*): Remove.
18026 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
18028 (fusion_addis_mem_combo_load): Ditto.
18029 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
18031 (rs6000_cpu_cpp_builtins): Ditto.
18032 * config/rs6000/rs6000-linux.c
18033 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
18034 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
18035 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
18037 (rs6000_setup_reg_addr_masks): Ditto.
18038 (rs6000_init_hard_regno_mode_ok): Ditto.
18039 (rs6000_option_override_internal): Ditto.
18040 (legitimate_lo_sum_address_p): Ditto.
18041 (rs6000_legitimize_address): Ditto.
18042 (rs6000_legitimize_reload_address): Ditto.
18043 (rs6000_legitimate_address_p): Ditto.
18044 (abi_v4_pass_in_fpr): Ditto.
18045 (setup_incoming_varargs): Ditto.
18046 (rs6000_gimplify_va_arg): Ditto.
18047 (rs6000_split_multireg_move): Ditto.
18048 (rs6000_savres_strategy): Ditto.
18049 (rs6000_emit_prologue_components): Ditto.
18050 (rs6000_emit_epilogue_components): Ditto.
18051 (rs6000_emit_prologue): Ditto.
18052 (rs6000_emit_epilogue): Ditto.
18053 (rs6000_elf_file_end): Ditto.
18054 (rs6000_function_value): Ditto.
18055 (rs6000_libcall_value): Ditto.
18056 * config/rs6000/rs6000.h: Ditto.
18057 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
18058 (TARGET_MINMAX): ... this. New.
18059 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
18060 * config/rs6000/rs6000.md: Remove Xilinx FP support.
18061 (*movsi_internal1_single): Delete.
18062 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
18063 mfpu=, mxilinx-fpu): Delete.
18064 * config/rs6000/singlefp.h: Delete.
18065 * config/rs6000/sysv4.h: Remove Xilinx FP support.
18066 * config/rs6000/t-rs6000: Ditto.
18067 * config/rs6000/t-xilinx: Delete.
18068 * config/rs6000/titan.md: Adjust for fp_type removal.
18069 * config/rs6000/vsx.md: Remove Xilinx FP support.
18070 (VStype_simple): Delete.
18071 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
18072 * config/rs6000/xfpu.h: Delete.
18073 * config/rs6000/xfpu.md: Delete.
18074 * config/rs6000/xilinx.h: Delete.
18075 * config/rs6000/xilinx.opt: Delete.
18076 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
18077 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
18079 2018-05-04 Tom de Vries <tom@codesourcery.com>
18082 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
18085 2018-05-04 Richard Biener <rguenther@suse.de>
18087 PR middle-end/85627
18088 * tree-complex.c (update_complex_assignment): We are always in SSA form.
18089 (expand_complex_div_wide): Likewise.
18090 (expand_complex_operations_1): Likewise.
18091 (expand_complex_libcall): Preserve EH info of the original stmt.
18092 (tree_lower_complex): Handle removed blocks.
18093 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
18094 on complex multiplication and division libcall builtins.
18096 2018-05-04 Richard Biener <rguenther@suse.de>
18098 PR middle-end/85574
18099 * fold-const.c (negate_expr_p): Restrict negation of operand
18100 zero of a division to when we know that can happen without
18102 (fold_negate_expr_1): Likewise.
18104 2018-05-04 Jakub Jelinek <jakub@redhat.com>
18107 * real.h (real_nextafter): Declare.
18108 * real.c (real_nextafter): New function.
18109 * fold-const-call.c (fold_const_nextafter): New function.
18110 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
18111 CASE_CFN_NEXTTOWARD.
18112 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
18113 even when arg1_mode is different from arg0_mode.
18115 2018-05-03 Nathan Sidwell <nathan@acm.org>
18117 * doc/extend.texi (Deprecated Features): Remove
18118 -ffriend-injection.
18119 (Backwards Compatibility): Likewise.
18120 * doc/invoke.texi (C++ Language Options): Likewise.
18121 (C++ Dialect Options): Likewise.
18123 2018-05-03 Jakub Jelinek <jakub@redhat.com>
18126 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
18127 _mm512_mask_mullox_epi64): New intrinsics.
18129 2018-05-03 Tom de Vries <tom@codesourcery.com>
18132 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
18133 dump files): Add offload-tree.
18135 2018-05-03 Richard Biener <rguenther@suse.de>
18137 PR tree-optimization/85615
18138 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
18139 to loops not nested in BBs loop father to avoid creating multi-entry
18142 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18144 PR tree-optimization/70291
18145 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
18146 arguments. Change return type to tree. Emit libcall as a new
18147 statement rather than replacing existing one when inplace_p is true.
18148 (expand_complex_multiplication_components): New function.
18149 (expand_complex_multiplication): Expand floating-point complex
18150 multiplication using the above.
18151 (expand_complex_division): Rename inner_type parameter to type.
18152 Update expand_complex_libcall call-site.
18153 (expand_complex_operations_1): Update expand_complex_multiplication
18154 and expand_complex_division call-sites.
18156 2018-05-02 Jakub Jelinek <jakub@redhat.com>
18159 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
18160 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
18161 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
18162 the highest significant bit of the shift count mask is clear. In
18163 check whether and[sq]i3 is needed verify that all significant bits
18164 of the shift count other than the highest are set.
18166 2018-05-02 Tom de Vries <tom@codesourcery.com>
18169 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
18170 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
18171 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
18172 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
18173 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
18174 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
18175 __builtin_goacc_parlevel_size.
18177 2018-05-02 Richard Biener <rguenther@suse.de>
18179 PR tree-optimization/85597
18180 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
18181 do not use split vect_get_vec_defs call but call vect_get_slp_defs
18184 2018-05-02 Tom de Vries <tom@codesourcery.com>
18187 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
18188 dump files): Add ltrans-tree.
18190 2018-05-02 Tom de Vries <tom@codesourcery.com>
18193 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
18194 dump files): Add wpa-ipa.
18196 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
18198 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
18199 powerpc*-*-linux*paired* target.
18200 * config/rs6000/750cl.h: Delete.
18201 * config/rs6000/paired.h: Delete.
18202 * config/rs6000/paired.md: Delete.
18203 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
18205 * config/rs6000/rs6000-builtin.def: Remove paired float support.
18206 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
18207 comment. Remove paired float support.
18208 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
18209 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
18211 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
18212 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
18214 * config/rs6000/rs6000.c: Remove paired float support.
18215 (paired_expand_vector_init, paired_expand_vector_move,
18216 paired_emit_vector_compare, paired_emit_vector_cond_expr,
18217 (paired_expand_lv_builtin, paired_expand_stv_builtin,
18218 paired_expand_builtin, paired_expand_predicate_builtin,
18219 paired_init_builtins): Delete.
18220 * config/rs6000/rs6000.h: Remove paired float support.
18221 * config/rs6000/rs6000.md: Remove paired float support.
18222 (move_from_CR_ov_bit): Delete.
18223 * config/rs6000/rs6000.opt (mpaired): Delete.
18224 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
18225 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
18227 2018-05-02 Richard Biener <rguenther@suse.de>
18229 PR middle-end/85567
18230 * gimplify.c (gimplify_save_expr): When in SSA form allow
18231 SAVE_EXPRs to compute to SSA vars.
18233 2018-05-02 Jakub Jelinek <jakub@redhat.com>
18236 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
18237 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
18238 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
18239 clobber operands[2], instead use a new pseudo. Formatting fixes.
18241 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
18243 PR tree-optimization/85586
18244 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
18245 exit early for statements in the same group if the accesses are
18248 2018-05-02 Tom de Vries <tom@codesourcery.com>
18251 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
18254 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
18256 PR tree-optimization/85143
18257 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
18259 2018-05-01 Tom de Vries <tom@codesourcery.com>
18262 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
18263 not found" error message.
18265 2018-05-01 Tom de Vries <tom@codesourcery.com>
18268 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
18269 * vec.c (test_ordered_remove_if): New function.
18270 (vec_c_tests): Call test_ordered_remove_if.
18271 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
18272 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
18273 * tree-vect-patterns.c (vect_pattern_recog_1): Use
18274 VEC_ORDERED_REMOVE_IF.
18276 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18278 PR tree-optimization/82665
18279 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
18280 pointer subtraction where arguments come from a memchr call.
18282 2018-05-01 Jakub Jelinek <jakub@redhat.com>
18284 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
18285 --push-state --as-needed and --pop-state instead of --as-needed and
18286 --no-as-needed if ld supports it.
18287 * configure: Regenerated.
18290 * doc/install.texi2html: Replace _002d with - and _002a with * in
18291 generated html files using sed.
18293 2018-04-30 David Malcolm <dmalcolm@redhat.com>
18296 * gcc-rich-location.c (blank_line_before_p): New function.
18297 (use_new_line): New function.
18298 (gcc_rich_location::add_fixit_insert_formatted): New function.
18299 * gcc-rich-location.h
18300 (gcc_rich_location::add_fixit_insert_formatted): New function.
18302 2018-04-30 David Malcolm <dmalcolm@redhat.com>
18304 * selftest.c (assert_streq): Rename "expected" and "actual" to
18305 "val1" and "val2". Extend NULL-handling to cover both inputs
18306 symmetrically, while still requiring both to be non-NULL for a pass.
18307 * selftest.h (assert_streq): Rename "expected" and "actual" to
18309 (ASSERT_EQ): Likewise.
18310 (ASSERT_EQ_AT): Likewise.
18311 (ASSERT_KNOWN_EQ): Likewise.
18312 (ASSERT_KNOWN_EQ_AT): Likewise.
18313 (ASSERT_NE): Likewise.
18314 (ASSERT_MAYBE_NE): Likewise.
18315 (ASSERT_MAYBE_NE_AT): Likewise.
18316 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
18317 the assertion to pass.
18318 (ASSERT_STREQ_AT): Likewise.
18320 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
18322 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
18323 interaction with -pie.
18325 2018-04-30 David Malcolm <dmalcolm@redhat.com>
18327 * selftest.h: Fix alphabetization of per-source-file selftest
18330 2018-04-30 Jason Merrill <jason@redhat.com>
18332 PR c++/61982 - dead stores to destroyed objects.
18333 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
18336 2018-04-30 Jason Merrill <jason@redhat.com>
18338 * tree.c (build_clobber): New.
18339 * tree.h: Declare it.
18340 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
18342 2018-04-30 David Malcolm <dmalcolm@redhat.com>
18344 * diagnostic-show-locus.c (layout::layout): Update for
18345 location_get_source_line returning a char_span.
18346 (struct char_span): Move to input.h.
18347 (struct correction): Update for fields in char_span becoming
18349 (struct source_line): Update for location_get_source_line
18350 returning a char_span.
18351 (layout::print_line): Likewise.
18352 * edit-context.c (edited_file::print_content): Likewise.
18353 (edited_file::print_diff_hunk): Likewise.
18354 (edited_file::print_run_of_changed_lines): Likewise.
18355 (edited_file::get_num_lines): Likewise.
18356 (edited_line::edited_line): Likewise.
18357 * final.c (asm_show_source): Likewise.
18358 * input.c (location_get_source_line): Convert return type
18359 from const char * to char_span, losing the final "line_len"
18361 (dump_location_info): Update for the above.
18362 (get_substring_ranges_for_loc): Likewise. Use a char_span
18363 when handling the literal within the line.
18364 (test_reading_source_line): Update for location_get_source_line
18365 returning a char_span.
18366 * input.h (class char_span): Move here from
18367 diagnostic-show-locus.c, converting from a struct to a class.
18368 Make data members private.
18369 (char_span::operator bool): New.
18370 (char_span::length): New.
18371 (char_span::get_buffer): New.
18372 (char_span::operator[]): New.
18373 (char_span::subspan): Make const.
18374 (char_span::xstrdup): New.
18375 (location_get_source_line): Convert return type from const char *
18376 to char_span, losing the final "line_size" param.
18378 2018-04-30 Jan Hubicka <jh@suse.cz>
18380 * lto-wrapper.c (ltrans_priorities): New static var.
18381 (cmp_priority): New.
18382 (run_gcc): Read priorities and if doing parallel build order
18383 the Makefile by them.
18385 2018-04-30 David Malcolm <dmalcolm@redhat.com>
18387 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
18389 2018-04-30 Richard Biener <rguenther@suse.de>
18391 * tree-cfg.c (verify_address): Remove base argument, add
18392 flag whether to check TREE_ADDRESSABLE and do that.
18393 (verify_expr): Remove.
18394 (verify_types_in_gimple_reference): Add pieces from verify_expr.
18395 (verify_gimple_assign_single): Likewise.
18396 (verify_gimple_switch): Likewise.
18397 (verify_expr_location_1): Dereference tp once. Add (disabled)
18398 piece from verify_expr.
18399 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
18401 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
18403 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
18405 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
18407 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
18408 (small_data_pattern): Likewise.
18409 (arc_rewrite_small_data): Likewise.
18410 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
18411 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
18412 (get_symbol_alignment): New function.
18413 (legitimate_small_data_address_p): Likewise.
18414 (legitimate_scaled_address): Update, call
18415 legitimate_small_data_address_p.
18416 (output_sdata): New static variable.
18417 (arc_print_operand): Update how we handle small data operands.
18418 (arc_print_operand_address): Likewise.
18419 (arc_legitimate_address_p): Update, use
18420 legitimate_small_data_address_p.
18421 (arc_rewrite_small_data_p): Remove.
18422 (arc_rewrite_small_data_1): Likewise.
18423 (arc_rewrite_small_data): Likewise.
18424 (small_data_pattern): Likewise.
18425 (compact_sda_memory_operand): Update to use
18426 legitimate_small_data_address_p and get_symbol_alignment.
18427 (prepare_move_operands): Don't rewite sdata pattern.
18428 (prepare_extend_operands): Remove.
18429 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
18431 (zero_extendqisi2): Likewise.
18432 (zero_extendhisi2): Likewise.
18433 (extendqihi2): Likewise.
18434 (extendqisi2): Likewise.
18435 (extendhisi2): Likewise.
18436 (addsi3): Likewise.
18437 (subsi3): Likewise.
18438 (andsi3): Likewise.
18439 * config/arc/constraints.md (Usd): Change it to memory constraint.
18441 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
18443 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
18444 as source of std instructions.
18445 * config/arc/arc.md (movsi_insn): Update pattern predicate to
18446 allow 6-bit constants as source for store instructions.
18447 (movdi_insn): Update instruction pattern to allow 6-bit constants
18448 as source for store instructions.
18450 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
18452 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
18454 2018-04-30 Nathan Sidwell <nathan@acm.org>
18455 Sandra Loosemore <sandra@codesourcery.com>
18457 * dumpfile.c (dump_open): Allow '-' for stdout.
18458 * doc/invoke.texi (Developer Options): Document dump filename
18459 determination early. Document stdin/stdout selection.
18461 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
18463 Microblaze Target: PIC data text relative
18465 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
18466 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
18468 * config/microblaze/microblaze.h (microblaze_constant_address_p):
18469 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
18470 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
18471 New addressing mode for data-text relative position indepenedent code.
18472 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
18473 'ADDRESS_SYMBOLIC_TXT_REL'.
18474 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
18475 (microblaze_legitimate_pic_operand): Exclude function calls from
18476 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
18477 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
18479 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
18480 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
18481 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
18482 for 'address + offset'.
18483 (microblaze_expand_prologue): Add new function prologue call for
18485 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
18486 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
18487 table in case of TARGET_PIC_DATA_TEXT_REL.
18488 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
18489 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
18490 Add new macros 'UNSPEC_TEXT',
18491 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
18492 + exclude function calls from 'UNSPEC_PLT' in case of data text
18494 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
18495 new target hook for generating address diff vector tables in case of
18497 * doc/tm.texi : Regenerate.
18498 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
18499 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
18500 of addr diff vector generation.
18501 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
18502 target hook definition.
18503 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
18504 Add default function for generate_pic_addr_diff_vec -> flag_pic.
18505 * doc/invoke.texi (Add new pic option): Add new microblaze pic
18506 option for data text relative.
18508 2018-04-30 Richard Biener <rguenther@suse.de>
18510 * tree-chrec.h (evolution_function_is_constant_p): Remove
18512 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
18514 2018-04-30 Richard Biener <rguenther@suse.de>
18517 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
18519 2018-04-30 Richard Biener <rguenther@suse.de>
18521 PR tree-optimization/28364
18522 PR tree-optimization/85275
18523 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
18524 copying first exit test.
18526 2018-04-28 Mark Wielaard <mark@klomp.org>
18528 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
18529 dwarf_version >= 5.
18530 (dwarf_AT): Handle DW_AT_addr_base.
18531 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
18533 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
18536 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
18537 (*ashl<dwi>3_doubleword_mask_1): Ditto.
18538 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
18539 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
18541 2018-04-28 Richard Biener <rguenther@suse.de>
18543 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
18544 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
18545 to reflect use. Only add interesting stmts.
18547 2018-04-27 Martin Jambor <mjambor@suse.cz>
18550 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
18551 the jump function allows for passing through aggregate values.
18553 2018-04-27 David Malcolm <dmalcolm@redhat.com>
18555 * input.h (in_system_header_at): Convert from macro to inline
18557 (from_macro_expansion_at): Likewise.
18558 (from_macro_definition_at): Likewise.
18560 2018-04-27 Jeff Law <law@redhat.com>
18562 * config.gcc: Mark tile* targets as deprecated/obsolete.
18564 2018-04-27 Richard Biener <rguenther@suse.de>
18566 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
18569 2018-04-27 Richard Biener <rguenther@suse.de>
18571 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
18573 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
18575 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
18576 with Yd constraint. Set "preferred_for_speed" attribute from
18577 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
18578 with Yd constraint.
18579 (*movdi_internal): Ditto.
18580 (movti_interunit splitters): Remove
18581 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
18582 (movdi_interunit splitters): Ditto.
18583 * config/i386/constraints.md (Ye): Remove.
18584 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
18586 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18589 * config/aarch64/constraints.md (Usg): Limit to 31.
18590 (Usj): Limit to 63.
18592 2018-04-27 Jakub Jelinek <jakub@redhat.com>
18594 PR tree-optimization/85529
18595 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
18596 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
18597 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
18598 zero extension or masking of the MSB bit.
18599 (optimize_range_tests): Add FIRST_BB argument, pass it through
18600 to optimize_range_tests_var_bound.
18601 (maybe_optimize_range_tests, reassociate_bb): Adjust
18602 optimize_range_tests callers.
18604 2018-04-26 Richard Biener <rguenther@suse.de>
18605 Jakub Jelinek <jakub@redhat.com>
18607 * cgraph.h (symbol_table): Just declare debug method here.
18608 * symtab.c (symbol_table::debug): Define.
18610 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
18612 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
18614 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
18616 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
18617 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
18618 (*movdi_internal): Substitute Yi and Yj constraint with x
18619 and Ym and Yn constraint with y constraint. Update "isa"
18620 attribute and set "preferred_for_speed" attribute from
18621 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
18622 (*movsi_internal): Ditto.
18623 (*movdf_internal): Ditto.
18624 (*movsf_internal): Ditto.
18625 (*zero_extendsidi2): Ditto.
18626 * config/i386/sse.md (vec_set<mode>_0): Ditto.
18627 (sse2_loadld): Ditto.
18628 (*vec_extract<ssevecmodelower>_0): Ditto.
18629 (*vec_extractv4si_0_zext_sse4): Ditto.
18630 (vec_concatv2di): Ditto.
18631 (*vec_dup<mode>): Ditto.
18632 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
18633 * config/i386/constraints.md (Yi): Remove.
18638 2018-04-26 Nathan Sidwell <nathan@acm.org>
18640 * dumpfile.c (dump_open): New.
18641 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
18642 (dump_finish): Detect stdio/stderr by value not name.
18644 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
18646 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
18648 2018-04-26 Tom de Vries <tom@codesourcery.com>
18651 * config/nvptx/nvptx.c (verify_neutering_jumps)
18652 (verify_neutering_labels): New function
18653 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
18655 2018-04-26 Tom de Vries <tom@codesourcery.com>
18658 * config/nvptx/nvptx.c (needs_neutering_p): New function.
18659 (nvptx_single): Use needs_neutering_p to skip over insns that do not
18662 2018-04-26 Richard Biener <rguenther@suse.de>
18663 Tom de Vries <tom@codesourcery.com>
18666 * lto-streamer-out.c (output_function): Fixup loops if required to match
18667 discovery done in the reader.
18669 2018-04-26 Richard Biener <rguenther@suse.de>
18671 PR tree-optimization/85116
18672 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
18673 have a loop exit from the single latch predecessor. Remove
18674 case of header with just condition.
18675 (ch_base::copy_headers): Exclude infinite loops from any
18677 (pass_ch::execute): Record exits.
18679 2018-04-26 Richard Biener <rguenther@suse.de>
18681 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
18682 prologue cost vector and pass it to vect_get_load_cost.
18683 (vect_get_peeling_costs_all_drs): Likewise.
18684 (vect_peeling_hash_get_lowest_cost): Likewise.
18685 (vect_enhance_data_refs_alignment): Likewise.
18687 2018-04-26 Richard Biener <rguenther@suse.de>
18689 PR middle-end/85450
18690 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
18691 checking of integer<->pointer conversions.
18692 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
18693 sign-/zero-extending pointer types.
18694 (expand_omp_for_static_chunk): Likewise.
18696 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
18697 Jean Lee <xiaoyur347@gmail.com>
18699 * config/mips/mips.c (mips_asan_shadow_offset): New function.
18700 (TARGET_ASAN_SHADOW_OFFSET): Define.
18701 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
18702 true for -fsanitize=address.
18704 2018-04-25 Mark Wielaard <mark@klomp.org>
18706 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
18709 2018-04-25 Jakub Jelinek <jakub@redhat.com>
18711 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
18712 than "alu", remove explicit "memory" and "imm_disp" attributes.
18713 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
18715 PR middle-end/85414
18716 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
18717 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
18718 gen_lowpart_no_emit.
18720 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
18723 * config/i386/i386.c (ix86_expand_builtin): Change memory
18724 operand to XI, extend p0 to Pmode.
18725 * config/i386/i386.md: Change unspec volatile and operand
18726 1 mode to XI, change operand 0 mode to P.
18728 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
18730 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
18731 GET_MODE_MASK before any checking.
18732 (nds32_can_use_bset_p): Likewise.
18733 (nds32_can_use_btgl_p): Likewise.
18735 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
18737 * config/nds32/nds32-doubleword.md: New define_split pattern for
18738 illegal register number.
18740 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
18742 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
18744 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
18746 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
18748 2018-04-25 Richard Biener <rguenther@suse.de>
18750 * lto-streamer.h (LTO_major_version): Bump to 8.
18752 2018-04-25 Jakub Jelinek <jakub@redhat.com>
18754 * BASE-VER: Set to 9.0.0.
18756 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
18758 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
18759 in __abskf2 and __powikf2.
18761 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18764 * config/aarch64/constraints.md (Usg, Usj): New constraints.
18765 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
18766 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
18767 Use the above on operand 2. Reindent.
18768 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
18770 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
18773 * common/config/i386/i386-common.c (ix86_handle_option): Don't
18775 * config/i386/i386.opt (mcet): Removed.
18776 * doc/install.texi: Remove -mcet documentation.
18777 * doc/invoke.texi: Likewise.
18779 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
18782 * doc/install.texi: Remove -mcet from bootstrap-cet.
18784 2018-04-24 Jakub Jelinek <jakub@redhat.com>
18787 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
18788 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
18792 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
18793 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
18794 containing a CONST_VECTOR.
18796 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
18798 * doc/install.texi: Update newlib dependency for nvptx.
18800 2018-04-24 Jakub Jelinek <jakub@redhat.com>
18803 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
18804 instead of INTVAL when shifting x left.
18806 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
18808 PR tree-optimization/85478
18809 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
18810 vect_grouped_store_supported for single element vectors.
18812 2018-04-24 Richard Biener <rguenther@suse.de>
18815 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
18816 load cost increase to the case of non-constant step.
18818 2018-04-24 Jakub Jelinek <jakub@redhat.com>
18821 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
18822 destination if any_malformed_asm.
18824 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
18826 PR middle-end/85496
18827 * expr.c (store_field): In the bitfield case, if the value comes from
18828 a function call and is returned in registers by means of a PARALLEL,
18829 do not change the mode of the temporary unless BLKmode and VOIDmode.
18831 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
18833 PR rtl-optimization/85423
18834 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
18835 dependencies to debug insns when the previous insn is non-debug.
18837 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
18839 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
18840 enums into a single definition.
18841 (fls): Fix predicates and printing.
18844 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
18846 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
18847 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
18848 and short u6 immediate.
18849 (check_if_valid_sleep_operand): Remove.
18850 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
18852 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
18854 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
18855 flag_always_save_lp condition.
18856 * config/nds32/nds32.opt (malways-save-lp): New option.
18858 2018-04-22 Shiva Chen <shiva0217@gmail.com>
18860 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
18861 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
18862 * config/nds32/nds32.h
18863 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
18864 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
18866 2018-04-22 Shiva Chen <shiva0217@gmail.com>
18868 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
18869 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
18871 2018-04-22 Shiva Chen <shiva0217@gmail.com>
18872 Chung-Ju Wu <jasonwucj@gmail.com>
18874 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
18876 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
18877 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
18879 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
18881 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
18883 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
18885 * config/nds32/nds32-protos.h (nds32_data_alignment,
18886 nds32_local_alignment): Declare.
18887 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
18888 nds32_local_alignment): New functions.
18889 (TARGET_CONSTANT_ALIGNMENT): Define.
18890 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
18892 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
18894 * config/nds32/nds32.c
18895 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
18896 (TARGET_MODES_TIEABLE_P): Likewise.
18898 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
18900 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
18901 level Ofast and Og.
18903 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
18904 Chung-Ju Wu <jasonwucj@gmail.com>
18906 * config/nds32/constants.md (unspec_volatile_element): Add enum values
18907 for unaligned access.
18908 * config/nds32/nds32-intrinsic.c: Implementation of expanding
18910 * config/nds32/nds32-intrinsic.md: Likewise.
18911 * config/nds32/nds32_intrinsic.h: Likewise.
18912 * config/nds32/nds32.h (nds32_builtins): Likewise.
18913 * config/nds32/nds32.opt (munaligned-access): New option.
18914 * config/nds32/nds32.c (nds32_asm_file_start): Display
18915 flag_unaligned_access status.
18917 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
18919 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
18920 -mno-relax is present.
18921 * config/riscv/linux.h (LINK_SPEC): Ditto.
18923 2018-04-20 Martin Sebor <msebor@redhat.com>
18926 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
18928 (gimple_fold_builtin_stxcpy_chk): Same.
18929 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
18931 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
18934 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
18935 __powikf2 when long double is IEEE 128-bit.
18937 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
18939 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
18940 step to make sure stack always aligned.
18942 2018-04-20 Carl Love <cel@us.ibm.com>
18945 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
18946 size check for arg0.
18948 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
18949 Tom de Vries <tom@codesourcery.com>
18952 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
18953 Emit insns for calls too.
18954 (nvptx_find_par): Always look for worker-level predecessor insn.
18955 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
18957 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
18958 (nvptx_process_pars): Propagate frames for calls.
18960 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
18963 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
18965 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
18966 (ix86_handle_option): Don't handle OPT_mibt.
18967 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
18969 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
18971 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
18972 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
18973 (ix86_target_macros): Define __CET__ with flag_cf_protection
18974 for -fcf-protection.
18975 * config/i386/i386.c (isa2_opts): Remove -mibt.
18976 * config/i386/i386.h (TARGET_IBT): Removed.
18977 (TARGET_IBT_P): Likewise.
18978 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
18979 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
18980 * config/i386/i386.opt (mcet): Update help message.
18981 (mshstk): Likewise.
18983 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
18984 -mcet as an alias for -mshstk.
18986 2018-04-20 Richard Biener <rguenther@suse.de>
18988 PR middle-end/85475
18989 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
18990 complexity by forcing a single use of the multiply operand.
18992 2018-04-20 Martin Jambor <mjambor@suse.cz>
18995 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
18996 recursion dependency to only apply to non-clones.
18998 2018-04-20 Martin Jambor <mjambor@suse.cz>
19001 * ipa-cp.c (create_specialized_node): Check that clones of
19002 self-recursive edges exist during IPA-CP.
19004 2018-04-19 Toon Moene <toon@moene.org>
19006 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
19009 2018-04-19 Jakub Jelinek <jakub@redhat.com>
19011 PR tree-optimization/85467
19012 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
19013 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
19014 VECTOR_CST element to type.
19016 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
19019 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
19020 * config/i386/i386.md (builtin_setjmp_setup): Removed.
19021 (builtin_longjmp): Likewise.
19022 (save_stack_nonlocal): New pattern.
19023 (restore_stack_nonlocal): Likewise.
19025 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
19028 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
19029 Replace ASM_OUTPUT_LABEL with fprintf.
19031 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
19034 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
19035 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
19036 * config/i386/i386-c.c (ix86_target_macros_internal): Also
19037 define __IBT__ and __SHSTK__ for -fcf-protection.
19038 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
19040 (ix86_trampoline_init): Likewise.
19041 (x86_output_mi_thunk): Likewise.
19042 (ix86_notrack_prefixed_insn_p): Likewise.
19043 (ix86_option_override_internal): Don't disallow -fcf-protection.
19044 * config/i386/i386.md (rdssp<mode>): Also enable for
19046 (incssp<mode>): Likewise.
19047 (nop_endbr): Likewise.
19048 * config/i386/i386.opt (mcet): Change help message to built-in
19051 (mshstk): Likewise.
19052 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
19053 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
19054 enable CET built-in functions.
19056 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
19058 * common/config/i386/i386-common.c
19059 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
19060 OPTION_MASK_ISA_MOVDIRI_UNSET,
19061 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
19062 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
19063 * config.gcc (movdirintrin.h): New header.
19064 * config/i386/cpuid.h (bit_MOVDIRI,
19065 bit_MOVDIR64B): New bits.
19066 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
19068 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
19069 (VOID, PVOID, PCVOID)): New function types.
19070 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
19071 __builtin_ia32_directstoreu_u64,
19072 __builtin_ia32_movdir64b): New builtins.
19073 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
19074 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
19076 (ix86_valid_target_attribute_inner_p): Ditto.
19077 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
19078 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
19079 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
19080 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
19081 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
19082 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
19083 (movdiri<mode>, movdir64b_<mode>): New.
19084 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
19085 * config/i386/immintrin.h: Include movdirintrin.h.
19086 * config/i386/movdirintrin.h: New file.
19087 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
19089 2018-04-19 Richard Biener <rguenther@suse.de>
19091 PR middle-end/85455
19092 * cfg.c (clear_bb_flags): When loop state says we have
19093 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
19095 2018-04-19 Richard Biener <rguenther@suse.de>
19097 PR tree-optimization/84737
19098 * tree-vect-data-refs.c (vect_copy_ref_info): New function
19099 copying restrict info.
19100 (vect_setup_realignment): Use it.
19101 * tree-vectorizer.h (vect_copy_ref_info): Declare.
19102 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
19103 the first DR to all generated stores.
19104 (vectorizable_load): Likewise for loads.
19106 2018-04-19 Jakub Jelinek <jakub@redhat.com>
19108 PR tree-optimization/85446
19109 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
19110 the integral and pointer types to have the same precision.
19112 * doc/install.texi: Document --disable-cet being the default and
19115 2018-04-18 Martin Liska <mliska@suse.cz>
19117 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
19120 2018-04-18 Martin Liska <mliska@suse.cz>
19123 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
19126 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
19127 arguments if they are comparable.
19129 2018-04-18 Martin Liska <mliska@suse.cz>
19132 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
19135 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
19138 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
19141 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
19142 ENDBR after calling __morestack.
19144 2018-04-18 David Malcolm <dmalcolm@redhat.com>
19147 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
19148 by using gcc_base_ver to generate a gcc_driver_version, and use
19149 it when generating GCC_DRIVER_NAME.
19150 * configure: Regenerate.
19152 2018-04-18 Jakub Jelinek <jakub@redhat.com>
19155 * config.gcc: Obsolete powerpc*-*-*spe*.
19157 2018-04-17 Jakub Jelinek <jakub@redhat.com>
19160 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
19161 (stabstr_D): Change type of unum from unsigned int to
19162 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
19165 2018-04-17 Jim Wilson <jimw@sifive.com>
19168 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
19169 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
19170 Set arg_pointer_offset after using pretend_args_size.
19172 2018-04-17 Jakub Jelinek <jakub@redhat.com>
19174 PR rtl-optimization/85431
19175 * dse.c (record_store): Ignore zero width stores.
19178 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
19179 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
19180 __builtin_stack_restore rather than after it.
19181 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
19182 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
19183 argument instead of virtual_dynamic_stack_rtx.
19185 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
19187 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
19189 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
19190 Add note to error message to explain internal mapping of overloaded
19191 built-in function name to non-overloaded built-in function name.
19192 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
19195 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
19198 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
19199 where the inputs overlap with the output.
19201 2018-04-17 Jakub Jelinek <jakub@redhat.com>
19203 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
19204 (=v, v) alternative and explicit "memory" attribute.
19205 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
19206 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
19208 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
19209 "sselog1" type instead of "sselog".
19210 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
19211 "sselog". Remove explicit "memory" attribute.
19212 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
19213 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
19215 (vec_extract_hi_v32hi): Merge all alternatives into one, use
19216 "sselog1" type instead of "sselog". Remove explicit "memory"
19218 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
19219 use "sselog1" type instead of "sselog". Remove explicit "memory"
19221 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
19222 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
19224 (vec_extract_hi_v64qi): Merge all alternatives into one, use
19225 "sselog1" type instead of "sselog". Remove explicit "memory"
19227 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
19228 use "sselog1" type instead of "sselog". Remove explicit "memory"
19232 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
19234 PR middle-end/85414
19235 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
19238 2018-04-17 Martin Jambor <mjambor@suse.cz>
19241 * ipa-cp.c (create_specialized_node): Call
19242 expand_all_artificial_thunks if necessary.
19244 2018-04-17 Martin Liska <mliska@suse.cz>
19247 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
19248 in message, remote space in between '_G' and '('.
19250 2018-04-17 Jakub Jelinek <jakub@redhat.com>
19253 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
19254 avx512f_vmcmp<mode>3<round_saeonly_name>,
19255 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
19256 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
19257 avx512f_rndscale<mode><round_saeonly_name>,
19258 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
19259 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
19260 Use %<iptr>2 instead of %2 for -masm=intel.
19261 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
19262 avx512f_vcvttss2usi<round_saeonly_name>,
19263 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
19265 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
19266 avx512f_vcvttsd2usi<round_saeonly_name>,
19267 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
19268 Use %q1 instead of %1 for -masm=intel.
19269 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
19270 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
19271 of %3 for -masm=intel.
19272 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
19274 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
19276 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
19278 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
19279 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
19281 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
19283 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
19284 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
19285 %g1 and one with %0 and %1.
19286 (avx512er_vmrcp28<mode><round_saeonly_name>,
19287 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
19288 %1 for -masm=intel.
19289 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
19290 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
19291 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
19292 of %0 and %{%4%} for -masm=intel.
19293 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
19294 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
19295 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
19296 order of %0 and %{%5%}%{z%} for -masm=intel.
19298 2018-04-17 Jan Hubicka <jh@suse.cz>
19301 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
19303 2018-04-17 Martin Liska <mliska@suse.cz>
19306 * multiple_target.c (create_dispatcher_calls): Set apostrophes
19307 for target_clone error message. Make default implementation
19308 clone to be a local declaration.
19309 (separate_attrs): Add new argument and check for an empty
19311 (expand_target_clones): Handle it.
19312 (ipa_target_clone): Make redirection just for target_clones
19315 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
19316 Tom de Vries <tom@codesourcery.com>
19318 PR middle-end/84955
19319 * omp-expand.c (expand_oacc_for): Add dummy false branch for
19320 tiled basic blocks without omp continue statements.
19322 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
19325 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
19326 vec_extract expression as having side effects to make sure it gets
19329 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
19332 * config/i386/i386.c (get_builtin_code_for_version): Check
19335 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
19338 * config.gcc: Support "skylake".
19339 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19341 * config/i386/i386.c (m_SKYLAKE): Define.
19342 (processor_target_table): Add "skylake".
19343 (ix86_option_override_internal): Add "skylake".
19344 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
19345 PROCESSOR_CANNONLAKE.
19346 (get_builtin_code_for_version): Fix priority for
19347 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
19348 PROCESSOR_SKYLAKE-AVX512.
19349 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
19350 (processor_type): Add PROCESSOR_SKYLAKE.
19352 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
19353 Jason Merrill <jason@redhat.com>
19356 * convert.c (convert_to_integer_1): Use direct recursion for
19357 enumeral types and types with a precision less than the number
19358 of bits in their mode.
19360 2018-04-16 Julia Koval <julia.koval@intel.com>
19363 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
19364 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
19366 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
19369 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
19370 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
19371 and -mno-direct-move.
19373 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
19376 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
19377 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
19378 Ensure negative shifts result in {0}.
19380 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
19382 PR rtl-optimization/79916
19383 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
19384 regs (if any) to define how to gnerate SD moves when LRA is in
19387 2018-04-13 Jakub Jelinek <jakub@redhat.com>
19389 PR rtl-optimization/85393
19390 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
19391 * except.c (expand_dw2_landing_pad_for_region): Make static.
19392 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
19393 a label and unconditional jump to old_bb, rather than
19394 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
19397 PR rtl-optimization/85376
19398 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
19399 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
19400 instead of a specific value.
19402 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
19403 Bin Cheng <bin.cheng@arm.com>
19405 PR tree-optimization/82965
19406 PR tree-optimization/83991
19407 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
19408 by_profile_only parameter.
19409 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
19410 information if the loop was predicted to iterate too many times.
19411 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
19413 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
19416 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
19419 2018-04-13 Martin Liska <mliska@suse.cz>
19420 Jakub Jelinek <jakub@redhat.com>
19422 PR middle-end/81657
19423 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
19424 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
19425 * builtins.c (expand_builtin_memory_copy_args): Use
19426 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
19427 handle dest_addr == pc_rtx.
19429 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
19432 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
19433 asked to not generate direct moves.
19434 (fix_trunc<mode>si2_stfiwx): Similar.
19435 (fix_trunc<mode>si2_internal): Similar.
19437 2018-04-12 Jakub Jelinek <jakub@redhat.com>
19440 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
19441 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
19442 lookup if dest in some wider mode is known to be const0_rtx and
19443 if so, record permanent equivalence for it to be ZERO_EXTEND of
19444 the narrower mode destination.
19446 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
19448 * lto-streamer-out.c (output_function): Revert 259346.
19449 * omp-expand.c (expand_oacc_for): Likewise.
19451 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
19453 PR rtl-optimization/85354
19454 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
19455 * sel-sched.c (sel_global_init): ... here.
19457 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
19460 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
19461 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
19462 mode for PE-COFF targets.
19463 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
19464 (i386_pe_asm_lto_end): Likewise.
19465 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
19466 (TARGET_ASM_LTO_END): Likewise.
19467 * config/i386/winnt.c (saved_debug_info_level): New static variable.
19468 (i386_pe_asm_lto_start): New function.
19469 (i386_pe_asm_lto_end): Likewise.
19471 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
19472 Richard Biener <rguenther@suse.de>
19474 PR middle-end/84955
19475 * lto-streamer-out.c (output_function): Fix CFG loop state before
19477 * omp-expand.c (expand_oacc_for): Handle calls to internal
19478 functions like regular functions.
19480 2018-04-12 Richard Biener <rguenther@suse.de>
19483 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
19484 for the early LTO debug to properly generate references to it
19485 during DIE emission. Do not re-use that for the skeleton for
19487 (dwarf2out_early_finish): Likewise.
19489 2018-04-12 Jakub Jelinek <jakub@redhat.com>
19492 * config/i386/sse.md
19493 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
19494 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
19495 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
19496 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
19497 and output is a reg, avoid creating invalid lowpart subreg, but
19498 instead split into a 512-bit move. Don't split if not AVX512VL,
19499 input is xmm16+ reg and output is a mem.
19500 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
19501 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
19502 xmm16+ reg and output is a mem.
19504 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
19506 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
19507 also for flag_dwarf2_cfi_asm.
19509 2018-04-12 Jakub Jelinek <jakub@redhat.com>
19511 PR rtl-optimization/85342
19512 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
19513 a bool scalar var inside of the loop instead. Don't try to update
19514 recog_data.operand after failed apply_change_group.
19516 2018-04-12 Tom de Vries <tom@codesourcery.com>
19519 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
19520 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
19521 array with flexible array member as array without given dimension.
19522 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
19523 argument for undefined param to true.
19525 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
19528 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
19529 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
19530 from PowerPC section.
19531 * config/rs6000/sysv4.opt (mcall-): Improve help text.
19532 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
19533 help text that is too long.
19534 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
19535 help text that is too long.
19536 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
19537 help text that is too long.
19539 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
19541 * config/alpha/alpha.md (stack_probe_internal): Rename
19542 from "probe_stack". Update all callers.
19544 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
19546 PR rtl-optimization/84566
19547 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
19548 sched_macro_fuse_insns.
19550 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
19553 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
19554 (compute_block_dependences): ... from here.
19556 2018-04-11 Jakub Jelinek <jakub@redhat.com>
19558 PR tree-optimization/85331
19559 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
19560 from int to HOST_WIDE_INT.
19562 2018-04-11 Martin Jambor <mjambor@suse.cz>
19565 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
19566 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
19567 not the same as the source val.
19568 (cgraph_edge_brings_value_p): New parameter.
19569 (gather_edges_for_value): Pass destination value to
19570 cgraph_edge_brings_value_p.
19571 (perhaps_add_new_callers): Likewise.
19572 (get_info_about_necessary_edges): Likewise and exclude values brought
19573 only by self-recursive edges.
19574 (create_specialized_node): Redirect only clones of self-calling edges.
19575 (+self_recursive_pass_through_p): New function.
19576 (find_more_scalar_values_for_callers_subset): Use it.
19577 (find_aggregate_values_for_callers_subset): Likewise.
19578 (known_aggs_to_agg_replacement_list): Removed.
19579 (decide_whether_version_node): Re-calculate known constants for all
19580 remaining context clones.
19582 2018-04-11 Richard Biener <rguenther@suse.de>
19585 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
19586 from early DWARF output.
19587 (dwarf2out_early_finish): Output line info unconditionally into
19588 early DWARF and add reference to it.
19590 2018-04-11 Jakub Jelinek <jakub@redhat.com>
19593 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
19594 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
19595 other than V2DFmode using iptr mode attribute.
19596 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
19598 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
19600 PR rtl-optimization/84659
19601 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
19603 2018-04-11 Jakub Jelinek <jakub@redhat.com>
19606 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
19608 (output_loc_list): Pass address of a dummy size variable even in the
19609 locview handling loop.
19610 (index_location_lists): Add comment on why skip_loc_list_entry can't
19613 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
19616 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
19619 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
19622 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
19623 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
19624 and -mstring-compare-inline-limit.
19626 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
19629 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
19630 for stack clash protection in a register whenever we need it to be in
19633 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
19635 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
19636 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
19638 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
19641 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
19643 (mlong-double-): Ditto.
19644 * config/rs6000/sysv4.opt (msdata=): Ditto.
19645 (mtls-size=): Ditto.
19647 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
19649 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
19650 erroneous entries for
19651 "vector int vec_ldl (int, long int *)", and
19652 "vector unsigned int vec_ldl (int, unsigned long int *)".
19653 Add comments and entries for
19654 "vector bool char vec_ldl (int, bool char *)",
19655 "vector bool short vec_ldl (int, bool short *)",
19656 "vector bool int vec_ldl (int, bool int *)",
19657 "vector bool long long vec_ldl (int, bool long long *)",
19658 "vector pixel vec_ldl (int, pixel *)",
19659 "vector long long vec_ldl (int, long long *)",
19660 "vector unsigned long long vec_ldl (int, unsigned long long *)".
19661 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
19662 type tree bool_long_long_type_node and correct definition of
19663 bool_V2DI_type_node to make reference to this new type tree.
19664 (rs6000_mangle_type): Replace erroneous reference to
19665 bool_long_type_node with bool_long_long_type_node.
19666 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
19667 comments to emphasize sign distinctions for char and int types and
19668 replace RS6000_BTI_bool_long constant with
19669 RS6000_BTI_bool_long_long constant. Also add comment to restrict
19670 use of RS6000_BTI_pixel.
19671 (bool_long_type_node): Remove this macro definition.
19672 (bool_long_long_type_node): New macro definition
19674 2018-04-10 Jakub Jelinek <jakub@redhat.com>
19676 PR rtl-optimization/85300
19677 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
19678 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
19679 simplify_unary_operation fails.
19681 2018-04-10 Martin Liska <mliska@suse.cz>
19683 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
19684 cgraph_edge and ipa_ref.
19686 2018-04-10 Jakub Jelinek <jakub@redhat.com>
19690 * config/i386/sse.md
19691 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
19692 computation of the VEC_MERGE selector from mask.
19693 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
19694 Fix decoding of the VEC_MERGE selector into mask.
19696 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
19698 PR tree-optimization/85286
19699 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
19701 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
19703 * final.c (final_1): Set insn_last_address as well as
19704 insn_current_address.
19706 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19709 * explow.c (emit_stack_probe): Call validize_mem on memory location
19710 before passing it to gen_probe_stack. Create address operand and
19711 legitimize it for the probe_stack_address case.
19713 2018-04-09 Jan Hubicka <jh@suse.cz>
19716 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
19717 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
19718 * tree.c (free_lang_data_in_type): Fix handling of binfos;
19720 (free_lang_data): Rebuild type inheritance graph.
19722 2018-04-09 Martin Sebor <msebor@redhat.com>
19724 * invoke.texi (-finline-small-functions): Mention other optimization
19726 (-findirect-inlining, -fpartial-inlining): Same.
19727 (-finline-functions-called-once): Same.
19728 (-freorder-blocks-and-partition): Same.
19730 2018-04-09 Jan Hubicka <jh@suse.cz>
19733 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
19734 jumps; choose last target that matches the criteria (i.e.
19735 no partition changes for non-crossing jumps).
19736 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
19737 support for redirecting crossing jumps to non-crossing.
19739 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
19741 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
19742 also for naked functions.
19744 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
19746 * config/arc/arc.md (add_shift): New pattern.
19747 (add_shift2): Likewise.
19748 (sub_shift): Likewise.
19749 (sub_shift_cmp0_noout): Likewise.
19750 (compare_si_ashiftsi): Likewise.
19751 (xbfu_cmp0_noout): New combine pattern.
19752 (xbfu_cmp0"): Likewise.
19753 (movsi_set_cc_insn): Place the predicable variant first.
19754 (commutative_binary_cmp0_noout): Remove clobber.
19755 (commutative_binary_cmp0): New pattern.
19756 (noncommutative_binary_cmp0): Likewise.
19757 (noncommutative_binary_cmp0_noout): Likewise.
19758 (noncommutative_binary_comparison_result_used): Removed.
19759 (rsub_cmp0): New pattern.
19760 (rsub_cmp0_noout): Likewise.
19761 (extzvsi): Changed, keep only meaningful variants.
19762 (SQH, SEZ): New iterators.
19763 (SQH_postfix): New mode attribute.
19764 (SEZ_prefix): New code attribute.
19765 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
19766 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
19767 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
19768 of numerical value.
19769 (noncommutative_operator): Check the availability of barrel
19772 2018-04-09 Richard Biener <rguenther@suse.de>
19774 PR tree-optimization/85284
19775 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
19776 Only use the niter constraining form of simple_iv when the exit
19777 is always executed.
19779 2018-04-09 Tom de Vries <tom@codesourcery.com>
19782 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
19783 (define_expand "*memory_barrier"): New define_expand.
19784 (define_insn "memory_barrier"): New insn.
19786 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
19788 PR rtl-optimization/80463
19789 PR rtl-optimization/83972
19790 PR rtl-optimization/83480
19792 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
19793 correct producer for the insn.
19794 (tidy_control_flow): Fixup seqnos in case of debug insns.
19796 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
19798 PR rtl-optimization/83913
19800 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
19801 different sched-times when merging exprs.
19803 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
19805 PR rtl-optimization/83962
19807 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
19808 tidy_fallthru_edge and tidy_control_flow.
19810 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
19812 PR rtl-optimization/83530
19814 * sel-sched.c (force_next_insn): New global variable.
19815 (remove_insn_for_debug): When force_next_insn is true, also leave only
19816 next insn in the ready list.
19817 (sel_sched_region): When the region wasn't scheduled, make another pass
19818 over it with force_next_insn set to 1.
19820 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
19822 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
19824 * config/nds32/constants.md (unspec_volatile_element): Add enum values
19825 for interrupt control.
19826 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
19827 functions for interrupt control.
19828 * config/nds32/nds32-intrinsic.md: Likewise.
19829 * config/nds32/nds32_intrinsic.h: Likewise.
19830 * config/nds32/nds32.h (nds32_builtins): Likewise.
19832 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
19834 * config/nds32/nds32.c (nds32_init_machine_status,
19835 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
19836 strict_aligned_p field.
19837 (nds32_expand_to_rtl_hook): New function.
19838 (TARGET_EXPAND_TO_RTL_HOOK): Define.
19839 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
19841 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
19842 Chung-Ju Wu <jasonwucj@gmail.com>
19844 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
19845 * config/nds32/nds32-n7.md: New file.
19846 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
19847 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
19849 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
19850 * config/nds32/nds32.md (pipeline_model): Add n7.
19851 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
19852 * config/nds32/pipelines.md: Include n7 settings.
19854 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
19855 Chung-Ju Wu <jasonwucj@gmail.com>
19857 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
19858 * config/nds32/nds32-e8.md: New file.
19859 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
19860 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
19862 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
19863 * config/nds32/nds32.md (pipeline_model): Add e8.
19864 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
19865 * config/nds32/pipelines.md: Include e8 settings.
19867 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
19868 Chung-Ju Wu <jasonwucj@gmail.com>
19870 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
19871 * config/nds32/nds32-n8.md: New file.
19872 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
19873 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
19875 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
19876 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
19877 * config/nds32/nds32.md (pipeline_model): Add n8.
19878 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
19879 * config/nds32/pipelines.md: Include n8 settings.
19881 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
19882 Chung-Ju Wu <jasonwucj@gmail.com>
19884 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
19885 * config/nds32/nds32-n9-2r1w.md: New file.
19886 * config/nds32/nds32-n9-3r2w.md: New file.
19887 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
19888 nds32_register_ports): New or modify for cpu n9.
19889 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
19891 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
19892 * config/nds32/nds32-utils.c: New file.
19893 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
19894 TARGET_MUL_SLOW): Define.
19895 * config/nds32/nds32.md (pipeline_model): New attribute.
19896 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
19897 New options that support cpu n9.
19898 * config/nds32/pipelines.md: Include n9 settings.
19899 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
19901 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
19903 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
19904 information if necessary.
19905 (output_cond_branch_compare_zero): Likewise.
19906 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
19907 (nds32_target_alignment): Refine for alignment.
19908 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
19909 (FUNCTION_BOUNDARY): Modify.
19910 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
19912 * config/nds32/nds32.opt (malways-align, malign-functions): New.
19914 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
19916 * config/nds32/constants.md (unspec_volatile_element): Add values for
19917 TLB operation and data prefetch.
19918 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
19919 functions for TLB operation and data prefetch.
19920 * config/nds32/nds32-intrinsic.md: Likewise.
19921 * config/nds32/nds32_intrinsic.h: Likewise.
19922 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
19923 (nds32_print_operand): Likewise.
19924 * config/nds32/nds32.h (nds32_builtins): Likewise.
19926 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
19927 Andrew Pinski <pinsika@gcc.gnu.org>
19929 PR middle-end/82976
19930 * match.pd: Use constant_boolean_node of correct type instead of
19931 boolean_true_node or boolean_false_node for simplifying
19932 pointer comparisons to zero.
19934 2018-04-07 Jakub Jelinek <jakub@redhat.com>
19936 PR tree-optimization/80021
19937 * tree.c (verify_type_variant): Make error call in verify_variant_match
19938 translatable and remove final full stop.
19940 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
19942 * config/nds32/constants.md (unspec_volatile_element): Add
19943 UNSPEC_VOLATILE_EH_RETURN.
19944 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
19945 nds32_output_stack_pop): Support dwarf exception handling process.
19946 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
19947 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
19948 exception handling process.
19949 (nds32_compute_stack_frame): Likewise.
19950 (nds32_return_addr_rtx): Likewise.
19951 (nds32_initial_elimination_offset): Likewise.
19952 (nds32_expand_prologue): Likewise.
19953 (nds32_expand_epilogue): Likewise.
19954 (nds32_dynamic_chain_address): New function.
19955 * config/nds32/nds32.h (machine_function): Add fields for dwarf
19956 exception handling.
19957 (DYNAMIC_CHAIN_ADDRESS): Define.
19958 (EH_RETURN_DATA_REGNO): Define.
19959 (EH_RETURN_STACKADJ_RTX): Define.
19960 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
19961 patterns for dwarf exception handling.
19963 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
19965 * config/nds32/nds32.h: Clean up obsolete macros.
19967 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
19969 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
19970 Add enum values for particular instructions.
19971 * config/nds32/nds32-intrinsic.c: Implementation of expanding
19972 particular intrinsic functions.
19973 * config/nds32/nds32-intrinsic.md: Likewise.
19974 * config/nds32/nds32_intrinsic.h: Likewise.
19975 * config/nds32/nds32.h (nds32_builtins): Likewise.
19976 * config/nds32/nds32.md (type): Add pbsad and pbsada.
19977 (btst, ave): New patterns for particular instructions.
19979 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
19981 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
19982 Add enum values for atomic load/store and memory sync.
19983 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
19985 * config/nds32/nds32-intrinsic.md: Likewise.
19986 * config/nds32/nds32_intrinsic.h: Likewise.
19987 * config/nds32/nds32.h (nds32_builtins): Likewise.
19989 2018-04-07 Jakub Jelinek <jakub@redhat.com>
19991 PR tree-optimization/85257
19992 * fold-const.c (native_encode_vector): If not all elts could fit
19993 and off is -1, return 0 rather than offset.
19994 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
19995 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
19996 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
19997 adjust buffer in native_interpret_expr call.
19999 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
20001 * config/nds32/constants.md (unspec_volatile_element): Add cache
20002 control enum values.
20003 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
20004 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
20005 * config/nds32/nds32.c (nds32_cctl_names): New.
20006 (nds32_print_operand): Handle cache control register names.
20007 * config/nds32/nds32.h (nds32_builtins): New enum values.
20008 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
20010 * config/nds32/nds32.md (type): Add mmu.
20011 * config/nds32/pipelines.md (simple_insn): Add mmu.
20013 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
20015 * config/nds32/nds32.md (type): Remove call.
20016 * config/nds32/pipelines.md (simple_insn): Likewise.
20018 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
20020 * config/nds32/constants.md (unspec_volatile_element): Add
20021 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
20022 UNSPEC_VOLATILE_FMFCFG.
20023 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
20024 description for fmfcfg and fmfcsr.
20025 (bdesc_1arg): Add fmtcsr.
20026 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
20027 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
20028 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
20029 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
20030 unspec_fmfcfg): New patterns.
20031 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
20032 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
20033 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
20034 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
20035 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
20036 __nds32__fmfcfg): Define.
20038 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
20040 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
20041 intrinsic register names.
20042 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
20043 intrinsic register enum values and macros.
20045 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
20047 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
20048 for load/store addressing form.
20049 (nds32_print_operand_address): Likewise.
20051 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
20054 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
20055 based on LABEL_REF. Remove useless assertion.
20056 (pic_address_needs_scratch): Fix formatting.
20057 (sparc_legitimize_pic_address): Minor tweaks.
20058 (sparc_delegitimize_address): Adjust assertion accordingly.
20059 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
20060 into symbolic_operand.
20061 (movsi_high_pic_label_ref): Likewise.
20062 (movsi_lo_sum_pic_label_ref): Likewise.
20063 (movdi_pic_label_ref): Likewise.
20064 (movdi_high_pic_label_ref): Likewise.
20065 (movdi_lo_sum_pic_label_ref): Likewise.
20067 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
20069 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
20070 custom LIB_SPEC setup.
20072 2018-04-06 Ruslan Bukin <br@bsdpad.com>
20073 Kito Cheng <kito.cheng@gmail.com>
20075 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
20076 * config/riscv/freebsd.h: New.
20078 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
20080 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
20081 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
20084 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
20085 Kito Cheng <kito.cheng@gmail.com>
20087 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
20088 nds32_output_call, nds32_symbol_binds_local_p): New functions.
20089 * config/nds32/nds32-protos.h (nds32_output_call,
20090 nds32_output_return): Declare.
20091 * config/nds32/nds32.md: Refine all the call and return patterns.
20093 2018-04-06 Jakub Jelinek <jakub@redhat.com>
20096 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
20097 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
20099 PR rtl-optimization/84872
20100 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
20101 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
20102 EDGE_CROSSING edge.
20104 2018-04-06 Tamar Christina <tamar.christina@arm.com>
20106 * expr.c (copy_blkmode_to_reg): Revert 254862.
20107 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
20109 2018-04-06 Richard Biener <rguenther@suse.de>
20111 PR middle-end/85244
20112 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
20113 after seeing a component reference with an adjacent field. Treat
20114 refs to arrays at struct end of external decls similar to
20115 refs to unconstrained commons.
20117 2018-04-06 Jakub Jelinek <jakub@redhat.com>
20120 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
20121 look through SAVE_EXPRs with non-side-effects argument. Adjust
20123 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
20126 2018-04-06 Richard Biener <rguenther@suse.de>
20128 PR middle-end/85180
20129 * alias.c (find_base_term): New wrapper around find_base_term
20130 unwinding CSELIB_VAL_PTR changes.
20131 (find_base_term): Do not restore CSELIB_VAL_PTR during the
20134 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
20136 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
20138 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
20139 constant definitions.
20140 ("nop"): lr 0,0 -> nopr r0
20141 ("nop_lr0", "nop_lr1"): New insn definitions.
20143 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
20145 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
20146 NDS32_V3PUSH_AVAILABLE_P macro.
20148 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
20149 Chung-Ju Wu <jasonwucj@gmail.com>
20151 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
20152 (nds32*-*-*): Add float and fpu_config into supported_defaults.
20153 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
20154 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
20155 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
20156 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
20157 * config/nds32/constraints.md: New constraints and checking for hard
20158 float configuration.
20159 * config/nds32/iterators.md: New mode iterator and attribute for hard
20160 float configuration.
20161 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
20163 * config/nds32/nds32-fpu.md: New file.
20164 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
20165 deal with hard float code generation.
20166 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
20168 (abi_type, float_reg_number): New enum type.
20169 * config/nds32/nds32-predicates.c: New predicates for hard float.
20170 * config/nds32/nds32-protos.h: Declare functions for hard float.
20171 * config/nds32/nds32.c: Implementation for hard float configuration.
20172 * config/nds32/nds32.h: Definitions for hard float configuration.
20173 * config/nds32/nds32.md: Include hard float machine description and
20174 modify patterns for hard float configuration.
20175 * config/nds32/nds32.opt: New options for hard float configuration.
20176 * config/nds32/predicates.md: New predicates for hard float
20179 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
20181 * common/config/nds32/nds32-common.c
20182 (nds32_option_optimization_table): Enable -mreleax-hint by default.
20184 2018-04-05 Jakub Jelinek <jakub@redhat.com>
20186 PR middle-end/85195
20187 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
20188 CONSTRUCTOR_ELT (ctor, ...)->value.
20190 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
20193 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
20195 2018-04-05 Tom de Vries <tom@codesourcery.com>
20198 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
20201 2018-04-05 Shiva Chen <shiva0217@gmail.com>
20202 Kito Cheng <kito.cheng@gmail.com>
20204 * config/nds32/constraints.md (U33): Fine-tune checking condition.
20205 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
20206 * config/nds32/nds32.h (nds32_16bit_address_type): Add
20207 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
20209 2018-04-05 Shiva Chen <shiva0217@gmail.com>
20210 Kito Cheng <kito.cheng@gmail.com>
20212 * config/nds32/constraints.md (Ufe): New memory constraint.
20213 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
20214 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
20215 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
20217 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
20218 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
20220 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
20222 * config/nds32/nds32.md: Use optimize_size in the condition for
20223 alu-shift instructions.
20225 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
20227 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
20229 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
20231 * config/nds32/nds32.md (negsi2): Refine pattern.
20233 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
20234 Chung-Ju Wu <jasonwucj@gmail.com>
20236 * config/nds32/iterators.md (shift_rotate): New code iterator.
20237 (shift): New code attribute.
20238 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
20239 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
20240 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
20241 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
20242 bit-wise operations.
20243 (andsi3, *andsi3): Ditto.
20244 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
20245 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
20246 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
20247 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
20248 nds32_ior_operand, nds32_xor_operand): New predicates.
20250 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
20252 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
20253 (addsi3, subsi3): ... this.
20255 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
20257 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
20259 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
20261 * config/nds32/nds32.md: Adjust indention.
20263 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
20265 * config/nds32/nds32.md (feature): New attribute.
20267 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
20269 * config/nds32/nds32.md (subtype): New attribute.
20271 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
20274 * config/arm/arm-builtins.c (arm_expand_builtin): Change
20275 expansion to perform a bitwise AND of the argument followed by a
20276 boolean negation of the result.
20278 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
20280 PR rtl-optimization/84878
20281 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
20282 the basic block. Assert the use reference is not artificial and that
20283 it has an associated insn.
20285 2018-04-04 Michael Matz <matz@suse.de>
20287 * builtins.c (compute_objsize): Pass correct operand
20288 to array_at_struct_end_p.
20290 2018-04-04 Richard Biener <rguenther@suse.de>
20293 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
20294 from contexts for DINFO_LEVEL_TERSE and below.
20296 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
20298 * config/nds32/nds32-doubleword.md (move_<mode>): Require
20299 resiter_operand condition.
20300 * config/nds32/nds32.md (*move<mode>): Ditto.
20302 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
20303 Monk Chiang <sh.chiang04@gmail.com>
20305 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
20307 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
20309 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
20311 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
20312 Kito Cheng <kito.cheng@gmail.com>
20314 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
20315 nds32_cond_code_str, output_cond_branch,
20316 output_cond_branch_compare_zero, nds32_expand_cbranch,
20317 nds32_expand_cstore, nds32_expand_movcc,
20318 nds32_output_cbranchsi4_equality_zero,
20319 nds32_output_cbranchsi4_equality_reg,
20320 nds32_output_cbranchsi4_equality_reg_or_const_int,
20321 nds32_output_cbranchsi4_greater_less_zero: New functions.
20322 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
20323 nds32_expand_cstore, nds32_expand_movcc,
20324 nds32_output_cbranchsi4_equality_zero,
20325 nds32_output_cbranchsi4_equality_reg,
20326 nds32_output_cbranchsi4_equality_reg_or_const_int,
20327 nds32_output_cbranchsi4_greater_less_zero): Declare.
20328 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
20329 nds32_rimm11s_operand): New predicates.
20330 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
20331 * config/nds32/nds32.md: Rewrite all the branch and conditional move
20334 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
20336 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
20337 * config/nds32/nds32.md: Ditto.
20338 * config/nds32/pipelines.md: Ditto.
20340 2018-04-04 Richard Biener <rguenther@suse.de>
20342 PR tree-optimization/85168
20343 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
20344 propagating abnormals.
20346 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
20348 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
20350 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
20351 Kito Cheng <kito.cheng@gmail.com>
20353 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
20354 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
20355 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
20356 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
20357 * config/nds32/nds32.md (sibcall_internal): New.
20358 (sibcall_register): Remove.
20359 (sibcall_immediate): Remove.
20360 (sibcall_value_internal): New.
20361 (sibcall_value_register): Remove.
20362 (sibcall_value_immediate): Remove.
20363 * config/nds32/predicates.md (nds32_general_register_operand): New.
20364 (nds32_call_address_operand): New.
20366 2018-04-03 Jakub Jelinek <jakub@redhat.com>
20368 PR rtl-optimization/85167
20369 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
20370 bb_defs if *split_p, instead preinitialize it to NULL.
20372 PR tree-optimization/85156
20373 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
20374 evaluating the argument multiple times.
20376 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
20378 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
20380 (_mm_cvtpd_ps): Likewise.
20381 (_mm_cvttpd_epi32): Likewise.
20382 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
20383 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
20384 vector, pixel, and bool following altivec.h include.
20386 2018-04-03 Martin Sebor <msebor@redhat.com>
20388 * doc/extend.texi (Common Function Attributes): Clarify.
20389 (const attribute): Likewise.
20390 (pure attribute): Likewise.
20392 2018-04-03 Jakub Jelinek <jakub@redhat.com>
20395 * config/i386/i386.c (ix86_expand_vector_set): Use
20396 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
20398 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
20400 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
20401 instructions when changing rounding bits to preserve precision bits
20402 in the x87 control word.
20404 2018-04-03 Martin Liska <mliska@suse.cz>
20406 PR tree-optimization/82491
20407 * rtl.h (strip_offset_and_add): Replace += suboffset with
20408 poly_uint64 () + suboffset.
20410 2018-03-29 Martin Liska <mliska@suse.cz>
20411 Martin Jambor <mjambor@suse.cz>
20414 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
20415 param_type is not an integral or pointer type.
20417 2018-04-03 Richard Biener <rguenther@suse.de>
20419 * sese.h (recompute_all_dominators): Remove.
20421 2018-04-02 Martin Sebor <msebor@redhat.com>
20423 * doc/invoke.texi (-Wrestrict): Fix typos.
20425 2018-04-02 Jim Wilson <jimw@sifive.com>
20427 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
20428 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
20429 (<optab>di3, <optab>si3_extend): Likewise.
20430 (<optab>si3_mask, <optab>si3_mask_1): New.
20431 (<optab>di3_mask, <optab>di3_mask_1): New.
20432 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
20433 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
20434 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
20436 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
20438 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
20441 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
20443 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
20444 (nds32_canonicalize_comparison): New function.
20446 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
20447 Kito Cheng <kito.cheng@gmail.com>
20448 Kuan-Lin Chen <kuanlinchentw@gmail.com>
20450 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
20451 * config/nds32/constants.md (unspec_volatile_element): Add
20452 UNSPEC_VOLATILE_RELAX_GROUP.
20453 * config/nds32/nds32-relax-opt.c: New file.
20454 * config/nds32/nds32-predicates.c
20455 (nds32_symbol_load_store_p): New function.
20456 * config/nds32/nds32-protos.h
20457 (nds32_symbol_load_store_p): Declare function.
20458 (make_pass_nds32_relax_opt): Declare new rtl pass function.
20459 * config/nds32/nds32.c
20460 (nds32_register_pass): New function to register pass.
20461 (nds32_register_passes): New function to register passes.
20462 * config/nds32/nds32.md (relax_group): New pattern.
20463 * config/nds32/nds32.opt (mrelax-hint): New option.
20464 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
20466 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
20468 * config/nds32/t-nds32: Modify files dependency.
20470 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
20472 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
20473 (PROFILE_HOOK): Define its implementation.
20475 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
20477 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
20478 type and 32-bit size.
20480 2018-04-01 Jakub Jelinek <jakub@redhat.com>
20482 PR middle-end/85090
20483 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
20484 (V_128_256): New mode iterator.
20485 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
20486 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
20487 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
20489 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
20490 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
20492 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
20495 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
20496 NaN inputs correctly.
20498 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
20501 * config/rs6000/vsx.md (??r): New mode attribute.
20502 (*vsx_mov<mode>_64bit): Use it.
20503 (*vsx_mov<mode>_32bit): Likewise.
20505 2018-03-30 Martin Sebor <msebor@redhat.com>
20507 PR tree-optimization/84818
20508 * builtins.c (check_access): Use warning_n.
20510 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
20513 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
20515 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
20518 2018-03-30 Julia Koval <julia.koval@intel.com>
20521 * x86-tune.def (movx, partial_reg_dependency): Enable for
20524 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
20526 PR inline-asm/84985
20527 * lra-constraints.c (process_alt_operands): Move setting
20528 this_alternative_matches below.
20530 2018-03-29 Martin Liska <mliska@suse.cz>
20533 * doc/invoke.texi: Document how LTO works with debug info.
20534 Describe auto-load support of binutils. Mention 'x86-64'
20535 as valid option value of -march option.
20537 2018-03-29 Jakub Jelinek <jakub@redhat.com>
20539 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
20542 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
20543 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
20544 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
20547 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
20550 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
20551 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
20552 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
20553 for RS6000_BTM_POWERPC64.
20554 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
20555 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
20556 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
20559 (DIVDEU): Likewise.
20561 2018-03-28 Carl Love <cel@us.ibm.com>
20564 2017-09-27 Carl Love <cel@us.ibm.com>
20566 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
20567 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
20568 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
20571 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20573 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
20574 instead of __vector bool.
20575 (_mm_max_pu8): Likewise.
20576 (_mm_min_pi16): Likewise.
20578 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
20581 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
20582 (DIVWEUO): Likewise.
20583 (DIVDEO): Likewise.
20584 (DIVDEUO): Likewise.
20585 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
20586 DIVWEUO and DIVDEUO.
20587 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
20588 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
20589 (div_extend): Likewise.
20590 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
20592 (__builtin_divweuo): Likewise.
20593 (__builtin_divdeo): Likewise.
20594 (__builtin_divdeuo): Likewise.
20596 2018-03-28 Jakub Jelinek <jakub@redhat.com>
20599 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
20600 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
20602 PR tree-optimization/82004
20603 * gimple-match-head.c (optimize_pow_to_exp): New function.
20604 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
20605 Don't fold to exp if optimize_pow_to_exp is false.
20607 2018-03-28 Martin Liska <mliska@suse.cz>
20610 * calls.c (initialize_argument_information): Fix trailing space.
20611 * common.opt: Fix typo and provide better explanation for
20612 -fsanitize-coverage option.
20613 * config/i386/i386.opt: Fix typo.
20615 2018-03-28 Jakub Jelinek <jakub@redhat.com>
20616 Martin Liska <mliska@suse.cz>
20619 * gimplify.c (asan_poison_variable): Don't do the check for
20620 gimplify_omp_ctxp here.
20621 (gimplify_decl_expr): Do it here.
20622 (gimplify_target_expr): Likewise.
20624 2018-03-28 Martin Liska <mliska@suse.cz>
20627 * config/i386/i386.c (ix86_function_arg_advance): Do not call
20628 chkp_type_bounds_count if MPX is not enabled.
20630 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
20632 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
20634 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
20637 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
20638 function to create the function decl for complex long double
20639 multiply and divide for -mabi=ieeelongdouble.
20640 (init_float128_ieee): Call it.
20642 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
20645 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
20646 -fcf-protection=branch -mibt.
20647 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
20649 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
20652 * config/arm/arm.c (arm_valid_symbolic_address): Handle
20653 arm_word_relocations.
20655 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
20658 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
20659 extern array declarations.
20661 2018-03-27 Richard Biener <rguenther@suse.de>
20663 PR middle-end/84067
20664 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
20665 explicit single_use checks.
20667 2018-03-27 Richard Biener <rguenther@suse.de>
20669 PR tree-optimization/85082
20670 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
20673 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
20675 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
20676 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
20677 Turn on fasynchronous-unwind-tables and funwind-tables.
20679 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
20682 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
20683 (*bmi_blsr_<mode>_ccz): Ditto.
20685 2018-03-26 Tom de Vries <tom@codesourcery.com>
20687 PR tree-optimization/85063
20688 * omp-general.c (offloading_function_p): New function. Factor out
20690 * omp-offload.c (pass_omp_target_link::gate): ... here.
20691 * omp-general.h (offloading_function_p): Declare.
20692 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
20693 with attribute omp declare target for offloading functions.
20695 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
20697 PR tree-optimization/84005
20698 * tree-data-ref.h (get_base_for_alignment): Declare.
20699 * tree-data-ref.c (get_base_for_alignment_1): New function.
20700 (get_base_for_alignment): Likewise.
20701 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
20702 get_base_for_alignment to find a suitable base object, instead
20703 of always using drb->base_address.
20705 2018-03-23 Jakub Jelinek <jakub@redhat.com>
20707 PR inline-asm/85022
20708 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
20709 known size by default.
20711 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
20713 PR inline-asm/85030
20714 * lra-constraints.c (process_alt_operands): Don't match BLKmode
20715 and non BLKmode operands.
20717 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20720 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
20721 Clean up attributes.
20723 2018-03-23 Richard Biener <rguenther@suse.de>
20726 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
20727 we are going to emit early debug for LTO.
20729 2018-03-23 Jakub Jelinek <jakub@redhat.com>
20731 PR inline-asm/85034
20732 * function.c (match_asm_constraints_1): Don't optimize if input
20733 doesn't satisfy general_operand predicate for output's mode.
20735 PR inline-asm/85022
20736 * alias.c (write_dependence_p): Don't require for x_canonicalized
20737 non-VOIDmode if x has VOIDmode.
20740 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
20741 just don't try to optimize it rather than assert it never happens.
20743 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
20745 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
20746 macro expansions for definition of ST_INTERNAL_<mode> and
20747 LD_INTERNAL_<mode> builtins.
20748 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
20750 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
20752 (altivec_expand_st_builtin): Likewise.
20753 (altivec_expand_builtin): Remove calls to deleted functions.
20754 (rs6000_address_for_altivec): Delete this function.
20755 * config/rs6000/vector.md: Remove expands for
20756 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
20758 2018-03-22 Sudakshina Das <sudi.das@arm.com>
20761 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
20762 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
20763 re-computing once computed.
20764 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
20765 (arm_init_machine_status): Initialize
20766 machine->static_chain_stack_bytes.
20768 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
20771 * doc/extend.texi: Add four new prototypes for vec_ld.
20772 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
20773 definitions for more logical presentation.
20774 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
20775 entries for V1TI variants of __builtin_altivec_ld builtin.
20776 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
20777 handling of V1TI variant of LVX icode pattern.
20778 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
20779 (rs6000_gimple_fold_builtin): Likewise.
20780 (altivec_init_builtins): Add code to define
20781 __builtin_altivec_lvx_v1ti function.
20783 2018-03-22 Jakub Jelinek <jakub@redhat.com>
20785 PR inline-asm/84941
20786 * function.c (match_asm_constraints_1): Don't do the optimization
20787 if input isn't a REG, SUBREG, MEM or constant.
20789 2018-03-22 Tom de Vries <tom@codesourcery.com>
20791 PR tree-optimization/84956
20792 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
20793 bb_has_abnormal_pred.
20795 2018-03-22 Jakub Jelinek <jakub@redhat.com>
20798 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
20799 DECL_INITIAL (decl) to decl at the end.
20800 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
20801 adjust the comment.
20803 2018-03-21 Joseph Myers <joseph@codesourcery.com>
20805 * doc/extend.texi (__builtin_tgmath): Document when complex
20806 integer types are treated as _Complex _Float64.
20808 2018-03-21 Tom de Vries <tom@codesourcery.com>
20810 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
20812 2018-03-21 Jakub Jelinek <jakub@redhat.com>
20814 PR tree-optimization/84960
20815 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
20816 if it is ENTRY block, move them into single succ of ENTRY in that case.
20818 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
20820 PR tree-optimization/84811
20821 * poly-int.h (poly_span_traits): Remove the T3 parameter and
20822 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
20823 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
20824 (known_subrange_p): Update accordingly. Cast each value involved
20825 in the size comparison, rather than casting the result of the
20828 2018-03-21 Jakub Jelinek <jakub@redhat.com>
20830 PR tree-optimization/84982
20831 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
20832 by flipping the least significant bit rather than all bits from
20833 bitpos to bitpos + bitsize - 1.
20835 2018-03-21 Nathan Sidwell <nathan@acm.org>
20837 * doc/extend.texi (Deprecated Features): Remove mention of
20838 long-deleted deprecations.
20840 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20843 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
20844 * configure: Regenerate.
20846 2018-03-21 Tom de Vries <tom@codesourcery.com>
20848 PR tree-optimization/83126
20849 * tree-parloops.c (num_phis): New function.
20850 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
20852 2018-03-21 Nathan Sidwell <nathan@acm.org>
20854 * doc/extend.texi (Deprecated Features): Update deprecated flags,
20855 mention anon-struct/union members and trailing attributes.
20857 2018-03-21 Bin Cheng <bin.cheng@arm.com>
20859 PR tree-optimization/84969
20860 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
20861 builtin memset partitions if they set different rhs values.
20863 2018-03-21 Jakub Jelinek <jakub@redhat.com>
20865 PR rtl-optimization/84989
20866 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
20867 VEC_DUPLICATE with scalar result mode.
20869 2018-03-21 Martin Liska <mliska@suse.cz>
20872 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
20873 not intended return statement.
20875 2018-03-21 Martin Liska <mliska@suse.cz>
20878 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
20879 (chkp_find_bound_slots_1): Limit number of iterations.
20881 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
20884 * Minor grammar fixes for x86 options.
20886 2018-03-20 Jakub Jelinek <jakub@redhat.com>
20889 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
20890 holding REG_CFA_RESTORE notes, instead turn them into a USE.
20892 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
20895 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
20896 (altivec_lvx_<mode>_1op): Likewise.
20897 (altivec_stvx_<mode>_2op): Likewise.
20898 (altivec_stvx_<mode>_1op): Likewise.
20899 (altivec_lvx_<VM2:mode>): New define_expand.
20900 (altivec_stvx_<VM2:mode>): Likewise.
20901 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
20902 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
20903 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
20904 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
20905 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
20906 (rs6000_gen_lvx): Likewise.
20907 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
20908 (altivec_expand_stv_builtin): Likewise.
20909 (altivec_expand_builtin): Likewise.
20910 * config/rs6000/vector.md: Likewise.
20912 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20915 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
20918 2018-03-20 Richard Biener <rguenther@suse.de>
20921 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
20922 sign-conversions as zero, fall back to standard scalar_stmt
20925 2018-03-20 Martin Liska <mliska@suse.cz>
20928 * predict.c (rebuild_frequencies): Handle case when we have
20929 PROFILE_ABSENT, but flag_guess_branch_prob is false.
20931 2018-03-20 Jakub Jelinek <jakub@redhat.com>
20934 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
20935 flag_section_anchors.
20936 * varasm.c (use_blocks_for_decl_p): Remove hack for
20937 dw2_force_const_mem.
20940 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
20942 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
20943 be created, use lowpart_subreg of operands[0] rather than operands[0]
20945 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
20946 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
20947 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
20948 and n constraint instead of aarch64_shift_imm_di and Usd.
20949 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
20950 (*aarch64_<optab>_reg_minus<mode>3): ... this.
20952 2018-03-20 Sudakshina Das <sudi.das@arm.com>
20955 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
20956 to favor GPR over NEON registers.
20957 (<shift>di3_neon): Likewise.
20959 2018-03-20 Tom de Vries <tom@codesourcery.com>
20962 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
20963 (nvptx_process_pars): Emit bar.sync asap and alap.
20965 2018-03-20 Tom de Vries <tom@codesourcery.com>
20968 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
20969 seen_label if seen_label is already set.
20971 2018-03-20 Jakub Jelinek <jakub@redhat.com>
20974 * config/i386/i386.c (fold_builtin_cpu): For features above 31
20975 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
20976 Use 1U instead of 1. Formatting fixes.
20979 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
20980 instead of TREE_TYPE (s1) for the return value.
20982 2018-03-19 Jakub Jelinek <jakub@redhat.com>
20984 PR tree-optimization/84946
20985 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
20986 bitsize + bitsize in poly_uint64 rather than poly_int64.
20989 * dwarf2asm.c: Include fold-const.c.
20990 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
20991 of decl rather than decl itself.
20993 PR rtl-optimization/84643
20994 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
20996 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
20999 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
21000 calling assemble_variable.
21002 2018-03-19 Sudakshina Das <sudi.das@arm.com>
21005 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
21006 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
21008 2018-03-19 Jim Wilson <jimw@sifive.com>
21011 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
21012 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
21013 (riscv_first_stack_step): Likewise.
21014 (riscv_option_override): Use STACK_BOUNDARY instead of
21015 MIN_STACK_BOUNDARY.
21016 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
21017 MIN_STACK_BOUNDARY.
21018 (BIGGEST_ALIGNMENT): Set to 128.
21019 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
21020 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
21023 2018-03-19 Richard Biener <rguenther@suse.de>
21025 PR tree-optimization/84933
21026 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
21027 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
21029 2018-03-19 Richard Biener <rguenther@suse.de>
21031 PR tree-optimization/84859
21032 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
21033 (cond_if_else_store_replacement): Perform sinking operation on
21034 single-store BBs regardless of MAX_STORES_TO_SINK setting.
21035 Generalize what a BB with a single eligible store is.
21037 2018-03-19 Richard Biener <rguenther@suse.de>
21039 PR tree-optimization/84929
21040 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
21041 chrec_is_positive against non-chrec arg.
21043 2018-03-19 Tamar Christina <tamar.christina@arm.com>
21046 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
21048 2018-03-18 Martin Liska <mliska@suse.cz>
21050 PR rtl-optimization/84635
21051 * regrename.c (build_def_use): Use matches_mode only when
21054 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
21056 PR tree-optimization/84913
21057 * tree-vect-loop.c (vectorizable_reduction): Don't try to
21058 vectorize chains of COND_EXPRs.
21060 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
21062 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
21064 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
21066 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
21068 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
21070 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
21072 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
21073 Kito Cheng <kito.cheng@gmail.com>
21075 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
21076 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
21077 (nds32_adjust_reg_alloc_order): New function.
21078 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
21080 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
21082 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
21083 nds32_print_operand, nds32_print_operand_address): Use
21084 HOST_WIDE_INT_PRINT_DEC instead.
21086 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
21088 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
21090 2018-03-17 Jakub Jelinek <jakub@redhat.com>
21093 * config/i386/i386.c (initial_ix86_tune_features,
21094 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
21095 unsigned long long.
21096 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
21097 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
21098 rather than 1u << ix86_tune. Formatting fix.
21099 (ix86_option_override_internal): Change ix86_arch_mask from
21100 unsigned int to unsigned HOST_WIDE_INT, initialize to
21101 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
21102 (ix86_function_specific_restore): Likewise.
21104 2018-03-16 Jakub Jelinek <jakub@redhat.com>
21107 * postreload.c (reload_combine_recognize_pattern): Perform
21108 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
21109 truncate_int_for_mode the result for the destination's mode.
21112 * hsa-gen.c (mem_type_for_type): Fix comment typo.
21113 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
21115 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
21118 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
21121 * lra-assigns.c (lra_split_hard_reg_for): Don't use
21122 regno_allocno_class_array and sorted_pseudos.
21123 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
21124 insns where regno is used.
21126 2018-03-16 Martin Liska <mliska@suse.cz>
21129 * multiple_target.c (create_dispatcher_calls): Redirect
21130 reference in the symbol table.
21132 2018-03-16 Martin Liska <mliska@suse.cz>
21135 * multiple_target.c (create_dispatcher_calls): Redirect also
21138 2018-03-16 Jakub Jelinek <jakub@redhat.com>
21142 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
21143 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
21144 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
21146 2018-03-16 Julia Koval <julia.koval@intel.com>
21148 * doc/invoke.texi (Skylake Server): Add CLWB.
21149 Cannonlake): Remove CLWB.
21151 2018-03-16 Jakub Jelinek <jakub@redhat.com>
21153 PR tree-optimization/84841
21154 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
21156 (FLOAT_ONE_CONST_TYPE): Define.
21157 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
21158 (sort_by_operand_rank): Put entries with higher constant_type last
21159 rather than first to match comments.
21161 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
21163 * config/nios2/nios2.md (movsi_internal): Fix thinko in
21166 2018-03-15 Jakub Jelinek <jakub@redhat.com>
21169 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
21170 check and use address of target always.
21172 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
21175 * config/i386/i386.c (indirect_thunk_needed): Update comments.
21176 (indirect_thunk_bnd_needed): Likewise.
21177 (indirect_thunks_used): Likewise.
21178 (indirect_thunks_bnd_used): Likewise.
21179 (indirect_return_needed): New.
21180 (indirect_return_bnd_needed): Likewise.
21181 (output_indirect_thunk_function): Add a bool argument for
21183 (output_indirect_thunk_function): Don't generate alias for
21184 function return thunk.
21185 (ix86_code_end): Call output_indirect_thunk_function to generate
21186 function return thunks.
21187 (ix86_output_function_return): Set indirect_return_bnd_needed
21188 and indirect_return_needed instead of indirect_thunk_bnd_needed
21189 and indirect_thunk_needed.
21191 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
21193 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
21194 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
21195 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
21197 2018-03-15 David Malcolm <dmalcolm@redhat.com>
21198 Paul Hua <paul.hua.gm@gmail.com>
21201 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
21203 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
21205 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
21206 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
21207 resp. SFmode cases.
21209 2018-03-15 Tamar Christina <tamar.christina@arm.com>
21212 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
21213 instead of GET_MODE_SIZE when comparing Units.
21215 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
21218 * varasm.c (hash_section): Return an unchangeble hash value
21219 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
21220 Return !aarch64_can_use_per_function_literal_pools_p ().
21222 2018-03-15 Jakub Jelinek <jakub@redhat.com>
21225 * optabs.c (emit_conditional_move): Pass address of cmode's copy
21226 rather than address of cmode as last argument to prepare_cmp_insn.
21228 2018-03-15 Julia Koval <julia.koval@intel.com>
21230 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
21231 F_AVX512VNNI, F_AVX512BITALG): New.
21233 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
21236 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
21237 insn for floating-point loads and stores.
21239 2018-03-14 Carl Love <cel@us.ibm.com>
21241 * config/rs6000/rs6000-c.c: Add macro definitions for
21242 ALTIVEC_BUILTIN_VEC_PERMXOR.
21243 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
21244 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
21245 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
21246 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
21248 * config/doc/extend.texi: Add prototypes for vec_permxor.
21250 2018-03-14 David Malcolm <dmalcolm@redhat.com>
21253 * diagnostic-show-locus.c (class layout_point): Convert m_line
21254 from int to linenum_type.
21255 (line_span::comparator): Use linenum "compare" function when
21256 comparing line numbers.
21257 (test_line_span): New function.
21258 (layout_range::contains_point): Convert param "row" from int to
21260 (layout_range::intersects_line_p): Likewise.
21261 (layout::will_show_line_p): Likewise.
21262 (layout::print_source_line): Likewise.
21263 (layout::should_print_annotation_line_p): Likewise.
21264 (layout::print_annotation_line): Likewise.
21265 (layout::print_leading_fixits): Likewise.
21266 (layout::annotation_line_showed_range_p): Likewise.
21267 (struct line_corrections): Likewise for field m_row.
21268 (line_corrections::line_corrections): Likewise for param "row".
21269 (layout::print_trailing_fixits): Likewise.
21270 (layout::get_state_at_point): Likewise.
21271 (layout::get_x_bound_for_row): Likewise.
21272 (layout::print_line): Likewise.
21273 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
21274 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
21275 * input.c (selftest::test_linenum_comparisons): New function.
21276 (selftest::input_c_tests): Call it.
21277 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
21278 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
21279 * selftest.h (ASSERT_GT): New macro.
21280 (ASSERT_GT_AT): New macro.
21281 (ASSERT_LT): New macro.
21282 (ASSERT_LT_AT): New macro.
21284 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
21286 PR rtl-optimization/84780
21287 * combine.c (distribute_links): Don't make a link based on pc_rtx.
21289 2018-03-14 Martin Liska <mliska@suse.cz>
21291 * tree.c (record_node_allocation_statistics): Use
21292 get_stats_node_kind.
21293 (get_stats_node_kind): New function extracted from
21294 record_node_allocation_statistics.
21295 (free_node): Use get_stats_node_kind.
21297 2018-03-14 Richard Biener <rguenther@suse.de>
21299 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
21300 that the value-set of ANTIC_IN doesn't grow.
21303 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
21305 (BB_VISITED_WITH_VISITED_SUCCS): New define.
21306 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
21308 2018-03-14 Julia Koval <julia.koval@intel.com>
21310 * config.gcc (icelake-client, icelake-server): New.
21312 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
21313 (initial_ix86_arch_features): Ditto.
21314 (PTA_SKYLAKE): Add SGX.
21315 (PTA_ICELAKE): Remove.
21316 (PTA_ICELAKE_CLIENT): New.
21317 (PTA_ICELAKE_SERVER): New.
21318 (ix86_option_override_internal): Split up icelake on icelake client and
21320 (get_builtin_code_for_version): Ditto.
21321 (fold_builtin_cpu): Ditto.
21322 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
21323 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
21324 * config/i386/i386.h (processor_type): Ditto.
21325 * doc/invoke.texi: Ditto.
21327 2018-03-14 Jakub Jelinek <jakub@redhat.com>
21330 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
21331 INTEGER_CST offset, add it together with bitpos / 8 and
21332 sign extend based on POINTER_SIZE.
21336 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
21339 * config/i386/constraints.md (Yc): New register constraint.
21340 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
21341 Use Yc constraint for alternative 2 of operand 0. Remove
21342 preferred_for_speed attribute.
21344 2018-03-14 Richard Biener <rguenther@suse.de>
21346 PR tree-optimization/84830
21347 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
21348 with the old one to avoid oscillations.
21350 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
21353 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
21355 (assign_by_spills): Return a flag of reload assignment failure.
21356 Do not process the reload assignment failures. Do not spill other
21357 reload pseudos if they has the same reg class. Update n if
21359 (lra_assign): Add a return arg. Set up from the result of
21360 assign_by_spills call.
21361 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
21362 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
21363 usage_insns if it is not NULL.
21364 (spill_hard_reg_in_range): New function.
21365 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
21366 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
21367 function prototypes.
21368 (lra_assign): Change prototype.
21369 * lra.c (lra): Add code to deal with fails by splitting hard reg
21372 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
21374 * config/riscv/riscv.opt (mrelax): New option.
21375 * config/riscv/riscv.c (riscv_file_start): Emit ".option
21376 "norelax" when riscv_mrelax is disabled.
21377 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
21379 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
21382 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
21383 reassociation for int modes.
21385 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
21387 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
21388 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
21390 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
21391 * config/aarch64/aarch64-sve.md
21392 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
21393 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
21394 (*extend<mode><Vwide>2): Rename to...
21395 (aarch64_sve_extend<mode><Vwide>2): ...this.
21396 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
21397 renaming the old pattern to...
21398 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
21400 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
21401 define_expand, renaming the old pattern to...
21402 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
21403 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
21404 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
21405 account when deciding which SVE instruction the optab should use.
21406 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
21408 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
21410 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
21411 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
21412 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
21413 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
21414 (tlsdesc_small_<mode>): Turn a define_expand and use
21415 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
21416 (tlsdesc_small_advsimd_<mode>): ...this.
21417 (tlsdesc_small_sve_<mode>): New pattern.
21419 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
21421 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
21422 (UNSPEC_UMUL_HIGHPART): New constants.
21423 (MUL_HIGHPART): New int iteraor.
21424 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
21425 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
21427 (*<su>mul<mode>3_highpart): New define_insn.
21429 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
21432 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
21435 2018-03-13 Martin Liska <mliska@suse.cz>
21438 * (sem_item_optimizer::sem_item_optimizer): Initialize new
21440 (sem_item_optimizer::~sem_item_optimizer): Release it.
21441 (sem_item_optimizer::merge_classes): Register variable aliases.
21442 (sem_item_optimizer::fixup_pt_set): New function.
21443 (sem_item_optimizer::fixup_points_to_sets): Likewise.
21444 * ipa-icf.h: Declare new variables and functions.
21446 2018-03-13 Jakub Jelinek <jakub@redhat.com>
21448 PR middle-end/84834
21449 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
21450 integer_pow2p@2 and test integer_pow2p in condition.
21451 (A < 0 ? C : 0): Similarly for @1.
21453 PR middle-end/84831
21454 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
21455 characters starting at p contain '\0' character, don't look beyond
21459 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
21460 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
21463 * reg-stack.c (change_stack): Change update_end var from int to
21464 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
21465 also call set_block_for_insn on the newly added insns and rescan.
21468 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
21469 on the last operand.
21472 * tree.c (stabilize_reference_1): Return save_expr (e) for
21473 STATEMENT_LIST even if it doesn't have side-effects.
21475 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
21477 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
21479 2018-03-12 Renlin Li <renlin.li@arm.com>
21481 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
21482 aarch64_output_scalar_simd_mov_immediate.
21484 2018-03-12 Martin Sebor <msebor@redhat.com>
21486 PR tree-optimization/83456
21487 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
21488 for perfectly overlapping calls to memcpy.
21489 (gimple_fold_builtin_memory_chk): Same.
21490 (gimple_fold_builtin_strcpy): Handle no-warning.
21491 (gimple_fold_builtin_stxcpy_chk): Same.
21492 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
21494 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
21496 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
21497 parameter. Use it for SFmode.
21498 (rs6000_function_arg_advance_1): Adjust.
21499 (rs6000_function_arg): Adjust.
21500 (rs6000_gimplify_va_arg): Pass false for that new parameter.
21502 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
21504 PR rtl-optimization/84169
21505 PR rtl-optimization/84780
21506 * combine.c (can_combine_p): Check for a 2-insn combination whether
21507 the destination register is used between the two insns, too.
21509 2018-03-12 Richard Biener <rguenther@suse.de>
21511 PR tree-optimization/84803
21512 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
21513 for refs DR analysis didn't process.
21515 2018-03-12 Richard Biener <rguenther@suse.de>
21517 PR tree-optimization/84777
21518 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
21519 force-vectorize loops ignore whether we are optimizing for size.
21521 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
21523 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
21524 (TARGET_MD_ASM_ADJUST): Define.
21526 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
21527 Kito Cheng <kito.cheng@gmail.com>
21528 Chung-Ju Wu <jasonwucj@gmail.com>
21530 * config/nds32/nds32.c (nds32_compute_stack_frame,
21531 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
21532 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
21533 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
21534 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
21535 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
21536 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
21537 * config/nds32/nds32.md (prologue, epilogue): Use macro
21538 NDS32_V3PUSH_AVAILABLE_P to do checking.
21540 2018-03-11 Jakub Jelinek <jakub@redhat.com>
21543 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
21544 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
21545 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
21546 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
21547 addition of most attributes on !orig_type_die or the attribute not
21548 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
21550 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
21551 Chung-Ju Wu <jasonwucj@gmail.com>
21553 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
21554 __NDS32_VH__ macro.
21555 * config/nds32/nds32.opt (mvh): New option.
21557 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
21558 Chung-Ju Wu <jasonwucj@gmail.com>
21560 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
21562 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
21563 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
21566 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
21567 Chung-Ju Wu <jasonwucj@gmail.com>
21569 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
21571 * config/nds32/nds32-multiple.md (strlensi): New pattern.
21572 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
21574 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
21575 Kito Cheng <kito.cheng@gmail.com>
21576 Chung-Ju Wu <jasonwucj@gmail.com>
21578 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
21579 UNSPEC_FFMISM and UNSPEC_FLMISM.
21580 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
21581 for ffb, ffmism and flmism.
21582 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
21583 (unspec_ffmism): Ditto.
21584 (unspec_flmism): Ditto.
21585 (nds32_expand_builtin_impl): Check if string extension is available.
21586 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
21587 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
21589 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
21592 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
21595 * lra-assigns.c (assign_by_spills): Return a flag of reload
21596 assignment failure. Do not process the reload assignment
21597 failures. Do not spill other reload pseudos if they has the same
21599 (lra_assign): Add a return arg. Set up from the result of
21600 assign_by_spills call.
21601 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
21602 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
21603 usage_insns if it is not NULL.
21604 (spill_hard_reg_in_range): New function.
21605 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
21606 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
21607 function prototypes.
21608 (lra_assign): Change prototype.
21609 * lra.c (lra): Add code to deal with fails by splitting hard reg
21612 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
21615 * config/i386/i386.opt: Replace Enforcment with Enforcement.
21617 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
21620 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
21621 (dw_val_node): Add val_symbolic_view.
21622 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
21623 (symview_upper_bound): New.
21624 (new_line_info_table): Initialize symviews_since_reset.
21625 (dwarf2out_source_line): Count symviews_since_reset and set
21626 symview_upper_bound.
21627 (dw_val_equal_p): Handle symview.
21628 (add_AT_symview): New.
21629 (print_dw_val): Handle symview.
21630 (attr_checksum, attr_checksum_ordered): Likewise.
21631 (same_dw_val_p, size_of_die): Likewise.
21632 (value_format, output_die): Likewise.
21633 (add_high_low_attributes): Use add_AT_symview for entry_view.
21634 (dwarf2out_finish): Reset symview_upper_bound, clear
21637 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
21640 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
21641 Add strict argument and use it.
21642 (rs6000_split_multireg_move): Update for new strict argument.
21643 (mem_operand_gpr): Disallow all non-offsettable addresses.
21644 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
21646 2018-03-09 Jakub Jelinek <jakub@redhat.com>
21649 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
21650 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
21651 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
21654 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
21655 decl, use remap_type if we want to use the type.
21657 2018-03-09 Martin Sebor <msebor@redhat.com>
21659 PR tree-optimization/84526
21660 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
21662 (builtin_access::generic_overlap): Be prepared to handle non-array
21665 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
21667 PR rtl-optimization/84682
21668 * lra-constraints.c (process_address_1): Check is_address flag
21669 for address constraints.
21670 (process_alt_operands): Likewise.
21671 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
21672 preprocess_constraints.
21673 * recog.h (preprocess_constraints): Add oploc parameter.
21675 * recog.c (preprocess_constraints): Test address_operand for
21676 CT_ADDRESS constraints.
21678 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
21681 * lra-assigns.c (assign_by_spills): Return a flag of reload
21682 assignment failure. Do not process the reload assignment
21683 failures. Do not spill other reload pseudos if they has the same
21685 (lra_assign): Add a return arg. Set up from the result of
21686 assign_by_spills call.
21687 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
21688 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
21689 usage_insns if it is not NULL.
21690 (spill_hard_reg_in_range): New function.
21691 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
21692 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
21693 function prototypes.
21694 (lra_assign): Change prototype.
21695 * lra.c (lra): Add code to deal with fails by splitting hard reg
21698 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21701 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
21702 Accept complain bool parameter. Only emit errors if it is true.
21703 (arm_parse_cpu_option_name): Likewise.
21704 (arm_target_thumb_only): Adjust callers of the above.
21705 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
21706 prototype to take a default true bool parameter.
21707 (arm_parse_arch_option_name): Likewise.
21709 2018-03-09 David Malcolm <dmalcolm@redhat.com>
21710 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
21714 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
21715 * configure: Regenerate.
21716 * configure.ac ("linker --version-script option"): New.
21717 ("linker soname option"): New.
21719 2018-03-09 Richard Biener <rguenther@suse.de>
21721 PR tree-optimization/84775
21722 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
21723 immediate uses of predicate stmts and mark them modified.
21726 PR tree-optimization/84178
21727 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
21729 (version_loop_for_if_conversion): Delay update_ssa call.
21730 (tree_if_conversion): Delay update_ssa until after predicate
21733 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
21736 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
21737 when the function accesses prior frames.
21739 2018-03-08 Jakub Jelinek <jakub@redhat.com>
21742 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
21743 gen_llsym, otherwise call maybe_gen_llsym.
21745 PR inline-asm/84742
21746 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
21747 has ',' character inside of it.
21749 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21752 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
21753 as clobbering CC_REGNUM.
21755 2018-03-08 Richard Biener <rguenther@suse.de>
21757 PR middle-end/84552
21758 * tree-scalar-evolution.c: Include tree-into-ssa.h.
21759 (follow_copies_to_constant): Do not follow SSA names registered
21762 2018-03-08 Richard Biener <rguenther@suse.de>
21764 PR tree-optimization/84178
21765 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
21767 (version_loop_for_if_conversion): Delay update_ssa call.
21768 (tree_if_conversion): Delay update_ssa until after predicate
21771 2018-03-08 David Malcolm <dmalcolm@redhat.com>
21773 PR tree-optimization/84178
21774 * tree-if-conv.c (release_bb_predicate): Remove the
21775 the assertion that the stmts have NULL use_ops.
21776 Discard the statements, asserting that they haven't
21777 yet been added to a BB.
21779 2018-03-08 Richard Biener <rguenther@suse.de>
21781 PR tree-optimization/84746
21782 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
21783 (phi_translate): Pass in destination ANTIC_OUT set.
21784 (phi_translate_1): Likewise. For a simplified result lookup
21785 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
21786 (phi_translate_set): Adjust.
21787 (do_pre_regular_insertion): Likewise.
21788 (do_pre_partial_partial_insertion): Likewise.
21790 2018-03-08 Martin Liska <mliska@suse.cz>
21792 PR gcov-profile/84735
21793 * doc/gcov.texi: Document usage of profile files.
21794 * gcov-io.h: Document changes in the format.
21796 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
21800 * dwarf2out.c (struct dw_line_info_table): Update comments for
21802 (FORCE_RESET_NEXT_VIEW): New.
21803 (FORCE_RESETTING_VIEW_P): New.
21804 (RESETTING_VIEW_P): Check for -1 too.
21805 (ZERO_VIEW_P): Likewise.
21806 (new_line_info_table): Force-reset next view.
21807 (dwarf2out_begin_function): Likewise.
21808 (dwarf2out_source_line): Simplify zero_view_p initialization.
21809 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
21810 view directly. Omit view when omitting .loc at line 0.
21812 2018-03-08 Jakub Jelinek <jakub@redhat.com>
21814 PR tree-optimization/84740
21815 * tree-switch-conversion.c (process_switch): Call build_constructors
21816 only if info.phi_count is non-zero.
21818 PR tree-optimization/84739
21819 * tree-tailcall.c (find_tail_calls): Check call arguments against
21820 DECL_ARGUMENTS (current_function_decl) rather than
21821 DECL_ARGUMENTS (func) when checking for tail recursion.
21823 2018-03-07 Jakub Jelinek <jakub@redhat.com>
21825 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
21826 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
21827 Volker Reichelt's entry and add entries for people that perform
21828 GCC fuzzy testing and report numerous bugs.
21830 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
21833 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
21834 readonly data in sdata, if that is disabled.
21835 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
21836 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
21837 -mreadonly-in-sdata option.
21839 2018-03-07 Martin Sebor <msebor@redhat.com>
21841 PR tree-optimization/84468
21842 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
21843 basic block when looking for nul assignment.
21845 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
21848 * except.h (output_function_exception_table): Adjust prototype.
21849 * except.c (output_function_exception_table): Remove FNNAME parameter
21850 and add SECTION parameter. Ouput one part of the table at a time.
21851 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
21852 the first part of the exception table and emit unwind directives.
21853 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
21854 (i386_pe_seh_cold_init): Likewise.
21855 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
21856 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
21857 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
21858 (ix86_output_call_insn): Emit a nop in one more case for SEH.
21859 * config/i386/winnt.c: Include except.h.
21860 (struct seh_frame_state): Add reg_offset, after_prologue and
21861 in_cold_section fields.
21862 (i386_pe_seh_end_prologue): Set seh->after_prologue.
21863 (i386_pe_seh_cold_init): New function.
21864 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
21865 to seh->in_cold_section.
21866 (seh_emit_push): Record the offset of the push.
21867 (seh_emit_save): Record the offet of the save.
21868 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
21869 Test seh->after_prologue to disregard the epilogue.
21870 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
21871 (i386_pe_end_cold_function): New function.
21873 2018-03-07 Jakub Jelinek <jakub@redhat.com>
21876 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
21877 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
21880 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
21882 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
21883 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
21885 PR middle-end/84723
21886 * multiple_target.c: Include tree-inline.h and intl.h.
21887 (expand_target_clones): Diagnose and fail if node->definition and
21888 !tree_versionable_function_p (node->decl).
21890 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
21892 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
21894 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
21895 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
21896 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
21898 2018-03-06 Jakub Jelinek <jakub@redhat.com>
21901 * combine.c (try_combine): Use reg_or_subregno instead of handling
21902 just paradoxical SUBREGs and REGs.
21904 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
21906 * config/arc/arc.c (arc_finalize_pic): Remove function.
21907 (arc_must_save_register): We use single base PIC register, remove
21908 checks to save/restore the PIC register.
21909 (arc_expand_prologue): Likewise.
21910 * config/arc/arc-protos.h (arc_set_default_type_attributes):
21912 (arc_verify_short): Likewise.
21913 (arc_attr_type): Likewise.
21914 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
21915 (walk_stores): Likewise.
21916 (arc_address_cost): Make it static.
21917 (arc_verify_short): Likewise.
21918 (branch_dest): Likewise.
21919 (arc_attr_type): Likewise.
21920 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
21921 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
21922 (arc_final_prescan_insn): Remove inserting the nops due to
21923 hardware hazards. It is done in reorg step.
21924 (insn_length_variant_t): Remove.
21925 (insn_length_parameters_t): Likewise.
21926 (arc_insn_length_parameters): Likewise.
21927 (arc_get_insn_variants): Likewise.
21928 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
21930 2018-03-06 Jakub Jelinek <jakub@redhat.com>
21932 PR inline-asm/84683
21933 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
21936 PR tree-optimization/84687
21937 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
21939 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
21941 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21943 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
21944 Rename to ppc_speculation_barrier.
21945 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
21946 __builtin_ppc_speculation_barrier.
21948 2018-03-05 Jakub Jelinek <jakub@redhat.com>
21951 * combine.c (combine_simplify_rtx): Don't try to simplify if
21952 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
21955 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
21957 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
21958 to 32 bytes when compiling for POWER9.
21960 2018-03-05 Jakub Jelinek <jakub@redhat.com>
21963 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
21964 regparm >= 3 with no arg reg available also for calls with
21965 flag_force_indirect_call. Pass decl to ix86_function_regparm.
21968 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
21970 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
21972 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
21975 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
21977 2018-03-05 Richard Biener <rguenther@suse.de>
21979 PR tree-optimization/84486
21980 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
21981 When inserting a __builtin_assume_aligned call set the LHS
21982 SSA name alignment info accordingly.
21984 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
21986 PR tree-optimization/84114
21987 * config/aarch64/aarch64.c (aarch64_reassociation_width)
21988 Avoid reassociation of FLOAT_MODE addition.
21990 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
21992 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
21993 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
21994 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
21995 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
21996 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
21997 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
21998 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
22000 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
22001 __builtin_ia32_wbinvd): New builtins.
22002 (SPECIAL_ARGS2): New.
22003 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
22004 (SPECIAL_ARGS2): New.
22005 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
22006 (ix86_valid_target_attribute_inner_p): Ditto.
22007 (ix86_init_mmx_sse_builtins): Add special_args2.
22008 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
22009 TARGET_WBNOINVD_P): New.
22010 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
22011 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
22012 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
22013 * config/i386/immintrin.h (_wbinvd): New intrinsic.
22014 * config/i386/pconfigintrin.h: New file.
22015 * config/i386/wbnoinvdintrin.h: Ditto.
22016 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
22018 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
22020 2018-03-05 Richard Biener <rguenther@suse.de>
22022 PR tree-optimization/84670
22023 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
22025 (BB_VISITED_WITH_VISITED_SUCCS): New define.
22026 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
22027 (compute_antic_aux): Only assert the number of values in ANTIC_IN
22028 doesn't grow if all successors (recursively) were visited at least
22031 2018-03-05 Richard Biener <rguenther@suse.de>
22033 PR tree-optimization/84650
22034 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
22035 if executed in the loop pipeline.
22037 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
22039 * doc/configfiles.texi (Configuration Files): Move info about
22040 conditionalizing $target-protos.h to...
22041 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
22042 differs from $target-protos.h.
22044 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
22045 Chung-Ju Wu <jasonwucj@gmail.com>
22047 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
22048 * config/nds32/nds32-multiple.md (setmemsi): Define.
22049 * config/nds32/nds32-memory-manipulation.c
22050 (nds32_gen_dup_4_byte_to_word_value): New.
22051 (emit_setmem_word_loop): New.
22052 (emit_setmem_byte_loop): New.
22053 (nds32_expand_setmem_loop): New.
22054 (nds32_expand_setmem_loop_v3m): New.
22055 (nds32_expand_setmem_unroll): New.
22056 (nds32_expand_setmem): New.
22058 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
22059 Chung-Ju Wu <jasonwucj@gmail.com>
22061 * config/nds32/nds32-memory-manipulation.c
22062 (nds32_emit_load_store): New.
22063 (nds32_emit_post_inc_load_store): New.
22064 (nds32_emit_mem_move): New.
22065 (nds32_emit_mem_move_block): New.
22066 (nds32_expand_movmemsi_loop_unknown_size): New.
22067 (nds32_expand_movmemsi_loop_known_size): New.
22068 (nds32_expand_movmemsi_loop): New.
22069 (nds32_expand_movmemsi_unroll): New.
22070 (nds32_expand_movmemqi): Rename ...
22071 (nds32_expand_movmemsi): ... to this.
22072 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
22073 (movmemsi): ... to this.
22074 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
22075 (nds32_expand_movmemsi): ... to this.
22077 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
22078 Monk Chiang <sh.chiang04@gmail.com>
22079 Chung-Ju Wu <jasonwucj@gmail.com>
22081 * config/nds32/nds32-protos.h
22082 (nds32_expand_load_multiple): New arguments.
22083 (nds32_expand_store_multiple): Ditto.
22084 (nds32_valid_multiple_load_store): Rename ...
22085 (nds32_valid_multiple_load_store_p): ... to this.
22086 * config/nds32/nds32-memory-manipulation.c
22087 (nds32_expand_load_multiple): Refine implementation.
22088 (nds32_expand_store_multiple): Ditto.
22089 * config/nds32/nds32-multiple.md
22090 (load_multiple): Update nds32_expand_load_multiple interface.
22091 (store_multiple): Update nds32_expand_store_multiple interface.
22092 * config/nds32/nds32-predicates.c
22093 (nds32_valid_multiple_load_store): Rename ...
22094 (nds32_valid_multiple_load_store_p): ... to this and refine
22096 * config/nds32/predicates.md
22097 (nds32_load_multiple_and_update_address_operation): New predicate.
22098 (nds32_store_multiple_and_update_address_operation): New predicate.
22100 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
22101 Chung-Ju Wu <jasonwucj@gmail.com>
22103 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
22104 (combo): New attribute.
22105 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
22107 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
22109 * config/nds32/nds32.opt: Change -mcmodel= default value.
22111 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
22112 Monk Chiang <sh.chiang04@gmail.com>
22113 Chung-Ju Wu <jasonwucj@gmail.com>
22115 * config/nds32/constants.md (unspec_element): New enum.
22116 * config/nds32/constraints.md (Umw): New constraint.
22117 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
22118 * config/nds32/nds32-intrinsic.md: Likewise.
22119 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
22120 (nds32_valid_smw_lwm_base_p): New.
22121 (nds32_output_smw_single_word): New.
22122 (nds32_output_lmw_single_word): New.
22123 (nds32_expand_unaligned_load): New.
22124 (nds32_expand_unaligned_store): New.
22125 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
22126 (nds32_output_smw_single_word): Declare.
22127 (nds32_output_lmw_single_word): Declare.
22128 (nds32_expand_unaligned_load): Declare.
22129 (nds32_expand_unaligned_store): Declare.
22130 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
22131 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
22132 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
22133 NDS32_BUILTIN_UASTORE_DW.
22134 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
22137 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
22138 Kito Cheng <kito.cheng@gmail.com>
22139 Chung-Ju Wu <jasonwucj@gmail.com>
22141 * config/nds32/nds32-intrinsic.c
22142 (nds32_expand_builtin_null_ftype_reg): Delete.
22143 (nds32_expand_builtin_reg_ftype_imm): Ditto.
22144 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
22145 (nds32_read_argument): New.
22146 (nds32_legitimize_target): Ditto.
22147 (nds32_legitimize_argument): Ditto.
22148 (nds32_check_constant_argument): Ditto.
22149 (nds32_expand_unop_builtin): Ditto.
22150 (nds32_expand_unopimm_builtin): Ditto.
22151 (nds32_expand_binop_builtin): Ditto.
22152 (nds32_builtin_decl_impl): Ditto.
22153 (builtin_description): Ditto.
22154 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
22155 (nds32_init_builtins_impl): Ditto.
22156 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
22157 (nds32_builtin_decl): New.
22158 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
22159 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
22161 2018-03-02 Jeff Law <law@redhat.com>
22163 * reorg.c (stop_search_p): Handle DEBUG_INSN.
22164 (redundant_insn, fill_simple_delay_slots): Likewise.
22165 (fill_slots_from_thread): Likewise.
22166 * resource.c (mark_referenced_resources): Likewise.
22167 (mark_set_resources, find_dead_or_set_registers): Likewise.
22169 2018-03-02 Jakub Jelinek <jakub@redhat.com>
22171 * substring-locations.h (format_warning_va): Formatting fix for
22172 ATTRIBUTE_GCC_DIAG.
22173 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
22175 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
22176 * substring-locations.c: Include intl.h.
22177 (format_warning_va): Turned into small wrapper around
22178 format_warning_n_va, renamed to ...
22179 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
22180 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
22182 (format_warning_at_substring_n): New function.
22183 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
22184 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
22185 format_warning_at_substring with just a shorter name instead of
22186 const function pointer.
22187 (fmtwarn_n): New function.
22188 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
22189 appropriate, get rid of all the fmtstr temporaries, move conditionals
22190 with G_() wrapped string literals directly into fmtwarn arguments,
22191 cast dir.len to (int), formatting fixes.
22193 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
22195 * doc/invoke.texi: Remove "Cilk Plus" references.
22197 2018-03-02 Jakub Jelinek <jakub@redhat.com>
22198 Richard Biener <rguenther@suse.de>
22201 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
22202 for error or warning attributes if CALL_FROM_THUNK_P is set.
22205 2018-03-02 Jakub Jelinek <jakub@redhat.com>
22208 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
22209 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
22212 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
22213 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
22215 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
22216 instead of -1U in last predictors element's probability member.
22218 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
22221 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
22222 arguments if they are comparable.
22224 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
22226 PR tree-optimization/84634
22227 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
22228 masks and masked_loop_p with a single loop_masks, making sure it's
22229 null for bb vectorization.
22231 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
22233 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
22234 (vect_analyze_data_ref_access): Use loop->safe_len rather than
22235 loop->force_vectorize to check whether there is no alias.
22237 2018-03-02 Jakub Jelinek <jakub@redhat.com>
22240 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
22242 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
22244 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
22245 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
22246 instead of a loop around prev_real_insn.
22247 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
22250 PR inline-asm/84625
22251 * config/i386/i386.c (ix86_print_operand): Use conditional
22252 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
22255 2018-03-02 Richard Biener <rguenther@suse.de>
22257 PR tree-optimization/84427
22258 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
22259 (bitmap_set_subtract_values): Rewrite to handle multiple
22262 (prune_clobbered_mems): Likewise.
22263 (phi_translate): Take edge instead of pred/phiblock.
22264 (phi_translate_1): Likewise.
22265 (phi_translate_set): Likewise. Insert all translated
22266 exprs for a value into the set, keeping possibly multiple
22267 expressions per value.
22268 (compute_antic_aux): Adjust for phi_translate changes.
22269 When intersecting union the expressions and prune those
22270 not in the final value set, keeping possibly multiple
22271 expressions per value. Do not use value-insertion
22272 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
22273 all expressions. Add verification that the value-sets
22274 only shrink during iteration.
22275 (compute_partial_antic_aux): Adjust for the phi_translate changes.
22276 (do_pre_regular_insertion): Likewise.
22277 (do_pre_partial_partial_insertion): Likewise.
22279 2018-03-02 Richard Biener <rguenther@suse.de>
22282 * config/darwin.c (saved_debug_info_level): New static global.
22283 (darwin_asm_lto_start): Disable debug info generation for LTO out.
22284 (darwin_asm_lto_end): Restore debug info generation settings.
22286 2018-03-01 Martin Liska <mliska@suse.cz>
22289 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
22290 volatile arguments.
22292 2018-03-01 Richard Biener <rguenther@suse.de>
22295 * dwarf2out.c (gen_variable_die): Properly handle late VLA
22296 type annotation with LTO when debug was disabled at compile-time.
22298 2018-03-01 Matthew Fortune <mfortune@gmail.com>
22300 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
22302 (mips_final_postscan_insn): Likewise.
22304 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
22306 PR rtl-optimization/84528
22307 * alias.c (init_alias_target): Add commentary.
22308 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
22309 a unique base value if the frame pointer is not eliminated
22310 to the stack pointer.
22312 2018-03-01 Tom de Vries <tom@codesourcery.com>
22314 PR rtl-optimization/83327
22315 * lra-int.h (hard_regs_spilled_into): Declare.
22316 * lra.c (hard_regs_spilled_into): Define.
22317 (init_reg_info): Init hard_regs_spilled_into.
22318 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
22319 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
22320 (process_bb_lives): Handle hard_regs_spilled_into.
22321 (lra_create_live_ranges_1): Before doing liveness propagation, clear
22322 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
22324 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
22326 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
22327 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
22328 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
22329 * config/rs6000/aix72.h: New file.
22331 2018-02-28 Jakub Jelinek <jakub@redhat.com>
22333 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
22334 instead of warning_at with conditional singular and plural messages
22338 * stor-layout.c (update_alignment_for_field): For
22339 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
22340 && !DECL_PACKED (field), do the alignment update, just use
22341 only desired_align instead of MAX (type_align, desired_align)
22343 (place_field): Don't do known_align < desired_align handling
22344 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
22345 is non-NULL, instead do it after rli->prev_field handling and
22346 only if not within a bitfield word. For DECL_PACKED (field)
22347 use type_align of BITS_PER_UNIT.
22349 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
22351 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
22352 superfluous parentheses and trailing spaces.
22354 2018-02-28 Richard Biener <rguenther@suse.de>
22356 PR tree-optimization/84584
22357 * graphite-scop-detection.c (scop_detection::add_scop): Discard
22358 SCoPs with fake exit edge.
22360 2018-02-28 Martin Liska <mliska@suse.cz>
22363 * timevar.c (timer::print): Fix format to properly print 100%
22366 2018-02-28 Richard Biener <rguenther@suse.de>
22368 PR middle-end/84607
22369 * genmatch.c (capture_info::walk_match): Do not mark
22370 captured expressions without operands as expr_p given
22371 they act more like predicates and should be subject to
22372 "lost tail" side-effect preserving.
22374 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
22376 PR rtl-optimization/81611
22377 * auto-inc-dec.c (attempt_change): Move dead note from
22378 mem_insn if it's the next use of regno
22379 (find_address): Take address use of reg holding
22380 non-incremented value. Add parm to limit search to the named
22382 (merge_in_block): Attempt to use a mem insn that is the next
22383 use of the original regno.
22385 2018-02-27 Martin Sebor <msebor@redhat.com>
22388 * doc/invoke.texi (-Wmissing-attributes): New option.
22389 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
22391 2018-02-27 Martin Sebor <msebor@redhat.com>
22393 PR translation/84207
22394 * diagnostic-core.h (warning_n, error_n, inform_n): Change
22395 n argument to unsigned HOST_WIDE_INT.
22396 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
22397 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
22398 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
22399 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
22401 2018-02-27 Richard Biener <rguenther@suse.de>
22403 PR tree-optimization/84512
22404 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
22405 Do not use the estimate returned from record_stmt_cost for
22406 the scalar iteration cost but sum properly using add_stmt_cost.
22408 2018-02-27 Richard Biener <rguenther@suse.de>
22410 PR tree-optimization/84466
22411 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
22412 Adjust last change to less strictly validate use operands.
22414 2018-02-27 Martin Liska <mliska@suse.cz>
22416 PR gcov-profile/84548
22417 * gcov.c (process_file): Allow partial overlap and consider it
22418 also as group functions.
22419 (output_lines): Properly calculate range of lines for a group.
22421 2018-02-27 Martin Liska <mliska@suse.cz>
22423 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
22424 'ggc' suffixes. Change first column width.
22425 (timer::print): Fix formatting of the column.
22427 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
22429 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
22430 preserve inline entry blocks for the sake of debug inline
22431 entry point markers alone.
22432 (remove_unused_locals): Suggest in comments a better place to
22433 force the preservation of inline entry blocks that are
22434 otherwise unused, but do not preserve them.
22436 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
22438 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
22440 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
22443 * config/i386/constraints.md (Bs): Replace
22444 ix86_indirect_branch_register with
22445 TARGET_INDIRECT_BRANCH_REGISTER.
22447 * config/i386/i386.md (indirect_jump): Likewise.
22448 (tablejump): Likewise.
22449 (*sibcall_memory): Likewise.
22450 (*sibcall_value_memory): Likewise.
22451 Peepholes of indirect call and jump via memory: Likewise.
22452 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
22453 (*sibcall_value_GOT_32): Likewise.
22454 * config/i386/predicates.md (indirect_branch_operand): Likewise.
22455 (GOT_memory_operand): Likewise.
22456 (call_insn_operand): Likewise.
22457 (sibcall_insn_operand): Likewise.
22458 (GOT32_symbol_operand): Likewise.
22459 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
22461 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
22463 PR rtl-optimization/83496
22464 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
22465 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
22466 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
22467 redundant insn, if any.
22468 (relax_delay_slots): Likewise.
22469 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
22471 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
22473 PR tree-optimization/83965
22474 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
22475 that grouped statements are part of a reduction chain. Return
22476 true if the statement is not marked as a reduction itself but
22477 is part of a group.
22478 (vect_recog_dot_prod_pattern): Don't check whether the statement
22479 is part of a group here.
22480 (vect_recog_sad_pattern): Likewise.
22481 (vect_recog_widen_sum_pattern): Likewise.
22483 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
22486 * final.c (rest_of_clean_state): Also look for calls inside sequences.
22488 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
22491 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
22493 (ix86_output_indirect_function_return): New prototype.
22494 (ix86_split_simple_return_pop_internal): Likewise.
22495 * config/i386/i386.c (indirect_return_via_cx): New.
22496 (indirect_return_via_cx_bnd): Likewise.
22497 (indirect_thunk_name): Handle return va CX_REG.
22498 (output_indirect_thunk_function): Create alias for
22499 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
22500 (ix86_output_indirect_jmp): Remove the bool argument.
22501 (ix86_output_indirect_function_return): New function.
22502 (ix86_split_simple_return_pop_internal): Likewise.
22503 * config/i386/i386.md (*indirect_jump): Don't pass false
22504 to ix86_output_indirect_jmp.
22505 (*tablejump_1): Likewise.
22506 (simple_return_pop_internal): Change it to define_insn_and_split.
22507 Call ix86_split_simple_return_pop_internal to split it for
22508 -mfunction-return=.
22509 (simple_return_indirect_internal): Call
22510 ix86_output_indirect_function_return instead of
22511 ix86_output_indirect_jmp.
22513 2018-02-26 Jakub Jelinek <jakub@redhat.com>
22516 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
22517 memset and value initialization afterwards.
22519 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
22521 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
22523 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
22526 * common/config/aarch64/aarch64-common.c
22527 (aarch_option_optimization_table[]): Switch
22528 off fomit-frame-pointer
22530 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
22531 Chung-Ju Wu <jasonwucj@gmail.com>
22533 * config/nds32/nds32-multiple.md (load_multiple): Disallow
22535 (store_multiple): Ditto.
22537 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
22539 * config.gcc: Add --with-cpu support for nds32 target.
22540 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
22541 * config/nds32/nds32.opt: Add -mcpu= option.
22543 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
22545 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
22546 isel=yes): Warn for these deprecated options.
22548 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
22550 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
22551 ISA_2_5_MASKS_EMBEDDED.
22553 2018-02-23 Jakub Jelinek <jakub@redhat.com>
22555 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
22556 p->max as pointers rather than using iterative_hash_expr.
22558 2018-02-23 Carl Love <cel@us.ibm.com>
22560 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
22561 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
22563 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
22564 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
22565 P8V_BUILTIN_VEC_VUNSIGNED2.
22567 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
22570 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
22571 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
22573 * lra-constraints.c (curr_insn_transform): Set up
22574 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
22576 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
22577 * lra-eliminations.c (spill_pseudos): Ditto.
22578 (process_insn_for_elimination): Ditto.
22579 * lra-lives.c (reg_early_clobber_p): Use the new macros.
22580 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
22581 LRA_NON_CLOBBERED_ALT.
22583 2018-02-22 Martin Sebor <msebor@redhat.com>
22585 PR tree-optimization/84480
22586 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
22587 to maybe_diag_stxncpy_trunc. Call it.
22588 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
22589 from gimple_fold_builtin_strcpy. Print inlining stack.
22590 (handle_builtin_stxncpy): Print inlining stack.
22591 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
22593 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
22596 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
22597 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
22598 and -fcheck-pointer-bounds are used together.
22599 (indirect_thunk_prefix): New enum.
22600 (indirect_thunk_need_prefix): New function.
22601 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
22602 "_nt" instead of "_bnd" for NOTRACK prefix.
22603 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
22604 (output_indirect_thunk_function): Likewise.
22606 (ix86_code_end): Update output_indirect_thunk_function calls.
22607 (ix86_output_indirect_branch_via_reg): Replace
22608 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
22609 (ix86_output_indirect_branch_via_push): Likewise.
22610 (ix86_output_function_return): Likewise.
22611 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
22612 incompatible with -fcf-protection=branch and
22613 -fcheck-pointer-bounds.
22615 2018-02-22 Steve Ellcey <sellcey@cavium.com>
22618 * config/aarch64/aarch64.c (aarch64_print_address_internal):
22619 Change gcc_assert call to output_operand_lossage.
22621 2018-02-22 Steve Ellcey <sellcey@cavium.com>
22623 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
22625 2018-02-22 DJ Delorie <dj@redhat.com>
22626 Sebastian Perta <sebastian.perta@renesas.com>
22627 Oleg Endo <olegendo@gcc.gnu.org>
22629 * config/rx/rx.c (rx_rtx_costs): New function.
22630 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
22632 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
22634 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
22636 2018-02-22 Martin Liska <mliska@suse.cz>
22639 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
22640 Add "native" as a possible value.
22642 2018-02-22 Martin Liska <mliska@suse.cz>
22645 * config/i386/i386.c (ix86_option_override_internal):
22646 Add "native" as a possible value for -march and -mtune.
22648 2018-02-22 Jakub Jelinek <jakub@redhat.com>
22651 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
22652 to all type variants.
22654 PR tree-optimization/84503
22655 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
22656 width as info->bitpos + info->bitsize - start.
22657 (merged_store_group::merge_overlapping): Simplify width computation.
22658 (check_no_overlap): New function.
22659 (imm_store_chain_info::try_coalesce_bswap): Compute expected
22660 start + width and last_order of the group, fail if check_no_overlap
22662 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
22663 to group if check_no_overlap fails.
22665 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
22667 * config/rs6000/altivec.md: Delete contraint arguments to
22668 define_expand, define_split, and define_peephole2, and in
22669 define_insn_and_split if always unused.
22670 * config/rs6000/darwin.md: Ditto.
22671 * config/rs6000/dfp.md: Ditto.
22672 * config/rs6000/rs6000.md: Ditto.
22673 * config/rs6000/sync.md: Ditto.
22674 * config/rs6000/vector.md: Ditto.
22675 * config/rs6000/vsx.md: Ditto.
22677 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
22679 * config/rs6000/altivec.md: Write output control strings as braced
22680 blocks instead of double-quoted strings.
22681 * config/rs6000/darwin.md: Ditto.
22682 * config/rs6000/rs6000.md: Ditto.
22683 * config/rs6000/vector.md: Ditto.
22684 * config/rs6000/vsx.md: Ditto.
22686 2018-02-21 Jason Merrill <jason@redhat.com>
22688 PR c++/84314 - ICE with templates and fastcall attribute.
22689 * attribs.c (build_type_attribute_qual_variant): Remove assert.
22691 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
22693 * ipa-cp.c (determine_versionability): Fix comment typos.
22695 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
22698 * ipa-cp.c (determine_versionability): Do not version functions caling
22701 2018-02-21 Martin Liska <mliska@suse.cz>
22704 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
22705 Add "native" as a possible value.
22706 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
22707 the macro when native cpu detection is available.
22709 2018-02-21 Martin Liska <mliska@suse.cz>
22712 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
22713 Add "native" as a possible value.
22714 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
22715 when native cpu detection is available.
22717 2018-02-21 Jakub Jelinek <jakub@redhat.com>
22718 Martin Sebor <msebor@redhat.com>
22720 PR tree-optimization/84478
22721 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
22723 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
22724 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
22725 support which is conservatively correct, for 2 only stay conservative
22726 for maxlen. Formatting and comment capitalization fixes. Add STRICT
22727 argument to the 2 argument get_range_strlen, adjust 6 arg
22728 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
22730 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
22731 (gimple_fold_builtin_strlen): Pass true as last argument to
22734 2018-02-20 Martin Sebor <msebor@redhat.com>
22736 PR middle-end/84095
22737 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
22738 (builtin_memref::set_base_and_offset): Same. Handle inner references.
22739 (builtin_memref::builtin_memref): Factor out parts into
22740 set_base_and_offset and call it.
22742 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
22744 PR middle-end/84406
22745 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
22746 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
22747 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
22748 search at the associated MODE_INT.
22750 2018-02-20 Jeff Law <law@redhat.com>
22752 PR middle-end/82123
22753 PR tree-optimization/81592
22754 PR middle-end/79257
22755 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
22756 for range data rather than using global data.
22757 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
22758 range data rather than using global data.
22759 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
22760 pass it to children as needed.
22761 (struct directive::fmtresult): Similarly.
22762 (struct directive::set_width): Similarly.
22763 (struct directive::set_precision): Similarly.
22764 (format_integer, format_directive, parse_directive): Similarly.
22765 (format_none): Accept unnamed vr_values parameter.
22766 (format_percent, format_floating, format_character): Similarly.
22767 (format_string, format_plain): Similarly.
22768 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
22769 the EVRP range analyzer for range data rather than using global data.
22770 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
22771 gimple-ssa-evrp-analyze.h
22772 (class sprintf_dom_walker): Add after_dom_children member function.
22773 Add evrp_range_analyzer member.
22774 (sprintf_dom_walker::before_dom_children): Call into the EVRP
22775 range analyzer as needed.
22776 (sprintf_dom_walker::after_dom_children): New member function.
22777 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
22779 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
22780 (evrp_range_analyzer::pop_to_marker): Likewise.
22782 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
22784 PR tree-optimization/84419
22785 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
22786 with the required type if its current type is compatible but
22789 2018-02-20 Jakub Jelinek <jakub@redhat.com>
22791 PR middle-end/82004
22792 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
22793 after vectorization.
22795 2018-02-20 Martin Liska <mliska@suse.cz>
22798 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
22799 possible values if we don't have a hint.
22801 2018-02-20 Martin Liska <mliska@suse.cz>
22805 * final.c (shorten_branches): Build align_tab array with one
22807 * opts.c (finish_options): Add alignment option limit check.
22808 (MAX_CODE_ALIGN): Likewise.
22809 (MAX_CODE_ALIGN_VALUE): Likewise.
22810 * doc/invoke.texi: Document maximum allowed option value for
22811 all -falign-* options.
22813 2018-02-19 Jakub Jelinek <jakub@redhat.com>
22816 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
22817 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
22818 * var-tracking.c (emit_note_insn_var_location): Remove all references
22819 to NOTE_INSN_CALL_ARG_LOCATION.
22820 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
22821 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
22822 Use copy_rtx_if_shared.
22823 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
22824 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
22825 (dwarf2out_var_location): Remove handling of
22826 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
22828 * final.c (final_scan_insn): Remove all references to
22829 NOTE_INSN_CALL_ARG_LOCATION.
22830 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
22831 before dumping final insns.
22832 * except.c (emit_note_eh_region_end): Remove all references to
22833 NOTE_INSN_CALL_ARG_LOCATION.
22834 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
22835 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
22836 * config/arc/arc.c (hwloop_optimize): Likewise.
22837 * config/arm/arm.c (create_fix_barrier): Likewise.
22838 * config/s390/s390.c (s390_chunkify_start): Likewise.
22839 * config/sh/sh.c (find_barrier): Likewise.
22840 * config/i386/i386.c (rest_of_insert_endbranch,
22841 ix86_seh_fixup_eh_fallthru): Likewise.
22842 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
22843 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
22844 * config/frv/frv.c (frv_function_prologue): Likewise.
22845 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
22847 (note_outside_basic_block_p): Remove all references to
22848 NOTE_INSN_CALL_ARG_LOCATION.
22849 * gengtype.c (adjust_field_rtx_def): Likewise.
22850 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
22852 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
22853 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
22856 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
22859 PR tree-optimization/84452
22860 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
22861 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
22864 2018-02-19 Martin Liska <mliska@suse.cz>
22867 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
22869 2018-02-19 Martin Liska <mliska@suse.cz>
22870 Richard Sandiford <richard.sandiford@linaro.org>
22872 PR tree-optimization/82491
22873 * gimple-fold.c (get_base_constructor): Make earlier bail out
22876 2018-02-19 Carl Love <cel@us.ibm.com>
22878 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
22879 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
22881 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
22882 P8V_BUILTIN_VEC_NEG.
22884 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
22886 * config/rl78/rl78.md (movdf): New define expand.
22888 2018-02-19 Martin Liska <mliska@suse.cz>
22891 * doc/invoke.texi: Fix typo.
22892 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
22894 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
22896 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
22897 handle rs6000_single_float and rs6000_double_float specially for
22900 2018-02-16 Jeff Law <law@redhat.com>
22902 * config/rx/rx.c (add_pop_cfi_notes): New function.;
22903 (pop_regs): Use it.
22905 2018-02-16 Jakub Jelinek <jakub@redhat.com>
22908 * ipa-inline.c (inline_small_functions): Fix a typo.
22910 2018-02-16 Nathan Sidwell <nathan@acm.org>
22912 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
22914 2018-02-16 Carl Love <cel@us.ibm.com>
22916 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
22917 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
22918 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
22919 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
22920 expansion to P8V_BUILTIN_VEC_FLOAT2.
22922 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
22924 PR rtl-optimization/70023
22925 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
22926 src_regno into account.
22928 2018-02-16 Carl Love <cel@us.ibm.com>
22930 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
22931 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
22932 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
22933 * config/rs6000/rs6000.c: Remove case statements for
22934 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
22935 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
22936 and P9V_BUILTIN_VEC_VINSERT4B.
22937 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
22938 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
22939 * config/rs6000/vsx.md:
22940 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
22943 2018-02-16 Carl Love <cel@us.ibm.com>
22945 * config/rs6000/altivec.h: Add builtin names vec_extract4b
22947 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
22949 * config/rs6000/rs6000-c.c: Add the definitions for
22950 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
22951 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
22952 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
22953 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
22954 definition for insert4b and define insn *insert3b_internal.
22955 * doc/extend.texi: Add documentation for vec_extract4b.
22957 2018-02-16 Nathan Sidwell <nathan@acm.org>
22959 * doc/extend.texi (Backwards Compatibility): Mention friend
22960 injection. Note for-scope is deprecated.
22961 * doc/invoke.texi (-ffriend-injection): Deprecate.
22963 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
22965 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
22966 that moved to I2, also allow destinations that are a paradoxical
22967 subreg (instead of a normal reg).
22969 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
22972 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
22975 2018-02-16 Richard Biener <rguenther@suse.de>
22977 PR tree-optimization/84037
22978 PR tree-optimization/84016
22980 * config/i386/i386.c (ix86_builtin_vectorization_cost):
22981 Adjust vec_construct for the fact we need additional higher latency
22982 128bit inserts for AVX256 and AVX512 vector builds.
22983 (ix86_add_stmt_cost): Scale vector construction cost for
22986 2018-02-16 Richard Biener <rguenther@suse.de>
22988 PR tree-optimization/84417
22989 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
22990 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
22991 (non_rewritable_lvalue_p): Likewise, use poly-ints.
22993 2018-02-16 Martin Liska <mliska@suse.cz>
22996 * internal-fn.def (ASAN_CHECK): Set proper flags.
22997 (ASAN_MARK): Likewise.
22999 2018-02-16 Julia Koval <julia.koval@intel.com>
23001 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
23002 from PTA_CANNONLAKE.
23004 2018-02-16 Jakub Jelinek <jakub@redhat.com>
23007 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
23008 Use ++iter rather than iter++ for std::list iterators.
23009 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
23010 defer deleting them until all nodes in the forest are processed. Do
23011 free even leaf nodes. Change to_process into auto_vec.
23014 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
23015 * vec.h (vec_default_construct): Use memset instead of placement new
23016 if BROKEN_VALUE_INITIALIZATION is defined.
23017 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
23018 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
23021 PR rtl-optimization/83723
23022 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
23023 * lra.c (lra_substitute_pseudo): Likewise. If true, use
23024 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
23026 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
23028 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
23030 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
23032 PR rtl-optimization/81443
23033 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
23034 from inner REGs to paradoxical SUBREGs.
23036 2018-02-16 Richard Biener <rguenther@suse.de>
23038 PR tree-optimization/84399
23039 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
23040 For operands we can analyze at their definition make sure we can
23041 analyze them at each use as well.
23043 2018-02-16 Richard Biener <rguenther@suse.de>
23045 PR tree-optimization/84190
23046 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
23047 volatile accesses if the decl isn't volatile.
23049 2018-02-15 Jason Merrill <jason@redhat.com>
23051 PR c++/84314 - ICE with templates and fastcall attribute.
23052 * attribs.c (build_type_attribute_qual_variant): Don't clobber
23053 TYPE_CANONICAL on an existing type.
23055 2018-02-15 Jakub Jelinek <jakub@redhat.com>
23057 PR tree-optimization/84383
23058 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
23059 dstoff nor call operand_equal_p if dstbase is NULL.
23061 PR tree-optimization/84334
23062 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
23063 also a CONSTANT_CLASS_P, punt.
23065 2018-02-14 Jim Wilson <jimw@sifive.com>
23067 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
23068 first SMALL_OPERAND check. New local min_second_step. Move assert
23069 to where locals are set. Add TARGET_RVC support.
23070 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
23072 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
23074 * doc/invoke.texi: Correct -Wformat-overflow code sample.
23076 2018-02-14 Martin Sebor <msebor@redhat.com>
23078 PR tree-optimization/83698
23079 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
23080 arrays constrain the offset range to their bounds.
23081 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
23082 (builtin_access::overlap): Avoid setting the size of overlap if it's
23084 (maybe_diag_overlap): Also consider arrays when deciding what values
23085 of offsets to include in diagnostics.
23087 2018-02-14 Martin Sebor <msebor@redhat.com>
23090 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
23091 that correspond to the kind of a declaration.
23093 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
23096 * config/pa/pa.md: Load address of PIC label using the linkage table
23097 if the label is nonlocal.
23099 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
23101 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
23102 warning message if user requests -maltivec=be.
23103 * doc/invoke.texi: Document deprecation of -maltivec=be.
23105 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
23108 * config/rs6000/rs6000-c.c: Update definitions for
23109 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
23110 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
23112 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
23115 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
23116 add _get_ssp intrinsics. Remove argument from
23117 __builtin_ia32_rdssp[d|q].
23118 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
23119 * config/i386/i386-builtin.def: Remove argument from
23120 __builtin_ia32_rdssp[d|q].
23121 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
23122 ix86_expand_special_args_builtin for _rdssp[d|q].
23123 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
23124 Clear register before usage.
23125 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
23126 Add documentation for new _get_ssp and _inc_ssp intrinsics.
23128 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
23130 PR tree-optimization/84357
23131 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
23132 operand 1 of an ARRAY_REF too.
23134 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
23137 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
23138 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
23140 (set_of_reg): New struct.
23141 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
23142 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
23143 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
23145 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
23146 Split into bitclr, bitset, bitinvert patterns if appropriate.
23147 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
23148 use rx_fuse_in_memory_bitop.
23149 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
23150 to named insn, correct maximum insn length.
23152 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
23155 * machmode.def: Define a complex mode for PARTIAL_INT.
23156 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
23158 * doc/rtl.texi: Document CSPImode.
23159 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
23161 (msp430_hard_regno_nregs_with_padding): Likewise.
23163 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
23166 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
23168 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
23170 PR rtl-optimization/84169
23171 * combine.c (try_combine): New variable split_i2i3. Set it to true if
23172 we generated a parallel as new i3 and we split that to new i2 and i3
23173 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
23174 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
23175 those to i2, not i1. Partially rewrite this scan code.
23177 2018-02-13 Jakub Jelinek <jakub@redhat.com>
23180 * stor-layout.c (place_field): For variable length fields, adjust
23181 offset_align afterwards not just based on the field's alignment,
23182 but also on the size.
23184 PR middle-end/84309
23185 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
23186 of exps and logs in the use_exp2 case.
23188 2018-02-13 Jeff Law <law@redhat.com>
23190 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
23191 entry for "vector".
23193 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
23196 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
23200 * common.opt (gas-loc-support, gas-locview-support): New.
23201 (ginline-points, ginternal-reset-location-views): New.
23202 * doc/invoke.texi: Document them. Use @itemx where intended.
23203 (gvariable-location-views): Adjust.
23204 * target.def (reset_location_view): New.
23205 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
23206 (TARGET_RESET_LOCATION_VIEW): New.
23207 * doc/tm.texi: Rebuilt.
23208 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
23209 (dwarf2out_default_as_locview_support): New.
23210 (output_asm_line_debug_info): Use option variables.
23211 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
23212 (output_loc_list): Likewise.
23213 (add_high_low_attributes): Check option variables.
23214 Don't output entry view attribute in strict mode.
23215 (gen_inlined_subroutine_die): Check option variables.
23216 (dwarf2out_inline_entry): Likewise.
23217 (init_sections_and_labels): Likewise.
23218 (dwarf2out_early_finish): Likewise.
23219 (maybe_reset_location_view): New, from...
23220 (dwarf2out_var_location): ... here. Call it.
23221 * debug.h (dwarf2out_default_as_loc_support): Declare.
23222 (dwarf2out_default_as_locview_support): Declare.
23223 * hooks.c (hook_int_rtx_insn_0): New.
23224 * hooks.h (hook_int_rtx_insn_0): Declare.
23225 * toplev.c (process_options): Take -gas-loc-support and
23226 -gas-locview-support from dwarf2out. Enable
23227 -gvariable-location-views by default only with locview
23228 assembler support. Enable -ginternal-reset-location-views by
23229 default only if the target defines the corresponding hook.
23230 Enable -ginline-points by default if location views are
23231 enabled; force it disabled if statement frontiers are
23233 * tree-inline.c (expand_call_inline): Check option variables.
23234 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
23236 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
23238 PR tree-optimization/84321
23239 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
23240 handling. Also check whether the anti-range contains any values
23241 that satisfy the mask; switch to a VR_RANGE if not.
23243 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
23246 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
23248 2018-02-13 Martin Jambor <mjambor@suse.cz>
23251 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
23252 of call statements, also set location of a load to a temporary.
23254 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
23256 * config/rl78/rl78.c (add_vector_labels): New function.
23257 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
23258 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
23259 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
23260 which checks that no arguments are passed.
23261 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
23262 * doc/extend.texi: Documentation for the new attribute.
23264 2018-02-13 Andreas Schwab <schwab@suse.de>
23266 * config/riscv/linux.h (CPP_SPEC): Define.
23268 2018-02-13 Jakub Jelinek <jakub@redhat.com>
23271 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
23272 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
23273 OPTION_MASK_ISA_AES as first argument to def_builtin_const
23274 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
23275 instead of OPTION_MASK_ISA_PCLMUL as first argument to
23276 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
23277 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
23278 temporarily for AES and PCLMUL builtins.
23280 PR tree-optimization/84339
23281 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
23282 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
23285 PR middle-end/84309
23286 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
23287 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
23288 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
23290 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
23292 * omp-simd-clone.h: New file.
23293 * omp-simd-clone.c: Include omp-simd-clone.h.
23294 (expand_simd_clones): No longer static.
23295 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
23296 cgraph.h and omp-simd-clone.h.
23297 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
23298 (vect_recog_widen_shift_pattern): Formatting fix.
23299 (vect_pattern_recog_1): Don't check optab for calls.
23302 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
23303 operands[2] into a REG before using gen_lowpart on it.
23305 2018-02-12 Jeff Law <law@redhat.com>
23308 * config/sh/sh.c (find_barrier): Consider a sibling call
23311 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
23312 successfully back substituting a reg.
23314 2018-02-12 Richard Biener <rguenther@suse.de>
23316 PR tree-optimization/84037
23317 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
23318 parameter, move visited init to caller.
23319 (vect_slp_analyze_operations): Separate cost from validity
23320 check, initialize visited once for all instances.
23321 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
23323 * tree-vect-stmts.c (vect_model_simple_cost): Make early
23325 (vect_model_promotion_demotion_cost): Likewise.
23326 (vectorizable_bswap): Guard cost modeling with !slp_node
23327 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
23329 (vectorizable_call): Likewise.
23330 (vectorizable_conversion): Likewise.
23331 (vectorizable_assignment): Likewise.
23332 (vectorizable_shift): Likewise.
23333 (vectorizable_operation): Likewise.
23334 (vectorizable_store): Likewise.
23335 (vectorizable_load): Likewise.
23336 (vectorizable_condition): Likewise.
23337 (vectorizable_comparison): Likewise.
23339 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
23342 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
23343 (ASAN_MARK): Fix fnspec to account for return value, change pointer
23344 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
23346 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
23348 PR middle-end/83665
23349 * params.def (inline-min-speedup): Increase from 8 to 15.
23350 (max-inline-insns-auto): Decrease from 40 to 30.
23351 * ipa-split.c (consider_split): Add some buffer for function to
23352 be considered inlining candidate.
23353 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
23356 2018-02-12 Richard Biener <rguenther@suse.de>
23358 PR tree-optimization/84037
23359 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
23360 matched stmts if we cannot swap the non-matched ones.
23362 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
23364 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
23365 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
23366 _mm_maskz_scalef_round_ss): New intrinsics.
23367 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
23368 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
23369 __builtin_ia32_scalefss_round): Remove.
23370 (__builtin_ia32_scalefsd_mask_round,
23371 __builtin_ia32_scalefss_mask_round): New intrinsics.
23372 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
23373 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
23374 ((match_operand:VF_128 2 "<round_nimm_predicate>"
23375 "<round_constraint>")): Changed to ...
23376 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
23377 "<round_scalar_constraint>")): ... this.
23378 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
23379 %0, %1, %2<round_op3>}"): Changed to ...
23380 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
23381 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
23382 %2<round_scalar_mask_op3>}"): ... this.
23383 * config/i386/subst.md (round_scalar_nimm_predicate): New.
23385 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
23387 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
23388 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
23389 (_mm_maskz_sqrt_round_ss): New intrinsics.
23390 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
23391 (__builtin_ia32_sqrtsd_mask_round)
23392 (__builtin_ia32_sqrtss_mask_round): New builtins.
23393 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
23394 (__builtin_ia32_sqrtss_round): Remove.
23395 (__builtin_ia32_sqrtsd_mask_round)
23396 (__builtin_ia32_sqrtss_mask_round): New builtins.
23397 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
23398 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
23399 ((match_operand:VF_128 1 "vector_operand"
23400 "xBm,<round_constraint>")): Changed to ...
23401 ((match_operand:VF_128 1 "vector_operand"
23402 "xBm,<round_scalar_constraint>")): ... this.
23403 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
23404 %0, %2, %<iptr>1<round_op3>}): Changed to ...
23405 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
23406 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
23407 %<iptr>1<round_scalar_mask_op3>}): ... this.
23408 ((set_attr "prefix" "<round_prefix>")): Changed to ...
23409 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
23411 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
23414 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
23415 Cast vec_cmpeq result to correct type.
23416 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
23417 Cast vec_cmpgt result to correct type.
23419 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
23421 * final.c (final_scan_insn_1): Renamed from...
23422 (final_scan_insn): ... this. New wrapper, to recover
23423 seen from the outermost call in recursive ones.
23424 * config/sparc/sparc.c (output_return): Drop seen from call.
23425 (output_sibcall): Likewise.
23426 * config/visium/visium.c (output_branch): Likewise.
23428 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
23430 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
23433 2018-02-10 Alan Modra <amodra@gmail.com>
23436 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
23437 Specify LR as an input.
23439 2018-02-10 Jakub Jelinek <jakub@redhat.com>
23442 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
23443 remove_member_access_dummy_vars): New functions.
23444 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
23445 lower_omp_1, execute_lower_omp): Use them.
23447 PR rtl-optimization/84308
23448 * shrink-wrap.c (spread_components): Release todo vector.
23450 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
23452 PR rtl-optimization/57193
23453 * ira-color.c (struct allocno_color_data): Add member
23454 conflict_allocno_hard_prefs.
23455 (update_conflict_allocno_hard_prefs): New.
23456 (bucket_allocno_compare_func): Add a preference based on
23457 conflict_allocno_hard_prefs.
23458 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
23459 (color_allocnos): Remove a dead code. Initiate
23460 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
23462 2018-02-09 Jakub Jelinek <jakub@redhat.com>
23465 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
23466 constraint from =wa to wa. Avoid a subreg on the output operand,
23467 instead use a pseudo and subreg it in a move.
23468 (p9_xxbrd_<mode>): Changed to ...
23469 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
23470 (p9_xxbrd_v2df): New expander.
23471 (p9_xxbrw_<mode>): Changed to ...
23472 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
23473 (p9_xxbrw_v4sf): New expander.
23475 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
23477 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
23479 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
23482 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
23483 multiply in 32-bit mode.
23484 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
23485 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
23488 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
23490 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
23491 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
23492 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
23493 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
23495 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
23498 * dwarf2out.c (is_trivial_indirect_ref): New function.
23499 (dwarf2out_late_global_decl): Do not generate a location
23500 attribute for variables that have a non-trivial DECL_VALUE_EXPR
23501 and that are not defined in the current unit.
23503 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
23505 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
23506 instead of a libcall for UNORDERED.
23508 2018-02-09 Tamar Christina <tamar.christina@arm.com>
23511 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
23512 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
23514 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
23517 * config/s390/s390.c (s390_set_current_function): Invoke
23518 s390_indirect_branch_settings also if fndecl didn't change.
23520 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
23522 * config/rs6000/rs6000.md (blockage): Set length to zero.
23524 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
23526 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
23528 2018-02-09 Jakub Jelinek <jakub@redhat.com>
23531 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
23532 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
23536 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
23537 PARALLEL incoming that failed vt_get_decl_and_offset check.
23539 PR middle-end/84237
23540 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
23541 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
23543 (get_variable_section): For decls in named .bss* sections pass true as
23544 second argument to bss_initializer_p.
23546 2018-02-09 Marek Polacek <polacek@redhat.com>
23547 Jakub Jelinek <jakub@redhat.com>
23550 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
23551 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
23552 Sync some changes from cxx_fold_indirect_ref.
23554 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
23556 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
23558 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
23559 (BLOCK_INLINE_ENTRY_LABEL): New.
23560 (dwarf2out_var_location): Disregard inline entry markers.
23561 (inline_entry_data): New struct.
23562 (inline_entry_data_hasher): New hashtable type.
23563 (inline_entry_data_hasher::hash): New.
23564 (inline_entry_data_hasher::equal): New.
23565 (inline_entry_data_table): New variable.
23566 (add_high_low_attributes): Add DW_AT_entry_pc and
23567 DW_AT_GNU_entry_view attributes if a pending entry is found
23568 in inline_entry_data_table. Add old entry_pc attribute only
23569 if debug nonbinding markers are disabled.
23570 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
23571 markers are enabled.
23572 (block_within_block_p, dwarf2out_inline_entry): New.
23573 (dwarf2out_finish): Check that no entries remained in
23574 inline_entry_data_table.
23575 * final.c (reemit_insn_block_notes): Handle inline entry notes.
23576 (final_scan_insn, notice_source_line): Likewise.
23577 (rest_of_clean_state): Skip inline entry markers.
23578 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
23580 * gimple.c (gimple_build_debug_inline_entry): New.
23581 * gimple.h (enum gimple_debug_subcode): Add
23582 GIMPLE_DEBUG_INLINE_ENTRY.
23583 (gimple_build_debug_inline_entry): Declare.
23584 (gimple_debug_inline_entry_p): New.
23585 (gimple_debug_nonbind_marker_p): Adjust.
23586 * insn-notes.def (INLINE_ENTRY): New.
23587 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
23588 inline entry marker notes.
23589 (print_insn): Likewise.
23590 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
23591 (INSN_DEBUG_MARKER_KIND): Likewise.
23592 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
23593 * tree-inline.c (expand_call_inline): Build and insert
23594 debug_inline_entry stmt.
23595 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
23596 inline entry blocks early, if nonbind markers are enabled.
23597 (dump_scope_block): Dump fragment info.
23598 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
23599 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
23600 (gimple_build_debug_inline_entry): New.
23601 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
23602 Enable/disable inline entry points too.
23603 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
23604 (DEBUG_INSN): Describe inline entry markers.
23606 * common.opt (gvariable-location-views): New.
23607 (gvariable-location-views=incompat5): New.
23608 * config.in: Rebuilt.
23609 * configure: Rebuilt.
23610 * configure.ac: Test assembler for view support.
23611 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
23612 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
23613 * dwarf2out.c (var_loc_view): New typedef.
23614 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
23615 (dwarf2out_locviews_in_attribute): New.
23616 (dwarf2out_locviews_in_loclist): New.
23617 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
23618 (enum dw_line_info_opcode): Add LI_adv_address.
23619 (struct dw_line_info_table): Add view.
23620 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
23621 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
23622 (zero_view_p): New variable.
23623 (ZERO_VIEW_P): New macro.
23624 (output_asm_line_debug_info): New.
23625 (struct var_loc_node): Add view.
23626 (add_AT_view_list, AT_loc_list): New.
23627 (add_var_loc_to_decl): Add view param. Test it against last.
23628 (new_loc_list): Add view params. Record them.
23629 (AT_loc_list_ptr): Handle loc and view lists.
23630 (view_list_to_loc_list_val_node): New.
23631 (print_dw_val): Handle dw_val_class_view_list.
23632 (size_of_die): Likewise.
23633 (value_format): Likewise.
23634 (loc_list_has_views): New.
23635 (gen_llsym): Set vl_symbol too.
23636 (maybe_gen_llsym, skip_loc_list_entry): New.
23637 (dwarf2out_maybe_output_loclist_view_pair): New.
23638 (output_loc_list): Output view list or entries too.
23639 (output_view_list_offset): New.
23640 (output_die): Handle dw_val_class_view_list.
23641 (output_dwarf_version): New.
23642 (output_compilation_unit_header): Use it.
23643 (output_skeleton_debug_sections): Likewise.
23644 (output_rnglists, output_line_info): Likewise.
23645 (output_pubnames, output_aranges): Update version comments.
23646 (output_one_line_info_table): Output view numbers in asm comments.
23647 (dw_loc_list): Determine current endview, pass it to new_loc_list.
23648 Call maybe_gen_llsym.
23649 (loc_list_from_tree_1): Adjust.
23650 (add_AT_location_description): Create view list attribute if
23651 needed, check it's absent otherwise.
23652 (convert_cfa_to_fb_loc_list): Adjust.
23653 (maybe_emit_file): Call output_asm_line_debug_info for test.
23654 (dwarf2out_var_location): Reset views as needed. Precompute
23655 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
23656 attribute. Set view.
23657 (new_line_info_table): Reset next view.
23658 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
23659 (dwarf2out_source_line): Likewise. Output view resets and labels to
23660 the assembler, or select appropriate line info opcodes.
23661 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
23662 (optimize_string_length): Catch it. Adjust.
23663 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
23664 dw_val_class_view_list, and remove it if no longer needed.
23665 (hash_loc_list): Hash view numbers.
23666 (loc_list_hasher::equal): Compare them.
23667 (optimize_location_lists): Check whether a view list symbol is
23668 needed, and whether the locview attribute is present, and
23669 whether they match. Remove the locview attribute if no longer
23671 (index_location_lists): Call skip_loc_list_entry for test.
23672 (dwarf2out_finish): Call output_asm_line_debug_info for test.
23673 Use output_dwarf_version.
23674 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
23675 (struct dw_val_node): Add val_view_list.
23676 * final.c (SEEN_NEXT_VIEW): New.
23677 (set_next_view_needed): New.
23678 (clear_next_view_needed): New.
23679 (maybe_output_next_view): New.
23680 (final_start_function): Rename to...
23681 (final_start_function_1): ... this. Take pointer to FIRST,
23682 add SEEN parameter. Emit param bindings in the initial view.
23683 (final_start_function): Reintroduce SEEN-less interface.
23684 (final): Rename to...
23685 (final_1): ... this. Take SEEN parameter. Output final pending
23686 next view at the end.
23687 (final): Reintroduce seen-less interface.
23688 (final_scan_insn): Output pending next view before switching
23689 sections or ending a block. Mark the next view as needed when
23690 outputting variable locations. Notify debug backend of section
23691 changes, and of location view changes.
23692 (rest_of_handle_final): Adjust.
23693 * toplev.c (process_options): Autodetect value for debug variable
23694 location views option. Warn on incompat5 without -gdwarf-5.
23695 * doc/invoke.texi (gvariable-location-views): New.
23696 (gvariable-location-views=incompat5): New.
23697 (gno-variable-location-views): New.
23699 2018-02-08 David Malcolm <dmalcolm@redhat.com>
23701 PR tree-optimization/84136
23702 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
23703 that the result of find_edge is non-NULL.
23705 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
23708 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
23709 storing integer register in SImode. Fix cost of 256 and 512
23710 byte aligned SSE register store.
23712 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
23714 * config/i386/i386.c (ix86_multiplication_cost): Fix
23715 multiplication cost for TARGET_AVX512DQ.
23717 2018-02-08 Marek Polacek <polacek@redhat.com>
23719 PR tree-optimization/84238
23720 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
23723 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
23725 PR tree-optimization/84265
23726 * tree-vect-stmts.c (vectorizable_store): Don't treat
23727 VMAT_CONTIGUOUS accesses as grouped.
23728 (vectorizable_load): Likewise.
23730 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
23732 PR tree-optimization/81635
23733 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
23734 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
23735 (test_round_for_mask): New functions.
23736 (wide_int_cc_tests): Call test_round_for_mask.
23737 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
23738 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
23739 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
23740 range returned by get_range_info.
23742 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
23745 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
23746 * symtab.c: Include builtins.h
23747 (symtab_node::output_to_lto_symbol_table_p): Move here
23748 from lto-streamer-out.c:output_symbol_p.
23749 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
23750 (output_symbol_p): Move all logic to symtab.c
23751 (produce_symtab): Update.
23753 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
23755 * config/s390/s390-opts.h (enum indirect_branch): Define.
23756 * config/s390/s390-protos.h (s390_return_addr_from_memory)
23757 (s390_indirect_branch_via_thunk)
23758 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
23759 (enum s390_indirect_branch_type): Define.
23760 * config/s390/s390.c (struct s390_frame_layout, struct
23761 machine_function): Remove.
23762 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
23763 (indirect_branch_table_label_no, indirect_branch_table_name):
23765 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
23766 (enum s390_indirect_branch_option): Define.
23767 (s390_return_addr_from_memory): New function.
23768 (s390_handle_string_attribute): New function.
23769 (s390_attribute_table): Add new attribute handler.
23770 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
23771 (s390_indirect_branch_via_thunk): New function.
23772 (s390_indirect_branch_via_inline_thunk): New function.
23773 (s390_function_ok_for_sibcall): When jumping via thunk disallow
23774 sibling call optimization for non z10 compiles.
23775 (s390_emit_call): Force indirect branch target to be a single
23776 register. Add r1 clobber for non-z10 compiles.
23777 (s390_emit_epilogue): Emit return jump via return_use expander.
23778 (s390_reorg): Handle JUMP_INSNs as execute targets.
23779 (s390_option_override_internal): Perform validity checks for the
23780 new command line options.
23781 (s390_indirect_branch_attrvalue): New function.
23782 (s390_indirect_branch_settings): New function.
23783 (s390_set_current_function): Invoke s390_indirect_branch_settings.
23784 (s390_output_indirect_thunk_function): New function.
23785 (s390_code_end): Implement target hook.
23786 (s390_case_values_threshold): Implement target hook.
23787 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
23789 * config/s390/s390.h (struct s390_frame_layout)
23790 (struct machine_function): Move here from s390.c.
23791 (TARGET_INDIRECT_BRANCH_NOBP_RET)
23792 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
23793 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
23794 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
23795 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
23796 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
23797 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
23798 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
23799 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
23800 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
23801 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
23802 (mnemonic attribute): Add values which aren't recognized
23804 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
23805 pattern for branch conversion. Fix mnemonic attribute.
23806 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
23807 indirect branch via thunk if requested.
23808 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
23809 ("*indirect_jump"): Disable for branch conversion using out of
23811 ("indirect_jump_via_thunk<mode>_z10")
23812 ("indirect_jump_via_thunk<mode>")
23813 ("indirect_jump_via_inlinethunk<mode>_z10")
23814 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
23815 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
23816 ("casesi_jump_via_inlinethunk<mode>_z10")
23817 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
23818 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
23819 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
23820 ("*indirect2_jump"): Disable for branch conversion.
23821 ("casesi_jump"): Turn into expander and expand patterns for branch
23823 ("return_use"): New expander.
23824 ("*return"): Emit return via thunk and rename it to ...
23825 ("*return<mode>"): ... this one.
23826 * config/s390/s390.opt: Add new options and and enum for the
23829 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
23831 * lra-constraints.c (match_reload): Unconditionally use
23832 gen_lowpart_SUBREG, rather than selecting between that
23833 and equivalent gen_rtx_SUBREG code.
23835 2018-02-08 Richard Biener <rguenther@suse.de>
23837 PR tree-optimization/84233
23838 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
23839 changed flag instead of boguously re-using phi_inserted.
23841 2018-02-08 Martin Jambor <mjambor@suse.cz>
23843 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
23844 static local variables.
23846 2018-02-08 Richard Biener <rguenther@suse.de>
23848 PR tree-optimization/84278
23849 * tree-vect-stmts.c (vectorizable_store): When looking for
23850 smaller vector types to perform grouped strided loads/stores
23851 make sure the mode is supported by the target.
23852 (vectorizable_load): Likewise.
23854 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
23856 * config/aarch64/aarch64.c (aarch64_components_for_bb):
23857 Increase LDP/STP opportunities by adding adjacent callee-saves.
23859 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
23861 PR rtl-optimization/84068
23862 PR rtl-optimization/83459
23863 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
23865 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
23867 PR tree-optimization/84224
23868 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
23869 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
23870 non-zero arguments.
23872 2018-02-07 Iain Sandoe <iain@codesourcery.com>
23875 * config/rs6000/altivec.md (*restore_world): Remove LR use.
23876 * config/rs6000/predicates.md (restore_world_operation): Adjust op
23877 count, remove one USE.
23879 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
23881 * doc/install.texi (Configuration): Document the
23882 --with-long-double-format={ibm,ieee} PowerPC configuration
23886 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
23887 Convert from define_expand to be define_insn_and_split. Rework
23888 float/double/_Float128 conversions to QI/HI/SImode to work with
23889 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
23890 conversions to QI/HImode types did a store and then a load to
23891 truncate the value. For conversions to VSX registers, don't split
23892 the insn, instead emit the code directly. Use the code iterator
23893 any_fix to combine signed and unsigned conversions.
23894 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
23895 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
23896 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
23897 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
23898 (fix_<mode>di2_hw): Likewise.
23899 (fixuns_<mode>di2_hw): Likewise.
23900 (fix_<mode>si2_hw): Likewise.
23901 (fixuns_<mode>si2_hw): Likewise.
23902 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
23903 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
23904 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
23905 fix<uns>_trunc<SFDF:mode>si2_p8.
23906 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
23907 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
23908 (fix<uns>_<mode>_mem): Likewise.
23909 (fctiw<u>z_<mode>_mem): Likewise.
23910 (fix<uns>_<mode>_mem): Likewise.
23911 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
23912 the register allocator from doing a direct move to the GPRs to do
23913 a store, and instead use the ISA 3.0 store byte/half-word from
23914 vector register instruction. For IEEE 128-bit floating point,
23915 also optimize stores of 32-bit ints.
23916 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
23918 2018-02-07 Alan Hayward <alan.hayward@arm.com>
23920 * genextract.c (push_pathstr_operand): New function to support
23922 (walk_rtx): Call push_pathstr_operand.
23923 (print_path): Support [a-zA-Z].
23925 2018-02-07 Richard Biener <rguenther@suse.de>
23927 PR tree-optimization/84037
23928 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
23929 (cse_and_gimplify_to_preheader): Declare.
23930 (vect_get_place_in_interleaving_chain): Likewise.
23931 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
23933 (_loop_vec_info::~_loop_vec_info): Delete it.
23934 (cse_and_gimplify_to_preheader): New function.
23935 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
23936 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
23937 (vectorizable_load): Likewise. For grouped stores always base
23938 the IV on the first element.
23939 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
23940 condition before gimplifying.
23942 2018-02-07 Jakub Jelinek <jakub@redhat.com>
23944 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
23945 *DIV_EXPR and *MOD_EXPR.
23947 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
23950 * config/i386/i386.c (ix86_option_override_internal): Mask out
23951 the CF_SET bit when checking -fcf-protection.
23953 2018-02-07 Tom de Vries <tom@codesourcery.com>
23956 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
23959 2018-02-07 Richard Biener <rguenther@suse.de>
23961 PR tree-optimization/84204
23962 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
23965 PR tree-optimization/84205
23966 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
23967 special-case isl_ast_op_zdiv_r.
23969 PR tree-optimization/84223
23970 * graphite-scop-detection.c (gather_bbs::before_dom_children):
23971 Only add conditions from within the region.
23972 (gather_bbs::after_dom_children): Adjust.
23974 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
23977 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
23978 * config/avr/avr.md: Only post-reload split REG-REG moves if
23979 either register is GENERAL_REG_P.
23981 2018-02-07 Jakub Jelinek <jakub@redhat.com>
23983 PR tree-optimization/84235
23984 * tree-ssa-scopedtables.c
23985 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
23986 if the subtraction is performed in floating point type where NaNs are
23987 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
23988 build 1. Formatting fix.
23990 2018-02-06 Jakub Jelinek <jakub@redhat.com>
23993 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
23994 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
23995 and skip it regardless of bb boundaries. Use CALL_P macro,
23996 don't test INSN_P (insn) together with CALL_P or JUMP_P check
23997 unnecessarily, formatting fix.
23999 2018-02-06 Michael Collison <michael.collison@arm.com>
24001 * config/arm/thumb2.md:
24002 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
24003 (*thumb_mov_notscc): Ditto.
24005 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
24008 * config/rs6000/rs6000.md (su code attribute): Use "u" for
24009 unsigned_fix, not "s".
24011 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
24013 * configure.ac (gcc_fn_eh_frame_ro): New function.
24014 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
24015 correct .eh_frame permissions.
24016 * configure: Regenerate.
24018 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
24020 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
24021 irrelevant options.
24023 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
24025 * config/rs6000/rs6000.c (rs6000_option_override_internal):
24026 Display warning message for -mno-speculate-indirect-jumps.
24028 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
24030 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
24032 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
24034 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
24036 PR tree-optimization/84225
24037 * tree-eh.c (find_trapping_overflow): Only call
24038 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
24040 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
24043 * config/i386/i386.c: Reimplement the check of possible options
24044 -mibt/-mshstk conbination. Change error messages.
24045 * doc/invoke.texi: Fix a typo: remove extra '='.
24047 2018-02-06 Marek Polacek <polacek@redhat.com>
24049 PR tree-optimization/84228
24050 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
24052 2018-02-06 Tamar Christina <tamar.christina@arm.com>
24055 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
24056 emitted arch directives.
24057 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
24058 __ARM_FEATURE_COPROC before changing architectures.
24060 2018-02-06 Richard Biener <rguenther@suse.de>
24062 * config/i386/i386.c (print_reg): Fix typo.
24063 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
24065 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
24067 * configure: Regenerate.
24069 2018-02-05 Martin Sebor <msebor@redhat.com>
24071 PR tree-optimization/83369
24072 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
24075 2018-02-05 Martin Liska <mliska@suse.cz>
24077 * doc/invoke.texi: Cherry-pick upstream r323995.
24079 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
24081 * ira.c (ira_init_register_move_cost): Adjust comment.
24083 2018-02-05 Martin Liska <mliska@suse.cz>
24085 PR gcov-profile/84137
24086 * doc/gcov.texi: Fix typo in documentation.
24088 2018-02-05 Martin Liska <mliska@suse.cz>
24090 PR gcov-profile/83879
24091 * doc/gcov.texi: Document necessity of --dynamic-list-data when
24092 using dlopen functionality.
24094 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
24096 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
24097 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
24098 _mm_maskz_range_ss, _mm_mask_range_round_ss,
24099 _mm_maskz_range_round_ss): New intrinsics.
24100 (__builtin_ia32_rangesd128_round)
24101 (__builtin_ia32_rangess128_round): Remove.
24102 (__builtin_ia32_rangesd128_mask_round,
24103 __builtin_ia32_rangess128_mask_round): New builtins.
24104 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
24105 __builtin_ia32_rangess128_round): Remove.
24106 (__builtin_ia32_rangesd128_mask_round,
24107 __builtin_ia32_rangess128_mask_round): New builtins.
24108 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
24109 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
24110 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
24111 "<round_saeonly_constraint>")): Changed to ...
24112 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
24113 "<round_saeonly_scalar_constraint>")): ... this.
24114 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
24115 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
24116 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
24117 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
24118 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
24120 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
24122 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
24124 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
24125 Remove all values except native, 8540 and 8548.
24127 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
24129 * config/i386/i386.c (ix86_output_function_return): Pass
24130 INVALID_REGNUM, instead of -1, as invalid register number to
24131 indirect_thunk_name and output_indirect_thunk.
24133 2018-02-02 Julia Koval <julia.koval@intel.com>
24135 * config.gcc: Add -march=icelake.
24136 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
24137 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
24138 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
24139 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
24140 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
24141 (processor_target_table): Add icelake.
24142 (ix86_option_override_internal): Handle new PTAs.
24143 (get_builtin_code_for_version): Handle icelake.
24144 (M_INTEL_COREI7_ICELAKE): New.
24145 (fold_builtin_cpu): Handle icelake.
24146 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
24147 * doc/invoke.texi: Add -march=icelake.
24149 2018-02-02 Julia Koval <julia.koval@intel.com>
24151 * config/i386/i386.c (ix86_option_override_internal): Change flags type
24152 to wide_int_bitmask.
24153 * wide-int-bitmask.h: New.
24155 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
24158 * config/i386/i386.md: Replace Pmode with word_mode in
24159 builtin_setjmp_setup and builtin_longjmp to support x32.
24161 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
24165 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
24167 (rs6000_supported_cpu_names): New static variable.
24168 (linux_cpu_translation_table): Likewise.
24169 (elf_platform) <cpu>: Define new static variable and use it.
24170 Translate kernel AT_PLATFORM name to canonical name if needed.
24171 Error if platform name is unknown.
24173 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
24176 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
24178 2018-02-01 Jeff Law <law@redhat.com>
24181 * config/i386/i386.c (release_scratch_register_on_entry): Add new
24182 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
24183 the scratch if RELEASE_VIA_POP is false.
24184 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
24185 If we have to save a temporary register, decrement SIZE appropriately.
24186 Pass new arguments to release_scratch_register_on_entry.
24187 (ix86_adjust_stack_and_probe): Likewise.
24188 (ix86_emit_probe_stack_range): Pass new arguments to
24189 release_scratch_register_on_entry.
24191 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
24193 PR rtl-optimization/84157
24194 * combine.c (change_zero_ext): Use REG_P predicate in
24195 front of HARD_REGISTER_P predicate.
24197 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
24199 * config/avr/avr.c (avr_option_override): Move disabling of
24200 -fdelete-null-pointer-checks to...
24201 * common/config/avr/avr-common.c (avr_option_optimization_table):
24204 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
24206 PR tree-optimization/81635
24207 * tree-data-ref.c (split_constant_offset_1): For types that
24208 wrap on overflow, try to use range info to prove that wrapping
24211 2018-02-01 Renlin Li <renlin.li@arm.com>
24214 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
24215 TAILCALL_ADDR_REGS.
24216 (aarch64_register_move_cost): Likewise.
24217 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
24218 TAILCALL_ADDR_REGS.
24219 (REG_CLASS_NAMES): Likewise.
24220 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
24221 TAILCALL_ADDR_REGS. Remove IP registers.
24222 * config/aarch64/aarch64.md (Ucs): Update register constraint.
24224 2018-02-01 Richard Biener <rguenther@suse.de>
24226 * domwalk.h (dom_walker::dom_walker): Add additional constructor
24227 for specifying RPO order and allow NULL for that.
24228 * domwalk.c (dom_walker::dom_walker): Likewise.
24229 (dom_walker::walk): Handle NULL RPO order.
24230 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
24232 (rewrite_update_dom_walker): Likewise.
24233 (mark_def_dom_walker): Likewise.
24235 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
24237 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
24238 (aarch64_maybe_expand_sve_subreg_move): Declare.
24239 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
24240 * config/aarch64/predicates.md (aarch64_any_register_operand): New
24242 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
24243 that are semantically a reverse operation.
24244 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
24245 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
24246 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
24248 (aarch64_can_change_mode_class): For big-endian, forbid changes
24249 between two SVE modes if they have different element sizes.
24251 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
24253 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
24254 the TImode handling for big-endian targets.
24256 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
24258 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
24259 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
24261 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
24262 Remove BSWAP handing for big-endian targets and use the form of
24263 LD1RQ appropariate for the mode.
24265 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
24267 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
24268 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
24269 duplicated element.
24271 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
24274 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
24275 check for operands that need to go through aarch64_sve_reload_be.
24277 2018-02-01 Jakub Jelinek <jakub@redhat.com>
24279 PR tree-optimization/81661
24280 PR tree-optimization/84117
24281 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
24282 * tree-eh.c: Include gimplify.h.
24283 (find_trapping_overflow, replace_trapping_overflow,
24284 rewrite_to_non_trapping_overflow): New functions.
24285 * tree-vect-loop.c: Include tree-eh.h.
24286 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
24287 * tree-data-ref.c: Include tree-eh.h.
24288 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
24290 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
24292 PR rtl-optimization/84123
24293 * combine.c (change_zero_ext): Check if hard register satisfies
24294 can_change_dest_mode before calling gen_lowpart_SUBREG.
24296 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
24299 * ira.c (ira_init_register_move_cost): Remove assert.
24301 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
24303 PR rtl-optimization/84071
24304 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
24305 * doc/tm.texi: Regenerate.
24307 2018-01-31 Richard Biener <rguenther@suse.de>
24309 PR tree-optimization/84132
24310 * tree-data-ref.c (analyze_miv_subscript): Properly
24311 check whether evolution_function_is_affine_multivariate_p
24312 before calling gcd_of_steps_may_divide_p.
24314 2018-01-31 Julia Koval <julia.koval@intel.com>
24317 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
24318 * config/i386/i386.md (rdpid_rex64) New.
24319 (rdpid): Make 32bit only.
24321 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
24324 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
24325 an IDENTIFIER_NODE for FUNCTION_TYPE's.
24327 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
24330 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
24332 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
24334 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
24336 PR rtl-optimization/84071
24337 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
24338 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
24340 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
24342 * config/arc/arc.c (arc_handle_aux_attribute): New function.
24343 (arc_attribute_table): Add 'aux' attribute.
24344 (arc_in_small_data_p): Consider aux like variables.
24345 (arc_is_aux_reg_p): New function.
24346 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
24347 (arc_get_aux_arg): New function.
24348 (prepare_move_operands): Handle aux-register access.
24349 (arc_handle_aux_attribute): New function.
24350 * doc/extend.texi (ARC Variable attributes): Add subsection.
24352 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
24354 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
24355 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
24356 (arc_attribute_table): Add 'uncached' attribute.
24357 (arc_print_operand): Print '.di' flag for uncached memory
24359 (arc_in_small_data_p): Do not consider for small data the uncached
24361 (arc_is_uncached_mem_p): New function.
24362 * config/arc/predicates.md (compact_store_memory_operand): Check
24363 for uncached memory accesses.
24364 (nonvol_nonimm_operand): Likewise.
24365 * doc/extend.texi (ARC Type Attribute): New subsection.
24367 2018-01-31 Jakub Jelinek <jakub@redhat.com>
24370 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
24371 falign-loops=): Add Optimization flag.
24373 2018-01-30 Jeff Law <law@redhat.com>
24376 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
24377 INT_REGISTERS_SAVED. Check it prior to calling
24378 get_scratch_register_on_entry.
24379 (ix86_adjust_stack_and_probe): Similarly.
24380 (ix86_emit_probe_stack_range): Similarly.
24381 (ix86_expand_prologue): Corresponding changes.
24383 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
24386 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
24387 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
24389 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
24392 * lra-constraints.c (curr_insn_transform): Process AND in the
24395 2018-01-30 Jakub Jelinek <jakub@redhat.com>
24397 PR rtl-optimization/83986
24398 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
24399 dependence against last_pending_memory_flush in addition to
24400 pending_jump_insns.
24402 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
24404 PR tree-optimization/81611
24405 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
24408 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
24411 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
24414 2018-01-30 Richard Biener <rguenther@suse.de>
24415 Jakub Jelinek <jakub@redhat.com>
24417 PR tree-optimization/84111
24418 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
24419 inner loops added during recursion, as they don't have up-to-date
24422 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
24425 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
24426 (can_inline_edge_by_limits_p): ... here.
24427 (can_early_inline_edge_p, check_callers,
24428 update_caller_keys, update_callee_keys, recursive_inlining,
24429 add_new_edges_to_heap, speculation_useful_p,
24430 inline_small_functions,
24431 inline_small_functions, flatten_function,
24432 inline_to_all_callers_1): Update.
24434 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
24436 * profile-count.c (profile_count::combine_with_ipa_count): Handle
24439 2018-01-30 Richard Biener <rguenther@suse.de>
24441 PR tree-optimization/83008
24442 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
24443 invariant and constant vector uses in stmts when they need
24444 more than one stmt.
24446 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
24449 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
24450 * configure: Regenerate.
24452 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
24454 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
24456 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
24457 Use gen_rtx_REG rather than gen_lowpart.
24459 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
24461 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
24462 rather than 0 when creating partial subregs.
24464 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
24466 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
24469 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
24472 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
24473 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
24474 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
24475 flags. This restores the settings used before the 2017-07-24.
24476 Turning off pre increment/decrement/modify allows IVOPTS to
24477 optimize DF/SF loops where the index is an int.
24479 2018-01-29 Richard Biener <rguenther@suse.de>
24480 Kelvin Nilsen <kelvin@gcc.gnu.org>
24483 * tree-vect-stmts.c (vectorizable_call): Don't call
24484 targetm.vectorize_builtin_md_vectorized_function if callee is
24487 2018-01-22 Carl Love <cel@us.ibm.com>
24489 * doc/extend.tex: Fix typo in second arg in
24490 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
24492 2018-01-29 Richard Biener <rguenther@suse.de>
24494 PR tree-optimization/84086
24495 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
24496 (flush_ssaname_freelist): When SSA names were released reset
24497 the SCEV hash table.
24499 2018-01-29 Richard Biener <rguenther@suse.de>
24501 PR tree-optimization/84057
24502 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
24503 removed paths when removing edges.
24505 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
24507 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
24508 -mfunction-return=@var{choice}.
24510 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
24512 PR diagnostic/84034
24513 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
24514 Handle CR like TAB.
24515 (layout::print_source_line): Likewise.
24516 (test_get_line_width_without_trailing_whitespace): Add test cases.
24518 2018-01-27 Jakub Jelinek <jakub@redhat.com>
24520 PR middle-end/84040
24521 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
24524 2018-01-26 Jim Wilson <jimw@sifive.com>
24526 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
24528 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
24531 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24533 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
24534 and CMP + SUB-immediate -> SUBS.
24536 2018-01-26 Martin Sebor <msebor@redhat.com>
24538 PR tree-optimization/83896
24539 * tree-ssa-strlen.c (get_string_len): Rename...
24540 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
24541 Avoid assuming length is constant.
24542 (handle_char_store): Use HOST_WIDE_INT for string length.
24544 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
24547 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
24548 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
24550 2018-01-26 Richard Biener <rguenther@suse.de>
24552 PR rtl-optimization/84003
24553 * dse.c (record_store): Only record redundant stores when
24554 the earlier store aliases at least all accesses the later one does.
24556 2018-01-26 Jakub Jelinek <jakub@redhat.com>
24558 PR rtl-optimization/83985
24559 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
24560 REG_CFA_RESTORE insns.
24561 (delete_unmarked_insns): Don't ignore separate shrink wrapping
24562 REG_CFA_RESTORE insns here.
24565 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
24566 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
24568 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
24570 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
24571 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
24572 (arc_init): Likewise.
24573 (arc_override_options): Likewise.
24574 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
24576 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
24578 * config/arc/arc.h (TARGET_DBNZ): Define.
24579 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
24580 properly set the tune attribute.
24581 (dbnz): Use TARGET_DBNZ guard.
24582 * config/arc/arc.opt (mtune): Add core3 option.
24584 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
24586 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
24587 recognize new pic like addresses.
24588 (arc_delegitimize_address): Clean up.
24590 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
24592 * config/arc/arc-arches.def: Option mrf16 valid for all
24594 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
24595 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
24596 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
24597 * config/arc/arc-tables.opt: Regenerate.
24598 * config/arc/arc.c (arc_conditional_register_usage): Handle
24599 reduced register file case.
24600 (arc_file_start): Set must have build attributes.
24601 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
24602 mrf16 option value.
24603 * config/arc/arc.opt (mrf16): Add new option.
24604 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
24605 * config/arc/genmultilib.awk: Handle new mrf16 option.
24606 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
24607 * config/arc/t-multilib: Regenerate.
24608 * doc/invoke.texi (ARC Options): Document mrf16 option.
24610 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
24612 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
24613 * config/arc/arc.c (arc_handle_secure_attribute): New function.
24614 (arc_attribute_table): Add 'secure_call' attribute.
24615 (arc_print_operand): Print secure call operand.
24616 (arc_function_ok_for_sibcall): Don't optimize tail calls when
24618 (arc_is_secure_call_p): New function. * config/arc/arc.md
24619 (call_i): Add support for sjli instruction.
24620 (call_value_i): Likewise.
24621 * config/arc/constraints.md (Csc): New constraint.
24623 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
24624 John Eric Martin <John.Martin@emmicro-us.com>
24626 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
24627 * config/arc/arc.c (_arc_jli_section): New struct.
24628 (arc_jli_section): New type.
24629 (rc_jli_sections): New static variable.
24630 (arc_handle_jli_attribute): New function.
24631 (arc_attribute_table): Add jli_always and jli_fixed attribute.
24632 (arc_file_end): New function.
24633 (TARGET_ASM_FILE_END): Define.
24634 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
24635 (arc_add_jli_section): New function.
24636 (jli_call_scan): Likewise.
24637 (arc_reorg): Call jli_call_scan.
24638 (arc_output_addsi): Remove 'S' from printing asm operand.
24639 (arc_is_jli_call_p): New function.
24640 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
24642 (movhi_insn): Likewise.
24643 (movsi_insn): Likewise.
24644 (movsi_set_cc_insn): Likewise.
24645 (loadqi_update): Likewise.
24646 (load_zeroextendqisi_update): Likewise.
24647 (load_signextendqisi_update): Likewise.
24648 (loadhi_update): Likewise.
24649 (load_zeroextendhisi_update): Likewise.
24650 (load_signextendhisi_update): Likewise.
24651 (loadsi_update): Likewise.
24652 (loadsf_update): Likewise.
24653 (movsicc_insn): Likewise.
24654 (bset_insn): Likewise.
24655 (bxor_insn): Likewise.
24656 (bclr_insn): Likewise.
24657 (bmsk_insn): Likewise.
24658 (bicsi3_insn): Likewise.
24659 (cmpsi_cc_c_insn): Likewise.
24660 (movsi_ne): Likewise.
24661 (movsi_cond_exec): Likewise.
24662 (clrsbsi2): Likewise.
24663 (norm_f): Likewise.
24673 (call_i): Remove 'S' asm letter, add jli instruction.
24674 (call_value_i): Likewise.
24675 * config/arc/arc.op (mjli-always): New option.
24676 * config/arc/constraints.md (Cji): New constraint.
24677 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
24679 (subsf3_fpx): Likewise.
24680 (mulsf3_fpx): Likewise.
24681 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
24683 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
24684 function attrbutes.
24685 * doc/invoke.texi (ARC): Document mjli-always option.
24687 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
24689 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
24690 avoid addition with 0 and use incw and decw where possible.
24692 2018-01-26 Richard Biener <rguenther@suse.de>
24694 PR tree-optimization/81082
24695 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
24696 association if it requires casting to unsigned.
24697 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
24698 from fold_plusminus_mult_expr to catch important cases late when
24699 range info is available.
24701 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
24703 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
24704 * configure.ac (hidden_linkonce): New test.
24705 * configure: Regenerate.
24706 * config.in: Regenerate.
24708 2018-01-26 Julia Koval <julia.koval@intel.com>
24710 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
24711 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
24712 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
24713 _mm_mask_bitshuffle_epi64_mask): Fix type.
24714 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
24715 USI_FTYPE_V4DI_V4DI_USI): Remove.
24716 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
24717 __builtin_ia32_vpshufbitqmb256_mask,
24718 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
24719 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
24720 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
24722 2018-01-26 Alan Modra <amodra@gmail.com>
24725 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
24726 UNSPEC_VBPERMQ. Sort other unspecs.
24728 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
24730 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
24732 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
24734 PR middle-end/83055
24735 * predict.c (drop_profile): Do not push/pop cfun; update also
24737 (handle_missing_profiles): Fix logic looking for zero profiles.
24739 2018-01-25 Jakub Jelinek <jakub@redhat.com>
24741 PR middle-end/83977
24742 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
24743 on functions with #pragma omp declare simd or functions with simd
24745 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
24746 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
24747 Remove trailing \n from warning_at calls.
24749 2018-01-25 Tom de Vries <tom@codesourcery.com>
24752 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
24753 for neutered workers.
24755 2018-01-24 Joseph Myers <joseph@codesourcery.com>
24758 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
24759 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
24761 2018-01-24 Jeff Law <law@redhat.com>
24764 * i386.c (get_probe_interval): Move to earlier point.
24765 (ix86_compute_frame_layout): If -fstack-clash-protection and
24766 the frame is larger than the probe interval, then use pushes
24767 to save registers rather than reg->mem moves.
24768 (ix86_expand_prologue): Remove conditional for int_registers_saved
24771 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
24774 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
24775 min/max for never referenced object.
24777 2018-01-24 Jakub Jelinek <jakub@redhat.com>
24779 PR middle-end/83977
24780 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
24782 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
24783 attributes from DECL_ATTRIBUTES (decl) without affecting
24784 DECL_ATTRIBUTES (current_function_decl).
24785 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
24786 functions with non-NULL DECL_ABSTRACT_ORIGIN.
24788 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
24790 PR tree-optimization/83979
24791 * fold-const.c (fold_comparison): Use constant_boolean_node
24792 instead of boolean_{true,false}_node.
24794 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
24796 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
24799 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
24801 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
24802 Simplify the clause that sets the length attribute.
24803 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
24804 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
24805 clause that sets the length attribute.
24806 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
24808 2018-01-24 Tom de Vries <tom@codesourcery.com>
24811 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
24812 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
24813 Add strict parameter.
24814 (prevent_branch_around_nothing): Insert dummy insn between branch to
24815 label and label with no ptx insn inbetween.
24816 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
24818 2018-01-24 Tom de Vries <tom@codesourcery.com>
24821 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
24822 for neutered threads in warp.
24823 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
24825 2018-01-24 Richard Biener <rguenther@suse.de>
24827 PR tree-optimization/83176
24828 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
24831 2018-01-24 Richard Biener <rguenther@suse.de>
24833 PR tree-optimization/82819
24834 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
24835 code generating pluses that are no-ops in the target precision.
24837 2018-01-24 Richard Biener <rguenther@suse.de>
24839 PR middle-end/84000
24840 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
24842 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
24844 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
24845 to merge probabilities.
24846 * predict.c (probably_never_executed): Also mark as cold functions
24847 with global 0 profile and guessed local profile.
24848 * profile-count.c (profile_probability::combine_with_count): New
24850 * profile-count.h (profile_probability::operator*,
24851 profile_probability::operator*=, profile_probability::operator/,
24852 profile_probability::operator/=): Reduce precision to adjusted
24853 and set value to guessed on contradictory divisions.
24854 (profile_probability::combine_with_freq): Remove.
24855 (profile_probability::combine_wiht_count): Declare.
24856 (profile_count::force_nonzero):: Set to adjusted.
24857 (profile_count::probability_in):: Set quality to adjusted.
24858 * tree-ssa-tail-merge.c (replace_block_by): Use
24859 combine_with_count.
24861 2018-01-23 Andrew Waterman <andrew@sifive.com>
24862 Jim Wilson <jimw@sifive.com>
24864 * config/riscv/riscv.c (riscv_stack_boundary): New.
24865 (riscv_option_override): Set riscv_stack_boundary. Handle
24866 riscv_preferred_stack_boundary_arg.
24867 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
24868 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
24869 (STACK_BOUNDARY): Set to riscv_stack_boundary.
24870 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
24871 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
24872 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
24874 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
24877 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
24878 of struct ix86_frame.
24879 (ix86_expand_epilogue): Likewise. Add a local variable for
24880 the reg_save_offset field in struct ix86_frame.
24882 2018-01-23 Bin Cheng <bin.cheng@arm.com>
24884 PR tree-optimization/82604
24885 * tree-loop-distribution.c (enum partition_kind): New enum item
24886 PKIND_PARTIAL_MEMSET.
24887 (partition_builtin_p): Support above new enum item.
24888 (generate_code_for_partition): Ditto.
24889 (compute_access_range): Differentiate cases that equality can be
24890 proven at all loops, the innermost loops or no loops.
24891 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
24892 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
24893 (finalize_partitions, distribute_loop): Don't fuse partition of
24894 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
24895 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
24896 parloop is enabled.
24898 2018-01-23 Martin Liska <mliska@suse.cz>
24900 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
24901 order to ignore the predictor.
24902 (PRED_POLYMORPHIC_CALL): Likewise.
24903 (PRED_RECURSIVE_CALL): Likewise.
24905 2018-01-23 Martin Liska <mliska@suse.cz>
24907 * tree-profile.c (tree_profiling): Print function header to
24908 aware reader which function we are working on.
24909 * value-prof.c (gimple_find_values_to_profile): Do not print
24910 not interesting value histograms.
24912 2018-01-23 Martin Liska <mliska@suse.cz>
24914 * profile-count.h (enum profile_quality): Add
24915 profile_uninitialized as the first value. Do not number values
24916 as they are zero based.
24917 (profile_count::verify): Update sanity check.
24918 (profile_probability::verify): Likewise.
24920 2018-01-23 Nathan Sidwell <nathan@acm.org>
24922 * doc/invoke.texi (ffor-scope): Deprecate.
24924 2018-01-23 David Malcolm <dmalcolm@redhat.com>
24926 PR tree-optimization/83510
24927 * domwalk.c (set_all_edges_as_executable): New function.
24928 (dom_walker::dom_walker): Convert bool param
24929 "skip_unreachable_blocks" to enum reachability. Move setup of
24930 edge flags to set_all_edges_as_executable and only do it when
24931 reachability is REACHABLE_BLOCKS.
24932 * domwalk.h (enum dom_walker::reachability): New enum.
24933 (dom_walker::dom_walker): Convert bool param
24934 "skip_unreachable_blocks" to enum reachability.
24935 (set_all_edges_as_executable): New decl.
24936 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
24937 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
24939 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
24940 but converting true to REACHABLE_BLOCKS.
24941 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
24943 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
24944 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
24945 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
24947 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
24948 if check_all_array_refs will be called.
24950 2018-01-23 David Malcolm <dmalcolm@redhat.com>
24952 * tree.c (selftest::test_location_wrappers): Add more test
24955 2018-01-23 David Malcolm <dmalcolm@redhat.com>
24957 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
24958 (selftest::test_bit_in_range): Likewise.
24960 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
24963 * doc/sourcebuild.texi (vect_float): Say that the selector
24964 only describes the situation when -funsafe-math-optimizations is on.
24965 (vect_float_strict): Document.
24967 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
24969 PR tree-optimization/83965
24970 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
24971 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
24972 instead of checking only for a reduction.
24973 (vect_recog_widen_sum_pattern): Likewise.
24975 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
24977 * predict.c (probably_never_executed): Only use precise profile info.
24978 (compute_function_frequency): Skip after inlining hack since we now
24979 have quality checking.
24981 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
24983 * profile-count.h (profile_probability::very_unlikely,
24984 profile_probability::unlikely, profile_probability::even): Set
24985 precision to guessed.
24987 2018-01-23 Richard Biener <rguenther@suse.de>
24989 PR tree-optimization/83963
24990 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
24991 Properly terminate dominator walk when crossing the exit edge not
24992 when visiting its source block.
24994 2018-01-23 Jakub Jelinek <jakub@redhat.com>
24997 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
24998 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
25000 2018-01-22 Jakub Jelinek <jakub@redhat.com>
25002 PR tree-optimization/83957
25003 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
25004 semicolon after for body surrounded by braces.
25006 PR tree-optimization/83081
25007 * profile-count.h (profile_probability::split): New method.
25008 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
25009 Use profile_probability::split.
25010 (do_compare_rtx_and_jump): Fix adjustment of probabilities
25011 when splitting a single conditional jump into 2.
25013 2018-01-22 David Malcolm <dmalcolm@redhat.com>
25015 PR tree-optimization/69452
25016 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
25019 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
25021 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
25022 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
25023 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
25025 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
25027 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
25029 * config/rl78/rl78.md (movdi): New define_expand.
25030 * config/rl78/rl78.c (rl78_split_movdi): New function.
25032 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
25035 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
25037 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
25038 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
25039 128-bit to produce an UNSPEC move to get the double word with the
25040 signbit and then a shift directly to do signbit.
25041 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
25042 implementation with a new version that just does either a direct
25043 move or a regular move. Move memory interface to separate insns.
25044 Move insns so they are next to the expander.
25045 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
25046 with signbit move. Split big and little endian case.
25047 (signbit<mode>2_dm_mem_le): Likewise.
25048 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
25049 (signbit<mode>2_dm2): Likewise.
25051 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
25053 * config/rl78/rl78.md (anddi3): New define_expand.
25055 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
25057 * config/rl78/rl78.md (umindi3): New define_expand.
25059 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
25061 * config/rl78/rl78.md (smindi3): New define_expand.
25063 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
25065 * config/rl78/rl78.md (smaxdi3): New define_expand.
25067 2018-01-22 Carl Love <cel@us.ibm.com>
25069 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
25070 LVX_V1TI): Add macro expansion.
25071 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
25072 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
25073 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
25074 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
25075 Change check to determine if the instruction is a byte reversing
25076 entry. Fix typo in comment.
25077 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
25078 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
25079 Add def_builtin calls for new builtins.
25080 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
25081 Add define_insn expansion.
25083 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
25085 * config/rl78/rl78.md (umaxdi3): New define_expand.
25087 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
25089 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
25090 for non-QImode registers.
25092 2018-01-22 Richard Biener <rguenther@suse.de>
25094 PR tree-optimization/83963
25095 * graphite-scop-detection.c (scop_detection::get_sese): Delay
25096 including the loop exit block.
25097 (scop_detection::merge_sese): Likewise.
25098 (scop_detection::add_scop): Do it here instead.
25100 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25102 * doc/sourcebuild.texi (arm_softfloat): Document.
25104 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
25107 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
25108 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
25109 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
25111 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
25112 David Edelsohn <dje.gcc@gmail.com>
25115 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
25116 Change "crset eq" to "crset 2".
25117 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
25118 (*call_indirect_aix<mode>_nospec): Likewise.
25119 (*call_value_indirect_aix<mode>_nospec): Likewise.
25120 (*call_indirect_elfv2<mode>_nospec): Likewise.
25121 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
25122 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
25123 change assembly output from . to $.
25124 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
25125 (indirect_jump<mode>_nospec): Change assembly output from . to $.
25126 (*tablejump<mode>_internal1_nospec): Likewise.
25128 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
25131 * config/sh/sh_optimize_sett_clrt.cc:
25132 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
25134 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
25136 PR tree-optimization/83940
25137 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
25138 offset_dt to vect_constant_def rather than vect_unknown_def_type.
25139 (vect_check_load_store_mask): Add a mask_dt_out parameter and
25140 use it to pass back the definition type.
25141 (vect_check_store_rhs): Likewise rhs_dt_out.
25142 (vect_build_gather_load_calls): Add a mask_dt argument and use
25143 it instead of a call to vect_is_simple_use.
25144 (vectorizable_store): Update calls to vect_check_load_store_mask
25145 and vect_check_store_rhs. Use the dt returned by the latter instead
25146 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
25147 instead of calls to vect_is_simple_use. Pass the scalar rather
25148 than the vector operand to vect_is_simple_use when handling
25149 second and subsequent copies of an rhs value.
25150 (vectorizable_load): Update calls to vect_check_load_store_mask
25151 and vect_build_gather_load_calls. Use the cached mask_dt and
25152 gs_info.offset_dt instead of calls to vect_is_simple_use.
25154 2018-01-20 Jakub Jelinek <jakub@redhat.com>
25156 PR middle-end/83945
25157 * tree-emutls.c: Include gimplify.h.
25158 (lower_emutls_2): New function.
25159 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
25160 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
25161 it before further processing.
25164 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
25165 UINTVAL (trueop1) instead of INTVAL (op1).
25167 2018-01-19 Jakub Jelinek <jakub@redhat.com>
25171 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
25172 INCOMING_FRAME_SP_OFFSET if not defined.
25173 (scan_trace): Add ENTRY argument. If true and
25174 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
25175 emit a note to adjust the CFA offset.
25176 (create_cfi_notes): Adjust scan_trace callers.
25177 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
25178 INCOMING_FRAME_SP_OFFSET in the CIE.
25179 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
25180 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
25182 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
25183 * doc/tm.texi: Regenerated.
25185 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
25187 PR rtl-optimization/83147
25188 * lra-constraints.c (remove_inheritance_pseudos): Use
25189 lra_substitute_pseudo_within_insn.
25191 2018-01-19 Tom de Vries <tom@codesourcery.com>
25192 Cesar Philippidis <cesar@codesourcery.com>
25195 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
25197 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
25200 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
25201 spaces for function labels.
25203 2018-01-19 Martin Liska <mliska@suse.cz>
25205 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
25206 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
25207 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
25208 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
25209 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
25210 (PRED_CONST_RETURN): Change from 69 to 65.
25211 (PRED_NULL_RETURN): Change from 91 to 71.
25212 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
25213 (PRED_LOOP_GUARD): Change from 66 to 73.
25215 2018-01-19 Martin Liska <mliska@suse.cz>
25217 * predict.c (predict_insn_def): Add new assert.
25218 (struct branch_predictor): Change type to signed integer.
25219 (test_prediction_value_range): Amend test to cover
25220 PROB_UNINITIALIZED.
25221 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
25222 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
25223 (PRED_LOOP_ITERATIONS_MAX): Likewise.
25224 (PRED_LOOP_IV_COMPARE): Likewise.
25225 * predict.h (PROB_UNINITIALIZED): Define new constant.
25227 2018-01-19 Martin Liska <mliska@suse.cz>
25229 * predict.c (dump_prediction): Add new format for
25230 analyze_brprob.py script which is enabled with -details
25232 * profile-count.h (precise_p): New function.
25234 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
25236 PR tree-optimization/83922
25237 * tree-vect-loop.c (vect_verify_full_masking): Return false if
25238 there are no statements that need masking.
25239 (vect_active_double_reduction_p): New function.
25240 (vect_analyze_loop_operations): Use it when handling phis that
25241 are not in the loop header.
25243 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
25245 PR tree-optimization/83914
25246 * tree-vect-loop.c (vectorizable_induction): Don't convert
25247 init_expr or apply the peeling adjustment for inductions
25248 that are nested within the vectorized loop.
25250 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25252 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
25255 2018-01-18 Jakub Jelinek <jakub@redhat.com>
25259 * function.h (gimplify_parameters): Add gimple_seq * argument.
25260 * function.c: Include gimple.h and options.h.
25261 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
25262 for the added local temporaries if needed.
25263 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
25264 if there are any parameter cleanups, wrap whole body into a
25265 try/finally with the cleanups.
25267 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
25270 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
25271 Use GET_MODE_CLASS for scalar floating point.
25273 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
25277 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
25278 Fix call of call_cgraph_insertion_hooks.
25280 2018-01-18 Martin Sebor <msebor@redhat.com>
25282 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
25284 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
25287 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
25290 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
25293 * common.opt: (-ffile-prefix-map): New option.
25294 * opts.c (common_handle_option): Defer it.
25295 * opts-global.c (handle_common_deferred_options): Handle it.
25296 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
25297 * file-prefix-map.h: New file.
25298 (remap_debug_filename, add_debug_prefix_map): ...here.
25299 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
25300 * final.c (debug_prefix_map, add_debug_prefix_map
25301 remap_debug_filename): Move to...
25302 * file-prefix-map.c: New file.
25303 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
25304 generalize, get rid of alloca(), use strrchr() instead of strchr().
25305 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
25306 Implement in terms of add_prefix_map().
25307 (remap_macro_filename, remap_debug_filename): Implement in term of
25309 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
25310 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
25311 * dbxout.c: Include file-prefix-map.h.
25312 * varasm.c: Likewise.
25313 * vmsdbgout.c: Likewise.
25314 * xcoffout.c: Likewise.
25315 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
25316 * doc/cppopts.texi (-fmacro-prefix-map): Document.
25317 * doc/invoke.texi (-ffile-prefix-map): Document.
25318 (-fdebug-prefix-map): Update description.
25320 2018-01-18 Martin Liska <mliska@suse.cz>
25322 * config/i386/i386.c (indirect_thunk_name): Document that also
25324 (output_indirect_thunk): Document why both instructions
25325 (pause and lfence) are generated.
25327 2018-01-18 Richard Biener <rguenther@suse.de>
25329 PR tree-optimization/83887
25330 * graphite-scop-detection.c
25331 (scop_detection::get_nearest_dom_with_single_entry): Remove.
25332 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
25333 (scop_detection::merge_sese): Re-implement with a flood-fill
25334 algorithm that properly finds a SESE region if it exists.
25336 2018-01-18 Jakub Jelinek <jakub@redhat.com>
25339 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
25340 pointer_diff optimizations use view_convert instead of convert.
25342 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
25344 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
25345 Generate different code for -mno-speculate-indirect-jumps.
25346 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
25347 (*call_indirect_aix<mode>): Disable for
25348 -mno-speculate-indirect-jumps.
25349 (*call_indirect_aix<mode>_nospec): New define_insn.
25350 (*call_value_indirect_aix<mode>): Disable for
25351 -mno-speculate-indirect-jumps.
25352 (*call_value_indirect_aix<mode>_nospec): New define_insn.
25353 (*sibcall_nonlocal_sysv<mode>): Generate different code for
25354 -mno-speculate-indirect-jumps.
25355 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
25357 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
25359 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
25360 long double type, set the flags for noting the default long double
25361 type, even if we don't pass or return a long double type.
25363 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
25366 * ipa-inline.c (flatten_function): Do not overwrite final inlining
25369 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
25371 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
25372 support for merge[hl].
25373 (fold_mergehl_helper): New helper function.
25374 (tree-vector-builder.h): New #include for tree_vector_builder usage.
25375 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
25376 (altivec_vmrglw_direct): Add xxmrglw insn.
25378 2018-01-17 Andrew Waterman <andrew@sifive.com>
25380 * config/riscv/riscv.c (riscv_conditional_register_usage): If
25381 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
25383 2018-01-17 David Malcolm <dmalcolm@redhat.com>
25386 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
25387 call the lto_location_cache before reading the
25388 DECL_SOURCE_LOCATION of the types.
25390 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
25391 Richard Sandiford <richard.sandiford@linaro.org>
25393 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
25394 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
25395 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
25396 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
25397 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
25399 * config/aarch64/constraints.md (aarch64_movti_operand):
25401 * config/aarch64/predicates.md (Uti): Add new constraint.
25403 2018-01-17 Carl Love <cel@us.ibm.com>
25405 * config/rs6000/vsx.md (define_expand xl_len_r,
25406 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
25407 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
25409 (define_expand, define_insn): Move the shift left from the
25410 define_insn to the define_expand for lxvl and stxvl instructions.
25411 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
25412 and XL_LEN_R definitions to PURE.
25414 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
25416 * config/i386/i386.c (indirect_thunk_name): Declare regno
25417 as unsigned int. Compare regno with INVALID_REGNUM.
25418 (output_indirect_thunk): Ditto.
25419 (output_indirect_thunk_function): Ditto.
25420 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
25421 in the call to output_indirect_thunk_function.
25423 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
25425 PR middle-end/83884
25426 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
25427 rather than the size of inner_type to determine the stack slot size
25428 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
25430 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
25433 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
25436 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
25438 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
25439 endian Linux systems to optionally enable multilibs for selecting
25440 the long double type if the user configured an explicit type.
25441 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
25442 have no long double multilibs if not defined.
25443 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
25444 warn if the user used -mabi={ieee,ibm}longdouble and we built
25445 multilibs for long double.
25446 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
25447 appropriate multilib option.
25448 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
25450 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
25451 for building long double multilibs.
25452 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
25454 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
25456 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
25459 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
25461 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
25464 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
25467 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
25470 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
25472 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
25473 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
25475 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
25477 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
25478 different rtl trees depending on TARGET_64BIT.
25479 (rs6000_gen_lvx): Likewise.
25481 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
25483 * config/visium/visium.md (nop): Tweak comment.
25484 (hazard_nop): Likewise.
25486 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
25488 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
25489 -mspeculate-indirect-jumps.
25490 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
25491 for -mno-speculate-indirect-jumps.
25492 (*call_indirect_elfv2<mode>_nospec): New define_insn.
25493 (*call_value_indirect_elfv2<mode>): Disable for
25494 -mno-speculate-indirect-jumps.
25495 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
25496 (indirect_jump): Emit different RTL for
25497 -mno-speculate-indirect-jumps.
25498 (*indirect_jump<mode>): Disable for
25499 -mno-speculate-indirect-jumps.
25500 (*indirect_jump<mode>_nospec): New define_insn.
25501 (tablejump): Emit different RTL for
25502 -mno-speculate-indirect-jumps.
25503 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
25504 (tablejumpsi_nospec): New define_expand.
25505 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
25506 (tablejumpdi_nospec): New define_expand.
25507 (*tablejump<mode>_internal1): Disable for
25508 -mno-speculate-indirect-jumps.
25509 (*tablejump<mode>_internal1_nospec): New define_insn.
25510 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
25513 2018-01-16 Artyom Skrobov tyomitch@gmail.com
25515 * caller-save.c (insert_save): Drop unnecessary parameter. All
25518 2018-01-16 Jakub Jelinek <jakub@redhat.com>
25519 Richard Biener <rguenth@suse.de>
25522 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
25523 return early, inline manually is_gimple_sizepos. Make sure if we
25524 call gimplify_expr we don't end up with a gimple constant.
25525 * tree.c (variably_modified_type_p): Don't return true for
25526 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
25527 * gimplify.h (is_gimple_sizepos): Remove.
25529 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
25531 PR tree-optimization/83857
25532 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
25533 vectorizable_live_operation for pure SLP statements.
25534 (vectorizable_live_operation): Handle PHIs.
25536 2018-01-16 Richard Biener <rguenther@suse.de>
25538 PR tree-optimization/83867
25539 * tree-vect-stmts.c (vect_transform_stmt): Precompute
25540 nested_in_vect_loop_p since the scalar stmt may get invalidated.
25542 2018-01-16 Jakub Jelinek <jakub@redhat.com>
25545 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
25546 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
25547 If off is not INTEGER_CST, issue a may not be aligned warning
25548 rather than isn't aligned. Use isn%'t rather than isn't.
25549 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
25551 <case MULT_EXPR>: Improve the case when bottom and one of the
25552 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
25553 operand, in that case check if the other operand is multiple of
25554 bottom divided by the INTEGER_CST operand.
25556 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
25559 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
25560 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
25561 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
25562 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
25563 * config/pa/pa.c (pa_function_arg_advance): Likewise.
25564 (pa_function_arg, pa_arg_partial_bytes): Likewise.
25565 (pa_function_arg_size): New function.
25567 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
25569 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
25570 in a separate statement.
25572 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
25574 PR tree-optimization/83847
25575 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
25576 group gathers and scatters.
25578 2018-01-16 Jakub Jelinek <jakub@redhat.com>
25580 PR rtl-optimization/86620
25581 * params.def (max-sched-ready-insns): Bump minimum value to 1.
25583 PR rtl-optimization/83213
25584 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
25585 to last if both are JUMP_INSNs.
25587 PR tree-optimization/83843
25588 * gimple-ssa-store-merging.c
25589 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
25590 store_immediate_info for bswap/nop orig_stores.
25592 2018-01-15 Andrew Waterman <andrew@sifive.com>
25594 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
25596 <UDIV>: Increase cost if !TARGET_DIV.
25598 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
25600 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
25601 (define_attr "cr_logical_3op"): New.
25602 (cceq_ior_compare): Adjust.
25603 (cceq_ior_compare_complement): Adjust.
25604 (*cceq_rev_compare): Adjust.
25605 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
25606 (is_cracked_insn): Adjust.
25607 (insn_must_be_first_in_group): Adjust.
25608 * config/rs6000/40x.md: Adjust.
25609 * config/rs6000/440.md: Adjust.
25610 * config/rs6000/476.md: Adjust.
25611 * config/rs6000/601.md: Adjust.
25612 * config/rs6000/603.md: Adjust.
25613 * config/rs6000/6xx.md: Adjust.
25614 * config/rs6000/7450.md: Adjust.
25615 * config/rs6000/7xx.md: Adjust.
25616 * config/rs6000/8540.md: Adjust.
25617 * config/rs6000/cell.md: Adjust.
25618 * config/rs6000/e300c2c3.md: Adjust.
25619 * config/rs6000/e500mc.md: Adjust.
25620 * config/rs6000/e500mc64.md: Adjust.
25621 * config/rs6000/e5500.md: Adjust.
25622 * config/rs6000/e6500.md: Adjust.
25623 * config/rs6000/mpc.md: Adjust.
25624 * config/rs6000/power4.md: Adjust.
25625 * config/rs6000/power5.md: Adjust.
25626 * config/rs6000/power6.md: Adjust.
25627 * config/rs6000/power7.md: Adjust.
25628 * config/rs6000/power8.md: Adjust.
25629 * config/rs6000/power9.md: Adjust.
25630 * config/rs6000/rs64.md: Adjust.
25631 * config/rs6000/titan.md: Adjust.
25633 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
25635 * config/i386/predicates.md (indirect_branch_operand): Rewrite
25636 ix86_indirect_branch_register logic.
25638 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
25640 * config/i386/constraints.md (Bs): Update
25641 ix86_indirect_branch_register check. Don't check
25642 ix86_indirect_branch_register with GOT_memory_operand.
25644 * config/i386/predicates.md (GOT_memory_operand): Don't check
25645 ix86_indirect_branch_register here.
25646 (GOT32_symbol_operand): Likewise.
25648 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
25650 * config/i386/predicates.md (constant_call_address_operand):
25651 Rewrite ix86_indirect_branch_register logic.
25652 (sibcall_insn_operand): Likewise.
25654 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
25656 * config/i386/constraints.md (Bs): Replace
25657 ix86_indirect_branch_thunk_register with
25658 ix86_indirect_branch_register.
25660 * config/i386/i386.md (indirect_jump): Likewise.
25661 (tablejump): Likewise.
25662 (*sibcall_memory): Likewise.
25663 (*sibcall_value_memory): Likewise.
25664 Peepholes of indirect call and jump via memory: Likewise.
25665 * config/i386/i386.opt: Likewise.
25666 * config/i386/predicates.md (indirect_branch_operand): Likewise.
25667 (GOT_memory_operand): Likewise.
25668 (call_insn_operand): Likewise.
25669 (sibcall_insn_operand): Likewise.
25670 (GOT32_symbol_operand): Likewise.
25672 2018-01-15 Jakub Jelinek <jakub@redhat.com>
25674 PR middle-end/83837
25675 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
25676 type rather than type addr's type points to.
25677 (expand_omp_atomic_mutex): Likewise.
25678 (expand_omp_atomic): Likewise.
25680 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
25683 * config/i386/i386.c (output_indirect_thunk_function): Use
25684 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
25685 for __x86_return_thunk.
25687 2018-01-15 Richard Biener <rguenther@suse.de>
25689 PR middle-end/83850
25690 * expmed.c (extract_bit_field_1): Fix typo.
25692 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25695 * config/arm/iterators.md (VF): New mode iterator.
25696 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
25697 Remove integer-related logic from pattern.
25698 (neon_vabd<mode>_3): Likewise.
25700 2018-01-15 Jakub Jelinek <jakub@redhat.com>
25702 PR middle-end/82694
25703 * common.opt (fstrict-overflow): No longer an alias.
25704 (fwrapv-pointer): New option.
25705 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
25706 also for pointer types based on flag_wrapv_pointer.
25707 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
25708 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
25709 opts->x_flag_wrapv got set.
25710 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
25711 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
25712 POINTER_TYPE_OVERFLOW_UNDEFINED.
25713 * match.pd: Likewise in address comparison pattern.
25714 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
25716 2018-01-15 Richard Biener <rguenther@suse.de>
25719 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
25720 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
25721 Reset type names to their identifier if their TYPE_DECL doesn't
25722 have linkage (and thus is used for ODR and devirt).
25723 (save_debug_info_for_decl): Remove.
25724 (save_debug_info_for_type): Likewise.
25725 (add_tree_to_fld_list): Adjust.
25726 * tree-pretty-print.c (dump_generic_node): Make dumping of
25727 type names more robust.
25729 2018-01-15 Richard Biener <rguenther@suse.de>
25731 * BASE-VER: Bump to 8.0.1.
25733 2018-01-14 Martin Sebor <msebor@redhat.com>
25736 * builtins.c (check_access): Avoid warning when the no-warning bit
25739 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
25741 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
25742 * ira-color (allocno_hard_regs_compare): Likewise.
25744 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
25747 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
25748 Use .pushsection/.popsection.
25750 2018-01-14 Martin Sebor <msebor@redhat.com>
25753 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
25755 2018-01-14 Jakub Jelinek <jakub@redhat.com>
25757 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
25758 entry from extra_headers.
25759 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
25760 extra_headers, make the list bitwise identical to the i?86-*-* one.
25762 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
25764 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
25765 -mcmodel=large with -mindirect-branch=thunk,
25766 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
25767 -mfunction-return=thunk-extern.
25768 * doc/invoke.texi: Document -mcmodel=large is incompatible with
25769 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
25770 -mfunction-return=thunk and -mfunction-return=thunk-extern.
25772 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
25774 * config/i386/i386.c (print_reg): Print the name of the full
25775 integer register without '%'.
25776 (ix86_print_operand): Handle 'V'.
25777 * doc/extend.texi: Document 'V' modifier.
25779 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
25781 * config/i386/constraints.md (Bs): Disallow memory operand for
25782 -mindirect-branch-register.
25784 * config/i386/predicates.md (indirect_branch_operand): Likewise.
25785 (GOT_memory_operand): Likewise.
25786 (call_insn_operand): Likewise.
25787 (sibcall_insn_operand): Likewise.
25788 (GOT32_symbol_operand): Likewise.
25789 * config/i386/i386.md (indirect_jump): Call convert_memory_address
25790 for -mindirect-branch-register.
25791 (tablejump): Likewise.
25792 (*sibcall_memory): Likewise.
25793 (*sibcall_value_memory): Likewise.
25794 Disallow peepholes of indirect call and jump via memory for
25795 -mindirect-branch-register.
25796 (*call_pop): Replace m with Bw.
25797 (*call_value_pop): Likewise.
25798 (*sibcall_pop_memory): Replace m with Bs.
25799 * config/i386/i386.opt (mindirect-branch-register): New option.
25800 * doc/invoke.texi: Document -mindirect-branch-register option.
25802 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
25804 * config/i386/i386-protos.h (ix86_output_function_return): New.
25805 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
25806 set function_return_type.
25807 (indirect_thunk_name): Add ret_p to indicate thunk for function
25809 (output_indirect_thunk_function): Pass false to
25810 indirect_thunk_name.
25811 (ix86_output_indirect_branch_via_reg): Likewise.
25812 (ix86_output_indirect_branch_via_push): Likewise.
25813 (output_indirect_thunk_function): Create alias for function
25814 return thunk if regno < 0.
25815 (ix86_output_function_return): New function.
25816 (ix86_handle_fndecl_attribute): Handle function_return.
25817 (ix86_attribute_table): Add function_return.
25818 * config/i386/i386.h (machine_function): Add
25819 function_return_type.
25820 * config/i386/i386.md (simple_return_internal): Use
25821 ix86_output_function_return.
25822 (simple_return_internal_long): Likewise.
25823 * config/i386/i386.opt (mfunction-return=): New option.
25824 (indirect_branch): Mention -mfunction-return=.
25825 * doc/extend.texi: Document function_return function attribute.
25826 * doc/invoke.texi: Document -mfunction-return= option.
25828 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
25830 * config/i386/i386-opts.h (indirect_branch): New.
25831 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
25832 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
25833 with local indirect jump when converting indirect call and jump.
25834 (ix86_set_indirect_branch_type): New.
25835 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
25836 (indirectlabelno): New.
25837 (indirect_thunk_needed): Likewise.
25838 (indirect_thunk_bnd_needed): Likewise.
25839 (indirect_thunks_used): Likewise.
25840 (indirect_thunks_bnd_used): Likewise.
25841 (INDIRECT_LABEL): Likewise.
25842 (indirect_thunk_name): Likewise.
25843 (output_indirect_thunk): Likewise.
25844 (output_indirect_thunk_function): Likewise.
25845 (ix86_output_indirect_branch_via_reg): Likewise.
25846 (ix86_output_indirect_branch_via_push): Likewise.
25847 (ix86_output_indirect_branch): Likewise.
25848 (ix86_output_indirect_jmp): Likewise.
25849 (ix86_code_end): Call output_indirect_thunk_function if needed.
25850 (ix86_output_call_insn): Call ix86_output_indirect_branch if
25852 (ix86_handle_fndecl_attribute): Handle indirect_branch.
25853 (ix86_attribute_table): Add indirect_branch.
25854 * config/i386/i386.h (machine_function): Add indirect_branch_type
25855 and has_local_indirect_jump.
25856 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
25858 (tablejump): Likewise.
25859 (*indirect_jump): Use ix86_output_indirect_jmp.
25860 (*tablejump_1): Likewise.
25861 (simple_return_indirect_internal): Likewise.
25862 * config/i386/i386.opt (mindirect-branch=): New option.
25863 (indirect_branch): New.
25866 (thunk-inline): Likewise.
25867 (thunk-extern): Likewise.
25868 * doc/extend.texi: Document indirect_branch function attribute.
25869 * doc/invoke.texi: Document -mindirect-branch= option.
25871 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
25874 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
25876 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
25878 * ipa-inline.c (want_inline_small_function_p): Return false if
25879 inlining has already failed with CIF_FINAL_ERROR.
25880 (update_caller_keys): Call want_inline_small_function_p before
25882 (update_callee_keys): Likewise.
25884 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
25886 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
25888 (rs6000_quadword_masked_address_p): Likewise.
25889 (quad_aligned_load_p): Likewise.
25890 (quad_aligned_store_p): Likewise.
25891 (const_load_sequence_p): Add comment to describe the outer-most loop.
25892 (mimic_memory_attributes_and_flags): New function.
25893 (rs6000_gen_stvx): Likewise.
25894 (replace_swapped_aligned_store): Likewise.
25895 (rs6000_gen_lvx): Likewise.
25896 (replace_swapped_aligned_load): Likewise.
25897 (replace_swapped_load_constant): Capitalize argument name in
25898 comment describing this function.
25899 (rs6000_analyze_swaps): Add a third pass to search for vector loads
25900 and stores that access quad-word aligned addresses and replace
25901 with stvx or lvx instructions when appropriate.
25902 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
25903 New function prototype.
25904 (rs6000_quadword_masked_address_p): Likewise.
25905 (rs6000_gen_lvx): Likewise.
25906 (rs6000_gen_stvx): Likewise.
25907 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
25908 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
25909 when memory address is aligned.
25910 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
25911 this split to select lvx instruction when memory address is aligned.
25912 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
25913 instruction when memory address is aligned.
25914 (*vsx_le_perm_load_v16qi): Likewise.
25915 (four unnamed splitters): Modify to select the stvx instruction
25916 when memory is aligned.
25918 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
25920 * predict.c (determine_unlikely_bbs): Handle correctly BBs
25921 which appears in the queue multiple times.
25923 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25924 Alan Hayward <alan.hayward@arm.com>
25925 David Sherwood <david.sherwood@arm.com>
25927 * tree-vectorizer.h (vec_lower_bound): New structure.
25928 (_loop_vec_info): Add check_nonzero and lower_bounds.
25929 (LOOP_VINFO_CHECK_NONZERO): New macro.
25930 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
25931 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
25932 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
25933 fields. Make seg_len the distance travelled, not including the
25935 (dr_direction_indicator): Declare.
25936 (dr_zero_step_indicator): Likewise.
25937 (dr_known_forward_stride_p): Likewise.
25938 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
25940 (runtime_alias_check_p): Allow runtime alias checks with
25942 (operator ==): Compare access_size and align.
25943 (prune_runtime_alias_test_list): Rework for new distinction between
25944 the access_size and seg_len.
25945 (create_intersect_range_checks_index): Likewise. Cope with polynomial
25947 (get_segment_min_max): New function.
25948 (create_intersect_range_checks): Use it.
25949 (dr_step_indicator): New function.
25950 (dr_direction_indicator): Likewise.
25951 (dr_zero_step_indicator): Likewise.
25952 (dr_known_forward_stride_p): Likewise.
25953 * tree-loop-distribution.c (data_ref_segment_size): Return
25954 DR_STEP * (niters - 1).
25955 (compute_alias_check_pairs): Update call to the dr_with_seg_len
25957 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
25958 (vect_preserves_scalar_order_p): New function, split out from...
25959 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
25960 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
25961 (vect_vfa_access_size): New function.
25962 (vect_vfa_align): Likewise.
25963 (vect_compile_time_alias): Take access_size_a and access_b arguments.
25964 (dump_lower_bound): New function.
25965 (vect_check_lower_bound): Likewise.
25966 (vect_small_gap_p): Likewise.
25967 (vectorizable_with_step_bound_p): Likewise.
25968 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
25969 depencies if the vectorization factor is 1. Convert the checks
25970 for nonzero steps into checks on the bounds of DR_STEP. Try using
25971 a bunds check for variable steps if the minimum required step is
25972 relatively small. Update calls to the dr_with_seg_len
25973 constructor and to vect_compile_time_alias.
25974 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
25976 (vect_loop_versioning): Call it.
25977 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
25979 (vect_estimate_min_profitable_iters): Account for any bounds checks.
25981 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25982 Alan Hayward <alan.hayward@arm.com>
25983 David Sherwood <david.sherwood@arm.com>
25985 * doc/sourcebuild.texi (vect_scatter_store): Document.
25986 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
25988 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
25990 * genopinit.c (main): Add supports_vec_scatter_store and
25991 supports_vec_scatter_store_cached to target_optabs.
25992 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
25993 IFN_MASK_SCATTER_STORE.
25994 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
25996 * internal-fn.h (internal_store_fn_p): Declare.
25997 (internal_fn_stored_value_index): Likewise.
25998 * internal-fn.c (scatter_store_direct): New macro.
25999 (expand_scatter_store_optab_fn): New function.
26000 (direct_scatter_store_optab_supported_p): New macro.
26001 (internal_store_fn_p): New function.
26002 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
26003 IFN_MASK_SCATTER_STORE.
26004 (internal_fn_mask_index): Likewise.
26005 (internal_fn_stored_value_index): New function.
26006 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
26007 for scatter stores.
26008 * optabs-query.h (supports_vec_scatter_store_p): Declare.
26009 * optabs-query.c (supports_vec_scatter_store_p): New function.
26010 * tree-vectorizer.h (vect_get_store_rhs): Declare.
26011 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
26012 true for scatter stores.
26013 (vect_gather_scatter_fn_p): Handle scatter stores too.
26014 (vect_check_gather_scatter): Consider using scatter stores if
26015 supports_vec_scatter_store_p.
26016 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
26017 scatter stores too.
26018 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
26019 internal_fn_stored_value_index.
26020 (check_load_store_masking): Handle scatter stores too.
26021 (vect_get_store_rhs): Make public.
26022 (vectorizable_call): Use internal_store_fn_p.
26023 (vectorizable_store): Handle scatter store internal functions.
26024 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
26025 when deciding whether the end of the group has been reached.
26026 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
26027 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
26028 (mask_scatter_store<mode>): New insns.
26030 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26031 Alan Hayward <alan.hayward@arm.com>
26032 David Sherwood <david.sherwood@arm.com>
26034 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
26035 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
26036 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
26038 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
26039 Use vect_truncate_gather_scatter_offset if we can't treat the
26040 operation as a normal gather load or scatter store.
26041 (get_group_load_store_type): Take the gather_scatter_info
26042 as argument. Try using a gather load or scatter store for
26043 single-element groups.
26044 (get_load_store_type): Update calls to get_group_load_store_type
26045 and vect_use_strided_gather_scatters_p.
26047 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26048 Alan Hayward <alan.hayward@arm.com>
26049 David Sherwood <david.sherwood@arm.com>
26051 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
26052 optional tree argument.
26053 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
26055 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
26056 but continue to use the current value as a fallback.
26057 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
26058 to compare the updates.
26059 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
26060 (get_load_store_type): Use it when handling a strided access.
26061 (vect_get_strided_load_store_ops): New function.
26062 (vect_get_data_ptr_increment): Likewise.
26063 (vectorizable_load): Handle strided gather loads. Always pass
26064 a step to vect_create_data_ref_ptr and bump_vector_ptr.
26066 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26067 Alan Hayward <alan.hayward@arm.com>
26068 David Sherwood <david.sherwood@arm.com>
26070 * doc/md.texi (gather_load@var{m}): Document.
26071 (mask_gather_load@var{m}): Likewise.
26072 * genopinit.c (main): Add supports_vec_gather_load and
26073 supports_vec_gather_load_cached to target_optabs.
26074 * optabs-tree.c (init_tree_optimization_optabs): Use
26075 ggc_cleared_alloc to allocate target_optabs.
26076 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
26077 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
26079 * internal-fn.h (internal_load_fn_p): Declare.
26080 (internal_gather_scatter_fn_p): Likewise.
26081 (internal_fn_mask_index): Likewise.
26082 (internal_gather_scatter_fn_supported_p): Likewise.
26083 * internal-fn.c (gather_load_direct): New macro.
26084 (expand_gather_load_optab_fn): New function.
26085 (direct_gather_load_optab_supported_p): New macro.
26086 (direct_internal_fn_optab): New function.
26087 (internal_load_fn_p): Likewise.
26088 (internal_gather_scatter_fn_p): Likewise.
26089 (internal_fn_mask_index): Likewise.
26090 (internal_gather_scatter_fn_supported_p): Likewise.
26091 * optabs-query.c (supports_at_least_one_mode_p): New function.
26092 (supports_vec_gather_load_p): Likewise.
26093 * optabs-query.h (supports_vec_gather_load_p): Declare.
26094 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
26095 and memory_type field.
26096 (NUM_PATTERNS): Bump to 15.
26097 * tree-vect-data-refs.c: Include internal-fn.h.
26098 (vect_gather_scatter_fn_p): New function.
26099 (vect_describe_gather_scatter_call): Likewise.
26100 (vect_check_gather_scatter): Try using internal functions for
26101 gather loads. Recognize existing calls to a gather load function.
26102 (vect_analyze_data_refs): Consider using gather loads if
26103 supports_vec_gather_load_p.
26104 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
26105 (vect_get_gather_scatter_offset_type): Likewise.
26106 (vect_convert_mask_for_vectype): Likewise.
26107 (vect_add_conversion_to_patterm): Likewise.
26108 (vect_try_gather_scatter_pattern): Likewise.
26109 (vect_recog_gather_scatter_pattern): New pattern recognizer.
26110 (vect_vect_recog_func_ptrs): Add it.
26111 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
26112 internal_fn_mask_index and internal_gather_scatter_fn_p.
26113 (check_load_store_masking): Take the gather_scatter_info as an
26114 argument and handle gather loads.
26115 (vect_get_gather_scatter_ops): New function.
26116 (vectorizable_call): Check internal_load_fn_p.
26117 (vectorizable_load): Likewise. Handle gather load internal
26119 (vectorizable_store): Update call to check_load_store_masking.
26120 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
26121 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
26122 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
26123 (aarch64_gather_scale_operand_d): New predicates.
26124 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
26125 (mask_gather_load<mode>): New insns.
26127 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26128 Alan Hayward <alan.hayward@arm.com>
26129 David Sherwood <david.sherwood@arm.com>
26131 * optabs.def (fold_left_plus_optab): New optab.
26132 * doc/md.texi (fold_left_plus_@var{m}): Document.
26133 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
26134 * internal-fn.c (fold_left_direct): Define.
26135 (expand_fold_left_optab_fn): Likewise.
26136 (direct_fold_left_optab_supported_p): Likewise.
26137 * fold-const-call.c (fold_const_fold_left): New function.
26138 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
26139 * tree-parloops.c (valid_reduction_p): New function.
26140 (gather_scalar_reductions): Use it.
26141 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
26142 (vect_finish_replace_stmt): Declare.
26143 * tree-vect-loop.c (fold_left_reduction_fn): New function.
26144 (needs_fold_left_reduction_p): New function, split out from...
26145 (vect_is_simple_reduction): ...here. Accept reductions that
26146 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
26147 (vect_force_simple_reduction): Also store the reduction type in
26148 the assignment's STMT_VINFO_REDUC_TYPE.
26149 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
26150 (merge_with_identity): New function.
26151 (vect_expand_fold_left): Likewise.
26152 (vectorize_fold_left_reduction): Likewise.
26153 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
26154 scalar phi in place for it. Check for target support and reject
26155 cases that would reassociate the operation. Defer the transform
26156 phase to vectorize_fold_left_reduction.
26157 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
26158 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
26159 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
26161 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26163 * tree-if-conv.c (predicate_mem_writes): Remove redundant
26164 call to ifc_temp_var.
26166 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26167 Alan Hayward <alan.hayward@arm.com>
26168 David Sherwood <david.sherwood@arm.com>
26170 * target.def (legitimize_address_displacement): Take the original
26171 offset as a poly_int.
26172 * targhooks.h (default_legitimize_address_displacement): Update
26174 * targhooks.c (default_legitimize_address_displacement): Likewise.
26175 * doc/tm.texi: Regenerate.
26176 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
26177 as an argument, moving assert of ad->disp == ad->disp_term to...
26178 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
26179 Try calling targetm.legitimize_address_displacement before expanding
26180 the address rather than afterwards, and adjust for the new interface.
26181 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
26182 Match the new hook interface. Handle SVE addresses.
26183 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
26184 new hook interface.
26186 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26188 * Makefile.in (OBJS): Add early-remat.o.
26189 * target.def (select_early_remat_modes): New hook.
26190 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
26191 * doc/tm.texi: Regenerate.
26192 * targhooks.h (default_select_early_remat_modes): Declare.
26193 * targhooks.c (default_select_early_remat_modes): New function.
26194 * timevar.def (TV_EARLY_REMAT): New timevar.
26195 * passes.def (pass_early_remat): New pass.
26196 * tree-pass.h (make_pass_early_remat): Declare.
26197 * early-remat.c: New file.
26198 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
26200 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
26202 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26203 Alan Hayward <alan.hayward@arm.com>
26204 David Sherwood <david.sherwood@arm.com>
26206 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
26207 vfm1 with a bound_epilog parameter.
26208 (vect_do_peeling): Update calls accordingly, and move the prologue
26209 call earlier in the function. Treat the base bound_epilog as 0 for
26210 fully-masked loops and retain vf - 1 for other loops. Add 1 to
26211 this base when peeling for gaps.
26212 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
26213 with fully-masked loops.
26214 (vect_estimate_min_profitable_iters): Handle the single peeled
26215 iteration in that case.
26217 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26218 Alan Hayward <alan.hayward@arm.com>
26219 David Sherwood <david.sherwood@arm.com>
26221 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
26222 single-element interleaving even if the size is not a power of 2.
26223 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
26224 accesses for single-element interleaving if the group size is
26227 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26228 Alan Hayward <alan.hayward@arm.com>
26229 David Sherwood <david.sherwood@arm.com>
26231 * doc/md.texi (fold_extract_last_@var{m}): Document.
26232 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
26233 * optabs.def (fold_extract_last_optab): New optab.
26234 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
26235 * internal-fn.c (fold_extract_direct): New macro.
26236 (expand_fold_extract_optab_fn): Likewise.
26237 (direct_fold_extract_optab_supported_p): Likewise.
26238 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
26239 * tree-vect-loop.c (vect_model_reduction_cost): Handle
26240 EXTRACT_LAST_REDUCTION.
26241 (get_initial_def_for_reduction): Do not create an initial vector
26242 for EXTRACT_LAST_REDUCTION reductions.
26243 (vectorizable_reduction): Leave the scalar phi in place for
26244 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
26245 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
26246 epilogue code for EXTRACT_LAST_REDUCTION and defer the
26247 transform phase to vectorizable_condition.
26248 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
26250 (vect_finish_stmt_generation): ...here.
26251 (vect_finish_replace_stmt): New function.
26252 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
26253 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
26255 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
26257 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26258 Alan Hayward <alan.hayward@arm.com>
26259 David Sherwood <david.sherwood@arm.com>
26261 * doc/md.texi (extract_last_@var{m}): Document.
26262 * optabs.def (extract_last_optab): New optab.
26263 * internal-fn.def (EXTRACT_LAST): New internal function.
26264 * internal-fn.c (cond_unary_direct): New macro.
26265 (expand_cond_unary_optab_fn): Likewise.
26266 (direct_cond_unary_optab_supported_p): Likewise.
26267 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
26268 loops using EXTRACT_LAST.
26269 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
26270 (extract_last_<mode>): ...this optab.
26271 (vec_extract<mode><Vel>): Update accordingly.
26273 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26274 Alan Hayward <alan.hayward@arm.com>
26275 David Sherwood <david.sherwood@arm.com>
26277 * target.def (empty_mask_is_expensive): New hook.
26278 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
26279 * doc/tm.texi: Regenerate.
26280 * targhooks.h (default_empty_mask_is_expensive): Declare.
26281 * targhooks.c (default_empty_mask_is_expensive): New function.
26282 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
26283 if the target says that empty masks are expensive.
26284 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
26286 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
26288 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26289 Alan Hayward <alan.hayward@arm.com>
26290 David Sherwood <david.sherwood@arm.com>
26292 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
26293 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
26294 (vect_use_loop_mask_for_alignment_p): New function.
26295 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
26296 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
26297 niters_skip argument. Make sure that the first niters_skip elements
26298 of the first iteration are inactive.
26299 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
26300 Update call to vect_set_loop_masks_directly.
26301 (get_misalign_in_elems): New function, split out from...
26302 (vect_gen_prolog_loop_niters): ...here.
26303 (vect_update_init_of_dr): Take a code argument that specifies whether
26304 the adjustment should be added or subtracted.
26305 (vect_update_init_of_drs): Likewise.
26306 (vect_prepare_for_masked_peels): New function.
26307 (vect_do_peeling): Skip prologue peeling if we're using a mask
26308 instead. Update call to vect_update_inits_of_drs.
26309 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
26311 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
26312 alignment. Do not include the number of peeled iterations in
26313 the minimum threshold in that case.
26314 (vectorizable_induction): Adjust the start value down by
26315 LOOP_VINFO_MASK_SKIP_NITERS iterations.
26316 (vect_transform_loop): Call vect_prepare_for_masked_peels.
26317 Take the number of skipped iterations into account when calculating
26319 * tree-vect-stmts.c (vect_gen_while_not): New function.
26321 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26322 Alan Hayward <alan.hayward@arm.com>
26323 David Sherwood <david.sherwood@arm.com>
26325 * doc/sourcebuild.texi (vect_fully_masked): Document.
26326 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
26327 default value to 0.
26328 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
26330 (vect_analyze_loop_2): ...here. Don't check the vectorization
26331 factor against the number of loop iterations if the loop is
26334 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26335 Alan Hayward <alan.hayward@arm.com>
26336 David Sherwood <david.sherwood@arm.com>
26338 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
26339 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
26340 (dump_groups): Update accordingly.
26341 (iv_use::mem_type): New member variable.
26342 (address_p): New function.
26343 (record_use): Add a mem_type argument and initialize the new
26345 (record_group_use): Add a mem_type argument. Use address_p.
26346 Remove obsolete null checks of base_object. Update call to record_use.
26347 (find_interesting_uses_op): Update call to record_group_use.
26348 (find_interesting_uses_cond): Likewise.
26349 (find_interesting_uses_address): Likewise.
26350 (get_mem_type_for_internal_fn): New function.
26351 (find_address_like_use): Likewise.
26352 (find_interesting_uses_stmt): Try find_address_like_use before
26353 calling find_interesting_uses_op.
26354 (addr_offset_valid_p): Use the iv mem_type field as the type
26355 of the addressed memory.
26356 (add_autoinc_candidates): Likewise.
26357 (get_address_cost): Likewise.
26358 (split_small_address_groups_p): Use address_p.
26359 (split_address_groups): Likewise.
26360 (add_iv_candidate_for_use): Likewise.
26361 (autoinc_possible_for_pair): Likewise.
26362 (rewrite_groups): Likewise.
26363 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
26364 (determine_group_iv_cost): Update after split of USE_ADDRESS.
26365 (get_alias_ptr_type_for_ptr_address): New function.
26366 (rewrite_use_address): Rewrite address uses in calls that were
26367 identified by find_address_like_use.
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 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
26375 * gimple-expr.h (is_gimple_addressable: Likewise.
26376 * gimple-expr.c (is_gimple_address): Likewise.
26377 * internal-fn.c (expand_call_mem_ref): New function.
26378 (expand_mask_load_optab_fn): Use it.
26379 (expand_mask_store_optab_fn): Likewise.
26381 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26382 Alan Hayward <alan.hayward@arm.com>
26383 David Sherwood <david.sherwood@arm.com>
26385 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
26386 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
26387 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
26388 (cond_umax@var{mode}): Document.
26389 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
26390 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
26391 (cond_umin_optab, cond_umax_optab): New optabs.
26392 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
26393 (COND_IOR, COND_XOR): New internal functions.
26394 * internal-fn.h (get_conditional_internal_fn): Declare.
26395 * internal-fn.c (cond_binary_direct): New macro.
26396 (expand_cond_binary_optab_fn): Likewise.
26397 (direct_cond_binary_optab_supported_p): Likewise.
26398 (get_conditional_internal_fn): New function.
26399 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
26400 Cope with reduction statements that are vectorized as calls rather
26402 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
26403 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
26404 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
26405 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
26406 (UNSPEC_COND_EOR): New unspecs.
26407 (optab): Add mappings for them.
26408 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
26409 (sve_int_op, sve_fp_op): New int attributes.
26411 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26412 Alan Hayward <alan.hayward@arm.com>
26413 David Sherwood <david.sherwood@arm.com>
26415 * optabs.def (while_ult_optab): New optab.
26416 * doc/md.texi (while_ult@var{m}@var{n}): Document.
26417 * internal-fn.def (WHILE_ULT): New internal function.
26418 * internal-fn.h (direct_internal_fn_supported_p): New override
26419 that takes two types as argument.
26420 * internal-fn.c (while_direct): New macro.
26421 (expand_while_optab_fn): New function.
26422 (convert_optab_supported_p): Likewise.
26423 (direct_while_optab_supported_p): New macro.
26424 * wide-int.h (wi::udiv_ceil): New function.
26425 * tree-vectorizer.h (rgroup_masks): New structure.
26426 (vec_loop_masks): New typedef.
26427 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
26428 and fully_masked_p.
26429 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
26430 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
26431 (vect_max_vf): New function.
26432 (slpeel_make_loop_iterate_ntimes): Delete.
26433 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
26434 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
26435 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
26436 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
26437 internal-fn.h, stor-layout.h and optabs-query.h.
26438 (vect_set_loop_mask): New function.
26439 (add_preheader_seq): Likewise.
26440 (add_header_seq): Likewise.
26441 (interleave_supported_p): Likewise.
26442 (vect_maybe_permute_loop_masks): Likewise.
26443 (vect_set_loop_masks_directly): Likewise.
26444 (vect_set_loop_condition_masked): Likewise.
26445 (vect_set_loop_condition_unmasked): New function, split out from
26446 slpeel_make_loop_iterate_ntimes.
26447 (slpeel_make_loop_iterate_ntimes): Rename to..
26448 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
26449 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
26450 (vect_do_peeling): Update call accordingly.
26451 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
26453 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
26454 mask_compare_type, can_fully_mask_p and fully_masked_p.
26455 (release_vec_loop_masks): New function.
26456 (_loop_vec_info): Use it to free the loop masks.
26457 (can_produce_all_loop_masks_p): New function.
26458 (vect_get_max_nscalars_per_iter): Likewise.
26459 (vect_verify_full_masking): Likewise.
26460 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
26461 retries, and free the mask rgroups before retrying. Check loop-wide
26462 reasons for disallowing fully-masked loops. Make the final decision
26463 about whether use a fully-masked loop or not.
26464 (vect_estimate_min_profitable_iters): Do not assume that peeling
26465 for the number of iterations will be needed for fully-masked loops.
26466 (vectorizable_reduction): Disable fully-masked loops.
26467 (vectorizable_live_operation): Likewise.
26468 (vect_halve_mask_nunits): New function.
26469 (vect_double_mask_nunits): Likewise.
26470 (vect_record_loop_mask): Likewise.
26471 (vect_get_loop_mask): Likewise.
26472 (vect_transform_loop): Handle the case in which the final loop
26473 iteration might handle a partial vector. Call vect_set_loop_condition
26474 instead of slpeel_make_loop_iterate_ntimes.
26475 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
26476 (check_load_store_masking): New function.
26477 (prepare_load_store_mask): Likewise.
26478 (vectorizable_store): Handle fully-masked loops.
26479 (vectorizable_load): Likewise.
26480 (supportable_widening_operation): Use vect_halve_mask_nunits for
26482 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
26483 (vect_gen_while): New function.
26484 * config/aarch64/aarch64.md (umax<mode>3): New expander.
26485 (aarch64_uqdec<mode>): New insn.
26487 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26488 Alan Hayward <alan.hayward@arm.com>
26489 David Sherwood <david.sherwood@arm.com>
26491 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
26492 (reduc_xor_scal_optab): New optabs.
26493 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
26494 (reduc_xor_scal_@var{m}): Document.
26495 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
26496 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
26497 internal functions.
26498 * fold-const-call.c (fold_const_call): Handle them.
26499 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
26500 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
26501 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
26502 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
26503 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
26504 (UNSPEC_XORV): New unspecs.
26505 (optab): Add entries for them.
26506 (BITWISEV): New int iterator.
26507 (bit_reduc_op): New int attributes.
26509 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26510 Alan Hayward <alan.hayward@arm.com>
26511 David Sherwood <david.sherwood@arm.com>
26513 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
26514 * internal-fn.def (VEC_SHL_INSERT): New internal function.
26515 * optabs.def (vec_shl_insert_optab): New optab.
26516 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
26517 (duplicate_and_interleave): Likewise.
26518 * tree-vect-loop.c: Include internal-fn.h.
26519 (neutral_op_for_slp_reduction): New function, split out from
26520 get_initial_defs_for_reduction.
26521 (get_initial_def_for_reduction): Handle option 2 for variable-length
26522 vectors by loading the neutral value into a vector and then shifting
26523 the initial value into element 0.
26524 (get_initial_defs_for_reduction): Replace the code argument with
26525 the neutral value calculated by neutral_op_for_slp_reduction.
26526 Use gimple_build_vector for constant-length vectors.
26527 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
26528 but the first group_size elements have a neutral value.
26529 Use duplicate_and_interleave otherwise.
26530 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
26531 Update call to get_initial_defs_for_reduction. Handle SLP
26532 reductions for variable-length vectors by creating one vector
26533 result for each scalar result, with the elements associated
26534 with other scalar results stubbed out with the neutral value.
26535 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
26536 Require IFN_VEC_SHL_INSERT for double reductions on
26537 variable-length vectors, or SLP reductions that have
26538 a neutral value. Require can_duplicate_and_interleave_p
26539 support for variable-length unchained SLP reductions if there
26540 is no neutral value, such as for MIN/MAX reductions. Also require
26541 the number of vector elements to be a multiple of the number of
26542 SLP statements when doing variable-length unchained SLP reductions.
26543 Update call to vect_create_epilog_for_reduction.
26544 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
26545 and remove initial values.
26546 (duplicate_and_interleave): Make public.
26547 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
26548 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
26550 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26551 Alan Hayward <alan.hayward@arm.com>
26552 David Sherwood <david.sherwood@arm.com>
26554 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
26555 (can_duplicate_and_interleave_p): New function.
26556 (vect_get_and_check_slp_defs): Take the vector of statements
26557 rather than just the current one. Remove excess parentheses.
26558 Restriction rejectinon of vect_constant_def and vect_external_def
26559 for variable-length vectors to boolean types, or types for which
26560 can_duplicate_and_interleave_p is false.
26561 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
26562 (duplicate_and_interleave): New function.
26563 (vect_get_constant_vectors): Use gimple_build_vector for
26564 constant-length vectors and suitable variable-length constant
26565 vectors. Use duplicate_and_interleave for other variable-length
26566 vectors. Don't defer the update when inserting new statements.
26568 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26569 Alan Hayward <alan.hayward@arm.com>
26570 David Sherwood <david.sherwood@arm.com>
26572 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
26573 min_profitable_iters doesn't go negative.
26575 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26576 Alan Hayward <alan.hayward@arm.com>
26577 David Sherwood <david.sherwood@arm.com>
26579 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
26580 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
26581 * optabs.def (vec_mask_load_lanes_optab): New optab.
26582 (vec_mask_store_lanes_optab): Likewise.
26583 * internal-fn.def (MASK_LOAD_LANES): New internal function.
26584 (MASK_STORE_LANES): Likewise.
26585 * internal-fn.c (mask_load_lanes_direct): New macro.
26586 (mask_store_lanes_direct): Likewise.
26587 (expand_mask_load_optab_fn): Handle masked operations.
26588 (expand_mask_load_lanes_optab_fn): New macro.
26589 (expand_mask_store_optab_fn): Handle masked operations.
26590 (expand_mask_store_lanes_optab_fn): New macro.
26591 (direct_mask_load_lanes_optab_supported_p): Likewise.
26592 (direct_mask_store_lanes_optab_supported_p): Likewise.
26593 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
26595 (vect_load_lanes_supported): Likewise.
26596 * tree-vect-data-refs.c (strip_conversion): New function.
26597 (can_group_stmts_p): Likewise.
26598 (vect_analyze_data_ref_accesses): Use it instead of checking
26599 for a pair of assignments.
26600 (vect_store_lanes_supported): Take a masked_p parameter.
26601 (vect_load_lanes_supported): Likewise.
26602 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
26603 vect_store_lanes_supported and vect_load_lanes_supported.
26604 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
26605 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
26606 parameter. Don't allow gaps for masked accesses.
26607 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
26608 and vect_load_lanes_supported.
26609 (get_load_store_type): Take a masked_p parameter and update
26610 call to get_group_load_store_type.
26611 (vectorizable_store): Update call to get_load_store_type.
26612 Handle IFN_MASK_STORE_LANES.
26613 (vectorizable_load): Update call to get_load_store_type.
26614 Handle IFN_MASK_LOAD_LANES.
26616 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26617 Alan Hayward <alan.hayward@arm.com>
26618 David Sherwood <david.sherwood@arm.com>
26620 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
26622 * config/aarch64/aarch64-protos.h
26623 (aarch64_sve_struct_memory_operand_p): Declare.
26624 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
26625 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
26626 (VPRED, vpred): Handle SVE structure modes.
26627 * config/aarch64/constraints.md (Utx): New constraint.
26628 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
26629 (aarch64_sve_struct_nonimmediate_operand): New predicates.
26630 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
26631 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
26632 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
26633 structure modes. Split into pieces after RA.
26634 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
26635 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
26637 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
26638 SVE structure modes.
26639 (aarch64_classify_address): Likewise.
26640 (sizetochar): Move earlier in file.
26641 (aarch64_print_operand): Handle SVE register lists.
26642 (aarch64_array_mode): New function.
26643 (aarch64_sve_struct_memory_operand_p): Likewise.
26644 (TARGET_ARRAY_MODE): Redefine.
26646 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26647 Alan Hayward <alan.hayward@arm.com>
26648 David Sherwood <david.sherwood@arm.com>
26650 * target.def (array_mode): New target hook.
26651 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
26652 * doc/tm.texi: Regenerate.
26653 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
26654 * hooks.c (hook_optmode_mode_uhwi_none): New function.
26655 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
26656 targetm.array_mode.
26657 * stor-layout.c (mode_for_array): Likewise. Support polynomial
26660 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26661 Alan Hayward <alan.hayward@arm.com>
26662 David Sherwood <david.sherwood@arm.com>
26664 * fold-const.c (fold_binary_loc): Check the argument types
26665 rather than the result type when testing for a vector operation.
26667 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26669 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
26670 * doc/tm.texi: Regenerate.
26672 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26673 Alan Hayward <alan.hayward@arm.com>
26674 David Sherwood <david.sherwood@arm.com>
26676 * doc/invoke.texi (-msve-vector-bits=): Document new option.
26677 (sve): Document new AArch64 extension.
26678 * doc/md.texi (w): Extend the description of the AArch64
26679 constraint to include SVE vectors.
26680 (Upl, Upa): Document new AArch64 predicate constraints.
26681 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
26683 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
26684 (msve-vector-bits=): New option.
26685 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
26686 SVE when these are disabled.
26687 (sve): New extension.
26688 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
26689 modes. Adjust their number of units based on aarch64_sve_vg.
26690 (MAX_BITSIZE_MODE_ANY_MODE): Define.
26691 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
26692 aarch64_addr_query_type.
26693 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
26694 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
26695 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
26696 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
26697 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
26698 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
26699 (aarch64_simd_imm_zero_p): Delete.
26700 (aarch64_check_zero_based_sve_index_immediate): Declare.
26701 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
26702 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
26703 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
26704 (aarch64_sve_float_mul_immediate_p): Likewise.
26705 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
26706 rather than an rtx.
26707 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
26708 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
26709 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
26710 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
26711 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
26712 (aarch64_regmode_natural_size): Likewise.
26713 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
26714 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
26716 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
26717 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
26718 for VG and the SVE predicate registers.
26719 (V_ALIASES): Add a "z"-prefixed alias.
26720 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
26721 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
26722 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
26723 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
26724 (REG_CLASS_NAMES): Add entries for them.
26725 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
26726 and the predicate registers.
26727 (aarch64_sve_vg): Declare.
26728 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
26729 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
26730 (REGMODE_NATURAL_SIZE): Define.
26731 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
26733 * config/aarch64/aarch64.c: Include cfgrtl.h.
26734 (simd_immediate_info): Add a constructor for series vectors,
26735 and an associated step field.
26736 (aarch64_sve_vg): New variable.
26737 (aarch64_dbx_register_number): Handle VG and the predicate registers.
26738 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
26739 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
26740 (VEC_ANY_DATA, VEC_STRUCT): New constants.
26741 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
26742 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
26743 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
26744 (aarch64_get_mask_mode): New functions.
26745 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
26746 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
26747 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
26748 predicate modes and predicate registers. Explicitly restrict
26749 GPRs to modes of 16 bytes or smaller. Only allow FP registers
26750 to store a vector mode if it is recognized by
26751 aarch64_classify_vector_mode.
26752 (aarch64_regmode_natural_size): New function.
26753 (aarch64_hard_regno_caller_save_mode): Return the original mode
26755 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
26756 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
26757 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
26758 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
26760 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
26761 does not overlap dest if the function is frame-related. Handle
26763 (aarch64_split_add_offset): New function.
26764 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
26765 them aarch64_add_offset.
26766 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
26767 and update call to aarch64_sub_sp.
26768 (aarch64_add_cfa_expression): New function.
26769 (aarch64_expand_prologue): Pass extra temporary registers to the
26770 functions above. Handle the case in which we need to emit new
26771 DW_CFA_expressions for registers that were originally saved
26772 relative to the stack pointer, but now have to be expressed
26773 relative to the frame pointer.
26774 (aarch64_output_mi_thunk): Pass extra temporary registers to the
26776 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
26777 IP0 and IP1 values for SVE frames.
26778 (aarch64_expand_vec_series): New function.
26779 (aarch64_expand_sve_widened_duplicate): Likewise.
26780 (aarch64_expand_sve_const_vector): Likewise.
26781 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
26782 Handle SVE constants. Use emit_move_insn to move a force_const_mem
26783 into the register, rather than emitting a SET directly.
26784 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
26785 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
26786 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
26787 (offset_9bit_signed_scaled_p): New functions.
26788 (aarch64_replicate_bitmask_imm): New function.
26789 (aarch64_bitmask_imm): Use it.
26790 (aarch64_cannot_force_const_mem): Reject expressions involving
26791 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
26792 (aarch64_classify_index): Handle SVE indices, by requiring
26793 a plain register index with a scale that matches the element size.
26794 (aarch64_classify_address): Handle SVE addresses. Assert that
26795 the mode of the address is VOIDmode or an integer mode.
26796 Update call to aarch64_classify_symbol.
26797 (aarch64_classify_symbolic_expression): Update call to
26798 aarch64_classify_symbol.
26799 (aarch64_const_vec_all_in_range_p): New function.
26800 (aarch64_print_vector_float_operand): Likewise.
26801 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
26802 "vN" for FP registers with SVE modes. Handle (const ...) vectors
26803 and the FP immediates 1.0 and 0.5.
26804 (aarch64_print_address_internal): Handle SVE addresses.
26805 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
26806 (aarch64_regno_regclass): Handle predicate registers.
26807 (aarch64_secondary_reload): Handle big-endian reloads of SVE
26809 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
26810 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
26811 (aarch64_convert_sve_vector_bits): New function.
26812 (aarch64_override_options): Use it to handle -msve-vector-bits=.
26813 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
26814 rather than an rtx.
26815 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
26816 Handle SVE vector and predicate modes. Accept VL-based constants
26817 that need only one temporary register, and VL offsets that require
26818 no temporary registers.
26819 (aarch64_conditional_register_usage): Mark the predicate registers
26820 as fixed if SVE isn't available.
26821 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
26822 Return true for SVE vector and predicate modes.
26823 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
26824 rather than an unsigned int. Handle SVE modes.
26825 (aarch64_preferred_simd_mode): Update call accordingly. Handle
26827 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
26829 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
26830 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
26831 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
26832 (aarch64_sve_float_mul_immediate_p): New functions.
26833 (aarch64_sve_valid_immediate): New function.
26834 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
26835 Explicitly reject structure modes. Check for INDEX constants.
26836 Handle PTRUE and PFALSE constants.
26837 (aarch64_check_zero_based_sve_index_immediate): New function.
26838 (aarch64_simd_imm_zero_p): Delete.
26839 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
26840 vector modes. Accept constants in the range of CNT[BHWD].
26841 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
26842 ask for an Advanced SIMD mode.
26843 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
26844 (aarch64_simd_vector_alignment): Handle SVE predicates.
26845 (aarch64_vectorize_preferred_vector_alignment): New function.
26846 (aarch64_simd_vector_alignment_reachable): Use it instead of
26848 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
26849 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
26851 (MAX_VECT_LEN): Delete.
26852 (expand_vec_perm_d): Add a vec_flags field.
26853 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
26854 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
26855 (aarch64_evpc_ext): Don't apply a big-endian lane correction
26857 (aarch64_evpc_rev): Rename to...
26858 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
26859 (aarch64_evpc_rev_global): New function.
26860 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
26861 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
26863 (aarch64_evpc_sve_tbl): New function.
26864 (aarch64_expand_vec_perm_const_1): Update after rename of
26865 aarch64_evpc_rev. Handle SVE permutes too, trying
26866 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
26867 than aarch64_evpc_tbl.
26868 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
26869 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
26870 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
26871 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
26872 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
26873 (aarch64_expand_sve_vcond): New functions.
26874 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
26875 of aarch64_vector_mode_p.
26876 (aarch64_dwarf_poly_indeterminate_value): New function.
26877 (aarch64_compute_pressure_classes): Likewise.
26878 (aarch64_can_change_mode_class): Likewise.
26879 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
26880 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
26881 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
26882 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
26883 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
26884 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
26885 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
26886 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
26888 (Dn, Dl, Dr): Accept const as well as const_vector.
26889 (Dz): Likewise. Compare against CONST0_RTX.
26890 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
26891 of "vector" where appropriate.
26892 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
26893 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
26894 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
26895 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
26896 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
26897 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
26898 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
26899 (v_int_equiv): Extend to SVE modes.
26900 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
26902 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
26903 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
26904 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
26905 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
26906 (SVE_COND_FP_CMP): New int iterators.
26907 (perm_hilo): Handle the new unpack unspecs.
26908 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
26910 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
26911 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
26912 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
26913 (aarch64_equality_operator, aarch64_constant_vector_operand)
26914 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
26915 (aarch64_sve_nonimmediate_operand): Likewise.
26916 (aarch64_sve_general_operand): Likewise.
26917 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
26918 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
26919 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
26920 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
26921 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
26922 (aarch64_sve_float_arith_immediate): Likewise.
26923 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
26924 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
26925 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
26926 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
26927 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
26928 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
26929 (aarch64_sve_float_arith_operand): Likewise.
26930 (aarch64_sve_float_arith_with_sub_operand): Likewise.
26931 (aarch64_sve_float_mul_operand): Likewise.
26932 (aarch64_sve_vec_perm_operand): Likewise.
26933 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
26934 (aarch64_mov_operand): Accept const_poly_int and const_vector.
26935 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
26936 as well as const_vector.
26937 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
26938 in file. Use CONST0_RTX and CONSTM1_RTX.
26939 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
26940 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
26941 Use aarch64_simd_imm_zero.
26942 * config/aarch64/aarch64-sve.md: New file.
26943 * config/aarch64/aarch64.md: Include it.
26944 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
26945 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
26946 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
26947 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
26948 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
26949 (sve): New attribute.
26950 (enabled): Disable instructions with the sve attribute unless
26952 (movqi, movhi): Pass CONST_POLY_INT operaneds through
26953 aarch64_expand_mov_immediate.
26954 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
26955 CNT[BHSD] immediates.
26956 (movti): Split CONST_POLY_INT moves into two halves.
26957 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
26958 Split additions that need a temporary here if the destination
26959 is the stack pointer.
26960 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
26961 (*add<mode>3_poly_1): New instruction.
26962 (set_clobber_cc): New expander.
26964 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26966 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
26967 parameter and use it instead of GET_MODE_SIZE (innermode). Use
26968 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
26969 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
26970 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
26971 Change innermode from fixed_mode_size to machine_mode.
26972 (simplify_subreg): Update call accordingly. Handle a constant-sized
26973 subreg of a variable-length CONST_VECTOR.
26975 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
26976 Alan Hayward <alan.hayward@arm.com>
26977 David Sherwood <david.sherwood@arm.com>
26979 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
26980 (add_offset_to_base): New function, split out from...
26981 (create_mem_ref): ...here. When handling a scale other than 1,
26982 check first whether the address is valid without the offset.
26983 Add it into the base if so, leaving the index and scale as-is.
26985 2018-01-12 Jakub Jelinek <jakub@redhat.com>
26988 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
26989 fold_for_warn before checking if arg2 is INTEGER_CST.
26991 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
26993 * config/rs6000/predicates.md (load_multiple_operation): Delete.
26994 (store_multiple_operation): Delete.
26995 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
26996 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
26997 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
26998 guarded by TARGET_STRING.
26999 (rs6000_output_load_multiple): Delete.
27000 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
27001 OPTION_MASK_STRING / TARGET_STRING handling.
27002 (print_operand) <'N', 'O'>: Add comment that these are unused now.
27003 (const rs6000_opt_masks) <"string">: Change mask to 0.
27004 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
27005 (MASK_STRING): Delete.
27006 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
27008 (load_multiple): Delete.
27015 (store_multiple): Delete.
27022 (movmemsi_8reg): Delete.
27023 (corresponding unnamed define_insn): Delete.
27024 (movmemsi_6reg): Delete.
27025 (corresponding unnamed define_insn): Delete.
27026 (movmemsi_4reg): Delete.
27027 (corresponding unnamed define_insn): Delete.
27028 (movmemsi_2reg): Delete.
27029 (corresponding unnamed define_insn): Delete.
27030 (movmemsi_1reg): Delete.
27031 (corresponding unnamed define_insn): Delete.
27032 * config/rs6000/rs6000.opt (mno-string): New.
27033 (mstring): Replace by deprecation warning stub.
27034 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
27036 2018-01-12 Jakub Jelinek <jakub@redhat.com>
27038 * regrename.c (regrename_do_replace): If replacing the same
27039 reg multiple times, try to reuse last created gen_raw_REG.
27042 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
27043 main to workaround a bug in GDB.
27045 2018-01-12 Tom de Vries <tom@codesourcery.com>
27048 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
27050 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
27052 PR rtl-optimization/80481
27053 * ira-color.c (get_cap_member): New function.
27054 (allocnos_conflict_by_live_ranges_p): Use it.
27055 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
27056 (setup_slot_coalesced_allocno_live_ranges): Ditto.
27058 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
27061 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
27062 (*saddl_se_1): Ditto.
27063 (*ssubsi_1): Ditto.
27064 (*ssubl_se_1): Ditto.
27066 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
27068 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
27069 rather than wi::to_widest for DR_INITs.
27070 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
27071 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
27072 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
27074 (vect_analyze_group_access_1): Note that here.
27076 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
27078 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
27079 polynomial type sizes.
27081 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
27083 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
27084 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
27085 (gimple_add_tmp_var): Likewise.
27087 2018-01-12 Martin Liska <mliska@suse.cz>
27089 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
27090 (gimple_alloc_sizes): Likewise.
27091 (dump_gimple_statistics): Use PRIu64 in printf format.
27092 * gimple.h: Change uint64_t to int.
27094 2018-01-12 Martin Liska <mliska@suse.cz>
27096 * tree-core.h: Use uint64_t instead of int.
27097 * tree.c (tree_node_counts): Likewise.
27098 (tree_node_sizes): Likewise.
27099 (dump_tree_statistics): Use PRIu64 in printf format.
27101 2018-01-12 Martin Liska <mliska@suse.cz>
27103 * Makefile.in: As qsort_chk is implemented in vec.c, add
27104 vec.o to linkage of gencfn-macros.
27105 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
27106 passing the info to record_node_allocation_statistics.
27107 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
27109 * ggc-common.c (struct ggc_usage): Add operator== and use
27110 it in operator< and compare function.
27111 * mem-stats.h (struct mem_usage): Likewise.
27112 * vec.c (struct vec_usage): Remove operator< and compare
27113 function. Can be simply inherited.
27115 2018-01-12 Martin Jambor <mjambor@suse.cz>
27118 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
27119 * tree-ssa-math-opts.c: Include domwalk.h.
27120 (convert_mult_to_fma_1): New function.
27121 (fma_transformation_info): New type.
27122 (fma_deferring_state): Likewise.
27123 (cancel_fma_deferring): New function.
27124 (result_of_phi): Likewise.
27125 (last_fma_candidate_feeds_initial_phi): Likewise.
27126 (convert_mult_to_fma): Added deferring logic, split actual
27127 transformation to convert_mult_to_fma_1.
27128 (math_opts_dom_walker): New type.
27129 (math_opts_dom_walker::after_dom_children): New method, body moved
27130 here from pass_optimize_widening_mul::execute, added deferring logic
27132 (pass_optimize_widening_mul::execute): Moved most of code to
27133 math_opts_dom_walker::after_dom_children.
27134 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
27135 * config/i386/i386.c (ix86_option_override_internal): Added
27136 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
27138 2018-01-12 Richard Biener <rguenther@suse.de>
27141 * dwarf2out.c (gen_variable_die): Do not reset old_die for
27142 inline instance vars.
27144 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
27147 * config/rx/rx.c (rx_is_restricted_memory_address):
27148 Handle SUBREG case.
27150 2018-01-12 Richard Biener <rguenther@suse.de>
27152 PR tree-optimization/80846
27153 * target.def (split_reduction): New target hook.
27154 * targhooks.c (default_split_reduction): New function.
27155 * targhooks.h (default_split_reduction): Declare.
27156 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
27157 target requests first reduce vectors by combining low and high
27159 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
27160 (get_vectype_for_scalar_type_and_size): Export.
27161 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
27162 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
27163 * doc/tm.texi: Regenerate.
27164 * config/i386/i386.c (ix86_split_reduction): Implement
27165 TARGET_VECTORIZE_SPLIT_REDUCTION.
27167 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
27170 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
27171 in PIC mode except for TARGET_VXWORKS_RTP.
27172 * config/sparc/sparc.c: Include cfgrtl.h.
27173 (TARGET_INIT_PIC_REG): Define.
27174 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
27175 (sparc_pic_register_p): New predicate.
27176 (sparc_legitimate_address_p): Use it.
27177 (sparc_legitimize_pic_address): Likewise.
27178 (sparc_delegitimize_address): Likewise.
27179 (sparc_mode_dependent_address_p): Likewise.
27180 (gen_load_pcrel_sym): Remove 4th parameter.
27181 (load_got_register): Adjust call to above. Remove obsolete stuff.
27182 (sparc_expand_prologue): Do not call load_got_register here.
27183 (sparc_flat_expand_prologue): Likewise.
27184 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
27185 (sparc_use_pseudo_pic_reg): New function.
27186 (sparc_init_pic_reg): Likewise.
27187 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
27188 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
27190 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
27192 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
27193 Add item for branch_cost.
27195 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
27197 PR rtl-optimization/83565
27198 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
27199 not extend the result to a larger mode for rotate operations.
27200 (num_sign_bit_copies1): Likewise.
27202 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
27205 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
27207 Use values-Xc.o for -pedantic.
27208 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
27210 2018-01-12 Martin Liska <mliska@suse.cz>
27213 * ipa-devirt.c (final_warning_record::grow_type_warnings):
27215 (possible_polymorphic_call_targets): Use it.
27216 (ipa_devirt): Likewise.
27218 2018-01-12 Martin Liska <mliska@suse.cz>
27220 * profile-count.h (enum profile_quality): Use 0 as invalid
27221 enum value of profile_quality.
27223 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
27225 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
27226 -mext-string options.
27228 2018-01-12 Richard Biener <rguenther@suse.de>
27230 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
27231 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
27232 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
27234 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
27236 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
27238 * configure.ac (--with-long-double-format): Add support for the
27239 configuration option to change the default long double format on
27241 * config.gcc (powerpc*-linux*-*): Likewise.
27242 * configure: Regenerate.
27243 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
27244 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
27245 used without modification.
27247 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
27249 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
27250 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
27251 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
27252 MISC_BUILTIN_SPEC_BARRIER.
27253 (rs6000_init_builtins): Likewise.
27254 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
27256 (speculation_barrier): New define_insn.
27257 * doc/extend.texi: Document __builtin_speculation_barrier.
27259 2018-01-11 Jakub Jelinek <jakub@redhat.com>
27262 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
27263 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
27264 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
27266 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
27267 integral modes instead of "ss" and "sd".
27268 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
27269 vectors with 32-bit and 64-bit elements.
27270 (vecdupssescalarmodesuffix): New mode attribute.
27271 (vec_dup<mode>): Use it.
27273 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
27276 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
27277 frame if argument is passed on stack.
27279 2018-01-11 Jakub Jelinek <jakub@redhat.com>
27282 * ree.c (combine_reaching_defs): Optimize also
27283 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
27284 reg2=any_extend(exp); reg1=reg2;, formatting fix.
27286 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
27288 PR middle-end/83189
27289 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
27291 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
27293 PR middle-end/83718
27294 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
27295 after they are computed.
27297 2018-01-11 Bin Cheng <bin.cheng@arm.com>
27299 PR tree-optimization/83695
27300 * gimple-loop-linterchange.cc
27301 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
27302 reset cached scev information after interchange.
27303 (pass_linterchange::execute): Remove call to scev_reset_htab.
27305 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27307 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
27308 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
27309 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
27310 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
27311 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
27312 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
27313 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
27314 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
27315 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
27316 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
27317 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
27318 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
27319 (V_lane_reg): Likewise.
27320 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
27322 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
27323 (vfmal_lane_low<mode>_intrinsic,
27324 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
27325 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
27326 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
27327 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
27328 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
27329 vfmsl_lane_high<mode>_intrinsic): New define_insns.
27331 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27333 * config/arm/arm-cpus.in (fp16fml): New feature.
27334 (ALL_SIMD): Add fp16fml.
27335 (armv8.2-a): Add fp16fml as an option.
27336 (armv8.3-a): Likewise.
27337 (armv8.4-a): Add fp16fml as part of fp16.
27338 * config/arm/arm.h (TARGET_FP16FML): Define.
27339 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
27341 * config/arm/arm-modes.def (V2HF): Define.
27342 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
27343 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
27344 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
27345 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
27346 vfmsl_low, vfmsl_high): New set of builtins.
27347 * config/arm/iterators.md (PLUSMINUS): New code iterator.
27348 (vfml_op): New code attribute.
27349 (VFMLHALVES): New int iterator.
27350 (VFML, VFMLSEL): New mode attributes.
27351 (V_reg): Define mapping for V2HF.
27352 (V_hi, V_lo): New mode attributes.
27353 (VF_constraint): Likewise.
27354 (vfml_half, vfml_half_selector): New int attributes.
27355 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
27357 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
27358 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
27360 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
27361 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
27362 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
27363 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
27365 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
27366 Document new effective target and option set.
27368 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27370 * config/arm/arm-cpus.in (armv8_4): New feature.
27371 (ARMv8_4a): New fgroup.
27372 (armv8.4-a): New arch.
27373 * config/arm/arm-tables.opt: Regenerate.
27374 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
27375 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
27376 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
27377 Add matching rules for -march=armv8.4-a and extensions.
27378 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
27380 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
27383 * config/rx/rx.md (BW): New mode attribute.
27384 (sync_lock_test_and_setsi): Add mode suffix to insn output.
27386 2018-01-11 Richard Biener <rguenther@suse.de>
27388 PR tree-optimization/83435
27389 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
27390 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
27391 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
27393 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
27394 Alan Hayward <alan.hayward@arm.com>
27395 David Sherwood <david.sherwood@arm.com>
27397 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
27399 (aarch64_classify_address): Initialize it. Track polynomial offsets.
27400 (aarch64_print_address_internal): Use it to check for a zero offset.
27402 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
27403 Alan Hayward <alan.hayward@arm.com>
27404 David Sherwood <david.sherwood@arm.com>
27406 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
27407 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
27408 Return a poly_int64 rather than a HOST_WIDE_INT.
27409 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
27410 rather than a HOST_WIDE_INT.
27411 * config/aarch64/aarch64.h (aarch64_frame): Protect with
27412 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
27413 hard_fp_offset, frame_size, initial_adjust, callee_offset and
27414 final_offset from HOST_WIDE_INT to poly_int64.
27415 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
27416 to_constant when getting the number of units in an Advanced SIMD
27418 (aarch64_builtin_vectorized_function): Check for a constant number
27420 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
27422 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
27423 attribute instead of GET_MODE_NUNITS.
27424 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
27425 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
27426 GET_MODE_SIZE for fixed-size registers.
27427 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
27428 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
27429 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
27430 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
27431 (aarch64_print_operand, aarch64_print_address_internal)
27432 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
27433 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
27434 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
27435 Handle polynomial GET_MODE_SIZE.
27436 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
27437 wider than SImode without modification.
27438 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
27439 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
27440 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
27441 passing and returning SVE modes.
27442 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
27443 rather than GEN_INT.
27444 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
27445 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
27446 (aarch64_allocate_and_probe_stack_space): Likewise.
27447 (aarch64_layout_frame): Cope with polynomial offsets.
27448 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
27449 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
27450 polynomial offsets.
27451 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
27452 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
27453 poly_int64 rather than a HOST_WIDE_INT.
27454 (aarch64_get_separate_components, aarch64_process_components)
27455 (aarch64_expand_prologue, aarch64_expand_epilogue)
27456 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
27457 (aarch64_anchor_offset): New function, split out from...
27458 (aarch64_legitimize_address): ...here.
27459 (aarch64_builtin_vectorization_cost): Handle polynomial
27460 TYPE_VECTOR_SUBPARTS.
27461 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
27463 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
27464 number of elements from the PARALLEL rather than the mode.
27465 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
27466 rather than GET_MODE_BITSIZE.
27467 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
27468 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
27469 (aarch64_expand_vec_perm_const_1): Handle polynomial
27470 d->perm.length () and d->perm elements.
27471 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
27472 Apply to_constant to d->perm elements.
27473 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
27474 polynomial CONST_VECTOR_NUNITS.
27475 (aarch64_move_pointer): Take amount as a poly_int64 rather
27477 (aarch64_progress_pointer): Avoid temporary variable.
27478 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
27479 the mode attribute instead of GET_MODE.
27481 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
27482 Alan Hayward <alan.hayward@arm.com>
27483 David Sherwood <david.sherwood@arm.com>
27485 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
27486 x exists before using it.
27487 (aarch64_add_constant_internal): Rename to...
27488 (aarch64_add_offset_1): ...this. Replace regnum with separate
27489 src and dest rtxes. Handle the case in which they're different,
27490 including when the offset is zero. Replace scratchreg with an rtx.
27491 Use 2 additions if there is no spare register into which we can
27492 move a 16-bit constant.
27493 (aarch64_add_constant): Delete.
27494 (aarch64_add_offset): Replace reg with separate src and dest
27495 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
27496 Use aarch64_add_offset_1.
27497 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
27498 an rtx rather than an int. Take the delta as a poly_int64
27499 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
27500 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
27501 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
27502 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
27503 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
27504 and aarch64_add_sp.
27505 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
27506 aarch64_add_constant.
27508 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
27510 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
27511 Use scalar_float_mode.
27513 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
27515 * config/aarch64/aarch64-simd.md
27516 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
27517 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
27518 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
27519 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
27520 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
27521 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
27522 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
27523 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
27524 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
27525 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
27527 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
27530 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
27531 targ_options->x_arm_arch_string is non NULL.
27533 2018-01-11 Tamar Christina <tamar.christina@arm.com>
27535 * config/aarch64/aarch64.h
27536 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
27538 2018-01-11 Sudakshina Das <sudi.das@arm.com>
27541 * expmed.c (emit_store_flag_force): Swap if const op0
27542 and change VOIDmode to mode of op0.
27544 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
27546 PR rtl-optimization/83761
27547 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
27548 than bytes to mode_for_size.
27550 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
27552 PR middle-end/83189
27553 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
27554 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
27557 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
27559 PR middle-end/83575
27560 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
27561 when in layout mode.
27562 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
27563 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
27566 2018-01-10 Michael Collison <michael.collison@arm.com>
27568 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
27569 * config/aarch64/aarch64-option-extension.def: Add
27570 AARCH64_OPT_EXTENSION of 'fp16fml'.
27571 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
27572 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
27573 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
27574 * config/aarch64/constraints.md (Ui7): New constraint.
27575 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
27576 (VFMLA_SEL_W): Ditto.
27579 (VFMLA16_LOW): New int iterator.
27580 (VFMLA16_HIGH): Ditto.
27581 (UNSPEC_FMLAL): New unspec.
27582 (UNSPEC_FMLSL): Ditto.
27583 (UNSPEC_FMLAL2): Ditto.
27584 (UNSPEC_FMLSL2): Ditto.
27585 (f16mac): New code attribute.
27586 * config/aarch64/aarch64-simd-builtins.def
27587 (aarch64_fmlal_lowv2sf): Ditto.
27588 (aarch64_fmlsl_lowv2sf): Ditto.
27589 (aarch64_fmlalq_lowv4sf): Ditto.
27590 (aarch64_fmlslq_lowv4sf): Ditto.
27591 (aarch64_fmlal_highv2sf): Ditto.
27592 (aarch64_fmlsl_highv2sf): Ditto.
27593 (aarch64_fmlalq_highv4sf): Ditto.
27594 (aarch64_fmlslq_highv4sf): Ditto.
27595 (aarch64_fmlal_lane_lowv2sf): Ditto.
27596 (aarch64_fmlsl_lane_lowv2sf): Ditto.
27597 (aarch64_fmlal_laneq_lowv2sf): Ditto.
27598 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
27599 (aarch64_fmlalq_lane_lowv4sf): Ditto.
27600 (aarch64_fmlsl_lane_lowv4sf): Ditto.
27601 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
27602 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
27603 (aarch64_fmlal_lane_highv2sf): Ditto.
27604 (aarch64_fmlsl_lane_highv2sf): Ditto.
27605 (aarch64_fmlal_laneq_highv2sf): Ditto.
27606 (aarch64_fmlsl_laneq_highv2sf): Ditto.
27607 (aarch64_fmlalq_lane_highv4sf): Ditto.
27608 (aarch64_fmlsl_lane_highv4sf): Ditto.
27609 (aarch64_fmlalq_laneq_highv4sf): Ditto.
27610 (aarch64_fmlsl_laneq_highv4sf): Ditto.
27611 * config/aarch64/aarch64-simd.md:
27612 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
27613 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
27614 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
27615 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
27616 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
27617 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
27618 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
27619 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
27620 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
27621 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
27622 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
27623 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
27624 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
27625 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
27626 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
27627 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
27628 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
27629 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
27630 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
27631 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
27632 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
27633 (vfmlsl_low_u32): Ditto.
27634 (vfmlalq_low_u32): Ditto.
27635 (vfmlslq_low_u32): Ditto.
27636 (vfmlal_high_u32): Ditto.
27637 (vfmlsl_high_u32): Ditto.
27638 (vfmlalq_high_u32): Ditto.
27639 (vfmlslq_high_u32): Ditto.
27640 (vfmlal_lane_low_u32): Ditto.
27641 (vfmlsl_lane_low_u32): Ditto.
27642 (vfmlal_laneq_low_u32): Ditto.
27643 (vfmlsl_laneq_low_u32): Ditto.
27644 (vfmlalq_lane_low_u32): Ditto.
27645 (vfmlslq_lane_low_u32): Ditto.
27646 (vfmlalq_laneq_low_u32): Ditto.
27647 (vfmlslq_laneq_low_u32): Ditto.
27648 (vfmlal_lane_high_u32): Ditto.
27649 (vfmlsl_lane_high_u32): Ditto.
27650 (vfmlal_laneq_high_u32): Ditto.
27651 (vfmlsl_laneq_high_u32): Ditto.
27652 (vfmlalq_lane_high_u32): Ditto.
27653 (vfmlslq_lane_high_u32): Ditto.
27654 (vfmlalq_laneq_high_u32): Ditto.
27655 (vfmlslq_laneq_high_u32): Ditto.
27656 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
27657 (AARCH64_FL_FOR_ARCH8_4): New.
27658 (AARCH64_ISA_F16FML): New ISA flag.
27659 (TARGET_F16FML): New feature flag for fp16fml.
27660 (doc/invoke.texi): Document new fp16fml option.
27662 2018-01-10 Michael Collison <michael.collison@arm.com>
27664 * config/aarch64/aarch64-builtins.c:
27665 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
27666 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
27667 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
27668 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
27669 (AARCH64_ISA_SHA3): New ISA flag.
27670 (TARGET_SHA3): New feature flag for sha3.
27671 * config/aarch64/iterators.md (sha512_op): New int attribute.
27672 (CRYPTO_SHA512): New int iterator.
27673 (UNSPEC_SHA512H): New unspec.
27674 (UNSPEC_SHA512H2): Ditto.
27675 (UNSPEC_SHA512SU0): Ditto.
27676 (UNSPEC_SHA512SU1): Ditto.
27677 * config/aarch64/aarch64-simd-builtins.def
27678 (aarch64_crypto_sha512hqv2di): New builtin.
27679 (aarch64_crypto_sha512h2qv2di): Ditto.
27680 (aarch64_crypto_sha512su0qv2di): Ditto.
27681 (aarch64_crypto_sha512su1qv2di): Ditto.
27682 (aarch64_eor3qv8hi): Ditto.
27683 (aarch64_rax1qv2di): Ditto.
27684 (aarch64_xarqv2di): Ditto.
27685 (aarch64_bcaxqv8hi): Ditto.
27686 * config/aarch64/aarch64-simd.md:
27687 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
27688 (aarch64_crypto_sha512su0qv2di): Ditto.
27689 (aarch64_crypto_sha512su1qv2di): Ditto.
27690 (aarch64_eor3qv8hi): Ditto.
27691 (aarch64_rax1qv2di): Ditto.
27692 (aarch64_xarqv2di): Ditto.
27693 (aarch64_bcaxqv8hi): Ditto.
27694 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
27695 (vsha512h2q_u64): Ditto.
27696 (vsha512su0q_u64): Ditto.
27697 (vsha512su1q_u64): Ditto.
27698 (veor3q_u16): Ditto.
27699 (vrax1q_u64): Ditto.
27700 (vxarq_u64): Ditto.
27701 (vbcaxq_u16): Ditto.
27702 * config/arm/types.md (crypto_sha512): New type attribute.
27703 (crypto_sha3): Ditto.
27704 (doc/invoke.texi): Document new sha3 option.
27706 2018-01-10 Michael Collison <michael.collison@arm.com>
27708 * config/aarch64/aarch64-builtins.c:
27709 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
27710 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
27711 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
27712 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
27713 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
27714 (AARCH64_ISA_SM4): New ISA flag.
27715 (TARGET_SM4): New feature flag for sm4.
27716 * config/aarch64/aarch64-simd-builtins.def
27717 (aarch64_sm3ss1qv4si): Ditto.
27718 (aarch64_sm3tt1aq4si): Ditto.
27719 (aarch64_sm3tt1bq4si): Ditto.
27720 (aarch64_sm3tt2aq4si): Ditto.
27721 (aarch64_sm3tt2bq4si): Ditto.
27722 (aarch64_sm3partw1qv4si): Ditto.
27723 (aarch64_sm3partw2qv4si): Ditto.
27724 (aarch64_sm4eqv4si): Ditto.
27725 (aarch64_sm4ekeyqv4si): Ditto.
27726 * config/aarch64/aarch64-simd.md:
27727 (aarch64_sm3ss1qv4si): Ditto.
27728 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
27729 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
27730 (aarch64_sm4eqv4si): Ditto.
27731 (aarch64_sm4ekeyqv4si): Ditto.
27732 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
27733 (sm3part_op): Ditto.
27734 (CRYPTO_SM3TT): Ditto.
27735 (CRYPTO_SM3PART): Ditto.
27736 (UNSPEC_SM3SS1): New unspec.
27737 (UNSPEC_SM3TT1A): Ditto.
27738 (UNSPEC_SM3TT1B): Ditto.
27739 (UNSPEC_SM3TT2A): Ditto.
27740 (UNSPEC_SM3TT2B): Ditto.
27741 (UNSPEC_SM3PARTW1): Ditto.
27742 (UNSPEC_SM3PARTW2): Ditto.
27743 (UNSPEC_SM4E): Ditto.
27744 (UNSPEC_SM4EKEY): Ditto.
27745 * config/aarch64/constraints.md (Ui2): New constraint.
27746 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
27747 * config/arm/types.md (crypto_sm3): New type attribute.
27748 (crypto_sm4): Ditto.
27749 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
27750 (vsm3tt1aq_u32): Ditto.
27751 (vsm3tt1bq_u32): Ditto.
27752 (vsm3tt2aq_u32): Ditto.
27753 (vsm3tt2bq_u32): Ditto.
27754 (vsm3partw1q_u32): Ditto.
27755 (vsm3partw2q_u32): Ditto.
27756 (vsm4eq_u32): Ditto.
27757 (vsm4ekeyq_u32): Ditto.
27758 (doc/invoke.texi): Document new sm4 option.
27760 2018-01-10 Michael Collison <michael.collison@arm.com>
27762 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
27763 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
27764 (AARCH64_FL_FOR_ARCH8_4): New.
27765 (AARCH64_FL_V8_4): New flag.
27766 (doc/invoke.texi): Document new armv8.4-a option.
27768 2018-01-10 Michael Collison <michael.collison@arm.com>
27770 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
27771 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
27772 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
27773 * config/aarch64/aarch64-option-extension.def: Add
27774 AARCH64_OPT_EXTENSION of 'sha2'.
27775 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
27776 (crypto): Disable sha2 and aes if crypto disabled.
27777 (crypto): Enable aes and sha2 if enabled.
27778 (simd): Disable sha2 and aes if simd disabled.
27779 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
27781 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
27782 (TARGET_SHA2): New feature flag for sha2.
27783 (TARGET_AES): New feature flag for aes.
27784 * config/aarch64/aarch64-simd.md:
27785 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
27786 conditional on TARGET_AES.
27787 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
27788 (aarch64_crypto_sha1hsi): Make pattern conditional
27790 (aarch64_crypto_sha1hv4si): Ditto.
27791 (aarch64_be_crypto_sha1hv4si): Ditto.
27792 (aarch64_crypto_sha1su1v4si): Ditto.
27793 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
27794 (aarch64_crypto_sha1su0v4si): Ditto.
27795 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
27796 (aarch64_crypto_sha256su0v4si): Ditto.
27797 (aarch64_crypto_sha256su1v4si): Ditto.
27798 (doc/invoke.texi): Document new aes and sha2 options.
27800 2018-01-10 Martin Sebor <msebor@redhat.com>
27802 PR tree-optimization/83781
27803 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
27806 2018-01-11 Martin Sebor <msebor@gmail.com>
27807 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
27809 PR tree-optimization/83501
27810 PR tree-optimization/81703
27812 * tree-ssa-strlen.c (get_string_cst): Rename...
27813 (get_string_len): ...to this. Handle global constants.
27814 (handle_char_store): Adjust.
27816 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
27817 Jim Wilson <jimw@sifive.com>
27819 * config/riscv/riscv-protos.h (riscv_output_return): New.
27820 * config/riscv/riscv.c (struct machine_function): New naked_p field.
27821 (riscv_attribute_table, riscv_output_return),
27822 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
27823 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
27824 (riscv_compute_frame_info): Only compute frame->mask if not a naked
27826 (riscv_expand_prologue): Add early return for naked function.
27827 (riscv_expand_epilogue): Likewise.
27828 (riscv_function_ok_for_sibcall): Return false for naked function.
27829 (riscv_set_current_function): New.
27830 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
27831 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
27832 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
27833 * doc/extend.texi (RISC-V Function Attributes): New.
27835 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
27837 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
27838 check for 128-bit long double before checking TCmode.
27839 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
27840 128-bit long doubles before checking TFmode or TCmode.
27841 (FLOAT128_IBM_P): Likewise.
27843 2018-01-10 Martin Sebor <msebor@redhat.com>
27845 PR tree-optimization/83671
27846 * builtins.c (c_strlen): Unconditionally return zero for the empty
27848 Use -Warray-bounds for warnings.
27849 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
27850 for non-constant array indices with COMPONENT_REF, arrays of
27851 arrays, and pointers to arrays.
27852 (gimple_fold_builtin_strlen): Determine and set length range for
27853 non-constant character arrays.
27855 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
27857 PR middle-end/81897
27858 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
27861 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
27863 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
27865 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
27868 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
27869 VECTOR_MEM_ALTIVEC_OR_VSX_P.
27870 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
27871 indexed_or_indirect_operand predicate.
27872 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
27873 (*vsx_le_perm_load_v8hi): Likewise.
27874 (*vsx_le_perm_load_v16qi): Likewise.
27875 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
27876 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
27877 (*vsx_le_perm_store_v8hi): Likewise.
27878 (*vsx_le_perm_store_v16qi): Likewise.
27879 (eight unnamed splitters): Likewise.
27881 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
27883 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
27884 * config/rs6000/emmintrin.h: Likewise.
27885 * config/rs6000/mmintrin.h: Likewise.
27886 * config/rs6000/xmmintrin.h: Likewise.
27888 2018-01-10 David Malcolm <dmalcolm@redhat.com>
27891 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
27893 * tree.c (tree_nop_conversion): Return true for location wrapper
27895 (maybe_wrap_with_location): New function.
27896 (selftest::check_strip_nops): New function.
27897 (selftest::test_location_wrappers): New function.
27898 (selftest::tree_c_tests): Call it.
27899 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
27900 (maybe_wrap_with_location): New decl.
27901 (EXPR_LOCATION_WRAPPER_P): New macro.
27902 (location_wrapper_p): New inline function.
27903 (tree_strip_any_location_wrapper): New inline function.
27905 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
27908 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
27909 stack_realign_offset for the largest alignment of stack slot
27911 (ix86_find_max_used_stack_alignment): New function.
27912 (ix86_finalize_stack_frame_flags): Use it. Set
27913 max_used_stack_alignment if we don't realign stack.
27914 * config/i386/i386.h (machine_function): Add
27915 max_used_stack_alignment.
27917 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
27919 * config/arm/arm.opt (-mbranch-cost): New option.
27920 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
27923 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
27926 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
27927 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
27929 2018-01-10 Richard Biener <rguenther@suse.de>
27932 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
27933 early out so it also covers the case where we have a non-NULL
27936 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
27938 PR tree-optimization/83753
27939 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
27940 for non-strided grouped accesses if the number of elements is 1.
27942 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
27945 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
27946 * i386.h (TARGET_USE_GATHER): Define.
27947 * x86-tune.def (X86_TUNE_USE_GATHER): New.
27949 2018-01-10 Martin Liska <mliska@suse.cz>
27952 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
27953 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
27955 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
27956 CLEANUP_NO_PARTITIONING is not set.
27958 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
27960 * doc/rtl.texi: Remove documentation of (const ...) wrappers
27961 for vectors, as a partial revert of r254296.
27962 * rtl.h (const_vec_p): Delete.
27963 (const_vec_duplicate_p): Don't test for vector CONSTs.
27964 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
27965 * expmed.c (make_tree): Likewise.
27968 * common.md (E, F): Use CONSTANT_P instead of checking for
27970 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
27971 checking for CONST_VECTOR.
27973 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
27975 PR middle-end/83575
27976 * predict.c (force_edge_cold): Handle in more sane way edges
27977 with no prediction.
27979 2018-01-09 Carl Love <cel@us.ibm.com>
27981 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
27983 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
27984 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
27985 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
27986 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
27987 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
27988 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
27989 * config/rs6000/rs6000-protos.h: Add extern defition for
27990 rs6000_generate_float2_double_code.
27991 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
27993 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
27994 (float2_v2df): Add define_expand.
27996 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
27999 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
28000 op_mode in the force_to_mode call.
28002 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
28004 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
28005 instead of checking each element individually.
28006 (aarch64_evpc_uzp): Likewise.
28007 (aarch64_evpc_zip): Likewise.
28008 (aarch64_evpc_ext): Likewise.
28009 (aarch64_evpc_rev): Likewise.
28010 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
28011 instead of checking each element individually. Return true without
28013 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
28014 whether all selected elements come from the same input, instead of
28015 checking each element individually. Remove calls to gen_rtx_REG,
28016 start_sequence and end_sequence and instead assert that no rtl is
28019 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
28021 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
28022 order of HIGH and CONST checks.
28024 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
28026 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
28027 if the destination isn't an SSA_NAME.
28029 2018-01-09 Richard Biener <rguenther@suse.de>
28031 PR tree-optimization/83668
28032 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
28034 (canonicalize_loop_form): ... here, renamed from ...
28035 (canonicalize_loop_closed_ssa_form): ... this and amended to
28036 swap successor edges for loop exit blocks to make us use
28037 the RPO order we need for initial schedule generation.
28039 2018-01-09 Joseph Myers <joseph@codesourcery.com>
28041 PR tree-optimization/64811
28042 * match.pd: When optimizing comparisons with Inf, avoid
28043 introducing or losing exceptions from comparisons with NaN.
28045 2018-01-09 Martin Liska <mliska@suse.cz>
28048 * asan.c (shadow_mem_size): Add gcc_assert.
28050 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
28052 Don't save registers in main().
28055 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
28056 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
28057 * config/avr/avr.c (avr_set_current_function): Don't error if
28058 naked, OS_task or OS_main are specified at the same time.
28059 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
28061 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
28063 * common/config/avr/avr-common.c (avr_option_optimization_table):
28064 Switch on -mmain-is-OS_task for optimizing compilations.
28066 2018-01-09 Richard Biener <rguenther@suse.de>
28068 PR tree-optimization/83572
28069 * graphite.c: Include cfganal.h.
28070 (graphite_transform_loops): Connect infinite loops to exit
28071 and remove fake edges at the end.
28073 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
28075 * ipa-inline.c (edge_badness): Revert accidental checkin.
28077 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
28080 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
28081 symbols; not inline clones.
28083 2018-01-09 Jakub Jelinek <jakub@redhat.com>
28086 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
28087 hard registers. Formatting fixes.
28089 PR preprocessor/83722
28090 * gcc.c (try_generate_repro): Pass
28091 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
28092 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
28095 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
28096 Kito Cheng <kito.cheng@gmail.com>
28098 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
28099 (riscv_leaf_function_p): Delete.
28100 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
28102 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
28104 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
28106 (do_ifelse): New function.
28107 (do_isel): New function.
28108 (do_sub3): New function.
28109 (do_add3): New function.
28110 (do_load_mask_compare): New function.
28111 (do_overlap_load_compare): New function.
28112 (expand_compare_loop): New function.
28113 (expand_block_compare): Call expand_compare_loop() when appropriate.
28114 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
28115 option description.
28116 (-mblock-compare-inline-loop-limit): New option.
28118 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
28121 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
28122 Reverse order of second and third operands in first alternative.
28123 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
28124 of first and second elements in UNSPEC_VPERMR vector.
28125 (altivec_expand_vec_perm_le): Likewise.
28127 2018-01-08 Jeff Law <law@redhat.com>
28129 PR rtl-optimizatin/81308
28130 * tree-switch-conversion.c (cfg_altered): New file scoped static.
28131 (process_switch): If group_case_labels makes a change, then set
28133 (pass_convert_switch::execute): If a switch is converted, then
28134 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
28136 PR rtl-optimization/81308
28137 * recog.c (split_all_insns): Conditionally cleanup the CFG after
28140 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
28142 PR target/83663 - Revert r255946
28143 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
28144 generation for cases where splatting a value is not useful.
28145 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
28146 across a vec_duplicate and a paradoxical subreg forming a vector
28147 mode to a vec_concat.
28149 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28151 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
28152 -march=armv8.3-a variants.
28153 * config/arm/t-multilib: Likewise.
28154 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
28156 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
28158 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
28160 (cceq_ior_compare_complement): Give it a name so I can use it, and
28161 change boolean_or_operator predicate to boolean_operator so it can
28162 be used to generate a crand.
28163 (eqne): New code iterator.
28164 (bd/bd_neg): New code_attrs.
28165 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
28166 a single define_insn.
28167 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
28168 decrement (bdnzt/bdnzf/bdzt/bdzf).
28169 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
28170 with the new names of the branch decrement patterns, and added the
28171 names of the branch decrement conditional patterns.
28173 2018-01-08 Richard Biener <rguenther@suse.de>
28175 PR tree-optimization/83563
28176 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
28179 2018-01-08 Richard Biener <rguenther@suse.de>
28181 PR middle-end/83713
28182 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
28184 2018-01-08 Richard Biener <rguenther@suse.de>
28186 PR tree-optimization/83685
28187 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
28188 references to abnormals.
28190 2018-01-08 Richard Biener <rguenther@suse.de>
28193 * dwarf2out.c (output_indirect_strings): Handle empty
28194 skeleton_debug_str_hash.
28195 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
28197 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
28199 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
28200 (emit_store_direct): Likewise.
28201 (arc_trampoline_adjust_address): Likewise.
28202 (arc_asm_trampoline_template): New function.
28203 (arc_initialize_trampoline): Use asm_trampoline_template.
28204 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
28205 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
28206 * config/arc/arc.md (flush_icache): Delete pattern.
28208 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
28210 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
28211 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
28214 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
28217 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
28218 by not USED_FOR_TARGET.
28219 (make_pass_resolve_sw_modes): Likewise.
28221 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
28223 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
28226 2018-01-08 Richard Biener <rguenther@suse.de>
28228 PR middle-end/83580
28229 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
28231 2018-01-08 Richard Biener <rguenther@suse.de>
28233 PR middle-end/83517
28234 * match.pd ((t * 2) / 2) -> t): Add missing :c.
28236 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
28238 PR middle-end/81897
28239 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
28240 basic blocks with a small number of successors.
28241 (convert_control_dep_chain_into_preds): Improve handling of
28243 (dump_predicates): Split apart into...
28244 (dump_pred_chain): ...here...
28245 (dump_pred_info): ...and here.
28246 (can_one_predicate_be_invalidated_p): Add debugging printfs.
28247 (can_chain_union_be_invalidated_p): Improve check for invalidation
28249 (uninit_uses_cannot_happen): Avoid unnecessary if
28250 convert_control_dep_chain_into_preds yielded nothing.
28252 2018-01-06 Martin Sebor <msebor@redhat.com>
28254 PR tree-optimization/83640
28255 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
28256 subtracting negative offset from size.
28257 (builtin_access::overlap): Adjust offset bounds of the access to fall
28258 within the size of the object if possible.
28260 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
28262 PR rtl-optimization/83699
28263 * expmed.c (extract_bit_field_1): Restrict the vector usage of
28264 extract_bit_field_as_subreg to cases in which the extracted
28265 value is also a vector.
28267 * lra-constraints.c (process_alt_operands): Test for the equivalence
28268 substitutions when detecting a possible reload cycle.
28270 2018-01-06 Jakub Jelinek <jakub@redhat.com>
28273 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
28274 by default if flag_selective_schedling{,2}. Formatting fixes.
28276 PR rtl-optimization/83682
28277 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
28278 if it has non-VECTOR_MODE element mode.
28279 (vec_duplicate_p): Likewise.
28281 PR middle-end/83694
28282 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
28283 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
28285 2018-01-05 Jakub Jelinek <jakub@redhat.com>
28288 * config/i386/i386-builtin.def
28289 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
28290 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
28291 Require also OPTION_MASK_ISA_AVX512F in addition to
28292 OPTION_MASK_ISA_GFNI.
28293 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
28294 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
28295 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
28296 to OPTION_MASK_ISA_GFNI.
28297 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
28298 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
28299 OPTION_MASK_ISA_AVX512BW.
28300 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
28301 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
28302 addition to OPTION_MASK_ISA_GFNI.
28303 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
28304 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
28305 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
28306 to OPTION_MASK_ISA_GFNI.
28307 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
28308 a requirement for all ISAs rather than any of them with a few
28310 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
28312 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
28313 bitmasks to be enabled with 3 exceptions, instead of requiring any
28314 enabled ISA with lots of exceptions.
28315 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
28316 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
28317 Change avx512bw in isa attribute to avx512f.
28318 * config/i386/sgxintrin.h: Add license boilerplate.
28319 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
28320 to __AVX512F__ and __AVX512VL to __AVX512VL__.
28321 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
28322 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
28324 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
28325 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
28326 temporarily sse2 rather than sse if not enabled already.
28329 * config/i386/sse.md (VI248_VLBW): Rename to ...
28330 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
28331 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
28332 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
28333 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
28334 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
28335 mode iterator instead of VI248_VLBW.
28337 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
28339 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
28340 (record_modified): Skip clobbers; add debug output.
28341 (param_change_prob): Use sreal frequencies.
28343 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
28345 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
28346 punt for user-aligned variables.
28348 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
28350 * tree-chrec.c (chrec_contains_symbols): Return true for
28353 2018-01-05 Sudakshina Das <sudi.das@arm.com>
28356 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
28357 of (x|y) == x for BICS pattern.
28359 2018-01-05 Jakub Jelinek <jakub@redhat.com>
28361 PR tree-optimization/83605
28362 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
28363 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
28366 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
28368 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
28369 * config/epiphany/rtems.h: New file.
28371 2018-01-04 Jakub Jelinek <jakub@redhat.com>
28372 Uros Bizjak <ubizjak@gmail.com>
28375 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
28376 QIreg_operand instead of register_operand predicate.
28377 * config/i386/i386.c (ix86_rop_should_change_byte_p,
28378 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
28379 comments instead of -fmitigate[-_]rop.
28381 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
28384 * cgraphunit.c (symbol_table::compile): Switch to text_section
28385 before calling assembly_start debug hook.
28386 * run-rtl-passes.c (run_rtl_passes): Likewise.
28389 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
28391 * tree-vrp.c (extract_range_from_binary_expr_1): Check
28392 range_int_cst_p rather than !symbolic_range_p before calling
28393 extract_range_from_multiplicative_op_1.
28395 2018-01-04 Jeff Law <law@redhat.com>
28397 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
28398 redundant test in assertion.
28400 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
28402 * doc/rtl.texi: Document machine_mode wrapper classes.
28404 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
28406 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
28407 using tree_to_uhwi.
28409 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
28411 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
28412 the VEC_PERM_EXPR fold to fail.
28414 2018-01-04 Jakub Jelinek <jakub@redhat.com>
28417 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
28418 to switched_sections.
28420 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
28423 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
28424 test for d.testing.
28426 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
28429 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
28430 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
28432 2018-01-04 Jakub Jelinek <jakub@redhat.com>
28435 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
28436 is BLKmode and bitpos not zero or mode change is needed.
28438 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
28441 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
28444 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
28447 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
28448 instead of MULT rtx. Update all corresponding splitters.
28449 (*saddl_se): Ditto.
28450 (*ssub<modesuffix>): Ditto.
28451 (*ssubl_se): Ditto.
28452 (*cmp_sadd_di): Update split patterns.
28453 (*cmp_sadd_si): Ditto.
28454 (*cmp_sadd_sidi): Ditto.
28455 (*cmp_ssub_di): Ditto.
28456 (*cmp_ssub_si): Ditto.
28457 (*cmp_ssub_sidi): Ditto.
28458 * config/alpha/predicates.md (const23_operand): New predicate.
28459 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
28460 Look for ASHIFT, not MULT inner operand.
28461 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
28463 2018-01-04 Martin Liska <mliska@suse.cz>
28465 PR gcov-profile/83669
28466 * gcov.c (output_intermediate_file): Add version to intermediate
28468 * doc/gcov.texi: Document new field 'version' in intermediate
28469 file format. Fix location of '-k' option of gcov command.
28471 2018-01-04 Martin Liska <mliska@suse.cz>
28474 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
28476 2018-01-04 Jakub Jelinek <jakub@redhat.com>
28478 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
28480 2018-01-03 Martin Sebor <msebor@redhat.com>
28482 PR tree-optimization/83655
28483 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
28484 checking calls with invalid arguments.
28486 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28488 * tree-vect-stmts.c (vect_get_store_rhs): New function.
28489 (vectorizable_mask_load_store): Delete.
28490 (vectorizable_call): Return false for masked loads and stores.
28491 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
28492 instead of gimple_assign_rhs1.
28493 (vectorizable_load): Handle IFN_MASK_LOAD.
28494 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
28496 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28498 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
28500 (vectorizable_mask_load_store): ...here.
28501 (vectorizable_load): ...and here.
28503 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28505 * tree-vect-stmts.c (vect_build_all_ones_mask)
28506 (vect_build_zero_merge_argument): New functions, split out from...
28507 (vectorizable_load): ...here.
28509 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28511 * tree-vect-stmts.c (vect_check_store_rhs): New function,
28513 (vectorizable_mask_load_store): ...here.
28514 (vectorizable_store): ...and here.
28516 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28518 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
28520 (vectorizable_mask_load_store): ...here.
28522 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28524 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
28525 (vect_model_store_cost): Take a vec_load_store_type instead of a
28527 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
28528 (vect_model_store_cost): Take a vec_load_store_type instead of a
28530 (vectorizable_mask_load_store): Update accordingly.
28531 (vectorizable_store): Likewise.
28532 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
28534 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28536 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
28537 IFN_MASK_LOAD calls here rather than...
28538 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
28540 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28541 Alan Hayward <alan.hayward@arm.com>
28542 David Sherwood <david.sherwood@arm.com>
28544 * expmed.c (extract_bit_field_1): For vector extracts,
28545 fall back to extract_bit_field_as_subreg if vec_extract
28548 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28549 Alan Hayward <alan.hayward@arm.com>
28550 David Sherwood <david.sherwood@arm.com>
28552 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
28553 they are variable or constant sized.
28554 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
28555 slots for constant-sized data.
28557 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28558 Alan Hayward <alan.hayward@arm.com>
28559 David Sherwood <david.sherwood@arm.com>
28561 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
28562 handling COND_EXPRs with boolean comparisons, try to find a better
28563 basis for the mask type than the boolean itself.
28565 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28567 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
28568 is calculated and how it can be overridden.
28569 * genmodes.c (max_bitsize_mode_any_mode): New variable.
28570 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
28572 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
28575 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28576 Alan Hayward <alan.hayward@arm.com>
28577 David Sherwood <david.sherwood@arm.com>
28579 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
28580 Remove the mode argument.
28581 (aarch64_simd_valid_immediate): Remove the mode and inverse
28583 * config/aarch64/iterators.md (bitsize): New iterator.
28584 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
28585 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
28586 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
28587 aarch64_simd_valid_immediate.
28588 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
28589 (aarch64_reg_or_bic_imm): Likewise.
28590 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
28591 with an insn_type enum and msl with a modifier_type enum.
28592 Replace element_width with a scalar_mode. Change the shift
28593 to unsigned int. Add constructors for scalar_float_mode and
28594 scalar_int_mode elements.
28595 (aarch64_vect_float_const_representable_p): Delete.
28596 (aarch64_can_const_movi_rtx_p)
28597 (aarch64_simd_scalar_immediate_valid_for_move)
28598 (aarch64_simd_make_constant): Update call to
28599 aarch64_simd_valid_immediate.
28600 (aarch64_advsimd_valid_immediate_hs): New function.
28601 (aarch64_advsimd_valid_immediate): Likewise.
28602 (aarch64_simd_valid_immediate): Remove mode and inverse
28603 arguments. Rewrite to use the above. Use const_vec_duplicate_p
28604 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
28605 and aarch64_float_const_representable_p on the result.
28606 (aarch64_output_simd_mov_immediate): Remove mode argument.
28607 Update call to aarch64_simd_valid_immediate and use of
28608 simd_immediate_info.
28609 (aarch64_output_scalar_simd_mov_immediate): Update call
28612 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28613 Alan Hayward <alan.hayward@arm.com>
28614 David Sherwood <david.sherwood@arm.com>
28616 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
28617 (mode_nunits): Likewise CONST_MODE_NUNITS.
28618 * machmode.def (ADJUST_NUNITS): Document.
28619 * genmodes.c (mode_data::need_nunits_adj): New field.
28620 (blank_mode): Update accordingly.
28621 (adj_nunits): New variable.
28622 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
28624 (emit_mode_size_inline): Set need_bytesize_adj for all modes
28625 listed in adj_nunits.
28626 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
28627 listed in adj_nunits. Don't emit case statements for such modes.
28628 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
28629 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
28630 nothing if adj_nunits is nonnull.
28631 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
28632 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
28633 (emit_mode_fbit): Update use of print_maybe_const_decl.
28634 (emit_move_size): Likewise. Treat the array as non-const
28636 (emit_mode_adjustments): Handle adj_nunits.
28638 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28640 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
28641 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
28642 (VECTOR_MODES): Use it.
28643 (make_vector_modes): Take the prefix as an argument.
28645 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28646 Alan Hayward <alan.hayward@arm.com>
28647 David Sherwood <david.sherwood@arm.com>
28649 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
28650 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
28651 for MODE_VECTOR_BOOL.
28652 * machmode.def (VECTOR_BOOL_MODE): Document.
28653 * genmodes.c (VECTOR_BOOL_MODE): New macro.
28654 (make_vector_bool_mode): New function.
28655 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
28657 * lto-streamer-in.c (lto_input_mode_table): Likewise.
28658 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
28660 * stor-layout.c (int_mode_for_mode): Likewise.
28661 * tree.c (build_vector_type_for_mode): Likewise.
28662 * varasm.c (output_constant_pool_2): Likewise.
28663 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
28664 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
28665 for MODE_VECTOR_BOOL.
28666 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
28667 of mode class checks.
28668 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
28669 instead of a list of mode class checks.
28670 (expand_vector_scalar_condition): Likewise.
28671 (type_for_widest_vector_mode): Handle BImode as an inner mode.
28673 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28674 Alan Hayward <alan.hayward@arm.com>
28675 David Sherwood <david.sherwood@arm.com>
28677 * machmode.h (mode_size): Change from unsigned short to
28679 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
28680 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
28681 or if measurement_type is not polynomial.
28682 (fixed_size_mode::includes_p): Check for constant-sized modes.
28683 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
28684 return a poly_uint16 rather than an unsigned short.
28685 (emit_mode_size): Change the type of mode_size from unsigned short
28686 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
28687 (emit_mode_adjustments): Cope with polynomial vector sizes.
28688 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
28690 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
28692 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
28693 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
28694 * caller-save.c (setup_save_areas): Likewise.
28695 (replace_reg_with_saved_mem): Likewise.
28696 * calls.c (emit_library_call_value_1): Likewise.
28697 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
28698 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
28699 (gen_lowpart_for_combine): Likewise.
28700 * convert.c (convert_to_integer_1): Likewise.
28701 * cse.c (equiv_constant, cse_insn): Likewise.
28702 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
28703 (cselib_subst_to_values): Likewise.
28704 * dce.c (word_dce_process_block): Likewise.
28705 * df-problems.c (df_word_lr_mark_ref): Likewise.
28706 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
28707 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
28708 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
28709 (rtl_for_decl_location): Likewise.
28710 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
28711 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
28712 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
28713 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
28714 (expand_expr_real_1): Likewise.
28715 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
28716 (pad_below): Likewise.
28717 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
28718 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
28719 * ira.c (get_subreg_tracking_sizes): Likewise.
28720 * ira-build.c (ira_create_allocno_objects): Likewise.
28721 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
28722 (ira_sort_regnos_for_alter_reg): Likewise.
28723 * ira-costs.c (record_operand_costs): Likewise.
28724 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
28725 (resolve_simple_move): Likewise.
28726 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
28727 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
28728 (lra_constraints): Likewise.
28729 (CONST_POOL_OK_P): Reject variable-sized modes.
28730 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
28731 (add_pseudo_to_slot, lra_spill): Likewise.
28732 * omp-low.c (omp_clause_aligned_alignment): Likewise.
28733 * optabs-query.c (get_best_extraction_insn): Likewise.
28734 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
28735 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
28736 (expand_mult_highpart, valid_multiword_target_p): Likewise.
28737 * recog.c (offsettable_address_addr_space_p): Likewise.
28738 * regcprop.c (maybe_mode_change): Likewise.
28739 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
28740 * regrename.c (build_def_use): Likewise.
28741 * regstat.c (dump_reg_info): Likewise.
28742 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
28743 (find_reloads, find_reloads_subreg_address): Likewise.
28744 * reload1.c (eliminate_regs_1): Likewise.
28745 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
28746 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
28747 (simplify_binary_operation_1, simplify_subreg): Likewise.
28748 * targhooks.c (default_function_arg_padding): Likewise.
28749 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
28750 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
28751 (verify_gimple_assign_ternary): Likewise.
28752 * tree-inline.c (estimate_move_cost): Likewise.
28753 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
28754 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
28755 (get_address_cost_ainc): Likewise.
28756 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
28757 (vect_supportable_dr_alignment): Likewise.
28758 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
28759 (vectorizable_reduction): Likewise.
28760 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
28761 (vectorizable_operation, vectorizable_load): Likewise.
28762 * tree.c (build_same_sized_truth_vector_type): Likewise.
28763 * valtrack.c (cleanup_auto_inc_dec): Likewise.
28764 * var-tracking.c (emit_note_insn_var_location): Likewise.
28765 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
28766 (ADDR_VEC_ALIGN): Likewise.
28768 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28769 Alan Hayward <alan.hayward@arm.com>
28770 David Sherwood <david.sherwood@arm.com>
28772 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
28774 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
28775 or if measurement_type is polynomial.
28776 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
28777 * combine.c (make_extraction): Likewise.
28778 * dse.c (find_shift_sequence): Likewise.
28779 * dwarf2out.c (mem_loc_descriptor): Likewise.
28780 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
28781 (extract_bit_field, extract_low_bits): Likewise.
28782 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
28783 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
28784 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
28785 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
28786 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
28787 * reload.c (find_reloads): Likewise.
28788 * reload1.c (alter_reg): Likewise.
28789 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
28790 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
28791 * tree-if-conv.c (predicate_mem_writes): Likewise.
28792 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
28793 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
28794 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
28795 * valtrack.c (dead_debug_insert_temp): Likewise.
28796 * varasm.c (mergeable_constant_section): Likewise.
28797 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
28799 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28800 Alan Hayward <alan.hayward@arm.com>
28801 David Sherwood <david.sherwood@arm.com>
28803 * expr.c (expand_assignment): Cope with polynomial mode sizes
28804 when assigning to a CONCAT.
28806 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28807 Alan Hayward <alan.hayward@arm.com>
28808 David Sherwood <david.sherwood@arm.com>
28810 * machmode.h (mode_precision): Change from unsigned short to
28812 (mode_to_precision): Return a poly_uint16 rather than an unsigned
28814 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
28815 or if measurement_type is not polynomial.
28816 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
28817 in which the mode is already known to be a scalar_int_mode.
28818 * genmodes.c (emit_mode_precision): Change the type of mode_precision
28819 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
28821 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
28822 for GET_MODE_PRECISION.
28823 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
28824 for GET_MODE_PRECISION.
28825 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
28827 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
28828 (expand_field_assignment, make_extraction): Likewise.
28829 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
28830 (get_last_value): Likewise.
28831 * convert.c (convert_to_integer_1): Likewise.
28832 * cse.c (cse_insn): Likewise.
28833 * expr.c (expand_expr_real_1): Likewise.
28834 * lra-constraints.c (simplify_operand_subreg): Likewise.
28835 * optabs-query.c (can_atomic_load_p): Likewise.
28836 * optabs.c (expand_atomic_load): Likewise.
28837 (expand_atomic_store): Likewise.
28838 * ree.c (combine_reaching_defs): Likewise.
28839 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
28840 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
28841 * tree.h (type_has_mode_precision_p): Likewise.
28842 * ubsan.c (instrument_si_overflow): Likewise.
28844 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28845 Alan Hayward <alan.hayward@arm.com>
28846 David Sherwood <david.sherwood@arm.com>
28848 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
28849 polynomial numbers of units.
28850 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
28851 (valid_vector_subparts_p): New function.
28852 (build_vector_type): Remove temporary shim and take the number
28853 of units as a poly_uint64 rather than an int.
28854 (build_opaque_vector_type): Take the number of units as a
28855 poly_uint64 rather than an int.
28856 * tree.c (build_vector_from_ctor): Handle polynomial
28857 TYPE_VECTOR_SUBPARTS.
28858 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
28859 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
28860 (build_vector_from_val): If the number of units is variable,
28861 use build_vec_duplicate_cst for constant operands and
28862 VEC_DUPLICATE_EXPR otherwise.
28863 (make_vector_type): Remove temporary is_constant ().
28864 (build_vector_type, build_opaque_vector_type): Take the number of
28865 units as a poly_uint64 rather than an int.
28866 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
28868 * cfgexpand.c (expand_debug_expr): Likewise.
28869 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
28870 (store_constructor, expand_expr_real_1): Likewise.
28871 (const_scalar_mask_from_tree): Likewise.
28872 * fold-const-call.c (fold_const_reduction): Likewise.
28873 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
28874 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
28875 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
28876 (fold_relational_const): Likewise.
28877 (native_interpret_vector): Likewise. Change the size from an
28878 int to an unsigned int.
28879 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
28880 TYPE_VECTOR_SUBPARTS.
28881 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
28882 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
28883 duplicating a non-constant operand into a variable-length vector.
28884 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
28885 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
28886 * ipa-icf.c (sem_variable::equals): Likewise.
28887 * match.pd: Likewise.
28888 * omp-simd-clone.c (simd_clone_subparts): Likewise.
28889 * print-tree.c (print_node): Likewise.
28890 * stor-layout.c (layout_type): Likewise.
28891 * targhooks.c (default_builtin_vectorization_cost): Likewise.
28892 * tree-cfg.c (verify_gimple_comparison): Likewise.
28893 (verify_gimple_assign_binary): Likewise.
28894 (verify_gimple_assign_ternary): Likewise.
28895 (verify_gimple_assign_single): Likewise.
28896 * tree-pretty-print.c (dump_generic_node): Likewise.
28897 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
28898 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
28899 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
28900 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
28901 (vect_shift_permute_load_chain): Likewise.
28902 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
28903 (expand_vector_condition, optimize_vector_constructor): Likewise.
28904 (lower_vec_perm, get_compute_type): Likewise.
28905 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
28906 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
28907 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
28908 (vect_recog_mask_conversion_pattern): Likewise.
28909 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
28910 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
28911 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
28912 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
28913 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
28914 (vectorizable_shift, vectorizable_operation, vectorizable_store)
28915 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
28916 (supportable_widening_operation): Likewise.
28917 (supportable_narrowing_operation): Likewise.
28918 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
28920 * varasm.c (output_constant): Likewise.
28922 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28923 Alan Hayward <alan.hayward@arm.com>
28924 David Sherwood <david.sherwood@arm.com>
28926 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
28927 so that both the length == 3 and length != 3 cases set up their
28928 own permute vectors. Add comments explaining why we know the
28929 number of elements is constant.
28930 (vect_permute_load_chain): Likewise.
28932 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28933 Alan Hayward <alan.hayward@arm.com>
28934 David Sherwood <david.sherwood@arm.com>
28936 * machmode.h (mode_nunits): Change from unsigned char to
28938 (ONLY_FIXED_SIZE_MODES): New macro.
28939 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
28940 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
28941 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
28943 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
28944 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
28945 or if measurement_type is not polynomial.
28946 * genmodes.c (ZERO_COEFFS): New macro.
28947 (emit_mode_nunits_inline): Make mode_nunits_inline return a
28949 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
28950 Use ZERO_COEFFS when emitting initializers.
28951 * data-streamer.h (bp_pack_poly_value): New function.
28952 (bp_unpack_poly_value): Likewise.
28953 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
28954 for GET_MODE_NUNITS.
28955 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
28956 for GET_MODE_NUNITS.
28957 * tree.c (make_vector_type): Remove temporary shim and make
28958 the real function take the number of units as a poly_uint64
28959 rather than an int.
28960 (build_vector_type_for_mode): Handle polynomial nunits.
28961 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
28962 * emit-rtl.c (const_vec_series_p_1): Likewise.
28963 (gen_rtx_CONST_VECTOR): Likewise.
28964 * fold-const.c (test_vec_duplicate_folding): Likewise.
28965 * genrecog.c (validate_pattern): Likewise.
28966 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
28967 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
28968 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
28969 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
28970 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
28971 * rtlanal.c (subreg_get_info): Likewise.
28972 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
28973 (vect_grouped_load_supported): Likewise.
28974 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
28975 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
28976 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
28977 (simplify_const_unary_operation, simplify_binary_operation_1)
28978 (simplify_const_binary_operation, simplify_ternary_operation)
28979 (test_vector_ops_duplicate, test_vector_ops): Likewise.
28980 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
28981 instead of CONST_VECTOR_NUNITS.
28982 * varasm.c (output_constant_pool_2): Likewise.
28983 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
28984 explicit-encoded elements in the XVEC for variable-length vectors.
28986 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28988 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
28990 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28991 Alan Hayward <alan.hayward@arm.com>
28992 David Sherwood <david.sherwood@arm.com>
28994 * coretypes.h (fixed_size_mode): Declare.
28995 (fixed_size_mode_pod): New typedef.
28996 * builtins.h (target_builtins::x_apply_args_mode)
28997 (target_builtins::x_apply_result_mode): Change type to
28998 fixed_size_mode_pod.
28999 * builtins.c (apply_args_size, apply_result_size, result_vector)
29000 (expand_builtin_apply_args_1, expand_builtin_apply)
29001 (expand_builtin_return): Update accordingly.
29003 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29005 * cse.c (hash_rtx_cb): Hash only the encoded elements.
29006 * cselib.c (cselib_hash_rtx): Likewise.
29007 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
29008 CONST_VECTOR encoding.
29010 2018-01-03 Jakub Jelinek <jakub@redhat.com>
29011 Jeff Law <law@redhat.com>
29014 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
29015 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
29016 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
29017 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
29020 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
29021 explicitly probe *sp in a noreturn function if there were any callee
29022 register saves or frame pointer is needed.
29024 2018-01-03 Jakub Jelinek <jakub@redhat.com>
29027 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
29028 BLKmode for ternary, binary or unary expressions.
29031 * var-tracking.c (delete_vta_debug_insn): New inline function.
29032 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
29033 insns from get_insns () to NULL instead of each bb separately.
29034 Use delete_vta_debug_insn. No longer static.
29035 (vt_debug_insns_local, variable_tracking_main_1): Adjust
29036 delete_vta_debug_insns callers.
29037 * rtl.h (delete_vta_debug_insns): Declare.
29038 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
29039 instead of variable_tracking_main.
29041 2018-01-03 Martin Sebor <msebor@redhat.com>
29043 PR tree-optimization/83603
29044 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
29045 arguments past the endof the argument list in functions declared
29046 without a prototype.
29047 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
29048 Avoid checking when arguments are null.
29050 2018-01-03 Martin Sebor <msebor@redhat.com>
29053 * doc/extend.texi (attribute const): Fix a typo.
29054 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
29055 issuing -Wsuggest-attribute for void functions.
29057 2018-01-03 Martin Sebor <msebor@redhat.com>
29059 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
29060 offset_int::from instead of wide_int::to_shwi.
29061 (maybe_diag_overlap): Remove assertion.
29062 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
29063 * gimple-ssa-sprintf.c (format_directive): Same.
29064 (parse_directive): Same.
29065 (sprintf_dom_walker::compute_format_length): Same.
29066 (try_substitute_return_value): Same.
29068 2018-01-03 Jeff Law <law@redhat.com>
29070 PR middle-end/83654
29071 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
29072 non-constant residual for zero at runtime and avoid probing in
29073 that case. Reorganize code for trailing problem to mirror handling
29076 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
29078 PR tree-optimization/83501
29079 * tree-ssa-strlen.c (get_string_cst): New.
29080 (handle_char_store): Call get_string_cst.
29082 2018-01-03 Martin Liska <mliska@suse.cz>
29084 PR tree-optimization/83593
29085 * tree-ssa-strlen.c: Include tree-cfg.h.
29086 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
29087 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
29088 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
29090 (strlen_dom_walker::before_dom_children): Call
29091 gimple_purge_dead_eh_edges. Dump tranformation with details
29093 (strlen_dom_walker::before_dom_children): Update call by adding
29094 new argument cleanup_eh.
29095 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
29097 2018-01-03 Martin Liska <mliska@suse.cz>
29100 * cif-code.def (VARIADIC_THUNK): New enum value.
29101 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
29104 2018-01-03 Jan Beulich <jbeulich@suse.com>
29106 * sse.md (mov<mode>_internal): Tighten condition for when to use
29107 vmovdqu<ssescalarsize> for TI and OI modes.
29109 2018-01-03 Jakub Jelinek <jakub@redhat.com>
29111 Update copyright years.
29113 2018-01-03 Martin Liska <mliska@suse.cz>
29116 * ipa-visibility.c (function_and_variable_visibility): Skip
29117 functions with noipa attribure.
29119 2018-01-03 Jakub Jelinek <jakub@redhat.com>
29121 * gcc.c (process_command): Update copyright notice dates.
29122 * gcov-dump.c (print_version): Ditto.
29123 * gcov.c (print_version): Ditto.
29124 * gcov-tool.c (print_version): Ditto.
29125 * gengtype.c (create_file): Ditto.
29126 * doc/cpp.texi: Bump @copying's copyright year.
29127 * doc/cppinternals.texi: Ditto.
29128 * doc/gcc.texi: Ditto.
29129 * doc/gccint.texi: Ditto.
29130 * doc/gcov.texi: Ditto.
29131 * doc/install.texi: Ditto.
29132 * doc/invoke.texi: Ditto.
29134 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29136 * vector-builder.h (vector_builder::m_full_nelts): Change from
29137 unsigned int to poly_uint64.
29138 (vector_builder::full_nelts): Update prototype accordingly.
29139 (vector_builder::new_vector): Likewise.
29140 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
29141 (vector_builder::operator ==): Likewise.
29142 (vector_builder::finalize): Likewise.
29143 * int-vector-builder.h (int_vector_builder::int_vector_builder):
29144 Take the number of elements as a poly_uint64 rather than an
29146 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
29147 from unsigned int to poly_uint64.
29148 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
29149 (vec_perm_indices::new_vector): Likewise.
29150 (vec_perm_indices::length): Likewise.
29151 (vec_perm_indices::nelts_per_input): Likewise.
29152 (vec_perm_indices::input_nelts): Likewise.
29153 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
29154 number of elements per input as a poly_uint64 rather than an
29155 unsigned int. Use the original encoding for variable-length
29156 vectors, rather than clamping each individual element.
29157 For the second and subsequent elements in each pattern,
29158 clamp the step and base before clamping their sum.
29159 (vec_perm_indices::series_p): Handle polynomial element counts.
29160 (vec_perm_indices::all_in_range_p): Likewise.
29161 (vec_perm_indices_to_tree): Likewise.
29162 (vec_perm_indices_to_rtx): Likewise.
29163 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
29164 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
29165 (tree_vector_builder::new_binary_operation): Handle polynomial
29166 element counts. Return false if we need to know the number
29167 of elements at compile time.
29168 * fold-const.c (fold_vec_perm): Punt if the number of elements
29169 isn't known at compile time.
29171 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29173 * vec-perm-indices.h (vec_perm_builder): Change element type
29174 from HOST_WIDE_INT to poly_int64.
29175 (vec_perm_indices::element_type): Update accordingly.
29176 (vec_perm_indices::clamp): Handle polynomial element_types.
29177 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
29178 (vec_perm_indices::all_in_range_p): Likewise.
29179 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
29181 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
29182 polynomial vec_perm_indices element types.
29183 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
29184 * fold-const.c (fold_vec_perm): Likewise.
29185 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
29186 * tree-vect-generic.c (lower_vec_perm): Likewise.
29187 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
29188 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
29189 element type to HOST_WIDE_INT.
29191 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29192 Alan Hayward <alan.hayward@arm.com>
29193 David Sherwood <david.sherwood@arm.com>
29195 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
29196 rather than an int. Use plus_constant.
29197 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
29198 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
29200 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29201 Alan Hayward <alan.hayward@arm.com>
29202 David Sherwood <david.sherwood@arm.com>
29204 * calls.c (emit_call_1, expand_call): Change struct_value_size from
29205 a HOST_WIDE_INT to a poly_int64.
29207 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29208 Alan Hayward <alan.hayward@arm.com>
29209 David Sherwood <david.sherwood@arm.com>
29211 * calls.c (load_register_parameters): Cope with polynomial
29212 mode sizes. Require a constant size for BLKmode parameters
29213 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
29214 forces a parameter to be padded at the lsb end in order to
29215 fill a complete number of words, require the parameter size
29216 to be ordered wrt UNITS_PER_WORD.
29218 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29219 Alan Hayward <alan.hayward@arm.com>
29220 David Sherwood <david.sherwood@arm.com>
29222 * reload1.c (spill_stack_slot_width): Change element type
29223 from unsigned int to poly_uint64_pod.
29224 (alter_reg): Treat mode sizes as polynomial.
29226 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29227 Alan Hayward <alan.hayward@arm.com>
29228 David Sherwood <david.sherwood@arm.com>
29230 * reload.c (complex_word_subreg_p): New function.
29231 (reload_inner_reg_of_subreg, push_reload): Use it.
29233 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29234 Alan Hayward <alan.hayward@arm.com>
29235 David Sherwood <david.sherwood@arm.com>
29237 * lra-constraints.c (process_alt_operands): Reject matched
29238 operands whose sizes aren't ordered.
29239 (match_reload): Refer to this check here.
29241 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29242 Alan Hayward <alan.hayward@arm.com>
29243 David Sherwood <david.sherwood@arm.com>
29245 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
29246 that the mode size is in the set {1, 2, 4, 8, 16}.
29248 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29249 Alan Hayward <alan.hayward@arm.com>
29250 David Sherwood <david.sherwood@arm.com>
29252 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
29253 Use plus_constant instead of gen_rtx_PLUS.
29255 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29256 Alan Hayward <alan.hayward@arm.com>
29257 David Sherwood <david.sherwood@arm.com>
29259 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
29260 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
29261 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
29262 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
29263 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
29264 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
29265 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
29266 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
29267 * config/i386/i386.c (ix86_push_rounding): ...this new function.
29268 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
29270 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
29271 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
29272 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
29273 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
29274 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
29275 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
29276 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
29277 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
29278 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
29279 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
29281 * expr.c (emit_move_resolve_push): Treat the input and result
29282 of PUSH_ROUNDING as a poly_int64.
29283 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
29284 (emit_push_insn): Likewise.
29285 * lra-eliminations.c (mark_not_eliminable): Likewise.
29286 * recog.c (push_operand): Likewise.
29287 * reload1.c (elimination_effects): Likewise.
29288 * rtlanal.c (nonzero_bits1): Likewise.
29289 * calls.c (store_one_arg): Likewise. Require the padding to be
29290 known at compile time.
29292 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29293 Alan Hayward <alan.hayward@arm.com>
29294 David Sherwood <david.sherwood@arm.com>
29296 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
29297 Use plus_constant instead of gen_rtx_PLUS.
29299 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29300 Alan Hayward <alan.hayward@arm.com>
29301 David Sherwood <david.sherwood@arm.com>
29303 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
29304 rather than an int.
29306 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29307 Alan Hayward <alan.hayward@arm.com>
29308 David Sherwood <david.sherwood@arm.com>
29310 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
29311 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
29312 via stack temporaries. Treat the mode size as polynomial too.
29314 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29315 Alan Hayward <alan.hayward@arm.com>
29316 David Sherwood <david.sherwood@arm.com>
29318 * expr.c (expand_expr_real_2): When handling conversions involving
29319 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
29320 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
29321 as a poly_uint64 too.
29323 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29324 Alan Hayward <alan.hayward@arm.com>
29325 David Sherwood <david.sherwood@arm.com>
29327 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
29329 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29330 Alan Hayward <alan.hayward@arm.com>
29331 David Sherwood <david.sherwood@arm.com>
29333 * combine.c (can_change_dest_mode): Handle polynomial
29334 REGMODE_NATURAL_SIZE.
29335 * expmed.c (store_bit_field_1): Likewise.
29336 * expr.c (store_constructor): Likewise.
29337 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
29338 and polynomial REGMODE_NATURAL_SIZE.
29339 (gen_lowpart_common): Likewise.
29340 * reginfo.c (record_subregs_of_mode): Likewise.
29341 * rtlanal.c (read_modify_subreg_p): Likewise.
29343 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29344 Alan Hayward <alan.hayward@arm.com>
29345 David Sherwood <david.sherwood@arm.com>
29347 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
29348 numbers of elements.
29350 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29351 Alan Hayward <alan.hayward@arm.com>
29352 David Sherwood <david.sherwood@arm.com>
29354 * match.pd: Cope with polynomial numbers of vector elements.
29356 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29357 Alan Hayward <alan.hayward@arm.com>
29358 David Sherwood <david.sherwood@arm.com>
29360 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
29361 in a POINTER_PLUS_EXPR.
29363 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29364 Alan Hayward <alan.hayward@arm.com>
29365 David Sherwood <david.sherwood@arm.com>
29367 * omp-simd-clone.c (simd_clone_subparts): New function.
29368 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
29369 (ipa_simd_modify_function_body): Likewise.
29371 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29372 Alan Hayward <alan.hayward@arm.com>
29373 David Sherwood <david.sherwood@arm.com>
29375 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
29376 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
29377 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
29378 (expand_vector_condition, vector_element): Likewise.
29379 (subparts_gt): New function.
29380 (get_compute_type): Use subparts_gt.
29381 (count_type_subparts): Delete.
29382 (expand_vector_operations_1): Use subparts_gt instead of
29383 count_type_subparts.
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 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
29390 (vect_compile_time_alias): ...this new function. Do the calculation
29391 on poly_ints rather than trees.
29392 (vect_prune_runtime_alias_test_list): Update call accordingly.
29394 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29395 Alan Hayward <alan.hayward@arm.com>
29396 David Sherwood <david.sherwood@arm.com>
29398 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
29400 (vect_schedule_slp_instance): Likewise.
29402 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29403 Alan Hayward <alan.hayward@arm.com>
29404 David Sherwood <david.sherwood@arm.com>
29406 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
29407 constant and extern definitions for variable-length vectors.
29408 (vect_get_constant_vectors): Note that the number of units
29409 is known to be constant.
29411 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29412 Alan Hayward <alan.hayward@arm.com>
29413 David Sherwood <david.sherwood@arm.com>
29415 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
29416 of units as polynomial. Choose between WIDE and NARROW based
29419 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29420 Alan Hayward <alan.hayward@arm.com>
29421 David Sherwood <david.sherwood@arm.com>
29423 * tree-vect-stmts.c (simd_clone_subparts): New function.
29424 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
29426 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29427 Alan Hayward <alan.hayward@arm.com>
29428 David Sherwood <david.sherwood@arm.com>
29430 * tree-vect-stmts.c (vectorizable_call): Treat the number of
29431 vectors as polynomial. Use build_index_vector for
29432 IFN_GOMP_SIMD_LANE.
29434 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29435 Alan Hayward <alan.hayward@arm.com>
29436 David Sherwood <david.sherwood@arm.com>
29438 * tree-vect-stmts.c (get_load_store_type): Treat the number of
29439 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
29440 for variable-length vectors.
29441 (vectorizable_mask_load_store): Treat the number of units as
29442 polynomial, asserting that it is constant if the condition has
29443 already been enforced.
29444 (vectorizable_store, vectorizable_load): Likewise.
29446 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29447 Alan Hayward <alan.hayward@arm.com>
29448 David Sherwood <david.sherwood@arm.com>
29450 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
29451 of units as polynomial. Punt if we can't tell at compile time
29452 which vector contains the final result.
29454 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29455 Alan Hayward <alan.hayward@arm.com>
29456 David Sherwood <david.sherwood@arm.com>
29458 * tree-vect-loop.c (vectorizable_induction): Treat the number
29459 of units as polynomial. Punt on SLP inductions. Use an integer
29460 VEC_SERIES_EXPR for variable-length integer reductions. Use a
29461 cast of such a series for variable-length floating-point
29464 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29465 Alan Hayward <alan.hayward@arm.com>
29466 David Sherwood <david.sherwood@arm.com>
29468 * tree.h (build_index_vector): Declare.
29469 * tree.c (build_index_vector): New function.
29470 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
29471 of units as polynomial, forcibly converting it to a constant if
29472 vectorizable_reduction has already enforced the condition.
29473 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
29474 to create a {1,2,3,...} vector.
29475 (vectorizable_reduction): Treat the number of units as polynomial.
29476 Choose vectype_in based on the largest scalar element size rather
29477 than the smallest number of units. Enforce the restrictions
29480 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29481 Alan Hayward <alan.hayward@arm.com>
29482 David Sherwood <david.sherwood@arm.com>
29484 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
29485 number of units as polynomial.
29487 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29488 Alan Hayward <alan.hayward@arm.com>
29489 David Sherwood <david.sherwood@arm.com>
29491 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
29492 * target.def (autovectorize_vector_sizes): Return the vector sizes
29493 by pointer, using vector_sizes rather than a bitmask.
29494 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
29495 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
29496 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
29498 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
29499 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
29500 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
29501 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
29502 * omp-general.c (omp_max_vf): Likewise.
29503 * omp-low.c (omp_clause_aligned_alignment): Likewise.
29504 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
29505 * tree-vect-loop.c (vect_analyze_loop): Likewise.
29506 * tree-vect-slp.c (vect_slp_bb): Likewise.
29507 * doc/tm.texi: Regenerate.
29508 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
29510 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
29511 the vector size as a poly_uint64 rather than an unsigned int.
29512 (current_vector_size): Change from an unsigned int to a poly_uint64.
29513 (get_vectype_for_scalar_type): Update accordingly.
29514 * tree.h (build_truth_vector_type): Take the size and number of
29515 units as a poly_uint64 rather than an unsigned int.
29516 (build_vector_type): Add a temporary overload that takes
29517 the number of units as a poly_uint64 rather than an unsigned int.
29518 * tree.c (make_vector_type): Likewise.
29519 (build_truth_vector_type): Take the number of units as a poly_uint64
29520 rather than an unsigned int.
29522 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29523 Alan Hayward <alan.hayward@arm.com>
29524 David Sherwood <david.sherwood@arm.com>
29526 * target.def (get_mask_mode): Take the number of units and length
29527 as poly_uint64s rather than unsigned ints.
29528 * targhooks.h (default_get_mask_mode): Update accordingly.
29529 * targhooks.c (default_get_mask_mode): Likewise.
29530 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
29531 * doc/tm.texi: Regenerate.
29533 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29534 Alan Hayward <alan.hayward@arm.com>
29535 David Sherwood <david.sherwood@arm.com>
29537 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
29538 * omp-general.c (omp_max_vf): Likewise.
29539 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
29540 (expand_omp_simd): Handle polynomial safelen.
29541 * omp-low.c (omplow_simd_context): Add a default constructor.
29542 (omplow_simd_context::max_vf): Change from int to poly_uint64.
29543 (lower_rec_simd_input_clauses): Update accordingly.
29544 (lower_rec_input_clauses): Likewise.
29546 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29547 Alan Hayward <alan.hayward@arm.com>
29548 David Sherwood <david.sherwood@arm.com>
29550 * tree-vectorizer.h (vect_nunits_for_cost): New function.
29551 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
29552 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
29553 (vect_analyze_slp_cost): Likewise.
29554 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
29555 (vect_model_load_cost): Likewise.
29557 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29558 Alan Hayward <alan.hayward@arm.com>
29559 David Sherwood <david.sherwood@arm.com>
29561 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
29562 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
29563 from an unsigned int * to a poly_uint64_pod *.
29564 (calculate_unrolling_factor): New function.
29565 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
29567 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29568 Alan Hayward <alan.hayward@arm.com>
29569 David Sherwood <david.sherwood@arm.com>
29571 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
29572 from an unsigned int to a poly_uint64.
29573 (_loop_vec_info::slp_unrolling_factor): Likewise.
29574 (_loop_vec_info::vectorization_factor): Change from an int
29576 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
29577 (vect_get_num_vectors): New function.
29578 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
29579 (vect_get_num_copies): Use vect_get_num_vectors.
29580 (vect_analyze_data_ref_dependences): Change max_vf from an int *
29581 to an unsigned int *.
29582 (vect_analyze_data_refs): Change min_vf from an int * to a
29584 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
29585 than an unsigned HOST_WIDE_INT.
29586 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
29587 (vect_analyze_data_ref_dependence): Change max_vf from an int *
29588 to an unsigned int *.
29589 (vect_analyze_data_ref_dependences): Likewise.
29590 (vect_compute_data_ref_alignment): Handle polynomial vf.
29591 (vect_enhance_data_refs_alignment): Likewise.
29592 (vect_prune_runtime_alias_test_list): Likewise.
29593 (vect_shift_permute_load_chain): Likewise.
29594 (vect_supportable_dr_alignment): Likewise.
29595 (dependence_distance_ge_vf): Take the vectorization factor as a
29596 poly_uint64 rather than an unsigned HOST_WIDE_INT.
29597 (vect_analyze_data_refs): Change min_vf from an int * to a
29599 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
29600 vfm1 as a poly_uint64 rather than an int. Make the same change
29601 for the returned bound_scalar.
29602 (vect_gen_vector_loop_niters): Handle polynomial vf.
29603 (vect_do_peeling): Likewise. Update call to
29604 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
29605 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
29607 * tree-vect-loop.c (vect_determine_vectorization_factor)
29608 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
29609 (vect_get_known_peeling_cost): Likewise.
29610 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
29611 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
29612 (vect_transform_loop): Likewise. Use the lowest possible VF when
29613 updating the upper bounds of the loop.
29614 (vect_min_worthwhile_factor): Make static. Return an unsigned int
29615 rather than an int.
29616 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
29617 polynomial unroll factors.
29618 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
29619 (vect_make_slp_decision): Likewise.
29620 (vect_supported_load_permutation_p): Likewise, and polynomial
29622 (vect_analyze_slp_cost): Handle polynomial vf.
29623 (vect_slp_analyze_node_operations): Likewise.
29624 (vect_slp_analyze_bb_1): Likewise.
29625 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
29626 than an unsigned HOST_WIDE_INT.
29627 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
29628 (vectorizable_load): Handle polynomial vf.
29629 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
29631 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
29633 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29634 Alan Hayward <alan.hayward@arm.com>
29635 David Sherwood <david.sherwood@arm.com>
29637 * match.pd: Handle bit operations involving three constants
29638 and try to fold one pair.
29640 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29642 * tree-vect-loop-manip.c: Include gimple-fold.h.
29643 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
29644 niters_maybe_zero parameters. Handle other cases besides a step of 1.
29645 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
29646 Add a path that uses a step of VF instead of 1, but disable it
29648 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
29649 and niters_no_overflow parameters. Update calls to
29650 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
29651 Create a new SSA name if the latter choses to use a ste other
29652 than zero, and return it via niters_vector_mult_vf_var.
29653 * tree-vect-loop.c (vect_transform_loop): Update calls to
29654 vect_do_peeling, vect_gen_vector_loop_niters and
29655 slpeel_make_loop_iterate_ntimes.
29656 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
29657 (vect_gen_vector_loop_niters): Update declarations after above changes.
29659 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
29661 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
29662 128-bit round to integer instructions.
29663 (ceil<mode>2): Likewise.
29664 (btrunc<mode>2): Likewise.
29665 (round<mode>2): Likewise.
29667 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
29669 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
29670 unaligned VSX load/store on P8/P9.
29671 (expand_block_clear): Allow the use of unaligned VSX
29672 load/store on P8/P9.
29674 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
29676 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
29678 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
29679 swap associated with both a load and a store.
29681 2018-01-02 Andrew Waterman <andrew@sifive.com>
29683 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
29684 * config/riscv/riscv.md (clear_cache): Use it.
29686 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
29688 * web.c: Remove out-of-date comment.
29690 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
29692 * expr.c (fixup_args_size_notes): Check that any existing
29693 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
29694 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
29695 (emit_single_push_insn): ...here.
29697 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
29699 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
29700 (const_vector_encoded_nelts): New function.
29701 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
29702 (const_vector_int_elt, const_vector_elt): Declare.
29703 * emit-rtl.c (const_vector_int_elt_1): New function.
29704 (const_vector_elt): Likewise.
29705 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
29706 of CONST_VECTOR_ELT.
29708 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
29710 * expr.c: Include rtx-vector-builder.h.
29711 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
29712 directly on the tree encoding.
29713 (const_vector_from_tree): Likewise.
29714 * optabs.c: Include rtx-vector-builder.h.
29715 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
29716 sequence of "u" values.
29717 * vec-perm-indices.c: Include rtx-vector-builder.h.
29718 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
29719 directly on the vec_perm_indices encoding.
29721 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
29723 * doc/rtl.texi (const_vector): Describe new encoding scheme.
29724 * Makefile.in (OBJS): Add rtx-vector-builder.o.
29725 * rtx-vector-builder.h: New file.
29726 * rtx-vector-builder.c: Likewise.
29727 * rtl.h (rtx_def::u2): Add a const_vector field.
29728 (CONST_VECTOR_NPATTERNS): New macro.
29729 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
29730 (CONST_VECTOR_DUPLICATE_P): Likewise.
29731 (CONST_VECTOR_STEPPED_P): Likewise.
29732 (CONST_VECTOR_ENCODED_ELT): Likewise.
29733 (const_vec_duplicate_p): Check for a duplicated vector encoding.
29734 (unwrap_const_vec_duplicate): Likewise.
29735 (const_vec_series_p): Check for a non-duplicated vector encoding.
29736 Say that the function only returns true for integer vectors.
29737 * emit-rtl.c: Include rtx-vector-builder.h.
29738 (gen_const_vec_duplicate_1): Delete.
29739 (gen_const_vector): Call gen_const_vec_duplicate instead of
29740 gen_const_vec_duplicate_1.
29741 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
29742 (gen_const_vec_duplicate): Use rtx_vector_builder.
29743 (gen_const_vec_series): Likewise.
29744 (gen_rtx_CONST_VECTOR): Likewise.
29745 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
29746 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
29747 Build a new vector rather than modifying a CONST_VECTOR in-place.
29748 (handle_special_swappables): Update call accordingly.
29749 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
29750 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
29751 Build a new vector rather than modifying a CONST_VECTOR in-place.
29752 (handle_special_swappables): Update call accordingly.
29754 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
29756 * simplify-rtx.c (simplify_const_binary_operation): Use
29757 CONST_VECTOR_ELT instead of XVECEXP.
29759 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
29761 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
29762 the selector elements to be different from the data elements
29763 if the selector is a VECTOR_CST.
29764 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
29765 ssizetype for the selector.
29767 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
29769 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
29770 before testing each element individually.
29771 * tree-vect-generic.c (lower_vec_perm): Likewise.
29773 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
29775 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
29776 * selftest-run-tests.c (selftest::run_tests): Call it.
29777 * vector-builder.h (vector_builder::operator ==): New function.
29778 (vector_builder::operator !=): Likewise.
29779 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
29780 (vec_perm_indices::all_from_input_p): New function.
29781 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
29782 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
29783 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
29784 instead of reading the VECTOR_CST directly. Detect whether both
29785 vector inputs are the same before constructing the vec_perm_indices,
29786 and update the number of inputs argument accordingly. Use the
29787 utility functions added above. Only construct sel2 if we need to.
29789 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
29791 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
29792 the broadcast of the low byte.
29793 (expand_mult_highpart): Use an explicit encoding for the permutes.
29794 * optabs-query.c (can_mult_highpart_p): Likewise.
29795 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
29796 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
29797 (vectorizable_bswap): Likewise.
29798 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
29799 explicit encoding for the power-of-2 permutes.
29800 (vect_permute_store_chain): Likewise.
29801 (vect_grouped_load_supported): Likewise.
29802 (vect_permute_load_chain): Likewise.
29804 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
29806 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
29807 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
29808 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
29809 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
29810 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
29811 (vect_gen_perm_mask_any): Likewise.
29813 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
29815 * int-vector-builder.h: New file.
29816 * vec-perm-indices.h: Include int-vector-builder.h.
29817 (vec_perm_indices): Redefine as an int_vector_builder.
29818 (auto_vec_perm_indices): Delete.
29819 (vec_perm_builder): Redefine as a stand-alone class.
29820 (vec_perm_indices::vec_perm_indices): New function.
29821 (vec_perm_indices::clamp): Likewise.
29822 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
29823 (vec_perm_indices::new_vector): New function.
29824 (vec_perm_indices::new_expanded_vector): Update for new
29825 vec_perm_indices class.
29826 (vec_perm_indices::rotate_inputs): New function.
29827 (vec_perm_indices::all_in_range_p): Operate directly on the
29828 encoded form, without computing elided elements.
29829 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
29830 encoding. Update for new vec_perm_indices class.
29831 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
29832 the given vec_perm_builder.
29833 (expand_vec_perm_var): Update vec_perm_builder constructor.
29834 (expand_mult_highpart): Use vec_perm_builder instead of
29835 auto_vec_perm_indices.
29836 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
29837 vec_perm_indices instead of auto_vec_perm_indices. Use a single
29838 or double series encoding as appropriate.
29839 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
29840 vec_perm_indices instead of auto_vec_perm_indices.
29841 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
29842 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
29843 (vect_permute_store_chain): Likewise.
29844 (vect_grouped_load_supported): Likewise.
29845 (vect_permute_load_chain): Likewise.
29846 (vect_shift_permute_load_chain): Likewise.
29847 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
29848 (vect_transform_slp_perm_load): Likewise.
29849 (vect_schedule_slp_instance): Likewise.
29850 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
29851 (vectorizable_mask_load_store): Likewise.
29852 (vectorizable_bswap): Likewise.
29853 (vectorizable_store): Likewise.
29854 (vectorizable_load): Likewise.
29855 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
29856 vec_perm_indices instead of auto_vec_perm_indices. Use
29857 tree_to_vec_perm_builder to read the vector from a tree.
29858 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
29859 vec_perm_builder instead of a vec_perm_indices.
29860 (have_whole_vector_shift): Use vec_perm_builder and
29861 vec_perm_indices instead of auto_vec_perm_indices. Leave the
29862 truncation to calc_vec_perm_mask_for_shift.
29863 (vect_create_epilog_for_reduction): Likewise.
29864 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
29865 from auto_vec_perm_indices to vec_perm_indices.
29866 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
29867 instead of changing individual elements.
29868 (aarch64_vectorize_vec_perm_const): Use new_vector to install
29869 the vector in d.perm.
29870 * config/arm/arm.c (expand_vec_perm_d::perm): Change
29871 from auto_vec_perm_indices to vec_perm_indices.
29872 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
29873 instead of changing individual elements.
29874 (arm_vectorize_vec_perm_const): Use new_vector to install
29875 the vector in d.perm.
29876 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
29877 Update vec_perm_builder constructor.
29878 (rs6000_expand_interleave): Likewise.
29879 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
29880 (rs6000_expand_interleave): Likewise.
29882 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
29884 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
29885 to qimode could truncate the indices.
29886 * optabs.c (expand_vec_perm_var): Likewise.
29888 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
29890 * Makefile.in (OBJS): Add vec-perm-indices.o.
29891 * vec-perm-indices.h: New file.
29892 * vec-perm-indices.c: Likewise.
29893 * target.h (vec_perm_indices): Replace with a forward class
29895 (auto_vec_perm_indices): Move to vec-perm-indices.h.
29896 * optabs.h: Include vec-perm-indices.h.
29897 (expand_vec_perm): Delete.
29898 (selector_fits_mode_p, expand_vec_perm_var): Declare.
29899 (expand_vec_perm_const): Declare.
29900 * target.def (vec_perm_const_ok): Replace with...
29901 (vec_perm_const): ...this new hook.
29902 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
29903 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
29904 * doc/tm.texi: Regenerate.
29905 * optabs.def (vec_perm_const): Delete.
29906 * doc/md.texi (vec_perm_const): Likewise.
29907 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
29908 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
29909 expand_vec_perm for constant permutation vectors. Assert that
29910 the mode of variable permutation vectors is the integer equivalent
29911 of the mode that is being permuted.
29912 * optabs-query.h (selector_fits_mode_p): Declare.
29913 * optabs-query.c: Include vec-perm-indices.h.
29914 (selector_fits_mode_p): New function.
29915 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
29916 is defined, instead of checking whether the vec_perm_const_optab
29917 exists. Use targetm.vectorize.vec_perm_const instead of
29918 targetm.vectorize.vec_perm_const_ok. Check whether the indices
29919 fit in the vector mode before using a variable permute.
29920 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
29921 vec_perm_indices instead of an rtx.
29922 (expand_vec_perm): Replace with...
29923 (expand_vec_perm_const): ...this new function. Take the selector
29924 as a vec_perm_indices rather than an rtx. Also take the mode of
29925 the selector. Update call to shift_amt_for_vec_perm_mask.
29926 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
29927 Use vec_perm_indices::new_expanded_vector to expand the original
29928 selector into bytes. Check whether the indices fit in the vector
29929 mode before using a variable permute.
29930 (expand_vec_perm_var): Make global.
29931 (expand_mult_highpart): Use expand_vec_perm_const.
29932 * fold-const.c: Includes vec-perm-indices.h.
29933 * tree-ssa-forwprop.c: Likewise.
29934 * tree-vect-data-refs.c: Likewise.
29935 * tree-vect-generic.c: Likewise.
29936 * tree-vect-loop.c: Likewise.
29937 * tree-vect-slp.c: Likewise.
29938 * tree-vect-stmts.c: Likewise.
29939 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
29941 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
29942 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
29943 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
29944 (aarch64_vectorize_vec_perm_const): ...this new function.
29945 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
29946 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
29947 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
29948 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
29949 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
29950 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
29951 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
29953 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
29954 check for NEON modes.
29955 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
29956 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
29957 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
29958 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
29960 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
29961 the old VEC_PERM_CONST conditions.
29962 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
29963 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
29964 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
29965 (ia64_vectorize_vec_perm_const_ok): Merge into...
29966 (ia64_vectorize_vec_perm_const): ...this new function.
29967 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
29968 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
29969 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
29970 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
29971 * config/mips/mips.c (mips_expand_vec_perm_const)
29972 (mips_vectorize_vec_perm_const_ok): Merge into...
29973 (mips_vectorize_vec_perm_const): ...this new function.
29974 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
29975 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
29976 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
29977 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
29978 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
29979 (rs6000_expand_vec_perm_const): Delete.
29980 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
29982 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
29983 (altivec_expand_vec_perm_const_le): Take each operand individually.
29984 Operate on constant selectors rather than rtxes.
29985 (altivec_expand_vec_perm_const): Likewise. Update call to
29986 altivec_expand_vec_perm_const_le.
29987 (rs6000_expand_vec_perm_const): Delete.
29988 (rs6000_vectorize_vec_perm_const_ok): Delete.
29989 (rs6000_vectorize_vec_perm_const): New function.
29990 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
29991 an element count and rtx array.
29992 (rs6000_expand_extract_even): Update call accordingly.
29993 (rs6000_expand_interleave): Likewise.
29994 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
29995 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
29996 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
29997 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
29998 (rs6000_expand_vec_perm_const): Delete.
29999 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
30000 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
30001 (altivec_expand_vec_perm_const_le): Take each operand individually.
30002 Operate on constant selectors rather than rtxes.
30003 (altivec_expand_vec_perm_const): Likewise. Update call to
30004 altivec_expand_vec_perm_const_le.
30005 (rs6000_expand_vec_perm_const): Delete.
30006 (rs6000_vectorize_vec_perm_const_ok): Delete.
30007 (rs6000_vectorize_vec_perm_const): New function. Remove stray
30008 reference to the SPE evmerge intructions.
30009 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
30010 an element count and rtx array.
30011 (rs6000_expand_extract_even): Update call accordingly.
30012 (rs6000_expand_interleave): Likewise.
30013 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
30014 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
30016 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
30018 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
30020 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
30021 vector mode and that that mode matches the mode of the data
30023 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
30024 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
30025 directly using expand_vec_perm_1 when forcing selectors into
30027 (expand_vec_perm_var): New function, split out from expand_vec_perm.
30029 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
30031 * optabs-query.h (can_vec_perm_p): Delete.
30032 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
30033 * optabs-query.c (can_vec_perm_p): Split into...
30034 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
30035 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
30036 particular selector is valid.
30037 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
30038 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
30039 (vect_grouped_load_supported): Likewise.
30040 (vect_shift_permute_load_chain): Likewise.
30041 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
30042 (vect_transform_slp_perm_load): Likewise.
30043 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
30044 (vectorizable_bswap): Likewise.
30045 (vect_gen_perm_mask_checked): Likewise.
30046 * fold-const.c (fold_ternary_loc): Likewise. Don't take
30047 implementations of variable permutation vectors into account
30048 when deciding which selector to use.
30049 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
30050 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
30051 with a false third argument.
30052 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
30053 to test whether the constant selector is valid and can_vec_perm_var_p
30054 to test whether a variable selector is valid.
30056 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
30058 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
30059 * optabs-query.c (can_vec_perm_p): Likewise.
30060 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
30061 instead of vec_perm_indices.
30062 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
30063 (vect_gen_perm_mask_checked): Likewise,
30064 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
30065 (vect_gen_perm_mask_checked): Likewise,
30067 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
30069 * optabs-query.h (qimode_for_vec_perm): Declare.
30070 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
30071 (qimode_for_vec_perm): ...this new function.
30072 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
30074 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
30076 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
30077 does not have a conditional at the top.
30079 2018-01-02 Richard Biener <rguenther@suse.de>
30081 * ipa-inline.c (big_speedup_p): Fix expression.
30083 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
30086 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
30089 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
30093 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
30094 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
30095 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
30096 cond_taken_branch_cost 3->4.
30098 2018-01-01 Jakub Jelinek <jakub@redhat.com>
30100 PR tree-optimization/83581
30101 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
30102 TODO_cleanup_cfg if any changes have been made.
30104 PR middle-end/83608
30105 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
30106 convert_modes if target mode has the right side, but different mode
30109 PR middle-end/83609
30110 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
30111 last argument when extracting from CONCAT. If either from_real or
30112 from_imag is NULL, use expansion through memory. If result is not
30113 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
30114 the parts directly to inner mode, if even that fails, use expansion
30117 PR middle-end/83623
30118 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
30119 check for bswap in mode rather than HImode and use that in expand_unop
30122 Copyright (C) 2018 Free Software Foundation, Inc.
30124 Copying and distribution of this file, with or without modification,
30125 are permitted in any medium without royalty provided the copyright
30126 notice and this notice are preserved.